Mechanism of Cu/Pd-catalyzed decarboxylative cross-couplings: A DFT investigation

Article

pubs.acs.org/JACS

Mechanism of Cu/Pd-Catalyzed Decarboxylative Cross-Couplings:

A DFT Investigation

Andreas Fromm, Christoph van Wullen, Dagmar Hackenberger, and Lukas J. Gooßen*

̈

Fachbereich Chemie and Forschungszentrum OPTIMAS, TU Kaiserslautern, Erwin-Schrodinger-Straße, 67663 Kaiserslautern,

̈

Germany

S

* Supporting Information

ABSTRACT: The reaction mechanism of decarboxylative cross-couplings of

benzoates with aryl halides to give biaryls, which is cooperatively catalyzed by

copper/palladium systems, was investigated with DFT methods. The

geometries and energies of all starting materials, products, intermediates, and

transition states of the catalytic cycle were calculated for the two model

reactions of potassium 2- and 4-ﬂuorobenzoate with bromobenzene in the

presence of a catalyst system consisting of copper(I)/1,10-phenanthroline and

the anionic monophosphine palladium complex [Pd(PMe₃)Br]⁻. Several

neutral and anionic pathways were compared, and a reasonable catalytic

cycle was identiﬁed. The key ﬁnding is that the transmetalation has a

comparably high barrier as the decarboxylation, which was previously believed

to be solely rate-determining. The electronic activation energy of the

transmetalation is rather reasonable, but the free energy loss in the initial Cu/Pd adduct formation is high. These results

suggested that research aimed at further improving the catalyst should target potentially bridging bidentate ligands likely to assist

in the formation of bimetallic intermediates. Experimental studies conﬁrm this somewhat counterintuitive prediction. With a

bidentate, potentially bridging ligand, designed to support the formation of bimetallic adducts, the reaction temperature for

decarboxylative couplings was reduced by 70 °C to only 100 °C.

Mannich-type reactions.¹⁶Oxidative decarboxylative cross-

INTRODUCTION

■

couplings have been used for Heck-type vinylations,^1f−h,17

Decarboxylative couplings have become a powerful tool for the

selective formation of carbon−carbon and carbon−heteroatom

bonds. In these transformations, CO₂is extruded from a

C−H arylations,¹⁸couplings of amino acids,¹⁹and various C−

heteroatom bond formations.²⁰

The concept of decarboxylative couplings has distinct

carboxylic acid precursor, and the carbon atom that has carried

advantages over traditional cross-coupling reactions in the

the carboxylic group then forms a bond with an electrophilic

regiospeciﬁc formation of C−C bonds, because it relies on

inexpensive and broadly available carboxylate salts instead of

expensive and sensitive preformed organometallic reagents as

the source of carbon nucleophiles.

carbon or heteroatom. Following pioneering contributions by

Nilsson, Tsuji, and Myers,¹a redox-neutral decarboxylative

biaryl synthesis was discovered in 2006 in which aromatic

carboxylic acids are coupled with various aryl bromides in the

presence of a copper/palladium bimetallic catalyst system

(Scheme 1).²

Typical catalyst systems and the reaction scope are shown in

Scheme 2.

Especially in redox-neutral couplings of aromatic carboxylate

salts, bimetallic copper/palladium systems have proven to be

the most generally applicable catalysts. For these systems, a

mechanistic outline consisting of two interlinked catalytic cycles

has been proposed (Scheme 3).^2,5The ligand environment at

the copper center is designed to promote the extrusion of CO₂

with the highest possible eﬃciency, while the palladium catalyst

is independently optimized for the cross-coupling step. The

proposed catalytic cycle starts with an anion exchange at the

copper center, in the course of which the aromatic carboxylate

is transferred to the copper complex a to give the metal

carboxylate b. In the decarboxylation step, an organocopper

Scheme 1. Decarboxylative Coupling Reactions

Since then, bimetallic systems have been used for

decarboxylative couplings of benzoates and α-imino-³and α-

oxocarboxylates⁴with aryl iodides, bromides,⁵chlorides,⁶

triﬂates,⁷tosylates,⁸and mesylates.⁹Decarboxylative couplings

with monometallic catalysts include Cu-mediated arylations of

perﬂuorinated arenecarboxylic acids,¹⁰Rh-catalyzed decarbox-

ylative 1,2-additions,¹¹Pd-catalyzed decarboxylative allyla-

tions¹²and arylations of ﬁve-ring heteroarene carboxylates,¹³

of alkynylcarboxylates,¹⁴and of oxalic acid monoesters,¹⁵and

Received: April 2, 2014

A

dx.doi.org/10.1021/ja503295x | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society

Article

a

Scheme 2. Representative Decarboxylative Cross-Couplings with Bimetallic Cu/Pd Catalyst Systems

a

Ar = aryl, FG = functional group, hetAr = heteroaryl, Phen = phenanthroline, Quin = quinoline, X = O, NR.

calculation even of simpliﬁed models. Computational studies of

this reaction have to date been limited to the decarboxylation

step (b → c in Scheme 3).²²That this step was rate-

determining in the early reaction protocols became evident

from the observation that Cu-catalyzed protodecarboxylations

required temperatures that were at least as high as those of

decarboxylative cross-couplings (170 °C). In contrast, Pd-

mediated cross-couplings of aryl−copper species are known to

proceed at much lower temperatures.

Scheme 3. Mechanism Proposed for Decarboxylative Cross-

Couplings

These DFT calculations conﬁrmed²²that the extrusion of

CO₂from phenanthroline copper benzoates is endothermic by

9.9−28.8 kcal mol⁻¹and endergonic (0.8−17.5 kcal mol⁻¹) and

has a considerable activation barrier of 27.2−36.1 kcal mol⁻¹.^21c

Another correct prediction derived from these calculations was

that for a narrow range of substitution patterns, e.g. o-

methoxybenzoates, the energy barrier for the decarboxylation is

substantially lower when copper is replaced by silver, whereas

the majority of carboxylates lose CO₂more easily when

coordinated to copper rather than silver complexes.^7c

Pd-catalyzed cross-couplings of organometallic reagents have

been intensively studied by DFT calculations.²³For Suzuki-type

couplings of carboxylic anhydrides with arylboronic acids,

various catalytic cycles were computed starting from neutral

(PMe₃)₂Pd(0) (4), anionic tricoordinate Amatore−Jutand-type

[(PMe₃)₂Pd(0)OAc]⁻(5), and anionic dicoordinate [(PMe₃)-

PdOAc]⁻(6) (Figure 1).²⁴With the anionic dicoordinate

palladium complex 6, the energy proﬁle obtained was most

favorable overall so that this species appears to be the optimal

candidate for calculations of related Pd-catalyzed couplings.

species c is generated by extrusion of CO₂. A subsequent

transmetalation links the two catalytic cycles. In this step, an

aryl residue is transferred from the copper species c onto a

palladium species e that has been generated by oxidative

addition of the aryl electrophile to a low-valent palladium(0)

cocatalyst d, and the original copper species a is regenerated.

The resulting diaryl palladium species f undergoes reductive

elimination, in which the new carbon−carbon bond is formed,

the biaryl product is released, and the palladium(0) species d is

regenerated.

At present, most decarboxylative couplings still require

temperatures in excess of 160 °C, which, in practice, represents

the main limitation of this elegant reaction type. The rational

development of catalysts that promote decarboxylative

couplings at lower temperatures is, thus, in the focus of current

research activities. In this context, mechanistic investigations

and theoretical studies are of vital importance.²¹

However, DFT modeling of the mechanistically complex

decarboxylative couplings catalyzed by bimetallic systems poses

a considerable challenge, which has so far precluded the

Figure 1. Pd-species investigated in Suzuki-type cross-couplings.

B

dx.doi.org/10.1021/ja503295x | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society

Article

Scheme 4. DFT Investigations of Decarboxylative Cross-Couplings with Monometallic Catalysts

Scheme 5. Model Reactions of Potassium 2- and 4-Fluorobenzoate with Bromobenzene

As mentioned above, some particularly activated substrates

can be decarboxylatively coupled with aryl electrophiles by

using monometallic catalysts. The reduced complexity of such

monometallic systems has permitted in-depth mechanistic

investigations (see Scheme 4). DFT studies were reported for

Pd-catalyzed decarboxylative cross-couplings of oxalic acid

monoesters,¹⁵decarboxylative allylations of enol carbo-

nates,^12,25decarboxylative cross-couplings of polyﬂuoroben-

zoates with aryl electrophiles,²⁶Myers’ Pd-catalyzed decarbox-

ylative Heck vinylation,^17aand certain intramolecular decar-

boxylative couplings.²⁷

For the mechanistically more complex decarboxylative

couplings mediated by bimetallic catalyst systems, a similarly

detailed computational study would be of substantial value.

With state-of-the-art catalysts, the protodecarboxylation of o-

nitrobenzoic acids with a copper/phenanthroline system can

meanwhile be performed at 100 °C,²⁸while decarboxylative

cross-couplings with aryl triﬂates, in which copper/phenanthro-

line systems are employed as the decarboxylation cocatalysts,

still do not proceed below 150 °C.^7a,bThus, the decarbox-

ylation is not necessarily rate-determining, and it is no longer

suﬃcient to single out the decarboxylation step in computa-

tional studies to obtain a lead for further catalyst optimization.

Instead, computational modeling should cover the entire

catalytic cycle. A particular challenge lies in modeling the

transmetalation step, in which an organic residue is transferred

from a fully ligated copper complex to a phosphine-stabilized

arylpalladium species.²⁹

MODEL SYSTEM

■

We herein investigate the decarboxylative coupling of

bromobenzene (8) with (a) potassium 2-ﬂuorobenzoate (7a)

as an example of a highly reactive arenecarboxylate and (b)

potassium 4-ﬂuorobenzoate (7b) as an example of an

arenecarboxylate with low reactivity (Scheme 5).

The catalyst system employed in the calculations of these

representative model reactions is composed of 1,10-phenan-

throline copper(I) bromide, which is known to display a

substantially greater decarboxylation activity than simple

copper(I) salts,²and an anionic Pd(0) bromide catalyst with

the simpliﬁed but still reasonably realistic ligand PMe₃.

Comparative studies performed upfront, in combination with

related literature studies, conﬁrmed that the catalytic cycle

C

dx.doi.org/10.1021/ja503295x | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society

Article

Scheme 6. Catalytic Cycles Calculated for Decarboxylative Cross-Couplings

shown in Scheme 6 is a realistic mechanistic model. For

example, computational studies of Suzuki-type couplings led to

the conclusion that catalytic cycles starting from various Pd(0)

species, each of which may be present in solution, will have

comparable energy proﬁles also in couplings with other carbon

nucleophiles.²⁴Among them, a pathway starting from

[(PMe₃)PdOAc]⁻(6) (Figure 1) had proven to be the most

advantageous overall. It thus appeared to be reasonable to use

an analogous cycle based on similar structural units as a starting

point for computing the cycle of the Pd-catalyst in the

decarboxylative cross-coupling. For the Cu-catalyst, two

diﬀerent catalytic cycles were computed, both starting from

1,10-phenanthroline copper(I) bromide. The ﬁrst solely

involves neutral species, whereas in the second, an additional

bromide is bound to copper during the decarboxylation step.

This was calculated to ﬁnd a possible explanation for the

profound inﬂuence of bromide ions on the decarboxylation of

non-ortho-substituted benzoates.

