The oxo-synthesis catalyzed by cationic palladium complexes, selectivity control by neutral ligand and anion

www.elsevier.nl/locate/jorganchem

Journal of Organometallic Chemistry 593–594 (2000) 211–225

The oxo-synthesis catalyzed by cationic palladium complexes,

selectivity control by neutral ligand and anion

E. Drent *, P.H.M. Budzelaar ¹

Shell Research and Technology Centre, Badhuisweg 3, NL-1031 CM Amsterdam, The Netherlands

Received 4 June 1999; accepted 27 September 1999

Abstract

Catalyst systems consisting of a palladium(II) diphosphine complex with weakly or non-coordinating counterions are efﬁcient

catalysts for the hydrocarbonylation of both aliphatic and functionalized oleﬁns. Moreover, variations of ligand, anion and/or

solvent can be used to steer the reaction towards alcohols, aldehydes, ketones or oligoketones. Non-coordinating anions and

arylphosphine ligands produce primarily (oligo)ketones; increasing ligand basicity or anion coordination strength shifts selectivity

towards aldehydes and alcohols. For the mechanisms of the aldehyde-producing step, we propose heterolytic dihydrogen cleavage,

assisted by the anion. At high electrophilicity of the palladium center, selective ketone formation is observed. The reactions

Keywords: Oxo synthesis; Ketones; Aldehydes; Copolymerization; Cationic palladium complexes

1. Introduction

aldehyde and alcohol synthesis reactions (Eq. (1)).

RꢀCHꢁCH₂+CO+H₂RꢀCH₂ꢀCH₂ꢀCHO

The discovery of the oxo-synthesis in 1938 [1] by

Otto Roelen at Ruhrchemie has marked the birth of the

large scale industrial application of homogeneous catal-

ysis by organometallic complexes. The term oxo-syn-

thesis denotes the synthesis of oxygenates by

hydro-carbonylation of oleﬁns, i.e. the reaction of

oleﬁns with mixtures of carbon monoxide and hydro-

gen. It was based upon the original ﬁnding that with a

cobalt catalyst and ethene as oleﬁnic substrate both

propionaldehyde and diethylketone were produced.

Roelen expected [2] that ‘one day it would be possible

to convert oleﬁns in general to aldehydes or ketones at

choice’. Despite later observations that only ethene

gives rise to signiﬁcant formation of ketone product,

the term oxo-synthesis remained in use in the chemical

community. However, the term hydroformylation,

which originated from a paper by Atkins and Krsek [3],

is a more correct expression for the widely applicable

RꢀCHꢁCH₂+CO+2H₂RꢀCH₂ꢀCH₂ꢀCH₂OH (1)

Driven by the need to produce oxygenates from

petrochemical hydrocarbons, hydroformylation has be-

come the largest-scale industrial chemical application of

homogeneous catalysis by transition-metal complexes,

with a present worldwide volume of about 6.5 million

ton per annum of aldehydes and alcohol products.

Several advances in homogeneous catalysis have

made this development possible. In the 1960s, it was

discovered that ligand substitution in the original

cobalt carbonyl catalyst HCo(CO)₄could inﬂuence the

performance signiﬁcantly. The resulting catalyst

HCoL(CO)₃(in which L represents a tertiary alkylphos-

phine) is much more stable and can be used at higher

temperatures [4]. A unique feature of these catalysts is

that they possess a high oleﬁn isomerization activity

combined with a high regio-selectivity for terminal

oleﬁn hydroformylation. This allows the production of

predominantly linear alcohols from internal straight-

chain oleﬁns. This process was ﬁrst commercialized by

Shell.

* Corresponding author.

¹Present address: Department of Inorganic Chemistry, Catholic

University Nijmegen, Toernooiveld 1, 6525 ED Nijmegen, The

Netherlands.

PII: S0022-328X(99)00554-9

212

E. Drent, P.H.M. Budzelaar / Journal of Organometallic Chemistry 593–594 (2000) 211–225

Another major breakthrough in catalyst activity and

merization could be efﬁciently interrupted, resulting in

low-molecular weight products and ultimately in alde-

hydes and monoketones. In this paper, we will show

that with L₂PdX₂complexes as catalysts, aldehydes and

ketones can indeed be produced from oleﬁnic substrates

in general. Moreover, we will show that by a proper

choice of the ligand (L₂) and the anion (X⁻), hydrocar-

bonylation of oleﬁns can be tuned to proceed selectively

towards either aldehydes (alcohols) or ketones. The

reported class of catalysts, thus provides for the ﬁrst

time a means of control for the selective production of

both possible oxo-type products.

regio-selectivity was made possible by advances in

organometallic chemistry [5]. Based on this, workers at

Union Carbide, Johnson and Matthey, and Davy Pow-

ergas jointly developed a rhodium–phosphine based

hydroformylation process to produce n-butanal from

propene. This process was ﬁrst commercialized in 1976.

In the last 2 decades, numerous hydroformylation cata-

lysts, predominantly based on the metals Co and Rh

combined with a wide variety of ligands, have become

available. However, none of these catalysts has allowed

the full and general exploitation of the oxo-synthesis:

the selecti6e production of ketones has thus far re-

mained an elusive goal [6].

2. Results

Our interest in the palladium-catalysed hydrocarbony-

lation of oleﬁns was awakened by the discovery of a

class of highly efﬁcient cationic palladium catalysts for

the alternating copolymerization of oleﬁns with carbon

monoxide (Eq. (2)) [7].

The catalysts tested were formed by the combination

of a suitable bidentate ligand (L₂) with a palladium (II)

species, e.g. Pd(OAc)₂, and an acid containing the

weakly or non-coordinating anions (X⁻), via ligand

complexation–anion displacement reaction (Eq. (3))

[8,9].

(2)

The active catalysts for this Reppe-type reaction can

be represented by the general formula L₂PdX₂, in which

L₂stands for a bidentate ligand (e.g. phosphine, pyridyl

or thioether) and X represents a weakly or non-coordi-

nating anion. When these copolymerization catalyst

were exposed to oleﬁns and carbon monoxide in the

presence of small amounts of hydrogen in aprotic sol-

vents, ketonic end-groups were generally observed.

However, with some catalysts containing moderately

strongly coordinating anions, both ketone and aldehyde

end-groups could be observed, indicating that chain

transfer by H₂is possible both at the Pdꢀalkyl and the

Pdꢀacyl stage of the growing chain. It was hoped that

under sufﬁciently high hydrogen pressure relative to

that of carbon monoxide, chain growth of the copoly-

L₂+Pd(OAc)₂+2HXL₂PdX₂+2HOAc

(3)

The catalysts are conveniently generated in situ by

applying some excess of ligand and acid over the stoi-

chiometric quantities required by Eq. (3) (see Section

5).

We will describe the effects of variations of the ligand

and acid components of the catalysts on the chemo-se-

lectivity in oxo-synthesis. In addition, we will illustrate

their effects on the regio-isomeric product distribution,

both in aldehyde–alcohol and ketone formation. Steric

and electronic (basicity) properties of the diphosphine

ligands R₂P(CH₂)_nPR₂were varied by variation of R

and n (no active catalysts could be obtained with

monodentate ligands). In the text, we will use abbrevia-

tions for both ligands and acids; these abbreviations are

explained in Table 1.

Table 1

Abbreviations used for ligands and acids

2.1. Chemo-selecti6ity in palladium catalyzed

oxo-synthesis

Name

R

n

Ligands R₂P(CH₂)_nPR₂

DPPP Ph

DsBPE sec-Bu

DEPP Et

DnBPP n-Bu

It appears appropriate to distinguish between

aliphatic oleﬁnic and functionalized vinylic substrates.

3

2

3

2.1.1. Aliphatic oleﬁns as substrate

A series of catalyst systems examined for the chemo-

selectivity of hydrocarbonylation of propene and 1-

octene is given in Table 2. Inspection of the table shows

that variation of the ligand and the acid component of

the catalyst system resulted in very signiﬁcant changes

in chemo-selectivity. For one substrate (propene) and

one acid (HOTs), ligand variation can shift the chemo-

selectivity from simultaneous production of aldehydes,

monoketones and co-oligomers (with the DPPP ligand),

DsBPP

DtBPP

sec-Bu

tert-Bu

Acids

HOAc

TFA

HOTf

HOMs

HOtBs

HOTs

CH₃COOH

CF₃COOH

CF₃SO₃H

CH₃SO₃H

(CH₃)₃CSO₃H

Acetic acid

Triﬂuoroacetic acid

Triﬂuoromethanesulfonic acid

Methanesulfonic acid

t-Butylsulfonic acid

p-CH₃C₆H₄SO₃H p-Toluenesulfonic acid

E. Drent, P.H.M. Budzelaar / Journal of Organometallic Chemistry 593–594 (2000) 211–225

213

Table 2

Chemoselectivity in oleﬁn hydrocarbonylation ^a

Ligand

Acid

Oleﬁn

T (°C)

