Pd-leaching and Pd-removal in Pd/C-catalyzed Suzuki couplings

www.elsevier.com/locate/apcata

Applied Catalysis A: General 325 (2007) 76–86

Pd-leaching and Pd-removal in Pd/C-catalyzed Suzuki couplings

Jeng-Shiou Chen, Aleksey N. Vasiliev, Anthony P. Panarello, Johannes G. Khinast *

Department of Chemical and Biochemical Engineering, Rutgers, The State University of New Jersey, Piscataway, NJ 08854, USA

Received 3 November 2006; received in revised form 7 March 2007; accepted 11 March 2007

Available online 15 March 2007

Abstract

Pd-leaching in Pd/C-catalyzed Suzuki couplings was investigated using the model coupling reaction of biphenylacetic acid. The filtration test

was used to prove that oxidative addition of aryl-bromides is the main cause for Pd-leaching, which is independent of the reaction solvent and

temperature. In addition, the oxidative addition of aryl-borates is another cause for Pd-leaching. PVPy adsorption studies suggest that the activity of

Pd/C is mainly due to leached Pd. Furthermore, PVPy was proven to be a good reagent for complete removal of Pd-residuals from the reaction

mixture. Excess PVPy (3 equiv. to Pd) is sufficient to carry out the removal within 2 h. The influence of oxygen on reactions was also investigated.

The kinetic results suggest that exclusion of oxygen is not necessary. The presence of water in the solvent is required to promote the reaction. Water

may stabilize Pd-nanoparticles, which can act as a reservoir for active Pd species.

# 2007 Elsevier B.V. All rights reserved.

Keywords: Suzuki couplings; Palladium; Leaching; Oxidation addition; Biphenylacetic acid

1. Introduction

Suzuki couplings, i.e., the formation of carbon–carbon

bonds during the coupling of organo-borates with organo-

halides, organo-triflates or organo-tosylates, have become an

efficient and clean strategy for the preparation of biologically

active functionalized biphenyls, which are important building

blocks for pharmaceutical and agricultural compounds [1–5].

Typical examples are the tyrosine kinase [6] and the

phosphodiesterase IV inhibitors [7]. Suzuki couplings are also

important for the synthesis of functionalized materials and

supported catalysts, e.g., for the immobilization of organome-

tallic complexes onto heterogeneous supports [8].

Transition metals, such as palladium-, nickel- [9], and

platinum-complexes [10] are often the catalysts of choice.

Among these, palladium on carbon (Pd/C) is most frequently

used for industrial applications due to its high catalytic activity,

low costs and commercial availability. Scanning transmission

electron microscopy (STEM) pictures of commercial Pd/C

catalysts show irregular palladium clusters of different sizes,

* Corresponding author. Present address: Graz University of Technology,

Austria. Tel.: +43 316 873 7978 (Austria)/+1 732 445 2970 (USA);

fax: +1 732 445 8025 (USA).

E-mail addresses: [email protected], [email protected] (J.G. Khinast).

0926-860X/$ – see front matter # 2007 Elsevier B.V. All rights reserved.

doi:10.1016/j.apcata.2007.03.010

randomly located on the carbon surface [11]. Thus, Pd/C is

considered a heterogeneous catalyst, enabling straightforward

removal of the catalyst from the reaction mixture with

insignificant Pd-residuals [12,13]. This notion, however, has

been challenged by several groups during recent years. For

example, the group of Sun at Merck proposed that Pd/C is a

quasi-heterogeneous catalyst [14], where leached Pd catalyzes

the reaction and re-adsorbs onto the carbon support after

completion of the reaction.

Pd-leaching from Pd/C was also observed in Heck reactions

[15–18], hydrodechlorinations [19] and during the isobutanol

synthesis [20]. Particularly for Heck reactions, detailed studies

of the Pd-leaching from the heterogeneous catalyst

[17,18,21,22] and the Pd-reabsorption by the support [23,24]

have been conduct.

In contrast to Heck reactions, a detailed experimental study

of Pd-leaching during Suzuki couplings with Pd/C has not been

performed. The objective of this work is to achieve a detailed

understanding and offer experimental evidence of Pd-leaching

in Pd/C-catalyzed Suzuki couplings. Furthermore, we intended

to provide a strategy for homogeneous Pd (Pd-residual)

removal from the reaction mixture after the reaction has been

completed. This is of interest especially for pharmaceutical

applications, where carryover of metal impurities may cause

serious problems in the production of many formulations.

mailto:[email protected]

mailto:[email protected]

http://dx.doi.org/10.1016/j.apcata.2007.03.010

Table 1

Measurement of the molar enthalpy of the preparation of biphenylacetic acids in

an RC1

Runs Conditiona DH (kJ/mol)b Deviation (%)c

1 1�, 65 8C �225.85 0.8

2 2�, 65 8C �223.61 �0.2

3 4�, 65 8C �222.57 �0.6

a 1� refers to 0.029 mol 1, 0.036 mol 2, 0.068 mol of Na2CO3, and

0.45 mmol Pd/C in 500 ml mixture of 10% IPA in water. 2� and 4� stand

for two and four times the amounts of 1, 2 and Na2CO3 with the same amount of

Pd/C.b Molar enthalpy calculated from reaction enthalpy of the RC1.c Deviation from the average molar heat of reaction, which equals

�224.01 kJ/mol.

J.-S. Chen et al. / Applied Catalysis A: General 325 (2007) 76–86 77

2. Experimental

2.1. Materials

In our study, we used commercially available 4-bromophe-

nylacetic acid (Aldrich, 98%), 4-iodiobenzoic acid (Aldrich,

99%), phenylboronic acid (Aldrich, 95%), Pd/C (Aldrich,

5 wt%, Degussa type, particle size 28–34 mm), palladium(II)

acetate (Aldrich, 99.9%), sodium carbonate (Fisher), trimethy-

lamine (Aldrich, 99.5%), 1,4 dioxanes (Fisher), DMA (Aldrich,

99%), DMF (Acros, 99%), methanol (Fisher), 2-propanol

(Fisher), THF (Alfa Aesar, 99.8%), toluene (Fisher), PVPy

(Aldrich, 2% crosslinked, powder, 60 mesh). All materials did

not require further purification.

