Multisite catalysis: A mechanistic study of β-lactone synthesis from epoxides and CO - Insights into a difficult case of homogeneous catalysis

Article abstract of DOI:10.1002/chem.200390144

Carbonylation of epoxides with a combination of Lewis acids and cobalt carbonyls was studied by both theoretical and experimental methods. Only multisite catalysis opens a low-energy pathway for trans opening of oxirane rings. This ring-opening reaction is not easily achieved with a single-site metal catalyst due to structural and thermodynamic constraints. The overall reaction pathway includes epoxide ring opening, which requires both a Lewis acid and a tetracarbonylcobaltate nucleophile, yielding a cobalt alkyl - alkoxy-Lewis acid moiety. After CO insertion into the Co-C_alkyl bond, lactone formation results from a nucleophilic attack of the alkoxy Lewis acid entity on the acylium carbon atom. A theoretical study indicates a marked influence of the Lewis acid on both ring-opening and lactone-formation steps, but not on carbonylation. Strong Lewis acids induce fast ring opening, but slow lactone formation, and visa versa: a good balance of Lewis acidity would give the fastest catalytic cycle as all steps have low barriers. Experimentally, carbonylation of propylene oxide to β-butyrolactone was monitored by online ATR-IR techniques with a mixture of tetracarbonylcobaltate and Lewis acids, namely BF₃, Me₃Al, Et₂Al⁺ · diglyme, and a combination of Me₃Al/dicobaltoctacarbonyl. We found that the last two mixtures are extremely active in lactone formation.

Full text of DOI:10.1002/chem.200390144

FULL PAPER
Multisite Catalysis: A Mechanistic Study of b-Lactone Synthesis from
Epoxides and CO–Insights into a Difficult Case of Homogeneous Catalysis
[
a]
[b]
[c]
[c]
Ferenc Molnar,* Gerrit A. Luinstra,* Markus Allmendinger, and Bernhard Rieger*
Abstract: Carbonylation of epoxides
with a combination of Lewis acids and
cobalt carbonyls was studied by both
theoretical and experimental methods.
Only multisite catalysis opens a low-
energy pathway for trans opening of
oxirane rings. This ring-opening reaction
is not easily achieved with a single-site
metal catalyst due to structural and
thermodynamic constraints. The overall
reaction pathway includes epoxide ring
opening, which requires both a Lewis
acid and a tetracarbonylcobaltate nucle-
ophile, yielding a cobalt alkyl–alkoxy ±
Lewis acid moiety. After CO insertion
into the CoÀC
bond, lactone forma-
bonylation. Strong Lewis acids induce
fast ring opening, but slow lactone for-
mation, and visa versa: a good balance
of Lewis acidity would give the fastest
catalytic cycle as all steps have low
barriers. Experimentally, carbonylation
of propylene oxide to b-butyrolactone
was monitored by online ATR-IR tech-
niques with a mixture of tetracarbonyl-
alkyl
tion results from a nucleophilic attack of
the alkoxy Lewis acid entity on the
acylium carbon atom. A theoretical
study indicates a marked influence of
the Lewis acid on both ring-opening and
lactone-formation steps, but not on car-
cobaltate and Lewis acids, namely BF ,
3
Keywords: carbonylation
¥ homo-
‡
Me Al, Et Al ¥ diglyme, and a combi-
3
2
geneous catalysis ¥ multisite cataly-
sis ¥ polymerization ¥ reaction mech-
anisms
nation of Me Al/dicobaltoctacarbonyl.
3
We found that the last two mixtures are
extremely active in lactone formation.
Introduction
catalysts has its inherent limitations, that is, excellent catalysts
are available for olefin and CO transformations such as
metathesis,[ polymerization, hydrogenation, (hydro)car-
3]
[4]
[5]
Catalysis is the key to controlling chemical reactions in many
cases. Recognizing and understanding the functionality of
catalysts has without doubt been of tremendous importance
for progress in chemistry. In this context, homogeneous
catalysis has very much become a domain of organometallic
chemistry.[ Recently, combinatorial and parallel experimen-
tation techniques have been introduced into the field.
Encouraging results are emerging, but these do not necessa-
rily represent a systematic extension to our understanding of
catalysis.[ Catalysts have been developed for a fair number of
reactions. However, the scope of the so-called single-site
bonylation,[ epoxidation, and other processes in which
6]
[7]
individual elementary reactions of a sequence of steps are
[8]
thermodynamically accessible. Other processes, however,
like methane oxidation,[ nitrogen fixation,
9]
[10]
and also
1]
carbonylation of epoxides (vide infra) are not yet conven-
iently amenable.[
11]
This might be the consequence of an oversimplified
approach and too narrow a view of single-site catalysis for
™non-simple reactions∫: both in enzyme and heterogeneous
catalysis, multiple sites and multiple interactions are of
essential importance. One key element of the catalytic action
of enzymes is the ™favorable∫ orientation of the substrate in
both space and time with respect to some catalytic entity.
Minute structural and dynamic details of the multiple
interactions of the active site with the substrate are very
important for a highly selective and smooth transition from
reactants to products. In heterogeneous catalysis, in a differ-
ent and yet related way, several intermediates in metal
catalysts for carbon monoxide hydrogenation, for example,
are stabilized on the catalytic surface with its multiple binding
sites and types.