Both variants of the catalyst cycle were successfully

calculated. Minima with reasonable structures and energies

could be found for all intermediates, and transition states

connecting these intermediates could be located. The computa-

tional studies led us to conclude that decarboxylative cross-

couplings follow a neutral pathway consisting of the anion

exchange (step I) and the decarboxylation step (step II) at the

copper catalyst, the oxidative addition (step III) of the

bromobenzene to the Pd catalyst, the transmetalation process

(step IV) between the resulting palladium and copper species,

and the reductive elimination (step V) of the biaryl product

with regeneration of the Pd catalyst. Subsequently, the

computational study of an anionic variant of the Cu-cycle is

presented, which starts with the coordination (step VI) of the

carboxylate to the copper catalyst, followed by decarboxylation

(step VII) and bromide decoordination (step VIII). This

alternative cycle was found to be less favorable.

COMPUTATIONAL METHOD

■

All calculations were performed with use of the Gaussian 03 or

Gaussian 09 software packages.³⁰The B3LYP exchange-correlation

functional³¹was used in all cases. For the geometry optimizations, the

polarized split-valence 6-31+G(d) basis set³²was employed for all

atoms except Cu and Pd, where a scalar-relativistic eﬀective core

potential³³replaced 10 (Cu) or 28 (Pd) core electrons, together with

the valence basis sets³³in double-ζ quality. Spherical basis functions

(5D, 7F) were used in all cases. In contrast to our previous

studies,^21c,22some of the species investigated are anionic, so that

diﬀuse basis functions were included in the basis set. Note that the

valence basis sets for Cu and Pd already contain diﬀuse functions from

the outset.

All geometries of minima (intermediates) and transition states were

fully optimized for isolated (gas-phase) molecules. To locate the

transition states, we ﬁrst performed a relaxed potential energy surface

scan in which the reaction coordinate was kept ﬁxed at several deﬁned

distances, while all other degrees of freedom were optimized. In these

scans, a series of structures were optimized in which the reaction

coordinate was increased stepwise. The structure with the highest

energy was then used as a starting point for the synchronous transit-

guided quasi-Newton method³⁴to locate the transition state with a

molecular Hessian of exactly one negative eigenvalue. To identify the

minima connected to the transition states thus identiﬁed, the intrinsic

reaction coordinate (IRC)³⁵was followed downhill for 10 points in

both directions and the two resulting molecular geometries were used

as starting points for subsequent geometry optimizations. Sometimes,

additional intermediates were discovered this way. In these cases, the

process described above was repeated until the reaction pathway was

complete. For each molecular structure (intermediate or transition

state), the calculated electronic energy (but not the molecular

structure and harmonic vibrational frequencies) was improved by a

single-point calculation in which the 6-31+G(d) basis (for all atoms

except Cu and Pd) was replaced by the better polarized valence triple-

ζ type 6-311+G(2d,p).³⁶For Cu and Pd, the d functions of the original

basis sets were decontracted slightly from [411] to [3111], and a single

set of f functions (taken from the def2-TZVP basis sets,³⁷η_f(Cu) =

2.233, η_f(Pd) = 1.24629) was added for both elements.

The above calculations provided electronic energies under vacuum

(E_tot). To model the reaction energies more closely resembling

D

dx.doi.org/10.1021/ja503295x | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society

Article

Figure 2. B3LYP/6-31+G(d) optimized structures for the anion exchange of potassium 2-ﬂuorobenzoate and phenanthroline copper(I) bromide;

hydrogens are omitted for clarity. Energy and solution Gibbs energy proﬁles overlaid for 2- and 4-ﬂuorobenzoate. Color code: C, black; H, white; Br,

brown; Cu, orange; F, turquoise; K, purple; N, blue; O, red.

experimental conditions, further terms were added, namely zero-point

vibrational energy, thermal corrections to the energy, solvation and

entropic eﬀects, as well as empirical dispersion corrections. This way,

an approximation to the Gibbs energy in solution (G_sov) was obtained.

Solvent eﬀects were included by estimating the Gibbs energy of

solvation for all intermediates and transition states using the

conductor-like polarizable continuum model³⁸(CPCM) of Gaussian

09, which is an implementation of the conductor-like screening

solvation model³⁹(COSMO). This model is particularly well-suited

for polar solvents such as N-methyl-2-pyrrolidone (NMP, ε = 32.55),

which is the most eﬀective solvent for decarboxylative cross-couplings.

To calculate the zero-point vibrational energies, the harmonic

frequencies calculated (with the small basis) were scaled by a factor of

0.9613 as suggested by Wong.⁴⁰The thermal correction to the Gibbs

energy of each component depends on its concentration in the

solution. Note that in the Gaussian program one has to specify the

concentration as a pressure value, using the ideal gas law p_i= RTn_i/V,

where p_iis the pressure, R the gas constant, T the absolute

temperature, n_ithe molar quantity, and V the reaction volume.

Typical experiments, as modeled in our calculations, involve

approximately 1 mmol of starting material and product in a reaction

volume of 2 mL, which corresponds to a pressure of 1,844,115 Pa

(18.2 atm) at a reaction temperature of 443 K. The amount of catalyst

is lower by a factor of 20, so that we used a pressure of 101,325 Pa (1

atm) for all species involving Cu and/or Pd. The experimental

solubility data of CO₂in NMP⁴¹show that under the experimental

conditions, the molar fraction of CO₂/NMP is clearly less than 0.01,

most likely about 0.001 when extrapolating temperature and pressure,

which implies that most of the CO₂leaves the NMP solution. The

resulting CO₂concentration in the reaction mixture corresponds to a

partial pressure of 101,325 Pa (1 atm) for an ideal gas. Because of the

ideal gas approximation, the molar Gibbs energy for a given

temperature T and pressure p was computed from the results for p₀

= 1 atm as G(T,p) = G(T,p₀) + RT ln(p/p₀).

considerations are relevant in this context because we have two

coupled catalytic cycles. Reducing the catalyst load (both Cu and Pd)

reduces the reaction rate of the steps involving only Cu or only Pd

linearly, while the transmetalation step is aﬀected quadratically.

Therefore, the relative rates of the transmetalation compared to the

other elementary steps depend on the catalyst load, with lower loads

disfavoring the transmetalation. This is taken into account in our

Gibbs energy proﬁles.

It is known that standard DFT methods neglect or severely

underestimate London dispersion interactions.⁴²A robust and

numerically inexpensive way to cure this deﬁciency is to add an

empirical dispersion correction that sums up the contribution of all

atom pairs. We used Grimme’s D3 parametrization⁴³and calculated

the dispersion correction for all minima and transition states. These

values are included in our ﬁnal solution Gibbs energies G_sov. This

correction signiﬁcantly lowers the Gibbs energy of an encounter

complex with respect to the two fragments, but also lowers the Gibbs

energy of a tight transition state with respect to a loose intermediate.

Note that the solvation model as used in our calculations does not

include nonelectrostatic solute−solvent interactions: in a numerical

experiment, we found zero contributions when we set the dielectric

constant to zero. This means that the loss of solute−solvent dispersion

interaction upon formation of an adduct, which is a consequence of

reducing the surface of the solute(s), is not taken into account.

Therefore, we think that the dispersion interaction in solution is

somewhat overestimated in our calcualtions, although the combination

of empirical dispersion and solvent corrections has been reported to

work well.⁴⁴

In the Supporting Information, the electronic energies (with and

without solvation), the calculated Gibbs energies at 298 and 443 K (at

1 and 18.2 atm), the empirical dispersion corrections, and the ﬁnal

solution Gibbs energy including empirical dispersion corrections G_sov

are documented. The diagrams in this paper contain the electronic

energy E_tot(since molecular geometries and vibrational frequencies are

obtained at this level) and solution Gibbs energy G_sovproﬁles

including empirical dispersion corrections (which govern the

reactivity). Throughout the text, all bond lengths are given in Å and

For any sequence of elementary steps that involve a single species,

the reaction proﬁle does not depend on the concentration. This

changes if association and dissociation steps are involved. These

E

dx.doi.org/10.1021/ja503295x | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society

Article

relative energies are expressed in kcal mol⁻¹. All ball-and-stick models

are rendered with GaussView 5.⁴⁵

bond lengths remain almost unchanged and the K−O bond is

slightly elongated (2.74 Å).

For the o-ﬂuoro model system, intermediate 21 ﬁnally

dissociates into KBr (10) and the product of the anion

exchange, the phenanthroline copper 2-ﬂuorobenzoate (1a)

that decarboxylates subsequently. Cleaving the K−O bond

requires some energy (Δ_rE_tot= 25.3 kcal mol⁻¹), but the

entropically favorable dissociation into two solvated molecules

makes this step only slightly endergonic (Δ_rG_sov= 0.5 kcal

mol⁻¹). The copper atom in 1a is coordinated to only one of

the carboxylate oxygens by a short bond (Cu−O: 1.89 Å) in a

distorted trigonal planar environment. The nitrogen donor

atoms of the phenanthroline ligand transfer suﬃcient electron

density to the copper(I) cation so that this coordination mode

is preferred, while for the electron-poor potassium cation, a

coordination to both oxygen atoms is favorable (structure 7).

The calculations further revealed that for 4-ﬂuorobenzoate,

the energy proﬁle of the anion exchange is almost identical with

that of the 2-ﬂuorobenzoate (Figure 2, see also the Supporting

Information). This result came as a surprise since decarbox-

ylative cross-couplings of aryl halides with 2-ﬂuorobenzoates

are well-documented, while the analogous reactions of 4-

ﬂuorobenzoates have not yet been achieved. Only aryl

electrophiles with noncoordinating groups such as triﬂates

can be coupled with non-ortho-substituted benzoates. To

rationalize these experimental ﬁndings, the salt metathesis of

copper(I) halides with potassium benzoates had been

postulated to be favorable for ortho-substituted derivatives,

but unfavorable for non-ortho-substituted benzoates. This

hypothesis was supported by protodecarboxylation experiments

in which benzoic acids (22) were converted to the

corresponding arenes (23) in the presence of a phenanthroline

copper system according to Scheme 7. In the absence of

MODELING THE CATALYTIC CYCLE BY DFT

■

Anion Exchange. In the anion exchange step, the

potassium benzoate (7a/b) reacts with the phenanthroline

copper bromide catalyst 11 to aﬀord the corresponding

phenanthroline copper benzoate (1a/b) along with potassium

bromide (10). The energy and solution Gibbs energy proﬁles

for the model substrates, 2- and 4-ﬂuorobenzoate are depicted

in Figure 2, along with the optimized structures for the o-ﬂuoro

derivates.

The anion exchange starts with the formation of an

encounter complex 19 between the substrates 7 and 11. This

process is exothermic in the gas phase (Δ_rE_tot= −20.6 kcal

mol⁻¹) but endergonic by Δ_rG_sov= 4.2 kcal mol⁻¹. This

diﬀerence has two origins. First, entropy is lost when forming

the adduct, even more so because the copper catalyst 11 is

present only in low concentration compared to the substrate 7.

This entropy loss strongly contributes to the increase in Gibbs

energy at the elevated reaction temperature (170 °C). Second,

solvation disfavors adduct formation, since the combined

solvation energies for 7 and 11 are larger than that for the

adduct 19. In the adduct, the C(2)−H bond of the 1,10-

phenanthroline points toward one of the carboxylate oxygens,

resulting in a short C−H···O (2.03 Å) and a long K−Br

distance (3.26 Å). A Cu−O bond has not formed at this stage.

The K−O bond is slightly elongated to 2.64 Å compared to

2.55 Å in 7. Also, the Cu−Br bond is slightly increased to 2.30

Å compared to 2.26 Å in 11.