Rate ^b

Solvent

Products (%) ^c

Aldehydes/Alcohols

Ketones

DPPP

TFA

Propene

1-Octene

Propene

1-Octene

Propene

1-Octene

Propene

125

90

200

500

100

300

1000

500

60

100

120

100

80

Diglyme

Methanol

Diglyme

95/–

30/–

–/–

98/–

95/–

15/25

–/–

96/–

80/5

3/93

–/–

4

50 ^d

70 ^d

Trace

50

90

Trace

10

HOTs

HOTf

TFA

DnBPP

TFA

HOTs

HOTf

TFA

HOTs

HOTs ^e

HOTs

HOTf

HCl

90

125

115

125

100

125

115

125

90

Trace

95

–/–

–

98

–

DsBPP

0

HOAc

TFA

500

800

100

150

40

100

300

B10

30

98/–

84/2

–/–

96/2

3/95

–/–

98/–

–/–

–

Trace

7

HOTs

HOTf

TFA

95

Trace

98

Trace

93

HOTs^e

HOTf

TFA

DsBPE

DtBPP

HOTf

TFA

HOTs

HOTf

90

80

–

4

75

85/10

14/11

10

^aBatch experiments, 250 ml Hastelloy C autoclave, 45 ml solvent, P_CO=P_H=30 bar (at room temperature). Propene reactions: 20 ml propene,

0.1 mmol Pd(OAc)₂, 0.2 mmol ligand, 0.5 mmol acid. 1-Octene reactions: 20 m²l octene, 0.25 mmol Pd(OAc)₂, 0.6 mmol ligand, 1 mmol acid. Rates

averaged over B30% oleﬁn conversion.

^bTurnover/h (mol/mol/h).

^cRemaining products are diketones and/or keto-aldehydes.

^dHigher oligoketones formed.

^eP_CO=20 bar, P_H=40 bar.

2

(−0.7) [10]. Along the vertical axis the ligand cata-

lyst components have been ordered for increasing

basicity: DPPPBDnBPPBDsBPPBDtBPP. Three

different regions of chemo-selectivity can be distin-

guished: co-oligomerization, monoketone formation and

through simultaneous production of aldehydes and

monoketones (with the DnBPP ligand) to the selective

production of aldehydes (with the ligands DsBPP and

DtBPP).

Likewise, for one selected ligand, e.g. DnBPP, acid

variation can bring about a remarkable shift in chemo-

selectivity from selective aldehydes formation (with

TFA) towards selective monoketones formation (with

HOTf).

Generally, the monoketones formed are both satu-

rated- and a-b unsaturated ketones (vide infra, Table

4).

With hydrogen chloride and weak carboxylic acids

(e.g. HOAc) as acid catalyst components, no active

catalysts could be generated.

The variation in chemo-selectivity for propene hydro-

carbonylation under conditions speciﬁed in Table 2, has

been summarized schematically in a graphical represen-

tation given in Fig. 1. Along the horizontal axis the

acid catalyst components have been ordered for in-

creasing pK_a: HOTf (−5.1BHOTs (−2.7)BTFA

Fig. 1. Schematic representation of chemoselectivity as a function of

ligand and acid properties.

214

E. Drent, P.H.M. Budzelaar / Journal of Organometallic Chemistry 593–594 (2000) 211–225

Table 3

Chemoselectivity in hydrocarbonylation of functionalized oleﬁns ^a

Ligand

Acid

Oleﬁn (ml)

T (°C)

Rate ^b

Solvent

Products (%)

Aldehydes

Ketones ^c

DPPP

TFA

HOTf

TFA

HOTf

TFA

HOTf

Styrene (20)

Methyl acrylate (20)

Acrylic acid (20)

Acryl-amide (5 g)

Methyl acrylate (15)/1-octene (40)

Acrylic acid (10)/1-octene (40)

115

90

75

100

70

60

100

120

250

100

200

600

50

Anisole

Diglyme

Diglyme/methanol

Water/THF

Diglyme

THF

98

50

98

–

12

–

B1

50

B1

99

DEPP

DnBPP

DEPP

34 (54)

88 (12)

60 ^d

30 ^e

100

250

95 (5) ^f

80 ^g

THF

^aBatch experiments, 250 ml Hastelloy C autoclave, 0.25 mmol Pd(OAc)₂, 0.3 mmol ligand, 1 mmol acid, 50 ml solvent.

^bTurnover/h (mol/mol/h).

^cPercentage byproduct methylpropionate in parentheses.

^dDetected as dispirolactone (see text); byproduct propionic acid (40%).

^eDetected as dispirolactam (see text); byproduct propionamide (70%).

^fMixed ketone selectivity 90% based on acrylate, 95% based on 1-octene.

^gMixed ketone selectivity 75% based on acrylic acid, 92% based on 1-octene; byproducts propionic acid (10%), dispirolactone (10%).

aldehyde–alcohol. The regions are separated by fairly

polar methanol solvent. The same shift towards ketone

discrete boundary regions. For example, the DnBPP–

formationisobservedbychanging(inthediglymesolvent)

HOTs combination is located in the boundary region

between ketone and aldehyde–alcohol formation (50/

40%), while the combination of DsBPP with the same acid

the acid from HOTs to HOTf. Thus, HOTs in methanol

behaves similar as HOTf in diglyme.

With all catalysts, the rate of hydrocarbonylation with

shifts chemo-selectivity to within the region of selective

propene was considerably higher than with 1-octene. This

aldehyde–alcohol formation (86%). Likewise, the

is at least partly due to the fact that 1-octene was

DPPP–HOTf combination is located in the boundary

isomerized to internal octenes in the course of the

region between co-oligomerization and monoketone for-

hydrocarbonylation experiments. In separate experi-

mation, while the combinations of DnBPP and DsBPP

ments with internal oleﬁns as substrate, it was found that

with the same acid are located well within the region of

these are converted to the same products as the terminal

selectivemonoketoneformation(90–95%). Thecombina-

oleﬁns. However, the rate of conversion of internal oleﬁns

tion of all ligands investigated with TFA is located in the

was about a factor of eight lower than the initial rate (at

region of selective aldehyde formation; reduction of the

B15% conversion) observed with 1-octene.

aldehydes to alcohols does not take place to a signiﬁcant

extent, with this acid as catalyst component.

Thelocationofthechemo-selectivityboundaryregions,

2.1.2. Functionalized oleﬁns as substrate

Some results obtained with functionalized oleﬁns are

e.g. between monoketone and aldehyde–alcohol forma-

collected in Table 3. The catalyst systems studied in this

tion, depends on the oleﬁnic substrate as well as on the

case contained a near stoichiometric amount of ligand

reactionconditions. Thus, theabove-mentionedDnBPP–

combined with a small excess of acid. For styrene, the

HOTs combination becomes more selective for aldehyde

dependence of chemo-selectivity on acid variation is very

formation with 1-octene as the substrate. The DPPP–

similar to that for propene and 1-octene. With DPPP as

TFA combination is clearly located well within the region

ligand and TFA as acid component, aldehydes were

of selective aldehyde formation with 1-octene. At a higher

formed with high selectivity (98%), whereas a 1:1 mixture

H₂:CO ratio (of 2), and with the combination DnBPP–

of aldehydes and ketones could be produced with HOTs

HOTs, ketones formation is completely suppressed and

in apolar solvents such as anisole. With DEPP as ligand,

alcohols are formed selectively (93%).

selectivity could be directed towards aldehydes (98%)

The reaction solvent could also bring about a dramatic

usingTFA, ortowardsketones(99%)usingHOTf. Higher

effect on the chemo-selectivity. This is demonstrated with

oligo-ketones were not observed.

the DnBPP–HOTs combination and 1-octene as sub-

With acrylic oleﬁns, such as methyl acrylate, ketone

strate. Whereas this ligand–acid combination in diglyme

formationdominatedoveraldehydeformation. Evenwith

as solvent resulted in selective aldehydes–alcohols forma-

TFA, which in other cases gave very selective hydro-

tion (85%), exclusively ketones were formed in the more

formylation, the proportion of aldehydes did not rise over

E. Drent, P.H.M. Budzelaar / Journal of Organometallic Chemistry 593–594 (2000) 211–225

215

:15% under conditions speciﬁed in Table 3. With HOTf,

ketones were formed exclusively. However, we observed

that hydrogenation sometimes strongly competed with

hydrocarbonylation. This is noteworthy, since we never

observed more than 2% hydrogenation in the hydrocar-

bonylation of aliphatic oleﬁns. Hydrogenation could,

however, be suppressed by using HOTf as acid compo-

nent, in particular in a polar reaction medium containing

methanol as (co)-solvent. Thus, methyl acrylate could be

converted with good selectivity (:90%) to the di-ester

ketone, di-methyl-4-oxopimilate.

With acrylic oleﬁns containing active hydrogen atoms,

such as acrylic acid and acrylamide, formation of di-spiro-

lactone or di-spirolactam was observed. The initially

formed diacid and diamide ketones apparently undergo

an intramolecular double cyclo-condensation (see Sec-

tions 2.2.2 and 3.3).

The high propensity of acrylic oleﬁns for ketone

formation suggested that insertion of these oleﬁns in

palladiumꢀacyl intermediates is facile compared to that

of aliphatic oleﬁns. This prompted us to test hydrocar-

bonylation using mixtures of functionalized oleﬁns and

1-alkenes. Indeed, these reactions turned out to produce

mixed ketones with good selectivity. Thus, with DEPP

as ligand and HOTf as acid, 1-octene could be converted

under mild conditions with acrylic acid or methyl acrylate

to 4-oxo-dodecanoic acid or its methyl ester, respectively.