2]
[
a] Dr. F. Molnar
BASF Aktiengesellschaft, Polymer Research
7056 Ludwigshafen (Germany)
6
Fax : (‡49)621-60-92281
E-mail: ferenc.molnar@basf-ag.de
[
b] Dr. G. A. Luinstra
BASF Aktiengesellschaft, Polymer Research
7056 Ludwigshafen (Germany)
6
Fax : (‡49)621-60-20313
E-mail: gerrit.luinstra@basf-ag.de
[
c] Prof. B. Rieger, M. Allmendinger
Department of Materials and Catalysis, University of Ulm
Homogeneous hydrogenation of CO on the other hand is
still a challenging reaction.[ Although multisite catalysis has
been around in homogeneous catalysis for some time, it has
8
9069 Ulm (Germany)
12]
Fax : (‡49)731-50-23039
E-mail: bernhard.rieger@chemie.uni-ulm.de
Chem. Eur. J. 2003, 9, No. 6
¹ 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
0947-6539/03/0906-1273 $ 20.00+.50/0
1273

F. Molnar, G. A. Luinstra, B. Rieger, and M. Allmendinger
FULL PAPER
perhaps received too little attention thus far.[13] It seems to
play a role in reactions with relatively strong bonds: these
cannot be broken without sufficient compensation. Such
reactions that do not readily occur are, for example, the
oxidative addition of strong CÀH bonds in alkanes to
However, ring opening of ethylene oxide, which should be a
À
step in forming the lactone with ™naked∫ Co(CO) , was not
4
observed; slowly decreasing the distance between Co and a
C atom of ethylene oxide, while at each point relaxing the rest
of the molecular structure, eventually does lead to opening of
the epoxide. However, the resulting structure is not stable and
reverts to the cobaltate ion and ethylene oxide after the
distance constraint is relaxed.
(
electron-rich) metals or dehydrogenation of alcohols. Op-
portunities for homogeneous catalysis become wider if, as in
the case of enzymes, more than one center of a catalytic
system directs the reaction. This is perspicuous when consid-
ering the smooth activation of methane at room temperature
with a dinuclear rhodium complex,[ the action of rutheni-
um ± nitrogen complexes in the reduction of keto com-
For reaction to occur, it seems that an initial activation
(polarization) of ethylene oxide is required for two reasons: 1)
the energetic barrier to ring opening must be lowered by
stabilizing the transition state, and 2) the ring-opened
structure needs to be stabilized. The latter can be achieved,
for example, through interaction with the alkoxy oxygen atom
generated. An elegant way to arrive at such a situation is the
use of a two-site interaction. Polarization of ethylene oxide
could in general be affected by interaction of electron-
deficient species with the epoxide oxygen atom. On the one
hand, the use of a proton for the purpose of polarizing
ethylene oxide results in immediate ring opening of ethylene
oxide without direct participation of the carbonylation
catalyst, and the concomitant risk of side reactions taking
place. A weaker electrostatic interaction on the other hand,
such as with the tetramethylammonium ion, does not improve
14]
pounds,[ or the hydroxylation of olefins. In the first of
these, a hydrogen atom and the corresponding methyl group
end up at two different sites, and in the second case, two
hydrogen atoms are delivered from a ruthenium center and a
nitrogen functionality, respectively.
15]
[16]
We became interested in the catalytic conversion of
oxiranes,[ and performed studies on metal-centered reac-
tions involving single-site catalysts with these substrates,
which clearly behave quite differently to simple olefins. To
minimize time-consuming experiments, a broad theoretical
study was initiated. The theoretical results and their exper-
imental verification are reported herein. We demonstrate that
multisite homogeneous catalysis is key to controlling the
reactivity of epoxides. Recently, two publications appeared on
the carbonylation of oxiranes with a mixture of Lewis acids
and cobalt carbonyls.[18] Here also cooperation of two differ-
ent catalytic centers is conjecturable. Older reports by Inoue
and Aida on aluminum porphyrins show similar behavior.[
The notion of multisite catalysis is therefore not new in this
context, but it still has to establish its role in the field of
homogeneous catalysis. We therefore want to offer ™multisite
thinking∫ as a general concept for the future development of
tailor-made homogeneous catalytic systems.
17]
À
the situation compared to Co(CO)4 alone: instead of polar-
izing ethylene oxide, the cation tends to associate with the
cobaltate anion and to decrease the anion×s effective nucle-
ophilicity. A whole variety of Lewis acids with ™tunable∫ acid
strength are available in between the tetramethylammonium
ion and a proton. Lewis acids turn out to be ideal candidates
for polarizing epoxides for ring opening in combination with
cobalt carbonyl complexes and for eventually forming b-
lactones by multisite catalysis.
19]
The carbonylation reaction is thought to proceed through
consecutive steps of epoxide activation, ring opening, CO
insertion, and ring closure to eventually yield the lactone.
Various mechanisms for activation, ring opening, and CO
insertion have been put forward;[ however, details about
these reactions are still unknown. We set out to elucidate the
characteristics of the reaction pathway with the help of
quantum-chemical calculations. Density functional theory
(DFT) was used to determine the potential energy surfaces
and structures of the key intermediates in the proposed
reaction scheme (Scheme 2). Details of the methodology used
for the calculations can be found later.
20]
Results and Discussion
The formation of b-lactones through cobalt-catalyzed carbon-
ylation of epoxides has been observed in a number of cases
and as a side reaction in the synthesis of polyesters (vide
supra).[ It is this reaction (Scheme 1) that is in the focus of
our efforts.
17]
An active catalytic system comprises the combination of the
tetracarbonylcobaltate anion and a Lewis acid. The Lewis
acid polarizes (activates) the epoxide, and the cobaltate ion
facilitates ring opening through backside attack. The resulting
alkoxy group is stabilized by a Lewis acid. Thereafter, CO
Scheme 1. Generalized formation of lactones starting from epoxides and
CO.
insertion occurs into the CoÀC
bond. After the coordina-
alkyl
We started our theoretical investigation by considering the
interaction of ethylene oxide with single-site catalysts. How-
ever, in spite of numerous attempts, calculations (also
involving Zn and Pd catalysts, data not shown) did not reveal
a chemically attractive low-energy pathway for ethylene oxide
ring opening at the metal center. Following these initial
tion sphere of Co has again been completed by the addition of
a CO ligand, lactone ring closure with subsequent dissociation
of the lactone/Lewis acid complex and carbonylation unit
takes place.
Each of the following subsections discusses one of the
reaction steps of the proposed mechanism of Scheme 2.
attempts, we turned to cobalt carbonyl complexes, with the
Calculations for each step were carried out for the following
À
‡
cobaltate anion [Co(CO) ] as a model for the active species.