Bringing a carboxylic oxygen closer to the copper center

leads to the transition state [19−20]^‡. This process has a very

small barrier (Δ^‡E_tot= 3.8 kcal mol⁻¹). The coordination

sphere of the copper is distorted trigonal planar, with the

bromine atom only slightly out of the plane formed by the

phenanthroline backbone. The atoms Cu, Br, K, and O form an

almost planar 4-membered ring. The K−Br (3.16 Å) and Cu−

O (ortho: 2.77 Å) bonds are long, and the K−O (2.64 Å) and

Cu−Br bonds (2.33 Å) are short. The forward motion along

the reaction coordinate derived from the normal mode of the

imaginary frequency (21 i cm⁻¹) is dominated by the

shortening of the Cu−O distance. At the same time, the

bromine atom moves out of the plane of the phenanthroline

ligand.

Scheme 7. Inﬂuence of Halides on Protodecarboxylations

In intermediate 20, the copper atom is now in an almost

tetrahedral environment. The K−Br (3.11 Å) and Cu−O (2.15

Å) bonds are further shortened and the K−O (2.66 Å) and

Cu−Br bonds (2.41 Å) are elongated.

halides, this reaction proceeded well both for ortho- and non-

ortho-substituted benzoic acids. Upon the addition of a halide

salt, the decarboxylation of non-ortho-substituted benzoic acids

was fully suppressed, whereas the reactivity of the ortho-

substituted benzoic acids remained high.⁵

In a series of control experiments, similar observations were

made also for the decarboxylation of the two ﬂuorobenzoic acid

isomers. The addition of 1 equiv of potassium bromide only

slightly reduced the yield of the protodecarboxylation of 2-

ﬂuorobenzoic acid (22a) from 91% to 72% yield in the

presence of potassium bromide. 4-Fluorobenzoic acid (22b)

gave only 27% even in the absence of bromide ions, and when

potassium bromide was added, the decarboxylation was

completely suppressed.⁴⁶

In the second phase of the anion exchange, a bromide leaves

the copper center and eventually forms KBr. Stretching the

Cu−Br bond leads to the transition state [20−21]^‡. The

activation barrier is rather low (Δ^‡E_tot= 5.2 kcal mol⁻¹, Δ^‡G_sov

= 4.9 kcal mol⁻¹). The copper is in a distorted trigonal planar

environment with the oxygen atom only slightly out of the

plane of the phenanthroline backbone. The K−Br (3.04 Å) and

Cu−O (1.91 Å) bonds are rather short, whereas the K−O (2.71

Å) and Cu−Br bonds are long (3.50 Å). The forward motion

along the reaction coordinate derived from the normal mode of

the imaginary frequency (21 i cm⁻¹) is dominated by the

elongation of the Cu−Br bond and the motion of the oxygen

atom into the plane of the phenanthroline ligand.

According to our present DFT studies, the salt exchange

between the copper bromide species 11 and the ﬂuorobenzoate

isomers (7a and 7b) is energetically very similar. It is slightly

endergonic (ortho: Δ_rG_sov= 5.3 kcal mol⁻¹; para: Δ_rG_sov= 5.0

In the subsequent intermediate 21, the Cu−Br bond is now

fully broken while the K−Br (3.04 Å) and Cu−O (1.91 Å)

F

dx.doi.org/10.1021/ja503295x | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society

Article

Figure 3. B3LYP/6-31+G(d) optimized structures for the decarboxylation of phenanthroline copper 2-ﬂuorobenzoate; hydrogens are omitted for

clarity. Energy and solution Gibbs energy proﬁles overlaid for 2- and 4-ﬂuorobenzoate. Color code: C, black; Cu, orange; F, turquoise; N, blue; O,

red.

kcal mol⁻¹) with low activation barriers. These calculations do

not provide any rationale for the hypothesis that the presence

of bromide ions suppresses the formation speciﬁcally of copper

4-ﬂuorobenzoate complexes, so that the unique reactivity of

ortho-substituted benzoates must have another reason. A

precipitation of KBr, which would not be taken into account

by the DFT calculations, is not observed in the experiment.

We therefore carefully explored alternative geometries for the

salt exchange product phenanthroline copper 2-ﬂuorobenzoate

(1a), in which the carboxylate oxygen and the o-ﬂuorine would

both coordinate to the copper, potentially resulting in a more

stable structure. This might be expected to lead to a

substantially less endergonic anion exchange step for potassium

2-ﬂuorobenzoate compared to 4-ﬂuorobenzoate. However, no

minima for complexes with chelating coordination were found.

We also investigated whether the energy proﬁle is aﬀected by

leaving out the potassium counterion in the calculations. The

resulting energy barriers are twice as high, and again, no

diﬀerences of any potential consequence were found between

2- and 4-ﬂuorobenzoates. The energy proﬁles along with the

optimized structures can be found in Figure S1 in the

Supporting Information.

Overall, the calculations do not support the previously

postulated explanation for the eﬀect of halide ions on the

decarboxylation of non-ortho-substituted carboxylates. A

possible explanation is that the additional bromide ions do

not have an eﬀect on the diﬀerence in reactivity of the ortho-

and para-substituted benzoates in the anion exchange step but

in the decarboxylation step that will be discussed in the next

section. The presence of excess bromide salt shifts the

equilibrium of the anion exchange step to the side of the

copper bromide so that less copper carboxylate is available.

This reduces the eﬃciency of the entire protodecarboxylation

process. Since the decarboxylation of ortho-substituted

benzoates requires less energy than that of non-ortho-

substituted benzoates, high conversions of ortho-substituted

benzoic acids can be achieved at temperatures where non-ortho-

substituted derivatives do not decarboxylate.

Decarboxylation. The decarboxylation step has already

been the subject of previous DFT investigations,^21c,22because it

was initially believed to be the only rate-determining step. We

have now reinvestigated it using the high-quality basis set with

diﬀuse functions detailed above, included solvent eﬀects,

dispersion interactions, and taken into account the concen-

trations and reaction temperatures.

The extrusion of carbon dioxide from phenanthroline copper

2-ﬂuorobenzoate (1) was found to proceed via a concerted

mechanism involving only one transition state ([1−2]^‡, Figure

3). An extensive search did not reveal any further intermediates,

such as π- or η²-bound arene−copper complexes.

In the copper 2-ﬂuorobenzoate 1a, all atoms are coplanar.

The two nitrogen−copper bond lengths of the phenanthroline

ligand are slightly diﬀerent (2.0 and 2.2 Å), which indicates that

this ligand geometry is not ideal for Cu^I. The decarboxylation

step can be viewed as a substitution of the carboxylate group by

the copper center in a concerted fashion. The activation barrier

for the extrusion of CO₂from 2-ﬂuorobenzoate is relatively

high (Δ^‡E_tot= 31.4 kcal mol⁻¹, Δ^‡G_sov= 23.8 kcal mol⁻¹), but

within a realistic range for a reaction that requires a

temperature of 160 °C. In the transition state [1a−2a]^‡, the

copper atom is in a distorted tetrahedral environment formed

by the two phenanthroline nitrogens, the 2-ﬂuorophenyl and

the CO₂carbons. The Cu−N bonds have similar lengths (ca.

2.1 Å). The ﬂuorophenyl moiety binds to both the copper and

the CO₂carbon via its C(1) carbon. The aryl−CO₂bond is

short (1.94 Å) while the aryl−copper bond is relatively long

(2.02 Å), which is indicative for a relatively early transition

state. The forward motion along the reaction coordinate

derived from the normal mode of the imaginary frequency (229

G

dx.doi.org/10.1021/ja503295x | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society

Article

Figure 4. B3LYP/6-31+G(d) optimized structures for the oxidative addition of bromobenzene to [Pd(PMe₃)Br]⁻; hydrogens are omitted for clarity.

Color code: C, black; Br, brown; P, yellow; Pd, green.

i cm⁻¹) is dominated by the elongation of the aryl−CO₂bond

with a concomitant contraction of the aryl−copper bond.

The IRC calculation reveals that the transition state directly

leads to the decarboxylation product 2a. The 2-ﬂuorophenyl

ring remains in the plane of the phenanthroline backbone. A

weak interaction between a ﬂuorine lone pair and the hydrogen

at C(2) of the phenanthroline scaﬀold, which are separated by a

distance of only 2.30 Å, additionally stabilizes this geometry.

The phenanthroline coordination is even more asymmetrical

than prior to decarboxylation, with Cu−N bonds of 2.0 and 2.4

Å, respectively.

due to the entropically favorable liberation of CO₂, it is only

slightly endergonic (Δ_rG_sov= 5.4 kcal mol⁻¹). The structure of

the decarboxylation product 2b diﬀers from that for the ortho-

derivative 2a in that the 4-ﬂuorophenyl ring is rotated out of

the plane of the phenanthroline backbone at an angle of 55.7°,

which may be caused by the absence of a hydrogen−ﬂuorine

interaction. Overall, the decarboxylation of non-ortho-substi-

tuted benzoates is both kinetically and thermodynamically

signiﬁcantly less favorable than the analogous reaction of ortho-

substituted benzoates, a result that is in perfect agreement with

experimental observations of protodecarboxylations.^21c

Overall, the decarboxylation of 2-ﬂuorobenzoate is moder-

ately endothermic (Δ_rE_tot= 13.4 kcal mol⁻¹). Whereas in the

previous, simpliﬁed calculations, the decarboxylation had come

out endergonic, the present, more reﬁned calculations predict it

to be exergonic (Δ_rG_sov= −2.5 kcal mol⁻¹) at the calculated

reaction temperature of T = 443.15 K. Among other eﬀects, the

entropy gained by releasing CO₂contributes more strongly to

the Gibbs energy because we take into account the higher

reaction temperature.

Oxidative Addition. The catalytic cycle of palladium starts

with the oxidative addition of the aryl halide to a Pd⁰species.

When starting from the anionic palladium monophosphine

complex [Pd(PMe₃)Br]⁻(12), this reaction step proceeds with

particular ease (Figure 4). Upon bringing the bromobenzene

(8) closer to the palladium center, an η²-π-complex (24) forms

via the transition state [12−24]^‡. The activation barrier for this

step is rather low for a process in which a bond forms between

two separately solubilized molecules (Δ^‡E_tot= 9.9 kcal mol⁻¹,

Δ^‡G_sov= 14.3 kcal mol⁻¹). This diﬀerence stems from entropy

and solvation as discussed above. In the transition state, a bond

between the palladium and the C(2) atom of bromobenzene

has already formed (Pd−C(2): 2.31 Å), and the previously

linear Br−Pd−P bond is bent at an angle of 125.4°. The

forward motion along the reaction coordinate derived from the

normal mode of the imaginary frequency (62 i cm⁻¹) is

dominated by the shortening of the Pd−C(1) and Pd−C(2)

distances with simultaneous decrease of the Br−Pd−P angle.

The IRC calculation leads to intermediate 24, in which the

palladium atom is coordinated almost symmetrically to the π-

bond between C(1) and C(2) (Pd−C(1): 2.11 Å; Pd−C(2):

2.18 Å). The C−Br bond is slightly elongated from 1.92 to 2.00

Å, and the Br−Pd−P angle has decreased to 90.8°.