2.2. Model reactions for Suzuki couplings

Model Suzuki couplings were carried out in a 100 ml

round-bottom glass reactor with a 1 cm-length stirring bar.

Both the reactor and the stirring bar were cleaned thoroughly

with acetone and 6 M hydrogen chloride solution and dried in

110 8C oven for 4 h before use. Phenylacetic acid (2.9 mmol),

phenylboronic acid (3.2 mmol) and sodium carbonate

(6.4 mmol) were added to 50 ml reaction solvent (IPA:water

1:9 volume ratio) at room temperature. After adding Pd/C

(0.01 mmol), the reaction mixture was heated to 65oC to carry

out the reaction. The reaction temperature was controlled by a

light mineral oil (Fisher) bath. Reactions were carried out at

least twice to verify reproducibility of the data. Variations

were less than 5% (in one experiment 7%) of the reported

values.

2.3. Analytical measurements

High performance liquid chromatograph (HPLC) analyses

were performed on a Shimadzu SP-10 liquid chromatograph

equipped with a UV–vis adsorption detector and an Aligent

zorbax eclipse XDB-C8 column. The HPLC mobile phases

were 40% deionized water with 0.05% formic acid (Aldrich,

96%) and 60% methanol (Fisher) in volume ratio. The flow rate

of the mobile phase was 1.2 ml/min. The wavelength of the

detection light was set to 208 nm. 0.1 ml samples were

extracted from the reactor over the course of the reaction. The

samples were diluted in 20 ml of a 60% methanol/40% water

solution and then filtered to remove the catalyst. The conversion

of the reaction was based on the consumption of the aryl-halide.

If cA is the concentration of the aryl-halide, the conversion Y is

defined as Y = [(cA,0 � cA)/cA,0].

2.4. Kinetic studies

Kinetic studies were carried out in a reaction calorimeter

(ASI/Mettler-Toledo RC1) equipped with a 1-l jacketed glass

reaction vessel containing a Hasteloy head and an anchor

impeller. Reaction temperature and agitation speed were held

constant at 65 � 0.02 8C and 250 rpm to ensure complete

suspension of the catalyst. Independence from the agitation

speed was observed above 200 rpm. The reaction scale was

500 ml. Initial concentrations of the reactants and reagents

used for kinetic studies in the RC1 were the same as those

given in Section 2.2. The reaction mixture was heated to

65 8C first, and then Pd/C was added to carry out the reaction.

Thermal data were collected every 2 s, and conversion was

estimated from the generated heat of reaction. Isothermal

conditions (dTr/dt = 0, Tr is the temperature inside the reactor)

were achieved by determining and calibrating all external

thermal influences (except for the heat of reaction), including

mixing events. Then, the heat flow q is directly proportional

to the reaction rate. The heat capacity of the reactor contents,

cp,r, was measured by subjecting the reactor contents to a

known temperature ramp. Then, the (measurable) heat flow q

through the reactor wall equals the enthalpy accumulation.

Comparison of cp,r and the heat transfer coefficient before and

after the reaction showed no significant changes, i.e.,

consistent conditions existed over the course of the reaction.

The conversion was calculated using the measured heat flow,

q(t), i.e.,

YðtiÞ ¼R ti

t0qðtÞ dt

R tE

t0qðtÞ dt

; (1)

where t0 is the start time, and tE is the time, after which no

thermal events were recorded any more.

In order to test the RC1, three reactions with different initial

concentration of the reagents were carried out, as shown in

Table 1. Good consistency was observed, with the average

molar enthalpy being �224.01 kJ/mol and the maximum

deviation of less than 1%. As the heat of reaction for this case is

not tabulated, DFT calculations in vacuo with the model

B3LYP/6-311+G(d,p) were carried out with Gaussian 03. A

molar heat of reaction of �233.64 kJ/mol was obtained, which

is in good agreement with the experimental data, given that no

detailed solvation model was used.

HPLC analysis was also used to validate the calorimetry

data. Since the reaction continued until the catalyst was

removed by filtration, the HPLC data had to be corrected by a 2-

min delay. Fig. 1 compares the time-corrected HPLC

conversion data with those generated by the RC1. As can be

seen, the agreement is excellent.

Fig. 1. Conversion of 4-bromophenylacetic acid as a function of time for

reactions at 55 and 65 8C from RC1 data and time-corrected HPLC measure-

ment. Fig. 2. Conversion of 4-bromophenylacetic acid as a function of time under

standard conditions (&) and removed Pd/C after 8 min (*).

J.-S. Chen et al. / Applied Catalysis A: General 325 (2007) 76–8678

3. Results and discussion

3.1. Pd-leaching of Pd/C

Biffis et al. [25] studied Heck reactions with Pd-supported

catalysts and observed Pd-leaching only in the presence of an

aryl-halide. On the basis of their work, Conlon et al. [14]

proposed the hypothesis that Pd-leaching from Pd/C occurs

after oxidative addition of the bromide. In our study, using the

model coupling reaction of biphenylacetic acid 3 (Scheme 1),

we provided experimental evidence to (partially) support this

hypothesis. Furthermore, we carried out a detailed investigation

of the mechanism of Pd-leaching from Pd/C during Suzuki

coupling reactions.

First, the existence of Pd-leaching during our model reaction

was verified using the filtration test, which is a straightforward

method to distinguish homogeneous and heterogeneous

catalytic activity, based on the comparison of the reaction

progress before and after removal of the solid phase [26]. If the

reaction proceeds after the removal of the solid catalyst – in this

case Pd/C – by filtration, this is clear evidence that leaching

forms homogeneous Pd in the filtrate which catalyzes the

reaction. As shown in Fig. 2, nearly complete conversion of 1

was observed around 15–20 min after the catalyst was removed

using 0.45 mm syringe filters, which guaranteed a complete

removal of solids (Pd/C particle size is 28–34 mm). In another

experiment, a reaction was carried out under standard

conditions. After 8 min, Pd/C was removed from the reaction

mixture, at a point when the conversion of 1 was 30%, but the

Scheme 1. The model reaction for Pd/C-catalyzed Suzuki couplings.

reaction continued and reached 91% conversion after another

32 min. Inductively coupled plasma (ICP) analysis indicated

that less than 1 ppm leached Pd existed in the filtrate after

catalyst removal. These results clearly show that (1) a minute

amount of leached Pd can catalyze a reaction and (2) Pd-

leaching does occur. Furthermore, this implies that leached Pd

plays an important role in the high activity of Pd/C for Suzuki

couplings.