Theseresultsclearlyillustratethatincationicpalladium

catalysts, variation of the ligand and acid components can

be used to control chemo-selectivity in hydrocarbonyla-

tion of oleﬁns.

2.2. Regio-selecti6ity in the palladium catalyzed

oxo-synthesis

Catalyst composition not only affects chemo-selectiv-

ity, but also inﬂuences the regio-selectivity of the palla-

dium-catalyzed hydrocarbonylation. In this section, we

will focus on regio-selectivity; the effects of ligand

variation are illustrated in Table 4, and the effect of acid

variation on regio- selectivity of hydroformylation is

summarized in Table 5.

2.2.1. Regio-selecti6ity in aldehyde–alcohol formation

With propene only two aldehydes can be formed: n-

and iso-butanal. From Table 4 it can be observed that,

with TFA or HOTs as acid component, hydroformylation

linearity increases with the steric bulk of the ligand from

60 to 65% with DnBPP to 84% with DsBPP. The

C₂-bridged ligand DsBPE afforded a signiﬁcantly lower

product linearity (76%) than the corresponding C₃-

bridged DsBPP ligand (84%).

The acid component also affects regio-selectivity, with

stronger acids affording lower product linearity. For

example, using the DsBPP ligand, linearity falls from 90%

with TFA to 77% with HOTs. The effect of acid strength

on product linearity was more pronounced at lower

reaction temperatures (Table 5).

Similar effects could be noted with 1-octene as sub-

Table 4

Regioisomer distribution in hydrocarbonylation ^a

b

Ligand

Acid

Oleﬁn

Aldehydes/alcohols

Ketones ^c

h to t

Percent linear

h to h

t to t

DPPP

TFA

Propene

1-Octene

Styrene

Methyl acrylate

Methyl acrylate/1-octene

72

62

–

68

60

–

30(85)

31(85)

–

6(20)

7(5)

–

2(10)

–

9(4)

–

B1

–

98(67)

–

84(4)

98(0)

92(0) ^d

HOTs

HOTf

TFA

HOTs

HOTf

TFA

HOTf

TFA

HOTf

TFA

HOTf

TFA

HOTs

TFA

HOTf

50(63)

52(79)

–

83(16)

82(5)

–

92(10)

–

81(12)

–

98(10)

–

2(80)

–

16(5)

B2

8(0)

20(90)

17(91)

–

11(25)

11(24)

–

6(15)

–

10(15)

–

2

–

DnBPP

DsBPP

DsBPE

DnBPP

DPPP

DEPP

DnBPP

DEPP

84

–

76

–

70 ^e

–

85

83

72

–

^aReaction conditions as speciﬁed in Tables 2 and 3.

^bLinear aldehydes+alcohols as percentage of total aldehydes+alcohols.

^cSaturated+unsaturated ketones as percentage of total ketones; fraction of unsaturated ketones given in parentheses.

^dMixed ketone.

^e20% a-methyl branched, 10% a-ethyl and a-propyl branched aldehydes.

216

E. Drent, P.H.M. Budzelaar / Journal of Organometallic Chemistry 593–594 (2000) 211–225

Table 5

Effect of acid on regioisomer distribution of hydroformylation ^a

b

Acid

pK_a

Oleﬁn

Rate

Aldehyde selectivity (%) ^c

Linearity (%) ^d

HOTs

HOMs

HOtBs

TFA

HOTs

HOtBs

−2.7

−1.9

−1.2

−0.7

−2.7

−1.2

Propene

1-Octene

150

80

20

80

30

86 (11, 3)

94 (4.5, 1.5)

95 (3.5, 1)

99

98

77

84

88

90

78

85

^aBatch experiments, 250 ml Hastelloy C autoclave, P_CO=P_H=30 bar (at room temperature), solvent diglyme, 0.25 mmol Pd(OAc)₂, 0.6 mmol

2

ligand (DsBPP), 0.5 mmol acid, T=75°C.

^bTurnover/h (mol/mol/h).

^cByproducts CH₃CH₂CH₂COCH(CH₃)CH₂CHO and CH₃CH₂CH₂COC(CH₃)ꢁCH₂given in parentheses.

^dLinear aldehyde as percentage of total aldehyde.

strate. Mainly linear nonanals were formed under alde-

hyde formation conditions using TFA. With DnBPP as

the ligand, n-nonanal comprised 70% of the total amount

of aldehydes formed. In addition to the branched isomer

a-methyloctanal (20%), the product also contained

branched isomers derived from internal octenes (a-ethyl-

heptanal and a-propylhexanal, together 10%). These

latter products are a consequence of concomitant isomer-

ization of 1-octene under hydrocarbonylation conditions

as noted earlier. As with propene, the acid component

also affected aldehyde product linearity (Table 5). With

DsBPP as ligand, it changed from 78% with HOTs to 85%

with HOtBs.

Under aldehyde forming conditions with TFA as

catalyst component, similar phenomena could also be

noticed with styrene. The main product was the linear

aldehyde 3-phenylpropionaldehyde. Product linearity in-

creased with ligand bulk, e.g. from 72% for DEPP to 85%

for DPPP. Acid variation (TFA vs. HOTs) had little effect

on the regio-selectivity of aldehyde formation, possibly

because these reactions were carried out at a relatively

high temperature.

The observed distribution of regio-isomeric ketones,

under conditions speciﬁed in Table 2, is illustrated in

Table 4. Of the two different unsaturated h to t isomers,

iso-propenyl-n-propyl ketone was consistently formed in

much higher amounts than the alternative isomer n-

propenyl-iso-propyl ketone. Only traces of the latter

isomer could be observed with all catalysts containing

alkylphosphine ligands. With DPPP as ligand, however,

a small but non-negligible quantity could be detected

(about 5% of total ketones). The h to t ketone contents

given in Table 4, denote the combined amount of both

unsaturated h to t isomers.

It can be seen that the ligand structure has a very

signiﬁcant inﬂuence on the regio-isomer distribution of

the ketone products. With all ligands mentioned in Table

4, the h to t isomer was predominantly formed, but this

preference was considerably stronger with alkyl phosphi-

nes (up to 92%) than with the arylphosphine DPPP

(:50%). The preference for the h to t regio-isomer

increased with the steric bulk of the alkylphosphine, from

82% with DnBPP to 92% with DsBPP. The C₃-bridged

diphosphine DsBPP gave a signiﬁcantly higher preference

(92%) for h to t enchainment than the C₂-bridged

analogue DsBPE (81%).

2.2.2. Regio-selecti6ity in ketone formation

With a higher oleﬁn such as propene, a variety of

regio-isomeric ketones can, in principle, be obtained.

Three saturated monoketones with, respectively, h (head)

to h (head), h (head) to t (tail) and t (tail) to t (tail)

enchainment of the propyl groups via carbonyl can be

formed. Likewise, four unsaturated ketones can be pro-

duced with h to h, h to t (twice) and t to t enchainment

of the propyl and the propenyl moiety via carbonyl,

respectively.

No signiﬁcant inﬂuence of the anion on the isomeric

distribution of the ketones could be established. This was

illustrated in two instances, with DPPP and DnBPP as

ligand (Table 4). Whereas the hydrocarbonylation could

beshiftedfromsimultaneousproductionofaldehydesand

ketones towards exclusive production of ketones by

changing the acid from HOTs to HOTf, the regio-isomer

distribution of the ketone products remained virtually

unchanged.

The proportion of unsaturated ketones depends on the

ligand and changes from predominantly unsaturated

(:85%) with DPPP towards mainly saturated with

alkylphosphines (:80–85%). Generally, unsaturation

was found to decrease with increasing temperature and

increasing hydrogen pressure. It can be noted

E. Drent, P.H.M. Budzelaar / Journal of Organometallic Chemistry 593–594 (2000) 211–225

217

that the highest degree of unsaturation was consistently

found in the h to h coupled ketone isomer.

nucleophiles, such as oleﬁns, carbon monoxide and

alcohols to facilitate insertion reactions, but also allows

reactions of dihydrogen with intermediate Pdꢀalkyl

and/or Pdꢀacyl species. By a suitable choice of the

ligand and acid, the reaction with dihydrogen can efﬁ-

ciently prevent copolymerization chain growth. De-

pending on the ligand and acid, the chain growth can

be interrupted selectively at the Pdꢀacyl stage to pro-

duce aldehydes (and alcohols by further reduction) or

at the second Pdꢀalkyl stage to produce monoketones.

This class of catalysts thus provides a link between the

Reppe-type oleﬁn carbonylation chemistry and the

Roelen-type oxo-chemistry.

We consider it worthwhile to comment on the inter-

play between the catalysts’ ligand and anion and their

effect on the course of the hydrocarbonylation reac-

tions (chemo-selectivity). Subsequently, we will consider

their role in the mode of oleﬁn insertion at the palla-

dium center, both in aldehyde–alcohol and ketone

forming reactions (regio-selectivity).