The pathway for the decarboxylation of 4-ﬂuorobenzoate is

slightly diﬀerent. The activation barrier for the extrusion of

CO₂is considerably higher (Δ^‡E_tot= 35.9 kcal mol⁻¹, Δ^‡G_sov

=

27.1 kcal mol⁻¹), which is in good agreement with the

experimental ﬁndings that the protodecarboxylation of non-

ortho-substituted benzoic acids requires higher temperatures

and longer reaction times than that of ortho-substituted

derivatives.^21cIn contrast to the decarboxylation of 2-

ﬂuorobenzoate, the transition state originating from 4-

ﬂuorobenzoate [1b−2b]^‡has a long aryl−CO₂bond (2.06

Å) and a short aryl−copper bond (1.98 Å). This is indicative of

a later transition state, which suggests that the decarboxylation

is more endergonic for this isomer. The formation of 2b is

indeed strongly endothermic (Δ_rE_tot= 23.4 kcal mol⁻¹), but

H

dx.doi.org/10.1021/ja503295x | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society

Article

Figure 5. B3LYP/6-31+G(d) optimized structures for the transmetalation of the phenanthroline copper 2-ﬂuorophenyl complex (2) with

[Pd(PMe₃)(Ph)Br₂]⁻(13); hydrogens are omitted for clarity. Energy and solution Gibbs energy proﬁles overlaid for the 2- and 4-ﬂuorophenyl

derivative. Color code: C, black; Br, brown; Cu, orange; F, turquoise; N, blue; P, yellow; Pd, green.

Further shortening of the Pd−C(1) bond leads to another

transition state [24−13]^‡, that has a Pd−C(1) bond length of

2.03 Å, and in which a Pd−Br bond is starting to form (2.89 Å).

The C−Br bond is elongated to 2.22 Å. The forward motion

along the reaction coordinate derived from the normal mode of

the imaginary frequency (126 i cm⁻¹) is dominated by the

movement of the palladium atom into the plane of the aromatic

ring, resulting in a shortening of the Pd−C(1) bond, while the

bromine moves out of this plane with concomitant elongation

of the C−Br bond. The activation barrier for this step is rather

low (Δ^‡E_tot= 7.0 kcal mol⁻¹, Δ^‡G_sov= 6.2 kcal mol⁻¹).

In the ﬁnal oxidative addition product, the trimethylphos-

phino(phenyl)dibromopalladate 13, the palladium is in a

slightly distorted square-planar environment with the two

bromine ligands trans to each other. The Pd−Br(1) and Pd−

Br(2) bonds have almost the same lengths (2.54 and 2.52 Å),

and the Pd−C distance amounts to 2.03 Å.

The overall oxidative addition process is strongly exothermic

(Δ_rE_tot= −30.3 kcal mol⁻¹), and although two molecules are

converted to one product, it remains exergonic (Δ_rG_sov= −18.1

kcal mol⁻¹). Under the reaction conditions of the decarbox-

ylative coupling, the oxidative addition can be expected to

proceed very smoothly.

Transmetalation. In the transmetalation step, the phenan-

throline copper 2- or 4-ﬂuorophenyl complex 2, obtained by

decarboxylation of copper 2- or 4-ﬂuorophenyl benzoate, has to

approach the palladium complex 13, formed by the oxidative

addition, in order to transfer the ﬂuorophenyl group to the

palladium center. One would expect to ﬁnd that the actual

transmetalation step would be preceded by the formation of an

energetically favorable adduct between 2 and 13, which would

explain that this entropically unfavorable step takes place even

at the low experimental concentrations of the two species.

We ﬁrst sought for attractive interactions between 2a and 13

by positioning the two fragments close to each other at various

angles and orientations. Whereas no energetically favorable

interaction was found, e.g., when approaching the palladium

and the π-system of the ﬂuorophenyl ring, bringing palladium

complex 13 into close proximity to the copper atom in 2a via

its bromine ligand results in the formation of a stable adduct

(25a, Figure 5). The formation of a long Cu−Br(1) bond (2.82

Å) makes this step energetically favorable (Δ_rE_tot= −2.6 kcal

mol⁻¹), which makes up for some of the high entropic hurdle

and loss of solvation energy associated with forming an adduct

at high reaction temperatures. In the present case, the amount

of entropy lost is 33.4 cal K⁻¹mol⁻¹, which at 443 K increases

Δ_rG_sovby 14.8 kcal mol⁻¹. The dispersion interaction

additionally stabilizes the adduct by 13.1 kcal mol⁻¹, whereas

9.1 kcal mol⁻¹solvation energy is lost. Together, these factors

turn a slightly exothermic step into a moderately endergonic

one (Δ_rG_sov= 10.4 kcal mol⁻¹). The additional Cu−Br(1)

bond changes the coordination environment at the copper

center from a coordinatively unsaturated distorted trigonal

planar arrangement to a distorted tetrahedral geometry (∠Cu−

Br−Pd = 108.8°). The formation of the loose Cu−Br(1)

contact (2.82 Å) does not lead to a signiﬁcant elongation of the

Pd−Br(1) bond (2.53 Å), and the geometry of the palladium

fragment remains almost unchanged. The ﬂuorophenyl group is

bent outward slightly, but is not twisted. A transit scan from 25

with a stepwise elongation of the Cu−Br(1) bond revealed that

in the gas phase, the encounter complex 25 is formed without a

barrier.

Adduct 25 was found to be the entry point to a

transmetalation pathway in which the bromine ligand is

transferred from palladium to copper prior to transfer of the

ﬂuorophenyl group in the opposite direction. A search for

alternative transmetalation pathways in which, for example, the

transfer of the ﬂuorophenyl group from Cu to Pd precedes that

of the bromine ligand from Pd to Cu, did not yield any results.

I

dx.doi.org/10.1021/ja503295x | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society

Article

This was to be expected in the absence of attractive interactions

between the ﬂuorophenyl group and the copper.

transmetalation, which is therefore quite exothermic (Δ_rE_tot=

−22.2 kcal mol⁻¹) and even more exergonic (Δ_rG_sov= −32.9

kcal mol⁻¹), because of the substantial gain in entropy and

solvation energy.

Shortening the Cu−Br(1) bond of 25a to 2.54 Å leads to an

energetically reasonable transition state [25a−26a]^‡(Δ^‡E_tot

=

18.9 kcal mol⁻¹, Δ^‡G_sov= 14.5 kcal mol⁻¹). In [25a−26a]^‡, the

palladium is coordinated to the C(2) carbon of the

ﬂuorophenyl group over a rather long bond (2.59 Å), while

the Pd−Br(1) bond is almost broken (3.65 Å). The tetrahedral

environment at the copper is now more symmetrical than in

adduct 25a. The forward motion along the reaction coordinate

derived from the normal mode of the imaginary frequency (30 i

cm⁻¹) is dominated by the elongation of the Pd−Br(1) bond

with a simultaneous shortening of the Pd−C(2) and Pd−C(3)

distances. The IRC calculation conﬁrms that this transition

state leads to intermediate 26a, in which the palladium atom is

coordinated to the C(2)−C(3) π-bond in an almost sym-

metrical fashion (Pd−C(2): 2.48 Å; Pd−C(3): 2.52 Å). The

Cu−Br(1) bond is slightly shortened to 2.51 Å. Intermediate

26a is almost as high in energy as the transition state.

Starting from 25a, no trajectory could be found along which

the palladium would directly attack the C(1) carbon of the

ﬂuorophenyl group, so that the transfer of the bromine and of

the ﬂuorophenyl group would occur in a single step. This

ﬁnding corresponds well with those for several oxidative

addition and transmetalation pathways of other aromatic

substrates, which also have been found to involve the

intermediate formation of π-complexes to the C(2)−C(3)

bond of the arene rings.^24b,47

Shortening the Pd−C(1) distance leads to a transition state

[26a−14a]^‡across a rather low activation barrier (Δ^‡E_tot= 4.8

kcal mol⁻¹, Δ^‡G_sov= 0.4 kcal mol⁻¹). In [26a−14a]^‡, the

palladium atom has shifted to the C(1)−C(2) π-bond of the

ﬂuorophenyl group. The Pd−C(1) bond is slightly longer than

the Pd−C(2) bond (Pd−C(1): 2.82 Å; Pd−C(2): 2.60 Å). The

copper atom remains in the plane of the ﬂuorophenyl ring, and

the Cu−C(1) bond has the same length as in 26a (1.96 Å).

The distance between the copper and palladium atoms is

reduced to only 3.30 Å. The forward motion along the reaction

coordinate derived from the normal mode of the imaginary

frequency (59 i cm⁻¹) is dominated by a lengthening of the

Pd−C(2) bond with concomitant shortening of the Pd−C(1)

and Cu−Pd bonds, resulting in the transfer of the ﬂuorophenyl

group from copper to palladium.

At ﬁrst sight, it seemed counterintuitive that the copper atom

would remain in the plane of the arene ring in transition state

[26a−14a]^‡, and that the Cu−C(1) bond length would not

signiﬁcantly contribute to the normal mode corresponding to

the imaginary frequency. We would have expected a transition

state structure in which the copper would already have moved

out of the plane of the ﬂuorophenyl ring and the Cu−C(1)

bond would already have been elongated. We therefore

investigated the pathway from [26a−14a]^‡to the trans-

metalation products 14a + 11 in some detail by IRC

calculations, followed by geometry optimizations. On the way

downhill, we came across regions of the potential energy

surface with molecular geometries meeting our above expect-

ations, but found no stationary points. Instead, this process

conﬁrmed that the transition state [26a−14a]^‡indeed connects

intermediate 26a with the transmetalation products 14a + 11.

The somewhat surprising structural features of the transition

state are a consequence of its being very early, which conforms

to the Hammond postulate.⁴⁸Aryl transfer and separation into

two fragments occur simultaneously in this ﬁnal phase of the

The products of the transmetalation are the regenerated

phenanthroline copper bromide catalyst 11, which was already

described in the anion exchange chapter, and the square-planar

cis-diaryl palladium complex 14a. Its Pd−C(1) bond length is

further shortened to 2.04 Å, which is almost the same as that of

the Pd−C(1′) bond of 2.05 Å. Overall, the entire trans-

metalation starting from 2a and 13 and leading to 11 and 14a is

moderately exothermic (Δ_rE_tot= −8.2 kcal mol⁻¹) and

exergonic (Δ_rG_sov= −9.1 kcal mol⁻¹).

The reaction pathway for the transfer of the 4-ﬂuorophenyl

group from the aryl copper complex 2b to the palladium

complex 13 is diﬀerent from that of the 2-ﬂuorophenyl complex

2a. Both the geometries and the energies of several structures

involved are markedly diﬀerent. Moreover, an additional

intermediate is involved in the transmetalation for the 4-

ﬂuorophenyl derivative.

Adduct 25b and the ﬁrst transition state [25b−26b]^‡have

closely related structural features and relative energies to their

ortho-regioisomers. Compared to 26a, intermediate 26b is

lower in energy by −2.0 kcal mol⁻¹and the Gibbs energy is

lower by −0.6 kcal mol⁻¹. According to our calculations, this

can be attributed to the energy of solvation and explained with

greater structural diﬀerences between these intermediates. After

elongation of the Pd−Br contact in [25−26]^‡, the 4-

ﬂuorophenyl group rotates around the Cu−C bond, which

does not occur with the 2-ﬂuorophenyl group. The structures

of 26a and 26b provided in the Supporting Information are

displayed with the ﬂuorophenyl groups and coordinated

palladium fragments in the same orientation. In 26a, the

phenanthroline points to the left and the bromine to the right

of the copper. In 26b, the phenanthroline points downward and

the bromine upward. The Cu−Br bond in 26b is somewhat

longer (2.60 Å) than in [25b−26b]^‡(2.52 Å). The Pd−C(2)

bond in 26b is shorter by ca. 0.2 Å compared to the Pd−C(3)

bond (Pd−C(2): 2.41 Å; Pd−C(3): 2.60 Å).