(Note: there may be concern that vigorous mixing in the

reaction flask could produce small Pd/C fragments that may

pass through the syringe filter, thus biasing our results.

However, experiments carried out without some of the reactants

initially present (reported below) give no conversion after

filtration and after all reactants were added. Had small

fragments of Pd/C passed the syringe, the conversion would

have been high. Consequently, the syringe filter effectively

removes 100% of the solid phase and remaining activity is due

to leached Pd.)

Since Pd leaching was observed during the reaction, we

wanted to understand the cause of it. Various components may

affect Pd-leaching, such as the aqueous co-solvent, the aryl-

bromide, the aryl-borate and the base. All of these factors were

examined and screened one by one. The impact of the aqueous

co-solvent and of the reaction temperature was first examined.

In order to observe significant Pd-leaching, 1 g of Pd/C (i.e., 15

times the regular amount) was used in each experiment. Two

experiments were setup, where Pd/C was added to two flasks

with only 50 ml of IPA/water (1:9 volume ratio). These

mixtures were then stirred for one hour at room temperature

(RT) and at 75 8C. Then, Pd/C was removed by filtration and the

filtrate was used as the reaction solvent for a regular reaction

without adding any additional catalyst. If both experiments

showed a low conversion, this would mean that both the solvent

and the temperature have no effect on Pd-leaching, i.e.,

leaching is not caused by the dissolution of Pd atoms or clusters.

The conversions of the reactions at RT and at 75 8C runs after

2 h were 6% and 4% (Table 2), which indicates Pd-leaching

does not occur during the premixing. This strongly suggests

Table 2

Conversions for premixing schemes with different components of the reaction

mixture

Runs Premixing conditions with Pd/C Conversiona (%)/Reaction hours

1 IPA/water, at RT 6/2 h

2 IPA/water, at 75 8C 4/2 h

3 1, IPA/water, at 75 8C 96b/1 h

4 2, IPA/water, at 75 8C 98b/1 h

5 Na2CO3, IPA/water, at 75 8C 0/4 h

a Conversion refers to the conversion of 1.b Conversions were calculated based on different limiting reactants in each

run. In Run 3, 1 was ca. 93% recovered from the premixing and while 2 was ca.

52% recovered in Run 4; thus, the limiting reactant in Run 3 was 1 and that in

Run 4 was 2.

Scheme 2. Proposed catalytic cycle of Pd/C-catalyzed Suzuki couplings. I:

Oxidative addition; I0: Pd-leaching due to oxidative addition; II: metathesis; III:

transmetalation; IV: reductive elimination; V: arylborate activation; VI: Pd-

readsorption. The hydroxide ion, OH�, comes from dissociation of the base,

Na2CO3, NaOH, K2CO3, etc.


that Pd-leaching does not depend on the aqueous co-solvent and

the temperature.

These experiments clearly suggest that one or more of the

reactants, i.e., the aryl-halide, the aryl-borate or the base, are

the factors causing Pd-leaching. To investigate these possibi-

lities, we tested which reactant premixed with Pd/C in IPA/

water causes Pd-leaching. Three experiments were carried out,

where in each flask 1 g of Pd/C was added to 50 ml of IPA/water

(1:9 volume ratio) together with either phenylacetic acid 1,

phenylboronic acid 2, or Na2CO3. The mixtures were stirred for

1 h at 75 8C. Then, Pd/C was removed by filtration and the

filtrate was used as the reaction solvent for a new reaction by

adding the remaining reactants (for example, 2 and Na2CO3

were added, when 1 was premixed with Pd/C) without an

additional catalyst. Conversions of these runs are shown in

Table 2. The conversions of the runs of premixing Pd/C with 1

or 2 after 1 h were 96% and 98%, which shows that

homogeneous Pd exists in the filtrates and that Pd-leaching

occurs in the presence of both aryl-bromide and aryl-borate. A

0% conversion after premixing with Na2CO3 indicates that the

base did not cause Pd-leaching.

In the case of premixing Pd/C with 1 (initially, only Pd/C and

the aryl-bromide are in the reaction solvent), oxidation addition

of the bromide to the heterogeneous Pd clusters could occur.

Thus, this experiment indicates that oxidative addition of the

aryl-bromide is one critical step causing Pd-leaching, which

also supports Conlon et al.’s hypothesis. However, it was

interesting to see that Pd-leaching also occurred in the presence

of the aryl-borate, which is different from the previous case, as

in this system no step in the proposed catalytic cycle (see

Scheme 2 in Section 3.2) should occur. We thus further

analyzed the reaction mixture and found 42% yield of

biphenyls, which is the product of the self-coupling of 2. This

indicates that self-coupling of aryl-borates occurred during the

premixing phase. In accordance with Moreno-Manas et al.’s

proposed catalytic cycle for the self-coupling of aryl-borates

[27], this run demonstrates that oxidative addition of aryl-

borate is another factor causing Pd-leaching.

These results showed that Pd-leaching from Pd/C is mainly

caused by oxidative addition, which occurs independently both

with the bromide and the borates. This motivated us to test if the

self-coupling of aryl-borates occurs in parallel to the Suzuki

coupling. If it does, both 3 and the biphenyl should be seen after

completion of the reaction. However, HPLC analysis confirmed

the absence of biphenyls for our model system, which suggests

that self-coupling of the aryl-borates is either suppressed by the

competitive addition of the halide, or is significantly slower

than the Suzuki coupling pathway. Clearly, oxidative addition

of the aryl-bromides is the main factor causing Pd-leaching

during Suzuki couplings. However, when the aryl-bromide is

absent, self-coupling of aryl-borates becomes dominant, and

oxidative addition of aryl-borates is the main factor for

leaching. This has to be considered during the setup of reaction

protocols.