With 1-octene, the h to t regio-isomeric ketone was

almost exclusively formed under ketone formation con-

ditions with BnDPP and HOTf as catalyst components.

Thus, regio-speciﬁcity with the bulkier oleﬁn, 1-octene,

for h to t ketone formation is considerably higher than

observed with propene.

Ketone formation with styrene proceeded with a

remarkably high regio-selectivity to the t to t isomer

(98%) with DPPP and HOTs. Under these conditions

predominantly unsaturated ketones were formed. With

DEPP and HOTf, preference for t to t coupling was less

pronounced. As observed with propene and 1-octene,

mainly saturated ketones were formed with the alkyl

phosphine. It is noteworthy that the unsaturated frac-

tion of the h to t regio-isomer was exclusively 1,4

diphenylpent-1-ene-3-one with both types of ligand.

Acrylic oleﬁns also showed a high preference for t to

t coupling of the oleﬁnic moieties both with aryl- and

alkylphosphine ligands. This is exempliﬁed in Table 4

for methyl acrylate with DnBPP and HOTf. In this

case, saturated ketones are formed exclusively. The

formation of di-spirolactone and di-spirolactam from

acrylic acid and acrylamide can be explained by initial

t to t coupling followed by cyclocondensation (Eq. (4)).

3.1. Chemo-selecti6ity

3.1.1. Elementary steps in hydrocarbonylation

The catalyst systems we describe contain a square-

planar d⁸16-electron palladium(II) center surrounded

by a neutral cis-chelating diphosphine (L₂) and two

anionic ligands (X⁻). The anions are provided by the

acid component of the catalyst via the reaction given in

Eq. (3) [7].

The proposed aldehyde- and ketone-forming reac-

tions and their link to oleﬁn–CO copolymerization

have been summarized in Scheme 1. In analogy with

the term hydroformylation for aldehyde–alcohol form-

ing reactions, the ketone forming reactions can best be

termed as hydro-acylation of oleﬁns [11].

(4)

The actual active species in both hydroformylation

and hydro-acylation is thought to be a cationic

Pdꢀhydride complex L₂PdH⁺, formed by heterolytic

splitting of dihydrogen at the electrophilic palladium

center of the precursor L₂PdX₂(Eq. (5)).

In cross-hydrocarbonylation experiments with mix-

tures of acrylic and aliphatic oleﬁns, it was observed

that the mixed ketone products consisted predomi-

nantly of t to t regio-isomers. Thus, the mixed ketone

from methyl acrylate and 1-octene (using DEPP–

HOTf) contained 92% of methyl 4-oxo-dodecanoate.

The h to t regio-isomer by-product was exclusively

methyl 4-oxo-5-methylundecanoate. No unsaturated

mixed ketones were observed.

L₂Pd²⁺X₂⁻L₂PdH⁺X⁻+HX

(5)

The next step would involve coordination of the

oleﬁn to the Pd hydride, followed by migratory inser-

tion of the oleﬁn to generate a Pdꢀalkyl complex

L₂PdR⁺. Coordination and subsequent migratory in-

sertion of carbon monoxide then yields the Pdꢀacyl

complex L₂PdC(O)R⁺. It is at this stage that hydro-

formylation and hydro-acylation reactions are thought

to diverge.

In hydroformylation, hydrogenolysis of the Pdꢀacyl

bond takes place to give the aldehyde product and

regenerate the hydride L₂PdH⁺. In hydro-acylation, a

second oleﬁn molecule coordinates to the Pdꢀacyl, and

migratory insertion gives an internally coordinated

Pdꢀalkyl complex (see structure 1 below).

3. Discussion

The above reported results indicate that cationic

palladium(II) catalyst systems of the type L₂PdX₂, eas-

ily prepared from e.g. Pd(OAc)₂, a chelating ligand L₂

and an acid HX, have a wide applicability in oleﬁn

carbonylation chemistry, in addition to being efﬁcient

oleﬁn–carbon monoxide copolymerization catalysts [7].

The electrophilic palladium center can not only activate

218

E. Drent, P.H.M. Budzelaar / Journal of Organometallic Chemistry 593–594 (2000) 211–225

3.1.2. The interplay of ligands and anions in

electrophilicity

Inspection of Table 2 and Fig. 1 suggests that the

steric properties of the ligand and anions are probably

not a very crucial factor for the observed variation in

chemo-selectivity. No correlation between chemo-selec-

(1)

Stable species of this type have recently been ob-

tivity and ligand size, as manifested by the estimated

cone-angle [13] at the phosphorus atoms of the ligand,

could be established. However, the steric properties of

the ligand can indeed have a signiﬁcant effect on regio-

chemistry in hydrocarbonylation, as will be discussed in

Section 3.2.

Instead, Fig. 1 suggests that the electronic properties

of both the neutral ligand L₂and the anion X⁻deter-

mine the course of hydrocarbonylation. An increasing

ligand basicity should lead to a decreasing electrophilic-

ity of the palladium(II) center. Likewise, weaker acids

are generally associated with increasing coordination

strength of the anion to the palladium center and also

served spectroscopically in the studies of oleﬁn insertion

in L₂Pdꢀacyl complexes [12]. They are also thought to

play a key role as intermediates in the alternating

oleﬁn–CO copolymerization. Termination of hydro-

acylation can proceed by hydrogenolysis of complex 1

to form a saturated ketone and regenerate the hydride

L₂PdH⁺. Alternatively, complex 1 can undergo b-elimi-

nation to give the hydride and an unsaturated ketone.

These terminating reactions compete with further inser-

tion steps to give oligo- or poly-ketones.

We will now discuss the factors which determine the

fate of the crucial Pdꢀacyl intermediate.

Scheme 1. Proposed mechanisms of Pd-catalyzed hydroformylation, hydroacylation and oleﬁn–CO copolymerization.

E. Drent, P.H.M. Budzelaar / Journal of Organometallic Chemistry 593–594 (2000) 211–225

219

decrease the electrophilicity of the metal center. Appar-

ently, highly electrophilic complexes are efﬁcient co-

polymerization or hydroacylation catalysts, whereas

Pdꢀacyl bond will not occur for thermodynamic rea-

sons. However, once the oleﬁn enters the coordination

sphere, both a low barrier of insertion into the Pdꢀacyl

bond and a strong thermodynamic driving force for

oleﬁn insertion are expected. The insertion product is

stabilized by internal coordination of the b-carbonyl

group to the metal center (structure 1) [12]. Since part

of this stabilization will already be felt in the transition

state, the insertion barrier should be lower than for a

normal oleﬁn insertion. A more electrophilic metal

center should therefore favor oleﬁn insertion over

hydrogenolysis.

However, this is probably not the whole story. Gen-

erally, hydro-acylation (with non-coordinating anions)

proceeds with a somewhat lower rate than hydroformy-

lation (with weakly coordinating anions). It is reason-

able to expect that oleﬁn insertion in the Pdꢀacyl bond

represents the overall rate-determining step in hydro-

acylation. Therefore, it is thought that additional fac-

tors must come into play which make hydrogenolysis of

the Pdꢀacyl less probable with catalysts containing

non-coordinating anions. Hydrogenolysis of the

Pdꢀacyl bond is thought to require the electrophilic

activation of dihydrogen. One factor may be that

strong coordination of CO and/or oleﬁn under these

conditions prevents the approach of dihydrogen. In

addition, it could be that the coordinating anions giving

rise to hydroformylation play an acti6e role in the

hydrogenolysis reaction. The mechanism of the Pdꢀacyl

hydrogenolysis is not known, but we currently favor a

pathway involving heterolytic H₂dissociation at a sin-

gle Pd center. This would produce a Pd(hydride)(acyl)

complex, which could eliminate aldehyde and then react

with H⁺to regenerate a palladium hydride [14] (a

similar hydrogenolysis mechanism has been proposed

for hydrogenolysis of rutheniumꢀacyl complexes [15]).

The anion may assist the heterolytic dissociation by

(temporarily) binding H⁺. Non-coordinating anions

like OTf⁻are not basic enough to fulﬁll this role, and

also do not stay close to the Pd center.

less

electrophilic

complexes

give

rise

to

hydroformylation.

One essential requirement of the anions should be

their easy displacement by reactants from the coordina-

tion sites around palladium. Too strong coordination,

e.g. with halogen or weak carboxylic acids, leads to

inactive catalysts (Table 2). Even when the anions can

be displaced by the reactants, as evidenced by the

observed catalytic activity, their basicity still affects

chemoselectivity. It is thought that more basic anions

stay in closer proximity to the palladium(II) center than

less basic anions during the elementary steps of the

catalytic cycle. For example, they could remain at or

near a ﬁfth coordination site below or above the coor-

dination plane.

Cation–anion coordination, just like acidity, will de-

pend on the reaction solvent. Thus, it can be under-

stood that the OTs⁻anion, as far as chemo-selectivity

is concerned, behaves as a non-coordinating anion in a

polar solvent, i.e. very similar to the non-coordinating

OTf⁻anion in a less polar solvent under the same

conditions (see Table 2, experiments with DnBPP–1-

octene). Cation–anion dissociation in methanol is facil-

itated by solvation, whereas in diglyme such ion-pairs

stay in closer proximity.