The energetic hurdle between 26b and the transition state

[26b−27]^‡amounts to Δ^‡E_tot= 1.6 kcal mol⁻¹. This barrier is

only one-third as high as that for the ortho-substituted

derivative and can again be attributed to structural diﬀerences.

The Pd−C(2) bond is slightly shorter (2.46 Å) than in [26a−

14a]^‡(2.60 Å), and the distance between copper and palladium

of 3.55 Å in [26b−27]^‡is longer compared to 3.30 Å in [26a−

14a]^‡.

In contrast to the ortho-intermediate, 26b does not lead to

the transmetalation products 14b + 11 in a single step. It

involves an additional intermediate 27 in which the palladium

fragment is in a nearly square-planar environment. The

palladium atom has almost moved into the plane of the

ﬂuorophenyl ring, and the Pd−C(1) bond has shortened to

2.10 Å. The copper atom has moved below the plane of the

ﬂuorophenyl ring but is still connected to its ipso carbon over a

very long Cu−C(1) bond (2.32 Å). The Pd−C(1)−Cu angle of

77.8° is very small, so that the two transition metals are in close

proximity to each other. The Pd−Cu distance is as short as 2.77

Å, which is smaller than the sum of their van der Waals radii

(3.03 Å).⁴⁹Completing the transmetalation from intermediate

27 requires only little activation energy (transition state [27−

14b]^‡: Δ^‡E_tot= 0.0 kcal mol⁻¹, Δ^‡G_sov= 0.8 kcal mol⁻¹), so that

the existence or nonexistence of the intermediate virtually has

J

dx.doi.org/10.1021/ja503295x | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society

Article

Figure 6. B3LYP/6-31+G(d) optimized structures for the reductive elimination of [Pd(PMe₃)(2-F-Ph)(Ph)Br]⁻; hydrogens are omitted for clarity.

Energy and solution Gibbs energy proﬁles overlaid for the 2- and 4-ﬂuorophenyl derivative. Color code: C, black; Br, brown; F, turquoise; P, yellow;

Pd, green.

no eﬀect on the reactivity. Still, we invested some eﬀort in

ﬁnding a corresponding intermediate for the ortho-case. Thus,

the optimized structure 27 was transformed into its ortho-

derivative by replacing F by H and either of the two o-

hydrogens by ﬂuorine. However, for either of these structures,

new geometry optimizations directly led to the transmetalation

products. The existence of the intermediate 27 for the para- but

not the ortho-case may be explained with electronic eﬀects due

to the ﬂuorine substituent. The mesomeric eﬀects of the

ﬂuorine substituent are comparable for the ortho- and para-

positions, but the inductive eﬀect on the ipso carbon of the

ﬂuorine in the ortho-position is certainly higher due to the

smaller number of interjacent σ-bonds. In the 4-ﬂuorophenyl

ring, the electron-donating mesomeric eﬀect of the ﬂuorine

predominates over the inductive eﬀect and pushes suﬃcient

electron density into the ipso-carbon to bind both transition

metals simultaneously and lending stability to intermediate 27.

Contrarily, in the 2-ﬂuorophenyl ring, the short-range electron-

withdrawing inductive eﬀect outweighs the mesomeric eﬀect.

The ipso-carbon thus lacks the electron density required to

coordinate to palladium and copper at the same time.

Intermediate 27 features the shortest Pd−Cu distance (2.77

Å) observed in all these calculations. This begs the question

whether there is an interaction or even a bond between copper

and palladium. This is very diﬃcult to answer with DFT

calculations. Methods such as the analysis of the natural bond

order (NBO), the bond critical point (BCP), or the shared

electron number (SEN) with DFT are often regarded as

inconclusive. Therefore, we looked for literature on bimetallic

complexes with similarly short copper−palladium distances in

which metal−metal interactions and bonds are discussed.

Kawamura et al. crystallized [Pd₃(S₂CNⁿPr₂)₆Cu₂][PF₆]₂and

found slightly longer Pd−Cu distances of 2.864(4) and

2.896(4) Å.⁵⁰Their XPS binding studies suggest weak Pd−

Cu bonding interactions. However, the authors were unable to

prove the existence of direct Pd−Cu bonding. Peng and

Rohmer crystallized [Cu₂Pd(dpa)₄Cl₂] (dpa = dipyridylamine)

with somewhat shorter Pd−Cu distances of 2.4971(3) and

2.5022(3) Å.⁵¹For this complex, magnetic susceptibility

measurements revealed an antiferromagnetic coupling between

the two Cu^IImetal centers connected to the Pd atom, which

could be conﬁrmed by powder EPR experiments, DFT

calculations, and wave function theory calculations.⁵²Osakada

et al. observed Pd−Cu distances of 2.462(1) and 2.632(1) Å in

crystallized [Pd(μ-CuI)(Pd(dmpe))₃(μ₃-GePh₂)₃].⁵³¹H and

¹³C{¹H} NMR experiments of the complex at various

temperatures revealed that Cu forms a stable Pd−Cu bond to

the core palladium and a labile Pd−Cu bond to one edge

palladium. Because interactions or even bonds between copper

and palladium could be proven for these complexes, which all

possess short Pd−Cu distances in the same range as our

intermediate 27, we believe that attractive interactions between

the copper and palladium atoms are at play also in the

transmetalation step of decarboxylative cross-couplings.

Overall, our DFT results show that the transmetalation

process is exothermic (ortho: Δ_rE_tot= −8.2 kcal mol⁻¹; para:

Δ_rE_tot= −11.1 kcal mol⁻¹) as well as exergonic (ortho: Δ_rG_sov

=

−9.1 kcal mol⁻¹; para: Δ_rG_sov= −11.6 kcal mol⁻¹). However,

the process starts with the formation of a bimetallic adduct that

is endergonic (ortho: Δ_rG_sov= 10.4 kcal mol⁻¹; para: Δ_rG_sov

=

9.5 kcal mol⁻¹). Furthermore, the transmetalation faces high

activation barriers for its transition states that we will compare

with those for the other elementary steps at the end of this

section.

Reductive Elimination. In the reductive elimination step,

the biaryl product 9 is liberated from the square-planar cis-

diarylpalladium complex 14, regenerating the initial palladium

catalyst 12 (Figure 6). Both for the ortho- and the para-

substituted model compounds, the overall process is moder-

ately exothermic (ortho: Δ_rE_tot= −10.8 kcal mol⁻¹; para: Δ_rE_tot

= −14.4 kcal mol⁻¹) and strongly exergonic (ortho: Δ_rG_sov

=

K

dx.doi.org/10.1021/ja503295x | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society

Article

Scheme 8. Anionic Reaction Pathways Investigated for the Anion Exchange, Decarboxylation, and Adduct Formation

Figure 7. B3LYP/6-31+G(d) optimized structures for the anionic reaction pathway for anion exchange, decarboxylation, and adduct formation;

hydrogens are omitted for clarity. Energy and solution Gibbs energy proﬁles overlaid for 2- and 4-ﬂuorobenzoate. Color code: C, black; Br, brown;

Cu, orange; F, turquoise; N, blue; O, red; P, yellow; Pd, green.

−24.5 kcal mol⁻¹; para: Δ_rG_sov= −27.1 kcal mol⁻¹), since two

molecules form from one.

the biaryl product (Pd−C(1′): 2.25 Å; Pd−C(2′): 2.17 Å; P−

Pd−Br: 94.7°).

Starting from 14a, a ﬁrst transition state [14a−28a]^‡was

found by reducing the C(1) and C(1′) distance between the

two aromatic rings in the square-planar complex to 1.88 Å. The

barrier for its formation is comparatively low (Δ^‡E_tot= 20.4 kcal

mol⁻¹, Δ^‡G_sov= 17.2 kcal mol⁻¹). The Pd−C bond lengths and

the P−Pd−Br angle remain almost unchanged (Pd−C(1): 2.07

Å; Pd−C(1′): 2.11 Å; P−Pd−Br: 88.6°). The forward motion

along the reaction coordinate derived from the normal mode of

the imaginary frequency (306 i cm⁻¹) is dominated by a

shortening of the aryl−aryl bond, while the angle between the

planes of the aromatic rings is reduced.

The energy barrier for the release of the Pd(0) catalyst 12 is

very low (Δ^‡E_tot= 3.2 kcal mol⁻¹, Δ^‡G_sov= 0.03 kcal mol⁻¹). In

the transition state [28a−12a]^‡, which was found by increasing

the distance between palladium and the phenyl ring, the

palladium atom remains coordinated to C(2′) over a distance

of 2.35 Å, and the P−Pd−Br angle has increased to 126.1°. The

forward motion along the reaction coordinate derived from the

normal mode of the imaginary frequency (54 i cm⁻¹) is

dominated by a lengthening of the Pd−C bond with

simultaneous opening of the P−Pd−Br angle (Pd−C(2′):

2.35 Å; P−Pd−Br: 126.1).

An IRC calculation revealed that transition state [14a−28a]^‡

connects structure 14a with intermediate 28a. In 28a, the aryl−

aryl bond is fully formed with a ﬁnal length of 1.49 Å, but the

palladium remains η²-coordinated to the C(1′)−C(2′) bond of

A further IRC calculation showed that this transition state

connects directly to the linear anionic monophosphine

palladium complex 12 and the biaryl 9a. In 9a, the angle

between the planes of the two aromatic rings amounts to 44.3°.

L

dx.doi.org/10.1021/ja503295x | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society

Article

Figure 8. B3LYP/6-31+G(d) optimized structures for the simpliﬁed reaction proﬁle; hydrogens are omitted for clarity. Energy and solution Gibbs

energy proﬁles overlaid for the 2- and 4-ﬂuorophenyl derivative. Color code: C, black; Br, brown; Cu, orange; F, turquoise; K, purple; N, blue; O,

red; P: yellow; Pd, green.

As can be seen in Figure 6, the reaction pathway for the 4-

ﬂuorophenyl derivative is almost identical with that for the 2-

derivative. The main activation barrier (Δ^‡E_tot= 19.2 kcal

mol⁻¹, Δ^‡G_sov= 18.3 kcal mol⁻¹) and the imaginary frequency

of the transition state [14b−28b]^‡(316 i cm⁻¹) are almost

identical with those for the 2-derivative. In biaryl 9b, the

dihedral angle is slightly smaller than in 9a (41.4° versus 44.3°),

since the o-hydrogen requires less space than the ﬂuorine atom.

Anionic Copper Cycle. In the transformations discussed

above, all Cu^Iintermediates were neutral and tricoordinated.

An alternative pathway can be envisaged in which the

carboxylate anion adds to copper complex 11 leading to a

tetracoordinate anionic copper species, which after extrusion of

CO₂undergoes the transmetalation (Scheme 8). In this

alternative pathway, the equivalent to the anion exchange

step would be the coordination of the carboxylate 17 to the

copper complex 11. The resulting anionic copper complex 15

would decarboxylate to the anionic copper complex 16.

Complex 16 would then react with the palladium catalyst 13

in an associative substitution reaction to give adduct 25, which

would be the starting point for the transmetalation process.

Alternatively to this associative pathway, one could also imagine

that the bromide would decoordinate from 16 to give the

ﬂuorophenyl copper phenanthroline complex 2, which would

associate with 13 to give the adduct 25. The energy and

solution Gibbs energy proﬁles for the anionic pathway (11 →

15 → 16 → 25) are shown in Figure 7.