Lewis acids: Al(iPrO)
3
, AlMe
3
, AlMe
2
¥ diglyme, Bi(OMe)
3
,
4
1274
¹ 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
0947-6539/03/0906-1274 $ 20.00+.50/0
Chem. Eur. J. 2003, 9, No. 6

Multisite Catalysis
1273±1280
acids (cf. O ± B in Table 2), a
shorter bond length indicates a
stronger (interaction with the)
Lewis acid. Stronger Lewis acid
strength corresponds to lower
activation energies for the ring-
opening TS and to more exo-
thermic reaction energies. Ring
‡
opening involving AlMe2 ¥ di-
glyme, which is the strongest
Lewis acid in Table 2, repre-
sents an extreme case (very
exothermic) and behaves differ-
ently to the rest of the Lewis
acids, as will be discussed in
detail later.
Scheme 2. Reaction scheme for formation of lactones starting from epoxides and CO.
Bi(O CPh) , BF3, BF ¥ Et O, BH3, B(OMe)3, BMe3,
Table 2. Energies for EO ring opening with Lewis acid assistance.[a]
2
3
3
2
SnMe Cl, and SO . Ethylene oxide was used as the model
3
3
[kJmolÀ1][b]
[kJmolÀ1][c]
E
r
d [ä][d]
E
a
epoxide. Structures and energies for each step were calculated
as a function of Lewis acid strength. Figures are used to
illustrate molecular structures at critical points on the
Al(OiPr)
3
14
23
11
82
34
14
16
25
35
27
41
13
À 42
À 12
À 116
35
À 34
À 38
À 44
5
1.81
1.86
1.77
2.17
2.33
1.50
1.49
1.52
1.52
1.56
2.26
1.80
AlMe
AlMe
3
potential energy surface for the BF case.
‡
3
2
¥ diglyme
Bi(OMe)
Bi(O CPh)
3
2
3
Ring opening of epoxide and formation of Co ± alkyl com-
BF
BF
3
plexes: Figure 1 displays the ring opening of ethylene oxide by
3
2
¥ Et O
À
backside attack of Co(CO)4 with BF as the Lewis acid. From
3
BH3
B(OMe)
B(Me)
SnClMe
SO
the geometrical parameters presented in Table 1 it can be
inferred that a CoÀC interaction is present in the reactants
structure even before reaction; the CoÀC distance shortens in
the transition state (TS). In the product, a typical CoÀC bond
3
À 3
3
3
3
18
À 52
3
[
a] All energies were determined at the B-P86/TZVP//B-P86/SV(P) level
except for systems containing Bi, for which the B-P86/SV(P) level energies
are given. [b] E
length of approximately 2.2 ä is achieved. Ring opening is
also characterized by a growing CÀO bond length, on going
from reactant to product. The CÀO bond is already weak in
a
ˆ activation energy. [c] E ˆ Reaction energy. [d] Bond
r
length between the Lewis acid and the alkoxy oxygen atom.
the TS structure. In addition, based on bond length consid-
erations, the B ¥¥¥ O interaction is strongest in the product
structure, thus stabilizing the (formation of the) alkoxy-type
oxygen.
CO insertion: Figure 2 shows the details of the CO insertion
À
into the Co
BF
ÀC bond of the b-alkoxy species in the CoCO /
4
Table 2 gives an overview of the energetics of the ring-
opening reaction with various Lewis acids. Assuming that
Lewis acid strength correlates with bond length between the
Lewis acid and the oxygen atom within a set of related Lewis
3
system. A cobalt acyl species is formed in the process. The
CO inserted originates from the ligand sphere. The geo-
metrical data in Table 3 shows the changes in structure. The
length of the C
ÀCO bond formed in the reaction shortens
from 2.70 ä to a typical CÀC
value of 1.51 ä. The BÀO bond
length remains virtually unaf-
fected by this transformation.
The Lewis acid is remote from
the cobalt site, and therefore as
expected the energetics of the
CO insertion are more or less
Figure 1. Schematic representation of the ring-opening event. From left to right: reactants, transition state (TS),
and product structure. Color code: Co: dark blue; C: gray; O: red; H: white; B: pink; F: light blue.
independent of the type of
Lewis acid (Table 4). As in the
‡
case of ring opening, AlMe2
¥
Table 1. Geometry data: distances between atoms.[a,b]
diglyme represents a special case. The influence of the cation
is even felt by the insertion of CO into the remote CoÀC bond.
Initial
TS
Product
CoÀC
CÀO
BÀO
3.659
1.484
1.683
3.003
1.769
1.602
2.199
2.337
1.499
CO uptake and lactone formation: The formation of the free
lactone does not occur before uptake of additional CO
completes the coordination sphere of cobalt to [Co(CO)4-
[
a] Bond lengths [ä]. [b] Data corresponds to structures in Figure 2.
Chem. Eur. J. 2003, 9, No. 6
¹ 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
0947-6539/03/0906-1275 $ 20.00+.50/0
1275

F. Molnar, G. A. Luinstra, B. Rieger, and M. Allmendinger
FULL PAPER
marked effect on the transition
state structure (Table 5) and
energetics of lactone ring clo-
sure (Table 6) in this step.
Weaker Lewis acids lead to
smaller activation energies and
more exothermic reaction en-
Figure 2. Schematic representation of CO insertion into the CoÀC bond. Energetics of insertion is indifferent to
the nature of the Lewis acid. From left to right: reactants, transition state (TS), and product structure. Color code:
Co: dark blue; C: gray; O: red; H: white; B: pink; F: light blue.
ergies. They tend to bind less
strongly to the alkoxy group
and render it more reactive for
the formation of the final CÀO
Table 3. Geometry data: distances between atoms.[a,b]
bond. This effect can also be inferred by considering the OÀB
(
Lewis acid) bond length (Table 5). In the TS, the OÀB (Lewis
Initial
TS
Product
acid) bond is stretched from the initial structure. This
elongation will be easier to achieve with weaker Lewis acids
which is also expected intuitively. After ring closure to the
lactone is complete, the Co ¥¥¥ C interaction in the product can
be transformed to a Co ¥¥¥ H interaction between the tetra-
carbonylcobaltate and the lactone, which is energetically
more favorable (cf. Table 6; AlMe versus AlMe _H). For the
[
c]
CoÀC
1.798
2.698
1.498
1.765
2.007
1.502
1.888
1.511
1.499
[
d]
CÀCO
OÀB
[
[
a] Bond lengths [ä]. [b] Data corresponds to structures in Figure 3.
c] Distance between cobalt and carbonyl carbon atom. [d] Carbon ± car-
bon bond created in reaction.