3.1.3. Hydroformylation 6ersus hydro-acylation

Apparently, selective hydroformylation requires a not

too strongly electrophilic palladium center. This can be

achieved by using a very basic ligand (DsBPP, DtBPP)

and/or a not too poorly coordinating anion (e.g. TFA⁻).

It is interesting to note that, for one particular ligand

(DsBPP, Table 5), the selecti6ity for hydroformylation

increases with increasing coordination of the anion,

whereas the rate of hydroformylation decreases, pre-

sumably because anion displacement at the palladium

center becomes more difﬁcult.

Selective hydro-acylation requires a more strongly

electrophilic palladium center, such as achievable with

basic ligands (DnBPP through DtBPP) in combination

with a non-coordinating anion (OTf⁻). The even more

electrophilic palladium center obtained with OTf⁻and

the less basic DPPP ligand leads to oligo- or

polyketones.

Since the catalytic cycles for hydroformylation and

hydroacylation diverge at the Pdꢀacyl stage, the elec-

trophilicity of the Pd center apparently determines the

ratio of hydrogenolysis (to give aldehydes) and oleﬁn

insertion (to give ketones). A more electrophilic metal

center appears to favor oleﬁn insertion over hy-

drogenolysis. The substrate molecules (CO, oleﬁn) can

easily displace the weakly coordinating anion in

L₂Pd(acyl)(X). Carbon monoxide insertion in the

Thus, the ligand–anion combination affects the

catalysis in a number of ways:

By changing the electrophilicity of the metal center.

More electrophilic catalysts tend to give higher hy-

droformylation rates, but at high electrophilicity hy-

droacylation is preferred.

Coordinating anions block a coordination site and

thus result in lower insertion rates.

More basic anions assist hydrogenolysis and favor

aldehyde formation.

The low activity of catalysts containing the strongly

basic DtBPP as a ligand is noteworthy. With the more

coordinating TFA⁻anion, in particular, hydroformyla-

tion activity is negligible (Table 2). In this case, the

palladium center seems to be too weakly electrophilic

for the required heterolytic dihydrogen dissociation.

220

E. Drent, P.H.M. Budzelaar / Journal of Organometallic Chemistry 593–594 (2000) 211–225

The alternative possibility of steric hindrance by the

bulky phosphine during oleﬁn insertion can be almost

certainly excluded. When the hydroformylation of

propene or 1-octene was attempted in the presence of

methanol, methyl esters were produced exclusively and

with high rate [16]. This proves that intermediate Pdꢀacyl

species are not only formed rapidly but also undergo

rapid alcoholysis by the already pre-polarized CH₃O⁻

H⁺molecule. Thus, the problem in this case seems to be

hydrogenolysis of the Pdꢀacyl by the non-polar dihydro-

gen molecule. Apparently, electrophilicity of the palla-

dium center and anion basicity have to work together for

efﬁcient heterolytic dissociation of dihydrogen.

Finally, we comment on the terminating hydrogenol-

ysis of the Pdꢀalkyl bond to yield the monoketone

hydro-acylation products. The chelate formation in

structure 1 could affect the subsequent fate of the

Pdꢀalkyl intermediate in a number of ways. On one

hand, chelate formation could prevent or slow down

termination by b-hydride elimination (to yield unsatu-

rated ketones) or hydrogenolysis (to yield saturated

ketones). Inhibition of b-H elimination in metallacycles

is known [17]. For b-hydride elimination to occur the

b-H atom has to approach the palladium ion, but this

requires loss of PdꢀO coordination. For hydrogenolysis,

the dihydrogen molecule has to approach the palladium

center. This process might therefore also be inhibited by

strong carbonyl coordination to palladium. In the ex-

treme case of high electrophilicity, with DPPP as ligand

and OTf⁻as anion (Table 2 and Fig. 1), termination

indeed seems to be difﬁcult, since signiﬁcant amounts of

higher ketones were obtained. The monoketone products

in this case were mainly unsaturated, indicating termina-

tion by b-elimination.

On the other hand, it is also possible that chelate

formation assists hydrogenolysis of the Pdꢀalkyl bond.

This seems to be the case at intermediate electrophilicity,

and thus moderately strong coordination of the chelating

carbonyl group, with basic alkylphosphine ligands and

weakly to non-coordinating anions. Here efﬁcient and

fast hydrogenolysis produces mainly saturated monoke-

tones. The notion that chelating Pdꢀalkyl moieties can

undergo rapid hydrogenolysis is supported by observa-

tions made with acrylic substrates. The insertion of these

oleﬁns in Pdꢀhydride immediately affords a ﬁve-mem-

bered chelate ring similar to structure 1, but now

involving the ester, acid or amide carbonyl group. With

these substrates, unlike with aliphatic oleﬁns, hydrogena-

tion of the substrate itself strongly competes with hydro-

carbonylation.

the saturated ketone and then bind a proton to regener-

ate the cationic hydride. This assisted hydrogenolysis can

occur after the second oleﬁn insertion with aliphatic

oleﬁns, or already after the ﬁrst insertion with acrylic

oleﬁns. The ﬁrst insertion of aliphatic oleﬁns, however,

cannot follow this pathway, and therefore we ﬁnd only

a negligible amount of hydrogenation products with such

oleﬁns.

3.2. Regio-selecti6ity

In this section, we will consider the effects of the ligand

and anion on the mode of oleﬁn insertion at the

palladium center, both in hydroformylation and hydro-

acylation reactions.

3.2.1. Hydroformylation

For hydroformylation, the regio-selectivity of oleﬁn

insertion at palladium can directly be related to the

linearity of the product as given in Table 3. Obviously,

linear hydroformylation products can only be obtained

via 1,2-(n) insertion of oleﬁns in Pdꢀhydride intermedi-

ates, whereas branched products are only accessible via

2,1-(iso) insertion. It can be seen from Table 4 that the

n/iso insertion ratios of oleﬁns in Pdꢀhydride clearly

increase with increasing bulk of the ligand. For example,

the C₂-bridged bidentate DsBPE affords a signiﬁcantly

lower preference for 1,2-insertion than the C₃-bridged

ligand DsBPP.

These observations indicate that the mode of oleﬁn

insertion is determined primarily by the space available

at the palladium center. It is thought that both oleﬁn

insertion in Pdꢀhydride and CO insertion in the Pdꢀalkyl

bond are reversible [18] and that in subsequent aldehyde

formation, the respective acyl intermediate is trapped

irreversibly by hydrogenolysis (see also Section 3.2.2 for

additional support of this proposal). Thus, formation

and/or trapping of n-acyl intermediates is favored over

that of the sterically more demanding iso-acyl intermedi-

ates.

The results given in Tables 4 and 5, indicate that the

anion also has a distinct effect on the mode of oleﬁn

insertion at palladium. The stronger the basicity of

anion, as judged from the pK_aof the associated acid, the

greater is the observed product linearity both in propene

and 1-octene hydroformylation.

As described earlier, it is suggested that the measure

of coordination of the anion towards the palladium

center determines the proximity of the anion and cation

in apolar media. Therefore, the observed higher prefer-

ence for n-acyl- over iso-acyl intermediates with the more

coordinating anions could be a consequence of the

greater congestion at the palladium center. It could also

be that the anion-assisted hydrogenolysis reaction dis-

criminates between a Pdꢀn-acyl and Pdꢀiso-acyl species,

with more strongly coordinating anions favoring n-acyl

hydrogenolysis.

We suggest that the coordinated carbonyl group of the

chelate ring can assist heterolytic H₂dissociation by

temporarily binding H⁺. This would be similar to the

proposed intermolecular assistance of the (weakly) basic

anions as proton acceptors in Pdꢀacyl hydrogenolysis.

The resulting Pd(hydride)(alkyl) species can eliminate

E. Drent, P.H.M. Budzelaar / Journal of Organometallic Chemistry 593–594 (2000) 211–225

221

Table 6

Regioselectivity in hydrocarbonylation ^a

b

c

Ligand

Acid

Oleﬁn

Aldehydes

Ketones

1,2-Insertion (%)

1,2-Selectivity (%)

n-Acyl selectivity (%)

DPPP

TFA

Propene

72

62

–

60

–

84

–

76

–

70

–

85

84

72

–

65

64

93

92

–

98

–

90

–

99

–

0

–

0

–

74

78

88

–

94

–

89

–

98

–

98

–

84

98

92 ^e

HOTs

HOTf

HOTs

HOTf

TFA

HOTf

TFA

HOTf

TFA

HOTf

TFA

HOTs

TFA

HOTf

DnBPP

DsBPP

DsBPE

DnBPP

DPPP

Propene

1-Octene

Styrene

Methyl acrylate

Methyl acrylate/1-octene

DPPP

DEPP

DnBPP

DEPP

–

0

0 ^d

^aBased on the data in Table 4.

^bPercentage 1,2-insertion in PdꢀH, percentage linear aldehydes+alcohols of total aldehydes+alcohols.

^cPercentage 1,2-selectivity, percentage of 1,2-insertion in any Pdꢀacyl; percentage n-acyl selectivity, percentage of any insertion in a Pd n-acyl

(see text).