The anionic copper complex 15a can then directly

decarboxylate via transition state [15a−16a]^‡, instead of ﬁrst

releasing the bromine ligand. The Gibbs activation energy of

Δ^‡G_sov= 24.6 kcal mol⁻¹is higher than that for the

decarboxylation of the neutral phenanthroline copper 2-

ﬂuorobenzoate 1a (Δ^‡G_sov= 23.8 kcal mol⁻¹).

Two pathways are imaginable for the associative substitution

reaction of the anionic copper complex 16 with the palladium

catalyst 13 each leading to formation of a bromine bridge

between the two metals. One of the Pd-bound bromide ligands

may attack the copper center and replace its bromide (red

arrows, Scheme 8), or conversely the Cu-bound bromide may

attack the palladium center with loss of a Pd-bound bromide

(green arrows). For the ﬁrst pathway, we were unable to ﬁnd

any feasible pathway in our DFT calculations. For the second, a

transition state [16b−25b]^‡could be located only for the p-

ﬂuoro-substituted derivative. The activation barrier is very high

(Δ^‡E_tot= 51.2 kcal mol⁻¹, Δ^‡G_sov= 22.0 kcal mol⁻¹). For the

ortho-derivative 16a, the only pathway we found required

dissociation of the bromide. This dissociation of bromide is

thermodynamically downhill for both the ortho- and the para-

derivative (16a/16b), and leads to the phenanthroline copper

ﬂuorophenyl complex 2a/2b, so that the neutral pathway is re-

entered. When similar calculations were performed with

potassium ions, we found direct elimination of KBr and a

direct re-entry into the neutral pathway. Overall, the anionic

complexes 15a/b do not open up a pathway with a low energy

proﬁle that is more favorable for ortho- than for non-ortho-

substituted benzoates and could explain why the former

substrates react more readily in decarboxylative couplings.

Summary Assessment of the Catalytic Cycle. Improv-

ing a catalyst based on theoretical modeling requires a recipe

for extracting the catalyst eﬃciency from the plethora of

energetic data obtained in the calculations. The energetic span

concept by Amatore and Jutand⁵⁴says that the largest possible

The formation of adduct 15a from 2-ﬂuorobenzoate 17a and

the copper complex 11 is exothermic (Δ_rE_tot= −14.2 kcal

mol⁻¹) in the gas phase and proceeds via an encounter complex

29a and a transition state [29a−15a]^‡. Due to the loss of

entropy and solvation energy, this step is endergonic (Δ_rG_sov

=

7.7 kcal mol⁻¹), with a barrier (Δ^‡G_sov= 13.9 kcal mol⁻¹) that

is higher than in the neutral anion exchange pathway (see

Figure 2).

M

dx.doi.org/10.1021/ja503295x | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society

Article

rate and, correspondingly, the highest turnover number of the

catalytic cycle is obtained for the lowest Gibbs energy span,

which is the Gibbs energy diﬀerence between the highest

transition state and lowest intermediate of the entire cycle. This

concept has been reﬁned by Kozuch and Shaik⁵⁵by taking into

account overall exergonicity. In our case, the situation is more

complicated because of the two intertwined catalytic cycles,

which join and separate at the transmetalation. The

approximation behind the energy span concept is that the

concentration c_iof a catalyst intermediate i is given by c_i= c₀

exp(−ΔG_i/RT), where c₀is the overall catalyst concentration

and ΔG_ithe Gibbs energy of i with respect to the most stable

intermediate, where most of the catalyst is trapped. Both our

calculated energy proﬁle and our experience with related

catalytic reactions tell us that the palladium catalyst is capable

of providing the oxidative addition product 13 at a much faster

rate than the cross-coupling proceeds, and according to our

calculations, the oxidative addition is exergonic (Δ_rG = −18.1

kcal mol⁻¹). Therefore, we can simplify the kinetic model by

assuming that the stationary concentration of 13 is given by the

overall amount of palladium catalyst. In other words, we

consider a reaction with a single catalytic cycle in which one of

the educts is 13 at its catalytic concentration, instead of the aryl

bromide 8 at a stoichiometric concentration. We then apply the

energetic span concept to the energy proﬁle derived from this

single cycle, which involves four steps (anion exchange, CO₂

extrusion, transmetalation, reductive elimination). Since none

of these steps feature an intermediate lower in Gibbs energy

than either their educts or their products, we need only

consider a single eﬀective activation energy given by the highest

transition state of each step. This resulting simpliﬁed energy

diagram is shown in Figure 8.

The energy barriers of this simpliﬁed reaction proﬁle with the

highest barrier at Δ^‡G_sov= 36.2 kcal mol⁻¹are consistent with a

reaction that takes place at temperatures around 160 °C.

For the ortho-case, the largest Gibbs energy span, which

represents the apparent activation energy of the overall cycle, is

between the starting materials (7a + 11) and the decarbox-

ylation transition state [1a−2a]^‡(29.1 kcal mol⁻¹). For the

para-case, the largest Gibbs energy span is between the starting

materials (7b + 11) and the transmetalation transition state

[25b−26b]^‡(36.2 kcal mol⁻¹). This implies that for the ortho-

case, the decarboxylation transition state and for the para-case,

the transmetalation transition state is the rate-limiting transition

state. Considering each step individually, one would come to a

diﬀerent conclusion since for the ortho-case, the barrier is lower

in the decarboxylation (23.8 kcal mol⁻¹) compared to the

transmetalation (24.9 kcal mol⁻¹), whereas for the para-case

the barrier is higher in the decarboxylation (27.1 kcal mol⁻¹)

than in the transmetalation (25.8 kcal mol⁻¹). This contrast

stems from the exergonicity/endergonicity of the decarbox-

ylation step as its products are the starting point for the

subsequent transmetalation. For the ortho-case the decarbox-

ylation is exergonic by −2.5 kcal mol⁻¹which decreases the

starting point and the highest transition state of the

transmetalation, while for the para-case the decarboxylation is

endergonic by 5.4 kcal mol⁻¹, which increases the starting point

and the highest transition state of the transmetalation, which

explains the outcome of the overall reaction proﬁle. For the

ortho-case the diﬀerence between the heights of the

decarboxylation (29.1 kcal mol⁻¹) and transmetalation

transition state (27.7 kcal mol⁻¹) lies within the inaccuracy of

the computational method. It is safe to say that the

decarboxylation and the transmetalation are so similar in

energy that it probably depends on the individual substrate

which of these two steps will be rate-determining.

As discussed above, the reason for the high overall barrier of

the transmetalation is the entropic penalty associated with

forming an adduct of two species present only in catalytic

amounts, together with a loss in solvation energy. The accuracy

of some contributions to the activation energy such as solvent

eﬀects is diﬃcult to assess. However, such systematic errors are

likely to cancel out when calculating the para/ortho diﬀerence

in the Gibbs energy span. This diﬀerence is considerable (7.1

kcal mol⁻¹), and agrees with the experimental observation that

4-ﬂuorobenzoate is substantially less reactive than its ortho-

isomer. Interestingly, the largest contribution to this diﬀerence

stems from the exer-/endergonicity of the decarboxylation step,

where the diﬀerence between para and ortho is ΔΔ_rG_sov= 7.6

kcal mol⁻¹.

Based on this proﬁle, it can also be understood why the

presence of halide anions can aﬀect the reaction rate. The

presence of excess bromide shifts the equilibrium of the anion

exchange step to the side of the copper bromide so that less

copper carboxylate is available. If for a given substrate the

decarboxylation is the rate-determining step in a decarbox-

ylative cross-coupling, the presence of excess bromide should

have a similar eﬀect, because it increases the energy span

between 7 + 11 and [1−2]^‡+ 10 by about 3 kcal mol⁻¹.

Experimental Studies. The above DFT studies suggest

that the transmetalation is rate-limiting for substrates that

decarboxylate comparatively easily. Bidentate ligands designed

to bridge the two metals and bring them into close spatial

proximity can be expected to facilitate this step. To test this

hypothesis, we ﬁrst searched for a benzoic acid that

decarboxylates with particular ease. We performed a series of

protodecarboxylation experiments in which various benzoic

acids 22 were heated to 100 °C in the presence of 5 mol %

Cu₂O and 10 mol % 1,10-phenanthroline (Table 1). No

a

Table 1. Protodecarboxylation of Benzoic Acids

entry

carboxylic acid

R

t/h

product

yield/%

1

2

3

4

5

22a

22b

22c

22d

22c

2-F

6

23

30

3

0

4-F

2-NO₂

4-NO₂

2-NO₂

6

64

0

6

24

99

a

Reaction conditions: 0.5 mmol of carboxylic acid, 5 mol % Cu₂O, 10

mol % 1,10-Phen, 2.0 mL of NMP, 100 °C, GC yield after calibration.

conversion was observed at this low temperature for most

carboxylic acids tested, including 2- and 4-ﬂuorobenzoic acid

(22a or 22b) (entries 1 and 2). However, 2-nitrobenzoic acid

(22c) was smoothly converted into nitrobenzene (30) in near-

quantitative yield (entry 5).

It could thus be expected that the decarboxylation is not the

rate-limiting step in a decarboxylative cross-coupling of 2-

nitrobenzoic acid (22c) at 100 °C. We next performed various

decarboxylative couplings at this temperature. Among all aryl

electrophiles tested, triﬂates were the sole substrates to provide

N

dx.doi.org/10.1021/ja503295x | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society

Article

a

Table 2. Decarboxylative Cross-Couplings with Monodentate and Bridging Ligands

entry

carboxylate

R

temp/°C

L

product

yield/%

1

2

7c

7d

7c

2-NO₂

4-NO₂

2-NO₂

100

170

100

P(p-Tol)₃

32c

32d

32c

9

P(p-Tol)₃

0

b

3

P(p-Tol)₃

84 (91)^7a

4

5

6

7

L1

L2

L3

L4

0

40

88

c

8

a

Reaction conditions: 0.5 mmol of potassium carboxylate, 1 mmol of triﬂate, 5 mol % Cu₂O, 10 mol % 1,10-Phen, 2 mol % PdI₂, 6 mol % L, 2.0 mL

b

of NMP, 100 °C, 24 h, GC yield after calibration. 1 mmol of potassium carboxylate, 2 mmol of triﬂate, 5 mol % Cu₂O, 10 mol % 1,10-Phen, 2 mol

% PdI₂, 6 mol % P(p-Tol)₃, 4.0 mL of NMP, 170 °C, 1 h, GC yield after calibration, isolated yield in parentheses. 0.75 mmol of potassium

c

carboxylate, 0.5 mmol of triﬂate, 5 mol % Cu₂O, 10 mol % 1,10-Phen, 3 mol % Pd(acac)₂, 6 mol % L4, 4.0 mL of NMP, 100 °C, 24 h, GC yield after

calibration.

the coupling products in small amounts. For this reason, we

investigated the coupling of 4-chlorophenyl triﬂate (31) with

several potassium nitrobenzoates 7 as the model systems. With

the best known catalyst system (Table 2), no conversion was

observed in the coupling of potassium 4-nitrobenzoate (7d)

with 4-chlorophenyl triﬂate (31) (entry 2), and potassium 2-

nitrobenzoate (7c) gave less than 10% of 4-chloro-2′-

nitrobiphenyl (32c) (entry 1). According to the literature,

cross-couplings of aryl triﬂates with potassium benzoates with

Cu/Pd systems and monodentate p-tolylphosphine have to be

performed at 170 °C to give high yields.^7a,b

Conclusions and Outlook. The geometries and energies

of all starting materials, products, intermediates, and transition

states of the catalytic cycle were calculated for the

decarboxylative cross-coupling of potassium 2- and 4-

ﬂuorobenzoate with bromobenzene in the presence of a

catalyst system consisting of copper(I)/1,10-phenanthroline

and the anionic monophosphine palladium complex [Pd-

(PMe₃)Br]⁻. Among all pathways investigated, a catalytic cycle

via neutral copper complexes was found to be most favorable. It

consists of a carboxylate−bromide exchange at the copper

center to give an uncharged copper carboxylate, followed by a

decarboxylation step in which an organocopper species is

generated. A palladium species that has been formed by

oxidative addition of the aryl electrophile to a low-valent

palladium(0) species accepts the aryl group from the copper in

the subsequent transmetalation step, which links the catalytic

cycles of the copper and palladium catalysts. The resulting

diarylpalladium species undergoes reductive elimination,

releasing the biaryl product and regenerating the palladium(0)

species (Scheme 6).