3
3
‡
strong Lewis acid AlMe2 ¥ diglyme this switch from Co ¥¥¥ C to
Co ¥¥¥ H interaction can already occur in the transition state of
ring closure, and constitutes an alternative reaction channel.
[
a]
Table 4. Energies for CO insertion into CoÀC bond.
[kJmol ][b]
À1
E
r
[kJmol ][c]
À1
d [ä][d]
E
a
‡
Al(OiPr)
3
27
24
35
19
22
25
28
22
24
21
26
24
28
À 16
À 17
4
1.81
1.79
1.79
2.25
2.14
1.50
1.49
1.52
1.52
1.56
1.52
2.28
1.80
Since AlMe2 ¥ diglyme is a strong Lewis acid, a high barrier to
Al(CH
Al(CH
3
)
)
3
lactone formation would be expected based on the data for
the other Lewis acids. This is indeed found for the ™normal∫
‡
3
2
¥ diglyme
Bi(OCH
3
)
3
À 24
À 24
À 19
À 16
À 23
À 21
À 19
À 19
À 13
À 17
‡
reaction channel (cf. Table 6, AlMe2 ¥ diglyme). However,
Bi(O CPh)
2
3
BF
BF
3
3
switching from the Co ¥¥¥ C to Co ¥¥¥ H interaction in the
¥ Et
2
O
‡
transition state (c.f. Table 6, AlMe2 ¥ diglyme H) opens up an
BH
B(OCH
B(CH
CO
SnCl(CH
SO
3
alternative low-energy pathway for lactone formation.
3 3
)
3 3
)
Experimental investigation: The theoretical study provided
the insight that a multicomponent catalytic system based on
tetracarbonylcobaltate and various Lewis acids should induce
the carbonylation of epoxides. We set therefore out to screen
such combinations for catalytic activity. Bis(triphenylphos-
phoranylidene)ammonium (PPN) or tetraethylammonium
cobaltcarbonylate salts (PPN[Co(CO) ] or Et N[Co(CO) ])
2
3 3
)
3
[
a] All energies were determined at the B-P86/TZVP//B-P86/SV(P) level
expect for systems containing Bi, for which the B-P86/SV(P) level energies
are given. [b] E ˆ Reaction energy. [d] Bond
ˆ activation energy. [c] E
length between the Lewis acid and the alkoxy oxygen atom.
a
r
4
4
4
(
acyl)]. This became clear after
multiple attempts to locate
transition state structures failed
for a Co(CO) species. A com-
3
mon feature of all these fruit-
less attempts was that the lac-
tone resulting from a nucleo-
philic attack of the alkoxy
moiety at the acyl carbon was
bound too strongly to cobalt at
the vacant coordination site.
After prior CO uptake, mod-
Figure 3. Schematic representation of the ring-closure step. Weaker Lewis acids lead to easier CÀO bond
formation. From left to right: reactants, transition state (TS), and product structure. Color code: Co: dark blue; C:
gray; O: red; H: white; B: pink; F: light blue.
Table 5. Geometry data: distances between atoms.[a,b]
eled by a simple addition of CO to cobalt and relaxation of the
structure, the resulting [Co(CO) (acyl)] species readily con-
4
Initial
TS
Product
verts to propiolactone as the BF3 (or other Lewis acid)
complex.
Figure 3 displays this final step of the proposed reaction
pathway. Ring closure occurs by nucleophilic attack of the
alkoxy oxygen on the carbonyl carbon atom. During this
attack the CoÀCacyl bond is broken. The Lewis acid has a
[c]
CoÀC
2.127
2.904
1.502
2.664
1.860
1.565
4.929
1.459
1.751
[d]
CÀO
OÀB
[
[
a] Bond lengths [ä]. [b] Data corresponds to structures in Figure 4.
c] Distance between cobalt and carbonyl carbon atom. [d] Carbon ± oxy-
gen bond created in reaction.
1276
¹ 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
0947-6539/03/0906-1276 $ 20.00+.50/0
Chem. Eur. J. 2003, 9, No. 6

Multisite Catalysis
1273±1280
Table 6. Energies for lactone ring closure.[a]
A rather complicated IR spectrum was obtained by mixing
À
À1
E
a
[kJmol ][b]
E
r
[kJmol ][c]
À1
d [ä][d]
BF ¥ OEt and PPN[Co(CO) ]. Next to Co(CO)4 ( nÄ ˆ
3
2
4
À1
1
887 cm ), several absorptions of metal carbonyls were
Al(iPrO)
3
27
28
24
97
0
À 41
À 55
À 90
88
À 24
À 18
À 61
À 88
À 18
1.99
2.08
2.06
1.99
1.94
1.76
1.64
1.89
2.29
À1
present at 1800 ± 2200 cm , reminiscent of several cobalt
carbonyl species. Addition of propylene oxide and carbon
monoxide to this solution yielded a clean spectrum assigned to
Al(CH
Al(CH
Al(CH
Al(CH
3
)
3
)
3
)
3
)
3
3
2
2
H
‡
‡
¥ diglyme
¥ diglyme H
À
À1
a mixture of three species: Co(CO) ( nÄ ˆ 1887 cm ), b-
4
BF
BH
B(CH
SO
3
44
23
9
À1
butyrolactone ( nÄ ˆ 1829 cm ), and a tetracarbonyl cobaltate
3
)
3
acyl species with absorptions at 2107, 2043, 2024, 2005, and
3
À1
[21]
3
50
1710 cm (Figure 4).
Figure 5 represents an IR stack-plot of the catalytic
carbonylation reaction of propylene oxide with BF ¥ OEt as
Lewis acid to give b-butyrolactone. Apart from minor
[
[
a] All energies were determined at the B-P86/TZVP//B-P86/SV(P) level.
b] E
a
ˆ activation energy. [c] E ˆ Reaction energy. [d] Bond length be-
r
3
2
tween the Lewis acid and the alkoxy oxygen atom.
are not active on their own
(
Table 7, entries 1 and 2) as
expected, but are both suitable
À
sources for Co(CO)4 ions.