^dPercentage 1,2 acrylate insertion in Pdꢀnonoyl.

^ePercentage acrylate insertion in Pdꢀn-nonoyl.

3.2.2. Hydro-acylation

(h to t)+(h to h)×[1−0.01×(t to t)] (%).

Obtaining the regio-selectivity data from the isomer

distribution of ketone products is somewhat more com-

plicated. We will use propene hydroacylation as an

example. Both saturated and unsaturated ketones are

formed. The h to h ketones (both saturated and unsatu-

rated) can only be formed by 2,1-insertion in

Pdꢀhydride, followed by 1,2-insertion in Pdꢀacyl. Simi-

larly, t to t isomers must be formed by 1,2-insertion

followed by 2,1-insertion. The saturated h to t isomer,

however, can be formed either by two 1,2-insertions or

by two 2,1-insertions. Two different unsaturated iso-

mers, h to t and t to h, can be formed, respectively by

two 1,2-insertions and two 2,1-insertions. The percent-

age of all ketones formed via 1,2-insertion in any acyl is

thus given by the amounts of (h to h)+(h to t) — the

amount of saturated h to t isomer formed by double

2,1-insertions. From the structure of the unsaturated

isomers, which is predominantly of the iso-propenyl-

type for all catalysts, it can be concluded that there is a

clear preference for 1,2-insertions in Pdꢀacyl. The cor-

rection for double 2,1-insertion is small but not negligi-

ble. In particular, with catalysts based on DPPP we

have detected a small quantity of n-propenyl-iso-propy-

lketone, which must have been formed by two consecu-

tive 2,1-insertions. We approximate the percentage of

double 2,1-insertion leading to saturated ketone by

0.01×(h to h)×(t to t) (%), so that the overall selectiv-

ity for 1,2-insertion in any acyl becomes:

Similarly, the selectivity for any insertion in an n-acyl is

given by:

(h to t)+(t to t)×[1−0.01×(h to h)] (%).

The latter selectivity, of course, corresponds with the

apparent selectivity for 1,2 insertion in Pdꢀhydride of

the ﬁrst oleﬁnic fragment in ketone formation. The

selectivities calculated in this way from the data in

Table 4 have been collected in Table 6. For compari-

son, the selectivity for 1,2-insertion in Pdꢀhydride in

aldehyde formation is included in the table.

It can be seen from the results tabulated in Tables 4

and 6, that in hydro-acylation of aliphatic oleﬁns there

is a strong preference for h to t coupling of the oleﬁnic

fragments and this preference becomes larger with the

steric bulk of the ligand. Thus, oleﬁn insertion in both

the Pdꢀhydride and the Pdꢀacyl bond is preferentially

1,2. Bulkier ligands and higher oleﬁns (cf 1-octene vs.

propene) strongly increase this preference, due to in-

creased steric congestion at the palladium center. The

aryl phosphine ligand DPPP has a surprisingly low

regio-selectivity for 1,2-insertion in Pdꢀacyl, even

though it is larger than the smallest alkylphosphine

used in the examples. This suggests that not only steric,

but also electronic factors (e.g. electrophilicity) are

important. A two-methylene bridged ligand affords a

lower preference for 1,2-insertion in Pdꢀacyl than a

222

E. Drent, P.H.M. Budzelaar / Journal of Organometallic Chemistry 593–594 (2000) 211–225

three-methylene bridged ligand, consistent with the

larger space available around the palladium center.

It can be seen from a comparison of the results on

regio-selectivity in hydro-acylation with those in hydro-

formylation that the proportion of ketones derived

from n-acyl intermediates is consistently and substan-

tially higher than the proportion of n-aldehydes derived

from the same intermediates. Whereas anion variation

can shift the balance from the simultaneous production

of aldehydes and ketones to pure ketone formation, the

regio-selectivities for oleﬁn insertion in Pdꢀacyl remain

practically constant (Tables 4 and 6, see DPPP and

DnBPP examples). This is an indication that both the

oleﬁn insertions in the Pdꢀhydride bond and CO inser-

tions in the Pdꢀalkyl bonds are indeed reversible as

suggested above. The acyl intermediate can, thus, be

trapped either by hydrogenolysis to yield hydroformy-

lation products or by oleﬁn insertion to give an inter-

mediate like 1 and eventually a ketone. Apparently, the

oleﬁn insertion reaction is more sensitive than hy-

drogenolysis to the steric difference between n- and

iso-acyl intermediates, and therefore has a higher pref-

erence for trapping the n-acyl selectively. This prefer-

ence becomes even higher with bulkier ligands or larger

oleﬁns.

The degree of unsaturation of the product ketones

reﬂects the relative rates of b-H elimination and hy-

drogenolysis (Scheme 1). The highest degree of unsatu-

ration is consistently found in the h to h coupled ketone

isomer. This may be related to the strength of the

internal coordination of the carbonyl group in structure

1. Above, we have argued that internal coordination

stabilizes this structure and suppresses b-elimination.

Substituents in both oleﬁnic units of the intermediate

(1) could destabilize the chelate structure and so pro-

mote b-elimination. This effect should be the largest for

the h to h coupled ketone, where both substituents are

closest to the carbonyl group. Apparently, hydrogenol-

ysis is not hindered as much by chelate formation; as

suggested above, this reaction could even be promoted

by chelate coordination.

lectivity of insertion of styrene in the Pdꢀacyl bond

(Tables 4 and 6). We ﬁnd a high preference for t to t

enchainment of the styrene moieties via carbonyl. This

t to t ketone isomer can only be formed by 1,2-insertion

in Pdꢀhydride and a consecutive 2,1-insertion in

Pdꢀacyl intermediates. As the unsaturated h to t isomer

observed is exclusively 1,4 diphenylpent-1-ene-3-one, it

can be concluded that styrene insertion in Pdꢀacyl takes

place exclusively in the 2,1-fashion. Ketones produced

with similar catalysts containing bipyridyl type chelat-

ing ligands show the same selectivity [20], and 2,1-inser-

tion of styrene in Pdꢀacyl bonds is also observed in

styrene–CO copolymerization [19].

Thus, the most surprising aspect is that styrene,

unlike aliphatic oleﬁns, inserts into the Pdꢀacyl bond

exclusi6ely in a 2,1-fashion. One factor may be that the

steric hindrance between the acyl group and the phenyl

group in the 1,2-insertion transition state inhibits the

1,2 regio-chemistry. It could also be that the interaction

between the palladium center and the phenyl ring dur-

ing the insertion provides a lower-energy reaction path-

way. Brookhart et al. [21] have shown that in

styrene–CO copolymerization with bipyridyl contain-

ing catalysts, the intermediate after each styrene inser-

tion exists as a rapidly exchanging mixture of allylic

and chelate structures 2 and 3 (K:0.3 at −80°C).

This indicates that allylic coordination is, indeed,

quite strong and should favor 2,1-insertion in the

Pdꢀacyl bond. Such an allylic coordination would, of

course, also favor the initial 2,1-insertion of styrene in

the PdꢀH bond. However, if — as we believe — both

this ﬁrst insertion and the subsequent CO insertion are

reversible, this does not affect the apparent regio-chem-

istry of the ﬁrst insertion as observed in the ﬁnal

product.

3.3. Hydrocarbonylation of functionalized oleﬁns

It can be seen from Table 2, that acrylic oleﬁns

deviate from aliphatic oleﬁns in chemo-selectivity: hy-

dro-acylation dominates over hydroformylation even

with the more coordinating TFA⁻anions. Apparently,

these oleﬁns insert into the Pdꢀacyl bond very easily.

Again, a strong preference is observed for t to t cou-

pling. Thus, hydro-acylation of methyl acrylate pro-

duces 4-oxo-dimethylpimelate exclusively. The same

regioselectivity is observed with acrylic acid and acry-

lamide. In these cases the primary hydro-acylation

products can undergo a double cyclization, as indicated

in Eq. (5), to form a di-spirolactone and a di-spirolac-

The hydrocarbonylation of styrene shows no signiﬁ-

cant deviations from those of aliphatic oleﬁns, dis-

cussed above, as far as chemo-selectivity is concerned

(Table 3).

With respect to regio-selectivity, the proportion of

ketones produced via styrene insertion in Pdꢀn-acyl is

considerably higher than the proportion of linear alde-

hydes produced via the same Pdꢀn-acyl intermediate.

This indicates that styrene, like aliphatic oleﬁns, prefer-

entially traps the n-acyl intermediate, presumably for

steric reasons. However, a peculiar deviation from the

aliphatic alkenes discussed above is seen in the regio-se-

E. Drent, P.H.M. Budzelaar / Journal of Organometallic Chemistry 593–594 (2000) 211–225

223

tam, respectively. These cyclizations may have occurred

after hydroacylation itself. Alternatively, it is possible

that the spiro-products are directly formed by the cata-

lyst, similar to poly-spiro-ketal formation in the copoly-

merization of carbon monoxide with higher oleﬁns

using similar catalysts [22].

oxidative addition steps, which may be involved in

product generating termination steps like hydrogenoly-

sis. The electrophilic palladium center not only can

bind and activate nucleophilic molecules, such as

oleﬁns, carbon monoxide and alcohols, but also

dihydrogen.