We then replaced the p-tolylphosphine ligand by various

bidentate P,N-ligands (Figure 9).^9,56Among these, the

Based on realistic assumptions on the inherent accuracy of

these calculations, the energetic span model cannot unambig-

uously decide whether the decarboxylation or the trans-

metalation is rate-determining. This implies that depending

on the substrate, either step can pose the main limitation in the

catalyst performance. This is in sharp contrast to the current

opinion that it is only the decarboxylation step that needs to be

addressed in the catalyst improvement. The calculations also do

not support the hypothesis that the diﬀerence in reactivity

between ortho- and non-ortho-substituted carboxylates in the

reaction with aryl halides is caused by the anion exchange step.

The low reactivity of the latter is probably caused by the higher

barrier of the decarboxylation process, which is further

increased in relation to the starting material if excess halide

salt is present.

Figure 9. Bidentate P,N-ligands.

aminopyrimidinyl phosphines, especially ligand L4, developed

by Thiel et al. in the context of other coupling reactions were

most eﬀective (entry 7).^56cAfter slight modiﬁcations to the

reaction conditions, the coupling product 32c was detected in

high yield (88%) already at 100 °C (entry 8).

Lowering the reaction temperature by as much as 70 °C is a

decisive improvement over the state of the art, which we

attribute to the ability of this ligand to bring copper and

palladium into close proximity and thus assist the trans-

metalation step. The presence of bimetallic Cu−Pd complexes

was conﬁrmed by ESI-MS investigations of the reaction

mixture, suggesting that their formation is no longer rate-

determining.²⁸Further experimental studies with the bidentate

ligands, including the structural investigation of Cu/Pd adducts

detected by ESI-MS, are underway. We are convinced that the

concept of bringing the two metals together by bridging ligands

will allow reaching new levels of eﬃciency in decarboxylative

couplings of activated benzoates.

In the transmetalation, the electronic activation energy is not

overly high. It is the free energy loss in the initial adduct

formation that makes it diﬃcult. This result suggests that future

research aimed at further improving the catalyst should also

target the transmetalation and not only the decarboxylation

step. This was conﬁrmed by an experimental study, in which

O

dx.doi.org/10.1021/ja503295x | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society

Article

(11) Sun, Z. M.; Zhao, P. Angew. Chem. 2009, 121, 6854−6858;

Angew. Chem., Int. Ed. 2009, 48, 6726−6730.

the reaction temperature of a decarboxylative coupling was

lowered by 70 °C by a P,N-ligand designed to facilitate adduct

formation between copper and palladium. In combination with

these experimental ﬁndings, the mechanistic studies presented

herein are likely to induce a paradigm shift in the development

of more active catalyst generations for decarboxylative cross-

coupling reactions.

(12) Weaver, J. D.; Recio, A., III; Grenning, A. J.; Tunge, J. A. Chem.

Rev. 2010, 111, 1846−1913.

(13) (a) Peschko, C.; Winklhofer, C.; Steglich, W. Chem.Eur. J.

2000, 6, 1147−1152. (b) Forgione, P.; Brochu, M. C.; St-Onge, M.;

Thesen, K. H.; Bailey, M. D.; Bilodeau, F. J. Am. Chem. Soc. 2006, 128,

11350−11351. (c) Bilodeau, F.; Brochu, M. C.; Guimond, N.; Thesen,

K. H.; Forgione, P. J. Org. Chem. 2010, 75, 1550−1560. (d) Miyasaka,

M.; Fukushima, A.; Satoh, T.; Hirano, K.; Miura, M. Chem.Eur. J.

2009, 15, 3674−3677. (e) Miyasaka, M.; Hirano, K.; Satoh, T.; Miura,

M. Adv. Synth. Catal. 2009, 351, 2683−2688. (f) Arroyave, F. A.;

Reynolds, J. R. Org. Lett. 2010, 12, 1328−1331.

(14) (a) Moon, J.; Jeong, M.; Nam, H.; Ju, J.; Moon, J. H.; Jung, H.

M.; Lee, S. Org. Lett. 2008, 10, 945−948. (b) Moon, J.; Jang, M.; Lee,

S. J. Org. Chem. 2009, 74, 1403−1406. (c) Kima, H.; Lee, P. H. Adv.

Synth. Catal. 2009, 351, 2827−2832. (d) Park, K.; Bae, G.; Moon, J.;

Choe, J.; Song, K. H.; Lee, S. J. Org. Chem. 2010, 75, 6244−6251.

(e) Zhang, W.-W.; Zhang, X.-G.; Li, J.-H. J. Org. Chem. 2010, 75,

5259−5264. (f) Zhao, D.; Gao, C.; Su, X.; He, Y.; You, J.; Xue, Y.

Chem. Commun. 2010, 46, 9049−9051. (g) Park, K.; Bae, G.; Park, A.;

Kim, Y.; Choe, J.; Song, K. H.; Lee, S. Tetrahedron Lett. 2011, 52,

576−580. (h) Park, A.; Park, K.; Kim, Y.; Lee, S. Org. Lett. 2011, 5,

944−947.

ASSOCIATED CONTENT

* Supporting Information

■

S

Energies, graphical representations, geometrical parameters and

Cartesian coordinates of all DFT-optimized structures. This

material is available free of charge via the Internet at http://

pubs.acs.org.

AUTHOR INFORMATION

Corresponding Author

goossen@chemie.uni-kl.de; vanwullen@chemie.uni-kl.de

■

Notes

The authors declare no competing ﬁnancial interest.

ACKNOWLEDGMENTS

We thank W. R. Thiel for help with the DFT calculations and

for providing P,N-ligands, K. Gooßen for help proofreading the

■

(15) Shang, R.; Fu, Y.; Li, J. B.; Zhang, S. L.; Guo, Q. X.; Liu, L. J.

Am. Chem. Soc. 2009, 131, 5738−5739.

(16) Yin, L.; Kanai, M.; Shibasaki, M. J. Am. Chem. Soc. 2009, 131,

9610−9611.

manuscript, and M. F. Grunberg for technical assistance. We

̈

also thank the DFG (GO 853/5-2 and SFB-TRR 88 “3Met”)

(17) (a) Zhang, S. L.; Fu, Y.; Shang, R.; Guo, Q. X.; Liu, L. J. Am.

Chem. Soc. 2010, 132, 638−646. (b) Hu, P.; Kan, J.; Su, W. P.; Hong,

M. C. Org. Lett. 2009, 11, 2341−2344. (c) Fu, Z.; Huang, S.; Su, W.;

Hong, M. Org. Lett. 2010, 12, 4992−4995. (d) Sun, Z.-M.; Zhang, J.;

Zhao, P. Org. Lett. 2010, 12, 992−995. (e) Goossen, L. J.;

Zimmermann, B.; Knauber, T. Beilstein J. Org. Chem. 2010, 6, 43−51.

(18) (a) Voutchkova, A.; Coplin, A.; Leadbeater, N. E.; Crabtree, R.

H. Chem. Commun. 2008, 6312−6314. (b) Wang, C. Y.; Piel, I.;

Glorius, F. J. Am. Chem. Soc. 2009, 131, 4194−4195. (c) Cornella, J.;

Lu, P.; Larrosa, I. Org. Lett. 2009, 11, 5506−5509. (d) Zhang, F.;

Greaney, M. F. Angew. Chem. 2010, 122, 2828−2831; Angew. Chem.,

Int. Ed. 2010, 49, 2768−2771.

(19) (a) Gledhill, A. P.; McCall, C. J.; Threadgill, M. D. J. Org. Chem.

1986, 51, 3196−3201. (b) Huang, W. H.; Wang, M. L.; Yue, H.

Synthesis 2008, 1342−1344. (c) Bi, H.-P.; Zhao, L.; Liang, Y.-M.; Li,

C.-J. Angew. Chem. 2009, 121, 806−809; Angew. Chem., Int. Ed. 2009,

48, 792−795. (d) Bi, H.-P.; Chen, W.-W.; Liang, Y.-M.; Li, C.-J. Org.

Lett. 2009, 11, 3246−3249.

(20) (a) Duan, Z. Y.; Ranjit, S.; Zhang, P. F.; Liu, X. G. Chem.Eur.

J. 2009, 15, 3666−3669. (b) Jia, W.; Jiao, N. Org. Lett. 2010, 12,

2000−2003. (c) Ranjit, S.; Duan, Z.; Zhang, P.; Liu, X. Org. Lett. 2010,

12, 4134−4136.

(21) (a) President’s Information Technology Advisory Committee.

Computational Science: Ensuring America’s Competitiveness; Arlington,

VA, 2005. (b) Mpourmpakis, G.; Vlachos, D. G. MRS Bull. 2011, 36,

211−215. and references therein (c) Goossen, L. J.; Rodríguez, N.;

Linder, C.; Lange, P. P.; Fromm, A. ChemCatChem. 2010, 2, 430−442.

(d) Dudnik, A. S.; Xia, Y.; Li, Y.; Gevorgyan, V. J. Am. Chem. Soc.

2010, 132, 7645−7655.

and Landesgraduiertenforderung Rheinland-Pfalz (fellowship to

̈

A.F.) for funding.

REFERENCES

■

(1) (a) Nilsson, M. Acta Chem. Scan. 1966, 20, 423−426. (b) Nilsson,

M.; Ullenius, C. Acta Chem. Scand. 1968, 22, 1998−2002.

(c) Chodowska-Palicka, J.; Nilsson, M. Acta Chem. Scand. 1970, 24,

3353−3361. (d) Shimizu, I.; Yamada, T.; Tsuji, J. Tetrahedron Lett.

1980, 21, 3199−3202. (e) Tsuji, J.; Yamada, T.; Minami, I.; Yuhara,

M.; Nisar, M.; Shimizu, I. J. Org. Chem. 1987, 52, 2988−2995.

(f) Myers, A. G.; Tanaka, D.; Mannion, M. R. J. Am. Chem. Soc. 2002,

124, 11250−11251. (g) Tanaka, D.; Myers, A. G. Org. Lett. 2004, 6,

433−436. (h) Tanaka, D.; Romeril, S. P.; Myers, A. G. J. Am. Chem.

Soc. 2005, 127, 10323−10333.

(2) Goossen, L. J.; Deng, G.; Levy, L. M. Science 2006, 313, 662−664.

(3) (a) Rudolphi, F.; Song, B.; Goossen, L. J. Adv. Synth. Catal. 2011,

353, 337−342. (b) Collet, F.; Song, B.; Rudolphi, F.; Goossen, L. J.

Eur. J. Org. Chem. 2011, 6486−6501.