In addition, we studied Suzuki couplings of aryl-iodides and

aryl-borates. Instead of 1, 4-iodobenzoic acid was used together

with 2 in IPA/water and Na2CO3 under standard conditions.

The conversions of the regular reaction after 10 min and 1 h

were 42% and 100%, respectively. In the filtration test, the

conversion still reached 100% after 1 h, when Pd/C was

removed after 10 min. This indicates that Pd-leaching also

occurs during this reaction. Therefore, Pd-leaching can be

observed in Suzuki couplings of both aryl-bromide and aryl-

iodides.

3.2. Active material of Pd/C

As shown in the literature and confirmed by our experi-

ments, Pd/C can no longer be viewed as an entirely

heterogeneous catalyst. There are two possible scenarios: (1)

there is both a homogeneous (leached Pd) and a heterogeneous

(supported Pd) contribution or (2) only leached Pd is active and

catalyzes the reactions. The proposed mechanism of Pd/C-

catalyzed Suzuki coupling reactions by Ref. [14] is shown in

Scheme 2. Four steps occur in the main cycle: oxidative

addition (I), metathesis (II), transmetalation (III) and reductive


elimination (IV). Steps I0 and VI describe Pd-leaching due to

oxidative addition of the halides and the re-adsorption on the

support after completion of the reaction. Thus, the proposed

cycle implies that leached Pd is the only active material

catalyzing the reaction. However, there is a lack of experi-

mental evidence supporting this. Thus, in order to rule out the

‘‘two-pathway scenario’’, PVPy adsorption experiments were

carried out.

Poly(4-vinylpyridine) or PVPy, a solid polymer which is

insoluble under reaction conditions, has been used to

investigate Pd-leaching from Pd catalysts in Heck reactions

[17]. In their work PVPy was shown to deactivate free Pd(II)

(leached palladium) in the solution, bringing to a halt

homogeneous palladium activity. Furthermore, a mechanism

of Pd deactivation by PVPy was proposed [18] and it was

reported that high catalyst activity was observed in the presence

of molecular pyridine, while there was no reactivity in the

presence of PVPy. Because pyridines are known to easily bind

to Pd(II) [28] and molecular pyridine is liquid under reaction

conditions, Yu et al. [18] concluded that PVPy deactivates

homogeneous Pd by adsorbing and removing it from the

solution. Therefore, by impeding the homogeneous pathway,

PVPy can be helpful to distinguish between heterogeneous and

homogeneous contributions to the reaction.

In performing the experiments, we initially assumed that

both types of palladium, i.e., the leached and the heterogeneous

species, are active. Scheme 3 illustrates the heterogeneous

reaction pathway and the homogenous one. The heterogeneous

reaction occurs only on the surface of the palladium nano-

clusters adsorbed on the carbon support. In this pathway, the

surface atoms of the palladium nano-clusters participate in the

first step of the Suzuki couplings, the oxidative addition of aryl-

bromides. Palladium then becomes Pd(II) and catalyzes the rest

of the cycle. It is assumed that the carbon support and the

adjacent Pd(0) firmly bond the Pd(II) species and prevent it

from leaching during the course of the reaction. In the

Scheme 3. Schematic of heterogeneous and homogeneous Suzuki couplings,

homogenous pathway, palladium leaches from the support to

the solution, after becoming a soluble Pd(II) species [14], and

then catalyzes the remaining steps of the reaction.

Although PVPy is known to deactivate homogeneous Pd, a

priori there is the possibility that PVPy deactivates hetero-

geneous Pd as well. However, PVPy is in the form of large

insoluble polymer particles (250 mm), which may hardly

interact with the nanoclusters on the carbon support.

Furthermore, the light soluble reactants can react much faster

(based on diffusivity and steric arguments) than the PVPy.

Thus, we conclude that the possibility of PVPy interacting with

heterogeneous Pd is negligible. Scheme 4 illustrates our

proposed mechanism of PVPy poisoning of the Suzuki reaction,

indicating that only the homogeneous pathway is affected.

The mechanism of homogeneous Pd deactivation by PVPy

was further investigated by using the units of PVPy, i.e., 4-

ethylpyridine, instead of molecular pyridine. 4-Ethylpyridine is

liquid under reaction conditions. With an excess of 4-

ethylpyridine (150 equiv. to Pd), 0% conversion was obtained

after 2 h. The complete deactivation of Pd/C with 4-

ethylpyridine indicates that 4-ethylpridine deactivates Pd by

blocking the access to the metal center. This result is in contrast

to Yu et al.’s finding [18] that the Pd catalyst still showed

activity for Heck reactions in the presence of molecular

pyridine. In order to understand the difference between their

results and ours, we modeled both the 4-ethylpyridine-Pd and

pyridine-Pd complexes using Gaussian 03 (DFT with B3LYP/

(6-31G,SDD)). Leached Pd(II) species was assumed in the form

of R–Pd–Br, where R refers to the aryl-group of 1. Then, one,

two or four molecules of 4-ethylpyridine were added to Pd

center, respectively. DFT calculations yielded a molar heat of

reaction of �196.80 kJ/mol for the addition of one 4-

ethylpyridine molecule to Pd(II) species and �261.19 kJ/mol

for cis-addition of two molecules and �355.29 kJ/mol for

the trans-addition. However, we could not obtain any

optimal geometry of Pd(II) octahedral complexes with four

catalyzed by Pd/C. The proposed mechanisms for Pd-leaching is shown.

Scheme 4. Schematic of the proposed mechanism of Pd deactivation by PVPy for heterogeneous Pd and homogeneous Pd, respectively.


4-ethylpyridine molecules, which is consistent with the planar

coordination geometry of the Pd(II) complexes. The same trend

was observed for molecular pyridine (�192.32 kJ/mol for one

molecule addition, �255.81 kJ/mol for cis-addition of two

molecules, �348.76 kJ/mol for trans-addition). In both cases,

the results suggest trans-addition of two (ethyl)pyridine

molecules being the preferred form of the complex [29].