The carbonylation reactions
were monitored in an autoclave
TM
equipped with a ReactIR Si-
TM
Comp probe for in situ ATR-
IR measurements under high-
pressure conditions. In the
starting situation, we could de-
tect the characteristic absorp-
tion of the tetracarbonylcobal-
À1
tate ion ( nÄ ˆ 1887 cm ). Addi-
tion of propylene oxide to the
mixture did not result in a
change in the carbonyl region,
demonstrating that without
Lewis acids, no reaction takes
place. Next, combinations of
À
Figure 4. IR spectrum of the carbonylation reaction with BF
3
¥ OEt
2
as Lewis acid after 15 min [Co(CO)
4
( nÄ ˆ
À1
À1
1
1
887 cm ), b-butyrolactone ( nÄ ˆ 1829 cm ), tetracarbonylcobaltate acyl ( nÄ ˆ 2107, 2043, 2024, 2005, and
À1
710 cm )]
tetracarbonylcobaltate ion salts with a series of Lewis acids
were investigated for catalytic activity. We will present our
amounts of polyester no other species were detectable. While
À1
the concentration of b-butyrolactone ( nÄ ˆ 1829 cm ) grows
‡
findings for the Lewis acids BF , AlMe , and Et Al in
linearly with time during the first 2 ± 3 h, the carbonylation
reaction slows down somewhat afterwards. Conversion after
5 h was determined to be 40% (Table 7, entry 3).
3
3
2
combination with various sources of the tetracarbonylcobal-
tate ion. Diglyme was used as solvent.
Comparable
observations
Table 7. Carbonylation reactions: conversion of propylene oxide to b-butyrolactone.[a]
were made with AlMe3 as
Lewis acid in combination with
Entry
Catalyst
equiv)
Epoxide
(equiv)
Reaction
time [h]
Temperature
Yield
[%][
PPN[Co(CO) ]. Again, the IR
b]
4
(
[8C]
spectrum is indicative of the
same three compounds (Fig-
ure 6), albeit in different ratios.
The concentration of the tetra-
carbonylcobaltate is higher rel-
ative to the corresponding acyl
species when BF ¥ OEt is used
as the Lewis acid. This indicates
the impact of the Lewis acidity
on the catalytic reaction: the
acyl species is more reactive, or
in other words, the epoxide
ring-opening reaction is rela-
tively slow, resulting in a high
concentration of the tetracar-
bonylcobaltate resting state.
1
2
3
4
5
6
7
8
9
Et N[Co(CO)
4
4
] (1)
] (1)
] (1)/BF
] (1)/AlMe
PO (100)
PO (100)
PO (160)
PO (160)
PO (80)
20
20
20
20
20
4
5
4
2
16
75
75
75
75
75
75
75
75
95
95
0
0
PPN[Co(CO)
PPN[Co(CO)
PPN[Co(CO)
4
4
4
3
¥ OEt
2
(1)
90 (88)
65 (93)
10
70 (92)
100 (96)
100 (92)
100 (92)
80 (90)
3
(1)
Na[Co(CO)
Na[Co(CO)
4
] (1)
] (1)/ClAlEt
4
2
(1)
PO (80)
[Co
[Co
[Co
[Co
2
2
2
2
(CO)
(CO)
(CO)
(CO)
8
] (1)/AlMe
8
] (1)/AlMe
8
] (1)/AlMe
8
] (1)/AlMe
3
3
3
3
(2)
PO (160)
PO (160)
PO (160)
PO (1200)
3
2
(4)
(4)
(4)
1
0
[
(
a] Reaction conditions: 250-mL B¸chi autoclave, diglyme (50 mL), 60 bar CO, entries 1 ± 4: PPN[Co(CO)
1.65 g), Et N[Co(CO) ] (696 mg), BF ¥ OEt (300 mL), AlMe (2.3 mL of 2m solution in toluene), corresponding
amount of PO; entries 5,6: Na[Co(CO) ] (900 mg), ClAlEt (4.6 mL of 1m solution in heptane), corresponding
amount of PO; entries 7 ± 10: [Co (CO) ] (780 mg), Me Al (2.3 mL or 4.6 mL of 2m solution in toluene),
corresponding amount of PO, 60 bar CO (indeed, even much lower pressures of CO (40, 20, or 5 bar) do not
influence the activity and product distribution of the catalytic reaction). [b] First values represent the total
amount of carbonylated epoxide, the values in brackets are the percentage of b-butyrolactone formed. Side
products are in general low amounts of polyhydroxybutyrate and some acetone.
4
]
4
4
3
2
3
4
2
2
8
3
Chem. Eur. J. 2003, 9, No. 6
¹ 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
0947-6539/03/0906-1277 $ 20.00+.50/0
1277

F. Molnar, G. A. Luinstra, B. Rieger, and M. Allmendinger
bonylcobaltate was generated
FULL PAPER
by combining Et AlCl and Na-
2
Co(CO)4. The addition of
Et AlCl in heptane to a color-
2
less solution of NaCo(CO) in
4
diglyme and propylene oxide
results in
a
gradual color
change to deep red. Transfer-
ring this mixture to an auto-
clave and treatment with car-
bon monoxide gives the IR
spectrum presented in Figure 7.
With the mixture of Et AlCl
2
and Na[Co(CO) ], no tetracar-
4
bonylcobaltate absorption is
detectable; only the character-
istic absorptions of the cobalt
acyl intermediate are found
Figure 5. Stack-plot of IR spectra for the carbonylation reaction with BF
3
¥ OEt
2
/PPN[Co(CO)
4
] (b-butyrolac-
À1
À1
tone nÄ ˆ 1829 cm , polyhydroxybutyrate nÄ ˆ 1744 cm
)
(
1
nÄ ˆ 2107, 2043, 2024, 2007, and
À1
715 cm ). This behavior is
consistent with rapid ring open-
ing of the epoxide and corre-
lates well with the calculated
high reactivity of model
‡
AlMe2 ¥ diglyme (cf. Table 2).
In due course, the propylene
oxide is easily converted to
butyrolactone within 4 h at
7
58C (Table 7, entry 6).