(4) (a) Goossen, L. J.; Rudolphi, F.; Oppel, C.; Rodríguez, N. Angew.

Chem. 2008, 120, 3085−3088; Angew. Chem., Int. Ed. 2008, 47, 3043−

3045. (b) Grunberg, M. F.; Gooßen, L. J. J. Organomet. Chem. 2013,

̈

744, 140−143.

(5) Goossen, L. J.; Rodríguez, N.; Melzer, B.; Linder, C.; Deng, G.;

Levy, L. M. J. Am. Chem. Soc. 2007, 129, 4824−4833.

(6) Goossen, L. J.; Zimmermann, B.; Knauber, T. Angew. Chem.

2008, 120, 7211−7214; Angew. Chem., Int. Ed. 2008, 47, 7103−7106.

(7) (a) Goossen, L. J.; Rodriguez, N.; Linder, C. J. Am. Chem. Soc.

2008, 130, 15248−15249. (b) Goossen, L. J.; Linder, C.; Rodríguez,

N.; Lange, P. P. Chem.Eur. J. 2009, 15, 9336−9349. (c) Goossen, L.

J.; Lange, P. P.; Rodríguez, N.; Linder, C. Chem.Eur. J. 2010, 16,

3906−3909.

(22) Goossen, L. J.; Thiel, W. R.; Rodríguez, N.; Linder, C.; Melzer,

B. Adv. Synth. Catal. 2007, 349, 2241−2246.

(23) (a) Braga, A. A. C.; Ujaque, G.; Maseras, F. Organometallics

2006, 25, 3647−3658. and references therein (b) Kozuch, S.; Shaik,

S.; Jutand, A.; Amatore, C. Chem.Eur. J. 2004, 10, 3072−3080.

(c) Kozuch, S.; Amatore, C.; Jutand, A.; Shaik, S. Organometallics

2005, 24, 2319−2330.

(24) (a) Goossen, L. J.; Koley, D.; Hermann, H.; Thiel, W. J. Am.

Chem. Soc. 2005, 127, 11102−11114. (b) Goossen, L. J.; Koley, D.;

Hermann, H.; Thiel, W. Organometallics 2006, 25, 54−67.

(8) Goossen, L. J.; Rodríguez, N.; Lange, P. P.; Linder, C. Angew.

Chem. 2010, 122, 1129−1132; Angew. Chem., Int. Ed. 2010, 49, 1111−

1114.

(9) Song, B.; Knauber, T.; Goossen, L. J. Angew. Chem. 2013, 125,

3026−3030; Angew. Chem., Int. Ed. 2013, 52, 2954−2958.

(10) Shang, R.; Fu, Y.; Wang, Y.; Xu, Q.; Yu, H. Z.; Liu, L. Angew.

Chem. 2009, 121, 9514−9518; Angew. Chem., Int. Ed. 2009, 48, 9350−

9354.

P

dx.doi.org/10.1021/ja503295x | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society

Article

(25) Keith, J. A.; Behenna, D. C.; Mohr, J. T.; Ma, S.; Marinescu, S.

C.; Oxgaard, J.; Stoltz, B. M.; Goddard, W. A., III J. Am. Chem. Soc.

2007, 129, 11876−11877.

(47) (a) Goossen, L. J.; Koley, D.; Hermann, H.; Thiel, W. Chem.

Commun. 2004, 2141−2143. (b) Goossen, L. J.; Koley, D.; Hermann,

H.; Thiel, W. Organometallics 2005, 24, 2398−2410.

(48) Hammond, G. S. J. Am. Chem. Soc. 1955, 77, 334−338.

(49) Bondi, A. J. Phys. Chem. 1964, 68, 441−451.

(26) Shang, R.; Xu, Q.; Jiang, Y.-Y.; Wang, Y.; Liu, L. Org. Lett. 2010,

12, 1000−1003.

(50) Ebihara, M.; Tokoro, K.; Maeda, M.; Ogami, M.; Imaeda, K.;

Sakurai, K.; Masuda, H.; Kawamura, T. J. Chem. Soc., Dalton Trans.

1994, 3621−3635.

(27) Xie, H.; Lin, F.; Lei, Q.; Fang, W. Organometallics 2013, 32,

6957−6968.

(28) Hackenberger, D.; Song, B.; Grunberg, M. F.; Farsadpour, S.;

̈

(51) Liu, I. P.-C.; Lee, G.-H.; Peng, S.-M.; Ben

M. Inorg. Chem. 2007, 46, 9602−9608.

(52) Maynau, D.; Bolvin, H.; Van den Heuvel, W.; Ben

́

ard, M.; Rohmer, M.-

Taghizadeh Ghoochany, L.; Menges, F.; Burkhardt, L.; Niedner-

Schatteburg, G.; Thiel, W. R.; Goossen, L. J., unpublished results.

(29) Recently, related studies have investigated the transfer of an aryl

group from palladium to gold: Perez-Temprano, M. H.; Casares, J. A.;

de Lera, A. R.; Alvarez, R.; Espinet, P. Angew. Chem., Int. Ed. 2012, 51,

4917−4920.

́

ard, M.;

Rohmer, M.-M.; Ben Amor, N. C. R. Chim. 2012, 15, 170−175.

(53) Tanabe, M.; Ishikawa, N.; Chiba, M.; Ide, T.; Osakada, K.;

Tanase, T. J. Am. Chem. Soc. 2011, 133, 18598−18601.

(54) Amatore, C.; Jutand, A. J. Organomet. Chem. 1999, 576, 254−

278.

(30) (a) Gaussian 03, Revision E.01; Gaussian, Inc.: Wallingford,

CT, 2004. (b) Gaussian 09, Revision D.01; Gaussian, Inc.:

Wallingford, CT, 2013; for full citations see the Supporting

Information.

(31) (a) Stephens, P. J.; Devlin, F. J.; Chabalowski, C. F.; Frisch, M. J.

J. Phys. Chem. 1994, 98, 11623−11627. (b) Becke, A. D. J. Chem. Phys.

1993, 98, 5648−5652. (c) Becke, A. D. Phys. Rev. A 1988, 38, 3098−

3100. (d) Lee, C.; Yang, W.; Parr, R. G. Phys. Rev. B 1988, 37, 785−

789.

(55) Kozuch, S.; Shaik, S. Acc. Chem. Res. 2011, 44, 101−110.

(56) (a) Sun, Y.; Hienzsch, A.; Grasser, J.; Herdtweck, E.; Thiel, W.

R. J. Organomet. Chem. 2006, 691, 291−298. (b) Sarcher, C.;

Farsadpour, S.; Taghizadeh Ghoochany, L.; Sun, Y.; Thiel, W. R.;

Roesky, P. W. Dalton Trans. 2014, 43, 2397−2405. (c) Farsadpour, S.;

Taghizadeh Ghoochany, L.; Sun, Y.; Thiel, W. R. Eur. J. Inorg. Chem.

2011, 29, 4603−4609.

(32) (a) Ditchfield, R.; Hehre, W. J.; Pople, J. A. J. Chem. Phys. 1971,

54, 724−728. (b) Hehre, W. J.; Ditchfield, R.; Pople, J. A. J. Chem.

Phys. 1972, 56, 2257−2261. (c) Hariharan, P. C.; Pople, J. A. Theor.

Chim. Acta 1973, 28, 213−222. (d) Clark, T.; Chandrasekhar, J.;

Spitznagel, G. W.; Schleyer, P. v. R. J. Comput. Chem. 1983, 4, 294−

301. (e) Rassolov, V. A.; Pople, J. A.; Ratner, M. A.; Windus, T. L. J.

Chem. Phys. 1998, 109, 1223−1229. (f) Binning, R. C., Jr.; Curtiss, L.

A. J. Comput. Chem. 1990, 11, 1206−1216. (g) Dunning, T. H. J.

Chem. Phys. 1977, 66, 1382−1383.

(33) (a) Dolg, M.; Wedig, U.; Stoll, H.; Preuss, H. J. Chem. Phys.

1987, 86, 866−872. (b) Andrae, D.; Haußermann, U.; Dolg, M.; Stoll,

̈

H.; Preuss, H. Theor. Chim. Acta 1990, 77, 123−141.

(34) (a) Peng, C.; Schlegel, H. B. Isr. J. Chem. 1994, 33, 449−454.

(b) Peng, C.; Ayala, P. Y.; Schlegel, H. B.; Frisch, M. J. J. Comput.

Chem. 1996, 17, 49−56.

(35) (a) Fukui, K. Acc. Chem. Res. 1981, 14, 363−368. (b) Hratchian,

H. P.; Schlegel, H. B. In Theory and Applications of Computational

Chemistry: The First 40 Years; Dykstra, C. E., Frenking, G., Kim, K. S.,

Scuseria, G., Eds.; Elsevier: Amsterdam, The Netherlands, 2005; pp

195−249. (c) Hratchian, H. P.; Schlegel, H. B. J. Chem. Phys. 2004,

120, 9918−9924. (d) Hratchian, H. P.; Schlegel, H. B. J. Chem. Theory

Comput. 2005, 1, 61−69.

(36) (a) Krishnan, R.; Binkley, J. S.; Seeger, R.; Pople, J. A. J. Chem.

Phys. 1980, 72, 650−654. (b) Curtiss, L. A.; McGrath, M. P.;

Blaudeau, J.-P.; Davis, N. E.; Binning, R. C.; Radom, L. J. Chem. Phys.

1995, 103, 6104−6113. (c) Blaudeau, J.-P.; McGrath, M. P.; Curtiss, L.

A.; Radom, L. J. Chem. Phys. 1997, 107, 5016−5021.

(37) Weigend, F.; Ahlrichs, R. Phys. Chem. Chem. Phys. 2005, 7,

3297−3305.

(38) (a) Barone, V.; Cossi, M. J. Phys. Chem. A 1998, 102, 1995−

2001. (b) Cossi, M.; Rega, N.; Scalmani, G.; Barone, V. J. Comput.

Chem. 2003, 24, 669−681.

(39) (a) Klamt, A.; Schuurmann, G. J. Chem. Soc., Perkin Trans. 2

̈

1993, 799−805. (b) Schafer, A.; Klamt, A.; Sattel, D.; Lohrenz, J. C.

̈

W.; Eckert, F. Phys. Chem. Chem. Phys. 2000, 2, 2187−2193.

(40) Wong, M. W. Chem. Phys. Lett. 1996, 256, 391−399.

(41) Murrieta-Guevara, F.; Romero-Martinez, A.; Trejo, A. Fluid

Phase Equilib. 1988, 44, 105−115.

(42) Johnson, E. R.; Mackie, I. D.; DiLabio, G. A. J. Phys. Org. Chem.

2009, 22, 1127−1135.

(43) Grimme, S.; Antony, J.; Ehrlich, S.; Krieg, H. J. Chem. Phys.

2010, 132, 154104.

(44) Riley, K. E.; Vondrasek, J.; Hobza, P. Phys. Chem. Chem. Phys.

2007, 9, 5555−5560.

(45) GaussView 5.0.8; Gaussian, Inc.: Wallingford, CT, 2008.

(46) See the Supporting Information for details.

Q

dx.doi.org/10.1021/ja503295x | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article Doi

Mechanism of Cu/Pd-catalyzed decarboxylative cross-couplings: A DFT investigation

DOI: 10.1021/ja503295x

Source and publish data:

Authors:

Article abstract of DOI:10.1021/ja503295x

Full text of DOI:10.1021/ja503295x

Products guided by the article

R&D Labs maybe for 26969-40-4

Relevant to this article

Hot Product