Scheme 5 shows these structures. Comparison of the HOMO-

LUMO gap, bond lengths, bond angles, and dihedral angles of

the two complexes revealed little difference; in fact they were

almost identical. This suggests that molecular pyridine should

deactivate Pd in the same way as 4-ethylpyridine in Pd/C-

catalyzed Suzuki coupling reactions.

Thus, two additional experiments in the presence of

molecular pyridine were carried out. In the first experiment

with excess pyridine (150 equiv. to Pd), 0% conversion was

Scheme 5. Optimal geometries of the complexes of Pd(II) species and

obtained after 1 h. In the second experiment the conversion

remained at only 35% after 1 h, when the same amount of

excess molecular pyridine was added after 8 min. Both

experiments clearly indicate that molecular pyridine also

deactivates Pd(II).

The discrepancy between Yu et al.’s [18] and our results may

be due to the difference in reaction kinetics and the reaction

solvents. The Heck coupling of n-butyl acrylate and

iodobenzene is a much faster reaction, which reaches

completion in merely 10 min. A careful analysis of their

results shows that the presence of molecular pyridine did slow

down the reaction by 15 min (the reaction reached completion

in ca. 25 min). This implies that the fast Heck reaction is

competing with the complex formation (deactivation). In other

words, the observed activity of the catalyst in the presence of

molecular pyridine might be due to active leached Pd from the

pyridines by DFT calculation with model B3LYP/(6-31G, SDD).


catalyst before being poisoned. Another possible explanation

for the discrepancy of the results is the use of different solvents

in the both studies. In summary, it can be concluded that PVPy

deactivates Pd by forming complexes with homogeneous

Pd(II), thus inhibiting the catalytic cycle.

By comparing reactions in the presence and absence of

PVPy, we intended to quantify the ratio of the homogeneous

and heterogeneous catalytic pathway. Four experiments were

carried out in the RC1, one without PVPy addition and three

with different excess amounts of PVPy (PVPy:Pd is 100, 200

and 300 to 1 in molar equivalence, based on the molecular

weight of 4-vinylpyridine.) PVPy was added first, followed by

Pd/C to start the reaction. RC1 data indicated that the

conversion of biphenylacetic acid after 10 min was 99%, 12%,

1% and 4%, and the time to reach 99% conversions was 0.2, 1.6,

6.9 and 12.7 h, respectively. The conversion plots are shown in

Fig. 3. The plots clearly show that the more PVPy is added, the

slower reaction becomes. The fact that PVPy reduces

dramatically the activity of Pd/C indicates that the main

activity of Pd/C is due to the homogeneous pathway. In

addition, the initial lag phases (see Fig. 3b) suggest that at first a

certain homogeneous Pd concentration has to build up in order

Fig. 3. (a) RC1 conversion data of 4-bromophenylacetic acid as a function of

time for different amounts of PVPy. (b) RC1 conversion data of 4-bromophe-

nylacetic acid as a function of time for different amounts of PVPy (enlarged).

to catalyze the reaction. Nevertheless, even large excess

amounts PVPy are not able to stop the reaction completely,

which may be explained by three possible scenarios: (I)

heterogeneous Pd still catalyzes the reactions, albeit very

slowly, (II) Pd in the form of PVPy-Pd is still slightly active, or

(III) the adsorption of Pd is slow compared to the reaction, i.e.,

Pd molecules have a statistical time window to catalyze a few

cycles before being captured by PVPy.

Scenario II can be excluded, since there is a clear inverse

relation between reaction rate and PVPy concentration. Since

PVPy is in excess, its concentration should not influence the

rate, if indeed a PVPy-Pd complex catalyzes the reaction. In

order to discriminate between the other possible mechanisms,

additional tests were performed. However, this time Pd(OAc)2

was used. Pd(OAc)2, palladium(II) acetate, is a well-known

homogeneous catalyst for Suzuki couplings. Four reactions

catalyzed by Pd(OAc)2 were carried out as follows: (1) reaction

without PVPy; (2) 200 equiv. PVPy was added initially,

followed by Pd(OAc)2; (3) 200 equiv. PVPy and Pd(OAc)2

were premixed in the reaction solvent overnight. Then the

filtered and dried solid mixture of PVPy and Pd(OAc)2 was

used for a reaction; (4) the solid mixture from reaction 3 (used

PVPy + Pd(OAc)2) was recycled and used as the catalyst for

another reaction. RC1 studies indicated that the conversion

after 10 min was 99%, 99%, 96% and 0%, respectively. Also,

the reaction was very fast, i.e., the time required to reach 99%

conversion was 2.5, 4.5, 20 min and infinity. The trends are

shown in Fig. 4. From this series of reactions, three conclusions

can be drawn. First, reactions 1 and 2 show almost identical

reactivity, which indicates that the catalytic reaction dominates

Pd-capture by PVPy. This explains why PVPy cannot

completely poison the system, as the homogeneously catalyzed

Suzuki coupling is a fast reaction. Second, the high reactivity of

reaction 3 indicates that PVPy is not able to bind Pd(II) from

Pd(OAc)2 in the absence of a reaction occurring. This may be

due to the protection by the acetyl groups. Third, 0% conversion

of reaction 4 proves that Pd in the form of PVPy-Pd is entirely

inactive.

Fig. 4. RC1 conversion data of 4-bromophenylacetic acid as a function of time

for different PVPy addition schemes with Pd(OAc)2.

Fig. 5. Conversion of 4-bromophenylacetic acid as a function of time under

PVPy treatment at 65 8C (&) and at room temperature (*) when Pd/C was

removed after 8 min.