In any case, these products can only be formed via

1,2-insertion in Pdꢀhydride and 2,1-insertion in Pdꢀacyl

intermediates (Table 6). As with styrene, it is suggested

that both steric effects and the interaction of the elec-

tron-rich functionality with the electrophilic palladium

center contribute to the observed high selectivity to

a, v-functionalized products.

The cross-hydro-acylation between acrylic oleﬁns and

aliphatic oleﬁns also leads to the predominantly t to t

coupled linear mixed ketone, i.e. methyl 4-oxo-dode-

canoate with methyl acrylate and 1-octene (Tables 3

and 6, Eq. (8)).

The possibility of interrupting the chain-growth of

CO–oleﬁn co-polymerization by an efﬁcient chain-

transfer with dihydrogen either at the Pdꢀalkyl or the

Pdꢀacyl stage, forms the basis of the results on oleﬁn

hydrocarbonylation (oxo-synthesis) described in this

paper. With these palladium catalysts, the oxo-synthesis

can, thus, be fully exploited to produce, at choice,

aldehydes–alcohols by hydroformylation or ketones by

hydro-acylation.

We believe that the electrophilicity of the palladium

center is a key parameter for chemo-selectivity control

of the catalysis towards either reaction. Both the neu-

tral ligand and the anions can be used to adjust the

electrophilicity and the steric environment of the palla-

dium center with high precision. Apart from this, it

seems likely that the anions associated with the cationic

palladium center play an important role of their own.

Apparently anion basicity determines the efﬁciency of

hydrogenolysis at the Pdꢀacyl stage.

The interplay between the neutral chelating ligand

and the anion allows an almost perfect control over

selectivity. Both chemo- and regio-control can be

achieved; particularly in hydro-acylation, a nearly per-

fect selectivity to a single regio-isomeric ketone can be

obtained. We have demonstrated the principles with

only a limited number of ligands, anions and sub-

strates, but it is clear that they can be extended to many

other possible combinations. The application of chiral

neutral and/or anionic ligands would be another logical

extension. It can be expected that stereo-selective hy-

droacylation, in particular, is achievable, since stereose-

lective CO–oleﬁn copolymerization has already been

demonstrated [23].

(8)

In view of the results obtained with aliphatic and

acrylic oleﬁns separately, it is thought that the mixed t

to t coupled ketone must be formed by 1,2 insertion of

the aliphatic oleﬁn (present in excess) in Pdꢀhydride

followed by the very facile 2,1 insertion of the acrylic

oleﬁn in the PdꢀC(O)-n-C_nacyl intermediate. The ob-

served small quantity of h to t mixed ketone, methyl

4-oxo-5-methylundecanoate, must be formed by 2,1

insertion of 1-octene, also followed by 2,1 insertion of

methyl acrylate. The 2,1 insertion of methyl acrylate,

thus, selectively traps the n-C₈acyl intermediate from

an equilibrium mixture of n- and iso-C₈acyl intermedi-

ates. This may, again, be attributed to the greater steric

hindrance experienced by methyl acrylate for insertion

in a Pdꢀiso-C₈-acyl intermediate.

Although we have rationalized most of the observed

phenomena in a qualitative way it is clear that further

detailed studies of the elementary steps involved in the

catalysis are required to gain full insight into the factors

that control rate and selectivity in the palladium cata-

lyzed oxo-synthesis.

4. Conclusions

The cationic palladium complexes, L₂PdX₂, previ-

ously shown to be excellent catalysts for the alternating

copolymerization of oleﬁns with carbon monoxide [7],

owe their catalytic properties to the electrophilic nature

of the palladium(II) center. The metal has a square-pla-

nar environment made up of the cis-chelating neutral

ligand (L₂) and anionic ligands (X⁻). Cis-chelation by

the neutral ligand is considered essential for placing the

intermediate palladiumꢀhydride- and palladiumꢀcarbon

bonds cis to the fourth coordination site available to a

substrate molecule. This is an ideal situation for the

migratory insertion of the substrate molecules to gener-

ate intermediate Pdꢀalkyl- and Pdꢀacyl species. In addi-

tion, this arrangement favors reductive elimination and

5. Experimental

5.1. Analytical equipment

Product analysis of reaction mixtures was routinely

performed by gas–liquid chromatographic (GLC) anal-

ysis on a Perkin–Elmer 8500 equipped with two capil-

lary columns, Chrompack 50 m CP-sil-5 and 50 m

FFAP. Structural analysis was performed with GLC-

224

E. Drent, P.H.M. Budzelaar / Journal of Organometallic Chemistry 593–594 (2000) 211–225

mass spectroscopic (GC/MS) analysis on a Finnagin-

9610 gas chromatograph ﬁtted with the CP-sil-5 column

and coupled to a Finnigan-4000 triple-stage mass-spec-

trometer using electron impact ionization. This tech-

nique was applied to identify reaction products by

comparison with authentic samples.

autoclave from a high pressure ISCO pump. Subse-

quently, the autoclave was ﬁrst pressurized with 30 bar

of CO and then with 30 bar of hydrogen.

In about 10 min, the autoclave was heated to 70°C

and kept at this temperature by a Thermo-Electric 100

temperature control unit. The pressure was continu-

ously recorded by using a Transamerica Instruments

pressure transducer, series 2000. Activity data during

the experiment were calculated from the pressure de-

crease in time and from GLC analysis of the reaction

product at the end of a reaction period of 5 h. After

that time, the autoclave was allowed to cool, depressur-

ized and opened. Selectivity data were obtained from a

standard analysis of the ﬁnal reaction product by GLC.

From these analyses it appeared that n-butanal and

isobutanal were produced with a combined rate of

about 150 mol mol⁻¹Pd h⁻¹and selectivity of 86%.

The linearity of the aldehydes formed was determined

from the n/(n+iso) product ratio, being 77% in this

case. The byproducts, 4-oxo-3-methylheptanal and iso-

propenyl-n-propyl ketone (2-methylhex-1-ene-3-one)

were identiﬁed by GC/MS analysis and were produced

with a respective selectivity of 11 and 3%. Reaction

rates can vary over time with the level of substrate

conversion; the rate data in the tables are initial rates,

averaged over a period corresponding to B30%

conversion.

In the example above, the catalyst system was gener-

ated in situ by applying an excess of ligand over the

stoichiometric quantity required by Eq. (3). Under the

rigorous exclusion of oxygen, the same results could be

obtained with stoichiometric complexes prepared in situ

or in advance. We have observed that the application of

some excess of ligand (and acid) leads to more robust

catalysts, enabling easy handling under nitrogen blan-

keting to achieve reproducible, stable catalyst

performance.

¹³C-NMR spectra were recorded on a Bruker WM

250 spectrometer. ¹H- and ¹³P-NMR spectra were

recorded on a Bruker WM 250 spectrometer.

5.2. Materials

Palladium acetate, diglyme, THF, the Brønstedt

acids, p-toluenesulfonic acid, methanesulfonic acid,

triﬂuoroacetic acid and triﬂuoromethanesulfonic acid,

1-octene (p.a.), styrene (p.a.), methyl acrylate, acrylic

acid and acryl amide were all obtained from Merck and

were used as supplied. t-Butylsulfonic acid was pre-

pared in house by oxidation of t-butylthiol with hydro-

gen peroxide. Of the phosphine ligands, DPPP was

obtained from Strem, whereas DnBPP, DsBPP, Ds-

BPE, DEPP and DtBPP were all synthesized by stan-

dard preparation techniques [24], involving the reaction

between the respective di-alkylphosphine and 1,3 dibro-

mopropane or 1,2 dibromoethane to give the double

HBr salt of the respective di-phosphine. Neutralisation

with sodium hydroxide, and subsequent destillation

afforded the di-phosphine. Purity of the applied phos-

phines was always higher than 95% as observed with

¹H- and ³¹P-NMR. Phosphineꢀmonoxide was observed

as the main impurity. The enantiomeric composition of

DsBPP and DsBPE was statistical. The air-sensitive

phosphines were all stored and handled in a Glove box

under nitrogen. Propene of polymer grade quality was

used and was obtained from in house sources. Com-

mercial quality carbon monoxide (98%) and hydrogen

(\99%) were obtained from Air Products.

Whereas reaction product analysis in general was

performed with GLC-MS, some of the less common

products, such as dimethyl 4-oxopimelate, di-spirolac-

tone and di-spirolactam were isolated from reaction

media and characterized by ¹H- and ¹³C-NMR to

provide additional structural evidence.

5.3. Batch autocla6e procedure

All hydrocarbonylation experiments were carried out

in 250 or 300 ml magnetically stirred Hastelloy™ C

autoclaves, heated electrically. A typical experimental

procedure was as follows.