Our intention was to gener-
ate a highly active epoxide
carbonylation system by com-
bining [Co (CO) ] with two
2
8
equivalents of Me Al. Since
3
aluminum alkyls may act as
reducing agents, we anticipated
À
À1
that
Co(CO)4
a
catalyst based on
and the cation
Figure 6. IR spectrum of the carbonylation reaction with AlMe
3
after 40 min [Co(CO)
4
( nÄ ˆ 1887 cm ), b-
À1
À1
À
butyrolactone ( nÄ ˆ 1829 cm ), tetracarbonylcobaltate acyl ( nÄ ˆ 2107, 2043, 2024, 2005, and 1715 cm )]
This corresponds well with the
activation barriers for ring
opening given in Table 2: ring
opening with the model AlMe3
species has a higher activation
À1
barrier (23 kJmol ) than ring
opening
with
BF ¥ Et O
3 2
À1
(
16 kJmol ).
The next Lewis acid used was
‡
a diglyme adduct with Et Al
2
prepared in situ. According to
‡
our calculations, (alkyl) Al in
2
diglyme is most probably coor-
dinated by at least one diglyme
molecule, weakening the very
strong acidity of a ™free∫ cati-
onic aluminum center. The de-
2 4
Figure 7. IR spectrum of the carbonylation reaction with Et AlCl/Na[Co(CO) ] and CO/PO in the early stage
sired catalyst mixture of (al-
(
10 min); similar to that obtained for a reaction with [Co
2
(CO)
8
]/Me
3
Al as catalytic system [b-butyrolactone ( nÄ ˆ
‡
À1
À1
kyl) Al ¥ diglyme and tetracar-
1829 cm ), tetracarbonylcobaltate acyl ( nÄ ˆ 2107, 2043, 2024, 2007, and 1715 cm )]
2
1278
¹ 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
0947-6539/03/0906-1278 $ 20.00+.50/0
Chem. Eur. J. 2003, 9, No. 6

Multisite Catalysis
1273±1280
‡
Me Al ¥ diglyme would be formed in situ. Addition of Me Al
alents) we were able to optimize the performance of the
dinuclear catalytic system further. The rapid conversion of the
acyl species to lactones is explained by the results of the
2
3
in toluene to a solution of [Co (CO) ] in diglyme immediately
causes a significant change in the IR spectrum leading to two
main strong single absorptions (Figure 8). These correspond
2
8
‡
theoretical calculations (cf. Table 6). In solution the AlMe2
À1
to the tetracarbonylcobaltate ( nÄ ˆ 1887 cm ) and another
ion is coordinated by at least one diglyme molecule, which
helps to adjust its Lewis acidity. Therefore, tuning of the Lewis
acidic catalyst moiety is clearly one decisive factor in
controlling rapid lactone formation.
À1
carbonyl compound at nÄ ˆ 1980 cm which can be assigned to
‡
a tetracarbonylcobaltate coordinated to a Al(alkyl) species.
This assignment is based on theoretical calculations of
vibration frequencies (data not shown).
The addition of propylene oxide to this mixture followed by
treatment with carbon monoxide instantaneously yields an IR
Conclusion
spectrum that is identical to the one obtained with Et AlCl/
2
Na[Co(CO) ] (Figure 7). The b-butyrolactone product is
4
The present study emphasized the concept of multisite
catalysis by using the important example of lactone formation
from epoxides and CO. This reaction was shown to comprise a
Lewis acid controlled prepolarization of the epoxide ring with
formed rapidly and grows linearly over time (Figure 9).
Analysis of the reaction mixture after 4 ± 5 h by NMR
spectroscopy indicates quantitative consumption of the epox-
ide. Additionally, the carbonylation proceeds smoothly after
further addition of propylene oxide to the reaction mixture.
Side-products such as polyhydroxybutyrate are found in
amounts lower than 10% (Table 7, entries 7 ± 10). By fine-
tuning the reaction parameters (temperature, epoxide equiv-
À
a migratory CO insertion on the Co(CO)4 ion. The funda-
mental understanding of the basic reaction principles and the
design of an optimized Lewis acid fragment is essentially
guided by theoretical calculations that fit the portfolio of our
experimental findings with high precision.
We are convinced that a successful
combination of nucleophilic and elec-
trophilic (Lewis acid) sites will offer a
promising platform for a wide variation
of multisite catalysis effects. Similar
reaction principles already belong to
the established paradigms in enzymes
and heterogeneous transformations, but
are now on the verge of entering the
field of homogeneous catalysis.
Experimental Section
General methods: Dicobaltoctacarbonyl ([Co
CO) ]), Me Al (2m in toluene), Et ClAl (1m in
heptane), and BF ¥ OEt were obtained form
2
(-
8
3
2
3
2
Fluka or Aldrich and used without further
purification. Racemic propylene oxide (rac-
PO) was supplied by BASF AG (water
À1
Figure 8. IR spectrum of [Co
addition of 2 equivalents of Me
2
(CO)
8
] in diglyme before (black; nÄ ˆ 2072, 2040, and 1845 cm ) and after
À1
3
Al (red; nÄ ˆ 1887 and 1980 cm ).
<
15 ppm). Water-free diglyme was purchased
from Fluka and degassed before use. Na[Co(-
CO) ] was synthesized according to reference
22]. Et N[Co(CO) ] and PPN[Co(CO) ] were
both synthesized by the following one-pot
procedure under argon: [Co (CO) ] (4 g) was
4
[
4
4
4
2
8
added to water-free degassed THF (150 mL)
together with excess powdered NaOH (10 ±
2
0 equiv). After the solution had been stirred
for 20 ± 30 min, its color gradually changed from
deep red to slightly yellow. After addition of
PPNCl (10 g) or Et NCl (3.5 g), respectively,
4
stirring was continued for several hours. Filtra-
tion of the reaction mixture under argon gave a
slightly yellow solution. Removal of the THF in
vacuo afforded the crude white product. The
solid was extracted with degassed water and
isolated by filtration under argon. Finally, the
products were washed with cold degassed dieth-
yl ether and dried in vacuo. The desired salts
were obtained by this method in nearly quanti-
tative yields and were characterized by elemen-
tal analysis and IR.