In summary, our careful experiments indicate that Pd-

leaching is a critical step in the Pd/C-catalyzed Suzuki

couplings in aqueous solvents, and that Pd adsorption by PVPy

is not fast enough to entirely deactivate leached Pd, which is

consistent with Yu et al.’s finding [17] that PVPy cannot

completely poison the catalyst in Heck reactions. Therefore, we

conclude that the remaining reactivity we observed during the

Pd/C reactions with PVPy is due to leached, homogeneous Pd

that has not yet been captured by PVPy. Pd has two possible

pathways after leaching from the support: one is to react with

reactants and to contribute to the reaction, and the other is to

bind to PVPy and become inactive. These two mechanisms

compete with one other, as shown in Scheme 6. This

observation also implies that there should be a pseudo-steady

state between heterogeneous Pd on the support and homo-

geneous Pd in the solution, which is determined by the leaching

and adsorption rate.

3.3. Removal of leached homogeneous Pd

Pd/C is an increasingly important catalyst for the preparation

of fine chemicals and pharmaceuticals. However, only a few

ppm of Pd detected in the reaction mixture may seriously

contaminate the products, violating FDA guidelines. This is due

to the possibility of Pd forming complexes with drug

molecules, thus inactivating the drug, and to the general wish

to deliver heavy-metal-free drugs, as metals are considered

toxic and/or carcinogenic. Thus, it is important to develop a

technique for removing Pd-residual from the final reaction

mixture. The ability of PVPy to bind to homogeneous Pd is a

potential solution, as it is a straightforward method for

removing Pd-residuals from the reaction mixture. Two

experiments with PVPy at 65 8C and at room temperature

were carried out to demonstrate this approach. In both cases, the

reactions were setup under standard conditions. After 8 min Pd/

C was removed, and PVPy (200 equiv. to Pd by assuming

1 ppm leached Pd in the filtrate) was added and mixed at 65 8Cor RT, respectively, to observe PVPy effects on the leached-Pd-

catalyzed reaction. After another 1 h, PVPy was removed from

solution by filtration and the conversion was monitored for

another hour. The results are shown in Fig. 5. After removal of

Pd/C and addition of PVPy, the reactions in both cases stopped

within 10 min. Comparison to the reaction of removed Pd/C

(without adding PVPy) in Section 3.1 implies that PVPy is

adsorbing and deactivating leached/homogeneous palladium.

Taking a closer look at the 65 8C case, a slight increase of the

conversion during the first 10 min (29–33%) shows that at a

higher reaction temperature the adsorption of Pd by PVPy is

Scheme 6. The pathways of Pd after leaching. kl, kde are the leaching and

deactivation constants.

slow compared to the reaction. After removal of PVPy, the

reaction mixture was kept at 65 8C in both cases for one more

hour. In both cases, the conversion remained constant and no

further reaction occurred after removing PVPy. This indicates

that that PVPy can successfully and irreversibly remove

homogeneous palladium from the reaction mixture. Since Pd

adsorption by PVPy works in both cases, for economic and

practical consideration, using PVPy at RT is sufficient for

removing palladium.

Furthermore, the absorption capacity of PVPy for Pd was

investigated. A solution of 1001 ppm of Pd in a 5% HCl

solution (from Aldrich) was prepared. The orange-brown color

of the Pd solution showed abundant free Pd(II) in the solution.

First, two experiments were carried out, where twice 10 ml of

the Pd solution was added to two flasks. Excess PVPy (50 and

10 equiv. to Pd) was added to the flasks, respectively, and the

mixtures were stirred at RT for 3 h. Then, PVPy was filtered and

the color of the filtrate was observed. In both cases, the filtrates

were observed to be colorless, which implies that Pd was

completely removed. As a result, one more experiment was

carried out. The PVPy amount was reduced to 2 equiv. Pd and

the mixture was stirred for 2 h only. In this run, the filtrate still

had a light yellow-orange color. By ICP analysis of the light

yellow-orange filtrate, Pd amount was determined to be

172 ppm. This shows that the equilibrium molar ratio of Pd

removed to PVPy is 0.38/1 after 2 h, which is the maximum

ability of PVPy for Pd removal. Thus, for complete Pd removal

from the solution, approximately 3.0 equiv. of PVPy should be

sufficient. PVPy is commercially available for about 700 US$/

kg. Assuming a typical 400-gal batch with 10 ppm Pd-residual

in the solution, about 45 g of PVPy are required to completely

remove Pd. This would translate into costs of about US$ 30.

3.4. Oxygen effect on the Pd/C-catalyzed reaction

Dissolved gas from the headspace of the reactor is usually

considered an important factor that may affect the reaction

Fig. 6. Conversion of 4-bromophenylacetic acid as a function of time under

regular conditions (&), N2 purged (*), sparging Air during the entire reaction

(~), and sparging N2 during the entire reaction (!).


performance. Because palladium is an easily oxidized metal,

Pd-catalyzed reactions are typically carried out under inert gas

atmosphere [17,18] or vacuum [14] to prevent oxygen from

poisoning the Pd catalysts. For Pd/C-catalyzed Suzuki

couplings, the system is usually degassed before the experi-

mental study. However, Gala et al. [12] noted in their study of

Pd/C-catalyzed Suzuki couplings that degassing of the solvents

is unnecessary. This motivated us to investigate the impact of

oxygen on the reaction.

In one experiment, a regular reaction was setup under air

atmosphere. The flask was stirred for 20 min at RT and then

heated up to 60 8C to carry out the reaction. As shown in Fig. 6,

the reaction went to completion after about 20 min. Another

reaction was setup under the same conditions but purged with

N2 for 20 min, and then heated up to 60 8C. The conversion

plots show that the reaction in the absence of oxygen is very

similar (slightly slower), compared to the reaction in the

presence of oxygen. This implies that oxygen may have little or

no effect on the reactions. Moreover, under the same conditions

as the regular reaction at 60 8C, two more reactions were

carried out, one by sparging air and one by sparging N2 during

the entire reactions (gas flowrate: 650 ml/min). The conversion

plots, as shown in Fig. 6, also showed an almost identical

behavior. This confirms that oxygen has little or no effect on the

reactions. The reduced reaction rates in the sparged system

were due to cooling of the reaction mixture by the sparging gas

Table 3

Conversions of the Pd/C-catalyzed reaction in pure organic solvents and Na2CO3

Runs Organic solvent B

1 Dioxane N

2 Dioxane T

3 Dimethylacetimide (DMA) N

4 Dimethylacetimide (DMA) T

5 Tetrahydrofuran (THF) N

6 Tetrahydrofuran (THF) T

7 Toluene N

8 Toluene T

(RT). Therefore, we suggest that it is not necessary to exclude

air/oxygen during the reaction, i.e., that Pd/C catalyzed

reactions are not air-sensitive.