In a nitrogen glove box, the catalyst components,

0.25 mmol (56.1 mg) palladium acetate, 0.6 mmol

(199.6 mg) DsBPP were dissolved in 10 ml diglyme

contained in a small sealable bottle. The components

were allowed to react until no solid palladium acetate

was visible. The solution was transferred in the closed

bottle from the glove box and introduced in the auto-

clave preﬁlled with 40 ml degassed diglyme and 0.5

mmol (95.1 mg) p-toluenesulfonic acid (mono-hydrate),

which was blanketed under a continuous stream of

nitrogen. The autoclave was subsequently closed and

evacuated. Liquid propene (30 ml) was pumped in the

5.3.1. Di-methyl 4-oxo-pimelate

Following the procedure described above, a solution

of 30 ml of methylacrylate in 30 ml diglyme–20 ml

methanol was introduced in the autoclave. Subse-

quently, catalyst components, 0.25 mmol (56.1 mg)

palladium acetate, 0.3 mmol DnBPP and 0.5 mmol

(75.1 mg) HOTf were introduced under nitrogen. After

closure of the autoclave it was pressurized with 30 bar

CO and 30 bar H₂and heated to 75°C for 5 h. After

cooling to room temperature and slow depressurizing,

the autoclave was opened and its content was cooled to

−5°C overnight. Di-methyl 4-oxopimelate (12 g) was

E. Drent, P.H.M. Budzelaar / Journal of Organometallic Chemistry 593–594 (2000) 211–225

225

recovered by ﬁltration as colorless needles. ¹³C-NMR

(CDCl₃, 300Mhz: l(CH3ꢀO) 51.8 ppm, l(CH₃OꢀCꢁO)

173.5 ppm, l(CH₃OꢀC(O)ꢀCH₂ꢀ) 27.7 ppm,

l(ꢀCH₂ꢀC(O)ꢀ) 37.1 ppm; ¹H-NMR (CDCl₃,

200MHz): l(CH₃Oꢀ) 3.6 ppm (s), l(ꢀC(O)ꢀCH₂ꢀ) 2.5

ppm (t), l(ꢀCH₂ꢀC(O)ꢀ) 2.7 ppm (t).

[5] D. Evans, J.A. Osborn, G.J. Wilkinson, Chem. Soc. A (1968) 3133.

[6] B. Cornils, in: J. Falbe (Ed.), Hydroformylation, Oxo Synthesis,

Roelen Reaction in New Syntheses with Carbon Monoxide,

Springer-Verlag, Heidelberg, 1980.

[7] (a) E. Drent, J.A.M. van Broekhoven, M.J. Doyle, Organomet.

Chem. 417 (1991) 235. (b) E.Drent, P.H.M. Budzelaar, Chem.

Rev. 96 (1996) 613.

[8] (a) E. Drent, Eur. Pat. Appl. 220.767 (1985). (b) E. Drent, UK

Pat. Appl. 2.183.631 (1985). (c) E. Drent, Pure Appl. Chem. 62

(1990) 661.

5.3.2. Di-spirolactone

The same procedure with 10 ml of acrylic acid as the

substrate and 30 ml of diglyme as solvent, was followed

with the synthesis of di-spirolactone. After cooling and

addition of an equal volume of cyclohexane, 4.5 g of

di-spirolactone was recovered as a white solid: ¹³C-

NMR(CDCl₃, 300 MHz) l(ꢀOꢀC=O) 174.5 ppm,

l(ꢀOꢀC(O)ꢀCH₂ꢀ) 27.8 ppm, l(ꢀOꢀC(O)ꢀCH₂ꢀCH₂ꢀ)

32.3 ppm, l((ꢀCH₂ꢀ)₂Cspiro(ꢀOꢀ)₂) 112.7 ppm.

[9] E. Drent, E. Kragtwijk, D.H.L. Pello, Eur. Pat. Appl. 495.547

(1992).

[10] R. Stewart, in: H.H. Wasserman (ed.), The Proton; Applications

to Organic Chemistry in Organic Chemistry, vol. 46, Academic,

London, 1985.

[11] The term hydro-acylation has also been used to denote the reaction

of oleﬁns with aldehydes to produce ketones. Because of the

analogy with hydroformylation, however, we prefer to reserve this

term for the synthesis of ketones from oleﬁns, carbon monoxide

and hydrogen.

[12] G.P.C.M. Dekker, C.J. Elsevier, K. Vrieze, P.W.N.M. van

Leeuwen, C.F. Roobeek, J. Organomet. Chem. 430 (1992) 357.

[13] C.A. Tolman, Chem. Rev. 77 (1977) 313.

5.3.3. Di-spirolactam

Following the same procedure with 15 g of acryl

amide, instead of acrylic acid, yielded 7.4 g of di-spiro-

lactam as a white crystalline solid: ¹³C-NMR (D2O, 300

MHz), l(ꢀN(H)ꢀCꢁO) 180.6 ppm, l(ꢀN(H)ꢀC

(O)ꢀCH₂ꢀ) 25.7 ppm, l(ꢀN(H)ꢀC(O)ꢀCH₂ꢀCH₂ꢀ) 42.8

[14] An alternative possibility of intermediate Pd(IV)ꢀdihydride species

obtained via an oxidative pathway of Pd(II) and dihydrogen,

similar to that proposed for hydrogenolysis of Ir(I)ꢀacyl interme-

diates (P.P. Deutsch, R. Eisenberg, Organometallics 9 (1990) 709),

is considered unlikely. The generation of Pd(IV) from Pd(II)

species has, however, been demonstrated, but generally requires

hard ligands and strong oxidants, such as alkyl iodides (see for

example, P.K. Wong, J.K. Stille, J. Organomet. Chem. 70 (1974)

121). (a) D. Milstein, J.K. Stille, J. Am. Chem Soc. 101 (1972)

4992. (b) P.K. Byers, A.J. Canty, A.H. White, J. Chem. Soc. Chem.

Commun. (1986) 1722).

1

ppm, l((ꢀCH₂)₂ꢀCspiroꢀ(N(H)ꢀ)₂) 62.4 ppm; H-NMR

(D₂O, 200 MHz) l(ꢀN(H)ꢀC(O)ꢀ) 7.7 (weak due to

D/H exchange with D₂O), l(ꢀC(O)ꢀCH₂)ꢀ) 2.2 (br),

l(ꢀC(O)ꢀCH₂ꢀCH₂ꢀ) 1.6 (br).

[15] A.M. Joshi, B.R. James, Organometallics 9 (1990) 199.

[16] E. Drent, E. Kragtwijk, Eur. Pat. Appl. 495.548 (1992).

[17] (a) J.X. McDermott, J.F. White, G.M. Whitesides, J. Am. Chem.

Soc. 95 (1973) 4451. (b) ibid, 98 (1976) 6521.

[18] Strictly speaking, only one of the two CO insertions needs to be

reversible. There is precedent for a difference in reversibility

between n-acyl and iso-acyl formation in rhodium catalyzed

hydroformylation (see R. Lazzaroni, P. Pertici, G.J. Fabrici, J.

Mol. Catal. 58 (1990) 75).

[19] (a) E. Drent, Eur. Pat. Appl. 229.408 (1986). (b) P. Corradini, A.

de Rosa, A. Panunci, P. Pino, Chimia (1990) 4452. (c) M.

Barsacchi, G. Consiglio, L. Medici, U.W. Suter, Angew. Chem.

103 (1991) 992.

[20] (a) C. Pisano, G. Consiglio, A. Sironi, M.J. Moret, Chem. Soc.

Chem. Commun. (1991) 421. (b) C. Pisano, A. Mezzetti, G.

Consiglio, Organometallics 11 (1992) 20.

Acknowledgements

The authors are indebted to W.W. Jager, D.H.L.

Pello and E. Kragtwijk for their skillful technical assis-

tance. Thanks are also due to M.A. Nekkers, M.C. van

Grondelle and J.J. de Boer for performing GC/MS

analyses and to O. Sudmeijer for his assistance in NMR

measurements. Also appreciated is the support and

encouragement given to this project by Dr T.A.B.M.

Bolsman.

References

[21] M. Brookhart, F.C. Rix, J.M. DeSimone, J.C. Barborac, J. Am.

Chem. Soc. 114 (1992) 5894.

[1] O. Roelen, Deutsches Patent Schrift 849.548 (1938/1952), US

patent 2.327.066 (1943).

[22] P.K. Wong, J.A. van Doorn, E. Drent, O. Sudmeijer, H.A. Stil,

Ind. Eng. Chem. Res. 32 (1993) 986.

[2] O. Roelen, Chem. Exp. Didakt 3 (1977) 119.

[3] H. Adkins, G. Krsek, J. Am. Chem. Soc. 70 (1948) 383.

[4] L.H. Slaugh, R.D. Mullineaux, US patents 3.239.569 and

3.239.570 (1969).

[23] (a) P.K. Wong, Eur. Pat. Appl. 384.517 (1989). (b) A. Batistini,

G. Consiglio, U.W. Suter, Angew. Chem. 104 (1992) 306.

[24] J.A. van Doorn, Functionalised Phosphines, Thesis, University of

Amsterdam, The Netherlands, 1991.

.

Article Doi

The oxo-synthesis catalyzed by cationic palladium complexes, selectivity control by neutral ligand and anion

DOI: 10.1016/S0022-328X(99)00554-9

Source and publish data:

Authors:

Article abstract of DOI:10.1016/S0022-328X(99)00554-9

Full text of DOI:10.1016/S0022-328X(99)00554-9

Products guided by the article

R&D Labs maybe for 123-72-8

Relevant to this article

Hot Product