Figure 9. IR stack-plot of the [Co
2
(CO)
8
]/Me
3
Al carbonylation reaction (b-butyrolactone n ˆ
À1
1
829 cm ); after 4 ± 5 h, addition of further 160 equivalents of PO (red spectrum).
Chem. Eur. J. 2003, 9, No. 6
¹ 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
0947-6539/03/0906-1279 $ 20.00+.50/0
1279

F. Molnar, G. A. Luinstra, B. Rieger, and M. Allmendinger
FULL PAPER
General carbonylation procedure: Reactions were conducted in a 250-mL
[4] a) G. G. Hlatky, Coord. Chem. Rev. 2000, 199, 235 ± 329; b) N. E.
Grimmer, N. J. Conville, S. Afr. J. Chem. 2001, 54, 1 ± 112; c) Y.
Imanishi, N. Naga, Progr. Polym. Sci. 2001, 26, 1147 ± 1198; d) S. D.
Ittel, L. K. Johnson, M. Brookhart, Chem. Rev. 2000, 100, 1169 ± 1204;
e) A. Sommazzi, F. Garbassi, Progr. Polym. Sci. 1997, 22, 1547 ± 1605.
[5] F. J. MacQuillin, Homogeneous Hydrogenation in Organic Chemistry,
Kluwer, Dordrecht, 1976.
[6] a) Rhodium Catalyzed Hydroformylation (Eds.: P. W. N. M. van Leeu-
wen, C. Claver), Kluwer, Dordrecht, 2000; b) A. M. Trzeciak, J. J.
Ziolkowski, Coord. Chem. Rev. 1999, 192, 883 ± 900.
TM
TM
B¸chi reactor equipped with a ReactIR
SiComp
probe (Mettler
Toledo) for in situ ATR-IR measurements under high-pressure conditions.
All manipulations were performed under an argon atmosphere. For
preparation of the catalyst mixture the appropriate amount of diglyme
was transferred to the autoclave. The system was cooled to 58C with a
cryostat and the catalyst compounds together with the appropriate amount
of propylene oxide were introduced. After pressurizing the reactor with
6
0 bar CO the carbonylation reactions were carried out at the desired
temperatures for periods of 2 ± 20 h. The reactor was cooled and ventilated
to terminate the carbonylation reaction. Yields were determined by online
IR and NMR analysis of the resulting solutions.
[7] M. Palucki, N. S. Finney, P. J. Pospisil, M. L. Guler, T. Ishida, E. N.
Jacobsen, J. Am. Chem. Soc. 1998, 120, 948 ± 954.
[
8] a) A. Nakamura, M. Tsutsui, Principles and Applications of Homoge-
neous Catalysis, Wiley, New York, 1980; b) G. W. Parshall, Homoge-
neous Catalysis, The Applications and Chemistry of Catalysis by
Soluble Transition Metal Complexes, Wiley, New York, 1980.
9] A. E. Shilov, G. B. Shul'pin, Activation and Catalytic Reactions of
Saturated Hydrocarbons in the Presence of Metal Complexes, Kluwer,
Dordrecht, 2000.
Theoretical methods: The overall copolymerization reaction involving CO
and epoxides was conceptually split into several steps. For each the step, the
corresponding reaction mechanism was investigated by locating the
transition state (TS) and the associated reactants and products. The nature
of all transition states was verified (only one negative eigenvalue of the
hessian). Reactants and products were identified by inducing small
distortions in the TS structure along the eigenvector associated with the
negative eigenvalue. Distortions with positive and negative amplitude lead
to reactants and products after subsequent geometry optimization. All
calculations were performed with the quantum chemistry package TUR-
[
[
10] H. J. Evans, P. J. Bottomley, W. E. Newton, Nitrogen Fixation Research
Progress. Proceedings of the 6th International Symposium on Nitro-
gen Fixation, Corvallis/OR, August 4 ± 10, 1985. Nijhoff, Dordrecht,
1985.
11] The hydrocarbonylation of ethylene oxide to 1,3-propanediol is
probably also catalyzed in a multisite fashion, although no mechanistic
description is offered. For Co and Rh see: E.-I. Drent (Shell Int. Res.
Maatschappij), EP 478,850. For Co and Ru see: K. D. Allen, J. F.
Knifton, J. B. Powell, L. H. Slaugh, G. J. Talmadge, P. R. Wieder, T. S.
Williams (Shell Oil Company), WO 01/72675 and J. P. Arhancet, L. H.
Slaugh (Shell Oil Company), US 5,304,691.
12] a) G. Braca, Oxygenates by Homologation or CO Hydrogenation with
Metal Complexes, Kluwer, Dordrecht, 1994; b) M. Bowker, The Basis
and Applications of Heterogeneous Catalysis, Oxford University Press,
Oxford, 1998.
13] S. B. Marr, R. O. Carvel, D. T. Richens, H. J. Lee, M. Lane, P.
Stavropoulos, Inorg. Chem. 2000, 39(20), 4630 ± 4638.
[
23]
[24]
BOMOLE. DFT methodology was used at the B-P86/SV(P) level of
theory to locate all stationary points. Single-point energy calculations were
[
carried out with the TZVP[ basis set. Geometries were optimized on a
25]
2
0 processor Pentium II Linux cluster and a 64 CPU IBM SP3. Calculation
of hessians (vibrational spectra) was achieved with a modified version of
TURBOMOLE×s NumForce utility. The numerical calculation of second
derivatives requires 6N (N ˆ number of atoms) energy and gradient
calculations, which can be efficiently distributed and carried out in parallel
on a network of workstations. This calculation is achieved by using
TURBO-SERVER, an in-house development of BASF polymer research,
by harnessing the power of ordinary NT desktop PCs to carry out quantum-
[
chemical calculations ™at night∫.[ The effect of solvent (diglyme) was in
general not taken into account. In the case of very (bi)polar structures,
explicit solvation by diglyme molecules was considered.
26]
[
[
14] a) X.-X. Zhang, B. B. Wayland, Inorg. Chem. 2000, 39, 5318 ± 5325;
b) X.-X. Zhang, G. F. Parks, B. B. Wayland, J. Am. Chem. Soc. 1997,
1
19, 7938 ± 7944.