3.5. Pd-leaching in organic solvents

Pd-leaching of Pd/C-catalyzed Suzuki couplings in an

aqueous environment was carefully examined above. In

addition, we intended to explore whether Pd-leaching also

occurs in pure organic solvents and whether leached Pd

catalyzes the reactions, as well in such a system. Four organic

solvent systems, i.e., DMA, dioxane, THF and toluene, which

all were used in the past by different groups to carry out

homogeneous Pd-catalyzed Suzuki couplings [2,3] were

selected. Together with these four organic solvents, two

different bases were tested, one being the inorganic base,

Na2CO3, the other one the organic base, triethylamine (TEA).

The reactions were setup under regular conditions. Results are

shown in Table 3. Among the eight combinations of reaction

solvents and bases, only THF/Na2CO3, toluene/Na2CO3, and

toluene/TEA showed some reactivity, with the conversion after

4 h being 4%, 4% and 9%, respectively.

Since our studies suggest that the activity of Pd/C is mainly

due to leached Pd, the cases yielding 0% conversion may be due

to a lack of Pd-leaching. In order to examine whether Pd

leached during these reactions, the reaction mixture of DMA/

TEA was filtered and a sample after 4-h reaction was analyzed.

ICP analysis showed that the amount of Pd in DMA/TEA was

3 ppm, which is known to be sufficiently high for Suzuki

couplings in aqueous solvents according to Section 3.1. This

indicates that Pd-leaching from Pd/C also occurs in organic

solvents. Conlon et al. [14] observed the absence of Pd-leaching

in a mixture of only Pd/C and DMF at 80 8C. Thus, also in

organic solvents, presence of 1 and 2 in the organic solvent is

necessary for leaching to occur. Again the oxidative addition of

the aryl-bromide and/or the aryl-borate must be the cause for

this phenomenon. However, it is notable that leached Pd is

inactive or only slightly active under these conditions.

Homogeneous Pd-catalysts usually have ligands or there are

ligands dissolved in the solution. Therefore, we assumed that

the low activity of leached Pd is due to the lack of ligands in the

reaction mixture. In order to test this hypothesis, a homo-

geneous catalyst, tetrakis(triphenylphosphine) palladium

(Pd(PPh3)4), and the ligand triphenylphosphine (PPh3) were

used. Pd(PPh3)4 (0.01 g/10 ml solution) instead of Pd/C at

or triethylamine

ase Conversion (%)/Reaction 4 hours

a2CO3 0

riethylamine (TEA) 0

a2CO3 0


a2CO3 4


a2CO3 4



90 8C showed 8% conversion after 100 h for the model reaction

in DMA/TEA. This indicates that Pd(PPh3)4 is somewhat

active. Next, two experiments were carried out, where Pd/C was

added together with 1 and 2 in DMA. The two mixtures were

then stirred for 20 min at 80 8C. After that, in the first mixture

PPh3 was added (100 equiv. to Pd), followed by TEA after 2 h.

In the other mixture TEA was added first, followed by PPh3

after 2 h. In both runs, the leached palladium and PPh3 were

expected to form Pd(PPh3)4, which could catalyze the reaction.

Finding a non-zero conversion would indicate that leached Pd

in pure organic solvents needs a ligand to carry out the reaction.

Nevertheless, in both cases no desired product was found even

after 100 h.

In order to further explore the behavior of leached Pd in

different solvents, DFT calculations with the model B3LYP/(6-

31+G(d,p),SDD) were carried out. The fact that leached Pd is

active in the aqueous co-solvent but not active in the pure

organic solvent might be due to water which stabilizes Pd(II) in

solutions. By modeling the step of metathesis, the molar heat of

formation of �170.14 kJ/mol was obtained for R–Pd(II)–OH

without a hydration shell.�179.53 kJ/mol was obtained for the

formation of R–Pd(II)–OH with two associated H2O molecules.

Thus, the hydration of the complex has little effect on

stabilizing Pd(II). Therefore, the role of water is still somewhat

unclear, although it is known that Pd/C-catalyzed Suzuki

couplings are best carried out in a combination of organic

solvents and water [30].

A plausible explanation for the effect of water may stem

from the ability to stabilize Pd nanoparticles in water. Cassol

et al. [31] studied Heck reactions with Pd nanoparticles in an

ionic-liquid/organic-phase system. They observed active Pd

nanoparticles in the ionic liquid and inactive leached Pd in the

organic phase. However, no Pd nanoparticles were detected in

the organic phase. Thus, the authors concluded that the reaction

possibly proceeds on the surface of Pd nanoparticles, which are

stabilized in an ionic solvent. Thus, we conjecture that the

presence of water stabilizes Pd nanoparticles in our multi-

solvent system, which act as a reservoir for active dissolved Pd.

4. Conclusion

Pd-leaching in Pd/C-catalyzed Suzuki couplings was

investigated using a model coupling reaction to form

biphenylacetic acid. The main results of our work are:

� C
areful PVPy adsorption experiments show that the activity
of Pd/C is only due to leached Pd(II) species. There is no

heterogeneous contribution to the reaction.

� P
VPy adsorption is slow compared to the homogeneous
reaction. Thus, PVPy cannot completely suppress the

homogenous reaction. This also implies that there exists a

pseudo-steady state between heterogeneous Pd on the support

and homogeneous Pd in the solution, which is determined by

the leaching and adsorption rate.

� P
VPy was shown to be a good reagent to completely remove
Pd-residuals from the reaction mixture at very low costs.

Excess PVPy (approximately 3.0 equiv. to Pd) is sufficient.

� B
oth 4-ethylpyridine (the monomer of PVPy) and pyridine
bring to a stop the activity of Pd/C. Modeling studies suggest

the formation of a stable Pd-complex.