[
[
15] A. Fujii, S. Hashiguchi, N. Uematsu, T. Ikariya, R. Noyori, J. Am.
Chem. Soc. 1996, 118, 2521 ± 2522, and references therein.
16] A. J. DelMonte, J. Haller, K. N. Houk, K. B. Sharpless, D. A. Single-
ton, T. Strassner, A. A. Thomas, J. Am. Chem. Soc. 1997, 119, 9907 ±
Acknowledgements
Financial support from BMBF (Grant 03C0310) is gratefully acknowl-
edged.
9
90.
[
17] M. Allmendinger, R. Eberhardt, G. A. Luinstra, B. Rieger, J. Am.
Chem. Soc. 2002, 124, 5646 ± 5647.
[
1] a) B. Cornils, W. A. Herrmann, Applied Homogeneous Catalysis with
Organometallic Compounds, 2nd ed., Wiley, Weinheim, 1999; b) K.
Kirchner, W. Weissensteiner, Organometallic Chemistry and Catalysis,
Springer, Wien, 2001.
[18] a) J. T. Lee, P. J. Thomas, H. Alper, J. Org. Chem. 2001, 66, 5424 ± 26;
b) Y. D. Y. L. Getzler, E. B. Mahadevan, V. Lobkosky, G. W. Coates, J.
Am. Chem. Soc. 2002, 124(7), 1174 ± 1175, following an older patent by
Drent at Shell Chemical Company: (Shell) EP 577,206.
[
2] a) E. G. Derouane, F. Lemos, Combinatorial Catalysis and High
Throughput Catalyst Design and Testing, Proceedings of the NATO
Advanced Study, Vilamoura, Portugal, July 11 ± 24, 1999. Kluwer,
Dordrecht, 2000 X, 512 (NATO Science Series 560); b) A. Hagemey-
er, B. Jandeleit, Y. M. Liu, D. M. Poojary, H. W. Turner, A. F. Volpe,
W. H. Weinberg, Appl. Catal. A 2001, 221, 23 ± 43; c) B. Jandeleit, D. J.
Schaefer, T. S. Powers, H. W. Turner, W. H. Weinberg, Angew. Chem.
[19] T. Aida, S. Inoue, Acc. Chem. Res. 1996, 29, 39 ± 48.
[20] a) O. Rexin, R. M¸lhaupt, J. Polym. Sci. Part A 2002, 40, 864 ± 873;
b) Y. D. Y. L. Getzler, E. B. Mahadevan, V. Lobkosky, G. W. Coates, J.
Am. Chem. Soc. 2002, 124(7), 1174 ± 1175; c) T. Yashiro, K. Matsush-
ima, A. Kameyama, T. Nishikubo, Macromolecules 2001, 34, 3205 ±
3210; d) B. Antelmann, M. H. Chisholm, S. S. Iyer, J. C. Huffman, D.
Navarro ± Llobet, M. Pagel, Macromolecules 2001, 34, 3159 ± 3175;
[21] a) R. F. Heck, D. S. Breslow, J. Am. Chem. Soc. 1961, 83, 4023 ± 4027;
b) J. Kreisz, F. Ungv a¬ ry, A. Sisak, L. Mark o¬ , J. Organometal. Chem.
1991, 417, 89 ± 97.
[22] W. F. Edgell, J. Lyford, Inorg. Chem. 1970, 9, 1932 ± 1933.
[23] R. Ahlrichs, M. B‰r, M. H‰ser, H. Horn, C. Kˆlmel, Chem. Phys. Lett.
1989, 162, 165 ± 169.
[24] a) A. D. Becke, Phys. Rev. A 1988, 38, 3098 ± 3100; b) J. Perdew, Phys.
Rev. B 1986, 33, 8822 ± 8824; c) S. H. Vosko, L. Wilk, M. Nusair, Can. J.
Phys. 1980, 58, 1200 ± 1211; d) A. Sch‰fer, H. Horn, R. Ahlrichs, J.
Chem. Phys. 1992, 97, 2571 ± 2577.
1
999, 111, 2648 ± 2689; Angew. Chem. Int. Ed. 1999, 38, 2495 ± 2532;
d) I. E. Maxwell, Nature 1998, 394, 325 ± 326; e) S. Dahmen, S. Brase,
Synthesis 2001, 10, 1431 ± 1449; f) J. A. Loch, R. H. Crabtree, Pure
Appl. Chem. 2001, 73, 119 ± 128; g) B. Jandeleit, W. H. Weinberg,
Chem. Ind. 1998, 795 ± 798; h) M. B. Francis, T. F. Jamison, E. N.
Jacobsen, Curr. Opin. Chem. Biol. 1998, 2, 422 ± 428; i) C. Hoffmann,
A. Wolf, F. Sch¸th, Angew. Chem. 1999, 111, 2971 ± 2975; Angew.
Chem. Int. Ed. 1999, 38, 2800 ± 2803; j) F. Gennari, P. Seneci, S.
Miertus, Catal. Rev. Sci. Eng. 2000, 42, 385 ± 402.
[
3] a) K. J. Ivin, J. C. Mol, Olefin Metathesis and Metathesis Polymer-
ization, Academic Press, San Diego, 1997; b) A. Furstner, L.
Ackermann, B. Gabor, R. Goddard, C. W. Lehmann, R. Mynott, F.
Stelzer, O. R. Thiel, Chem. Eur. J. 2001, 7, 3236 ± 325; c) A. H.
Hoveyda, R. R. Schrock, Chem. Eur. J. 2001, 7, 945 ± 950; d) R. R.
Schrock, J. Chem. Soc. Dalton Trans. 2001, 2541 ± 2550.
[25] A. Sch‰fer, C. Huber, R. Ahlrichs, J. Chem. Phys. 1994, 100, 5829 ±
5835.
[26] H. Wei˚, H. Domgˆrgen, M. Grimm, S. Steiger, c't 2000, 22, 246 ± 251.
Received: August 21, 2002 [F4363]
1280
¹ 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
0947-6539/03/0906-1280 $ 20.00+.50/0
Chem. Eur. J. 2003, 9, No. 6