� B
y premixing Pd/C with different components of the reaction
mixture followed by filtration tests, the oxidation addition of

aryl-bromides was confirmed to be the main cause for Pd-

leaching. Dissolution of Pd clusters is not a factor in the

leaching mechanism.

� I
t was shown that the oxidative addition of aryl-borates to
form biphenyls is another cause for Pd-leaching in the self-

coupling of aryl-borates in the absence of aryl-bromides.

� T
he influence of oxygen on Pd/C-catalyzed Suzuki couplings
was proved to be negligible. Thus, this reaction is not air-

sensitive.

� W
ater as a co-solvent is required to obtain a certain catalytic
activity. Although leaching occurs in organic solvents, the

reaction is very slow without addition of water. This may be

due to the easier formation of Pd nanoparticles in the

presence of water.

Further studies will be concerned with establishing the

kinetics of the reaction and with elucidating the role of the

various solvents and additives.

Acknowledgements

The authors thank Merck Research Laboratories and Rutgers

Catalyst Manufacture Consortium and the EU Marie Curie

Chair program for financial support. The authors thank Prof.

Pedersen and Ms. N. Rodriguez-Pinto at Rutgers University for

access to HPLC and Prof. Krogh-Jespersen for many helpful

discussions. Furthermore, we want to acknowledge the helpful

suggestions made by two anonymous reviewers.

References

[1] N. Miyaura, T. Yanagi, A. Suzuki, Synth. Commun. 11 (1981) 513–519.

[2] A. Suzuki, J. Organomet. Chem. 576 (1999) 147–168.

[3] N. Miyaura, A. Suzuki, Chem. Rev. 95 (1995) 2457–2483.

[4] S.P. Stanforth, Tetrahedron 54 (1998) 263–303.

[5] A.F. Littke, C. Dai, G.C. Fu, J. Am. Chem. Soc. 122 (2000) 4020–4028.

[6] J.F. Payack, E. Vazquez, L. Matty, M.H. Kress, J. McNamara, J. Org.

Chem. 70 (2005) 175–178.

[7] J. Albaneze-Walker, J.A. Murry, A. Soheili, S. Ceglia, S.A. Springfield, C.

Bazaral, P.G. Dormer, D.L. Hughes, Tetrahedron 61 (2005) 6330–6336.

[8] A.P. Panarello, O. Vassylyev, J.G. Khinast, Tetrahedron Lett. 46 (2005)

1353–1536.

[9] B.H. Lipshutz, J.A. Sclafani, P.A. Blomgren, Tetrahedron 56 (2000) 2139–

2144.

[10] R.B. Bedford, S.L. Hazelwood, D.A. Albisson, Organometallics 21 (2002)

2599–2600.

[11] P.A. Midgley, M. Weyland, J.M. Thomas, P.L. Gai, E.D. Boyes, Angew.

Chem. 114 (2002) 3958–3961.

[12] D. Gala, A. Stamford, J. Jenkins, M. Kugelman, Org. Process Res. Dev. 1

(1997) 163–164.

[13] R.G. Heidenreich, K. Kohler, J.G.E. Krauter, J. Pietsch, Synlett 7 (2002)

1118–1122.

[14] D.A. Conlon, B. Pipik, S. Ferdinand, C.R. Leblond, J.R. Sowa Jr., B. Izzo,

P. Collin, G.-J. Ho, J.M. Williams, Y.-J. Shi, Y. Sun, Adv. Synth. Catal. 345

(2003) 931–935.

[15] M. Dams, L. Drijkoningen, B. Pauwels, G. Van Tendeloo, D.E. De Vos,

P.A. Jacobs, J. Catal. 209 (2002) 225–236.


[16] G.V. Ambulgekar, B.M. Bhanage, S.D. Samant, Tetrahedron Lett. 46

(2005) 2483–2485.

[17] K. Yu, W. Sommer, M. Weck, C.W. Jones, J. Catal. 226 (2004) 101–110.

[18] K. Yu, W. Sommer, J.M. Richardson, M. Weck, C.W. Jones, Adv. Synth.

Catal. 347 (2005) 161–171.

[19] G. Yuan, M.A. Keane, Appl. Catal. B: Environ. 52 (2004) 301–314.

[20] C. Carlini, M. Di Girolamo, A. Macinai, M. Marchionna, M. Noviello,

A.M.R. Galletti, G. Sbrana, J. Mol. Catal. A: Chem. 204 (2003) 721–728.

[21] F. Zhao, B.M. Bhanage, M. Shirai, M. Arai, Chem. Eur. J. 6 (2000) 843–

848.

[22] L. Djakovitch, K. Koehler, J. Mol. Catal. A: Chem. 142 (1999) 275–284.

[23] S. Muhkopadhyay, G. Rothenberg, A. Joshi, M. Baidossi, Y. Sasson, Adv.

Synth. Catal. 344 (2002) 348–354.

[24] R.G. Heidenreich, J.G.E. Krauter, J. Pietsch, K.J. Koehler, J. Mol. Catal.

A: Chem. 182–183 (2002) 499–509.

[25] A. Biffis, M. Zecca, M. Basato, Eur. J. Inorg. Chem. (2001) 1131–

1133.

[26] J.A. Widegren, R.G. Finke, J. Mol. Catal. A: Chem. 198 (2003) 317–341.

[27] M. Moreno-Manas, M. Perez, R. Pleixats, J. Org. Chem. 61 (1996) 2346–

2351.

[28] F.R. Hartley, Coord. Chem. Rev. 35 (1981) 143–209.

[29] R.M. Crabtree, The Organnometallic Chemistry of the Transition Metals,

Wiley-Interscience Publication, Connecticut, 2001, p. 150.

[30] F.-X. Felpin, T. Ayad, S. Mitra, Eur. J. Org. Chem. 12 (2006) 2679–2690.

[31] C.C. Cassol, A.P. Umpierre, G. Machado, S.I. Wolke, J. Dupont, J. Am.

Chem. Soc. 127 (2005) 3298–3299.

Pd-leaching and Pd-removal in Pd/C-catalyzed Suzuki couplings

Documents