Quantum mechanical modelling of alkene hydroformylation as catalyzed by xantphos-Rh complexes

Article abstract of DOI:10.1039/b108649a

Fundamental issues concerning the hydroformylation of 1-alkenes as catalyzed by Rh complexes ligated with the xantphos diphosphine ligand are explored using ONIOM calculations. In this study xantphos serves as a prototype of the large bite-angle ligands that are associated with high regioselectivity and rates in catalytic hydroformylation. Computations have been used to explore the thermodynamics of 56 unique isomers (e.g., cis vs. trans isomers of square planar complexes, diequatorial vs. axial-equatorial five-coordinate complexes) and conformers for intermediates along the reaction pathway. More than 20 transition states relevant to the catalyst mechanism have been determined. In terms of realistically modelling experiment, the computational results are mixed. In agreement with experiment, the computations yield a mixture of diequatorial and axial-equatorial isomers of HRh(xantphos)(CO)₂ as the catalyst resting state Dissociation of CO from these complexes is computed to be barrierless leading to a computed free energy for exchange of CO ligands around 15 kcal mol^-1, somewhat lower than the value of ca. 20 kcal mol^-1 derived from experimental data. The computed ratios of rates of propene insertion to form n-propyl and i-propyl Rh-alkyl (42 : 1) is in good agreement with experimental ratios of n-nonanal to i-nonanal (52:1) for 1-octene hydroformylation. Nonetheless, the computations dramatically overestimate the overall activation free energies for catalytic hydroformylation. Thus, at this stage computations do not provide useful insight into the the kinetics of hydroformylation and detailed mechanistic issuea It appears that much of this discrepancy between computed and experimental activation energies originates from the underestimation of propene bonding energies. The Royal Society of Chemistry 2002.

Full text of DOI:10.1039/b108649a

View Online / Journal Homepage / Table of Contents for this issue
Quantum mechanical modelling of alkene hydroformylation as
catalyzed by xantphos-Rh complexes†
Clark R. Landis* and Jamal Uddin
Department of Chemistry, University of Wisconsin, 1101 University Avenue, Madison, WI,
USA. E-mail: landis@chem.wisc.edu
Received 25th September 2001, Accepted 20th November 2001
First published as an Advance Article on the web 29th January 2002
Fundamental issues concerning the hydroformylation of 1-alkenes as catalyzed by Rh complexes ligated with the
xantphos diphosphine ligand are explored using ONIOM calculations. In this study xantphos serves as a prototype
of the large bite-angle ligands that are associated with high regioselectivity and rates in catalytic hydroformylation.
Computations have been used to explore the thermodynamics of 56 unique isomers (e.g., cis vs. trans isomers of
square planar complexes, diequatorial vs. axial-equatorial five-coordinate complexes) and conformers for
intermediates along the reaction pathway. More than 20 transition states relevant to the catalyst mechanism
have been determined. In terms of realistically modelling experiment, the computational results are mixed. In
agreement with experiment, the computations yield a mixture of diequatorial and axial-equatorial isomers of
HRh(xantphos)(CO)2 as the catalyst resting state. Dissociation of CO from these complexes is computed to be
barrierless leading to a computed free energy for exchange of CO ligands around 15 kcal molϪ1, somewhat lower
than the value of ca. 20 kcal molϪ1 derived from experimental data. The computed ratios of rates of propene
insertion to form n-propyl and i-propyl Rh-alkyl (42 : 1) is in good agreement with experimental ratios of n-nonanal
to i-nonanal (52 : 1) for 1-octene hydroformylation. Nonetheless, the computations dramatically overestimate the
overall activation free energies for catalytic hydroformylation. Thus, at this stage computations do not provide
useful insight into the the kinetics of hydroformylation and detailed mechanistic issues. It appears that much of this
discrepancy between computed and experimental activation energies originates from the underestimation of propene
bonding energies.
Introduction
Catalytic hydroformylation of simple alkenes, such as propene,
in the presence of Rh catalysts is one of the largest volume
applications of homogeneous, precious metal catalysts. Over
the last twenty years, continued development of new phosphine
speciation of catalyst among various isomers and complexes
that lie within vs. outside of the catalytic cycle, and the molecu-
larity (with respect to catalyst) of the aldehyde-producing
and phosphite ligands has driven significant advances in hydro-
formylation chemistry, especially with respect to catalyst select-
ivity and stability.1 A particularly useful development in the
step, are unclear at this point.1 Further compounding these
control of catalyst regioselectivity concerns the application of
bulky diphosphines and diphosphites with wide bite angles2
uncertainties is a paucity of reported kinetic data for the
catalytic process, particularly with the more effective large bite
angle bisphosphine and bisphosphite ligands.
(Scheme 1): the larger the bite angle the greater the selectivity
Many researchers have turned to computations in order to
fill the voids in our understanding of catalytic hydroformyl-
ation.7–12 The work of Morokuma and coworkers9 focussed on
complete quantum mechanical characterization of the hydro-
formylation of ethene in the presence of PH3-ligated Rh. Very
recently similar systems have been investigated by Decker and
Cundari10 using higher levels of theory (B3LYP/B3LYP and
CCSD(T)/B3LYP). The first detailed computational analyses
of bisphosphine ligands were reported by Gleich and co-
workers.11 These computations utilized a loosely coupled QM/
MM approach. More recently, and of most relevance to this
work, Carbó et al.12 have employed tightly coupled QM/MM
computational methods to examine the hydroformylation of
ethene and propene using benzoxantphos and homoxanthpos
ligands. These investigations provide the most detailed inform-
ation to date on a realistic model catalyst system and the origins
of regioselectivity.
Scheme 1 Some large bite-angle diphosphines.
for linear aldehydes. New chiral diphosphines,3 diphosphites,4
and mixed systems5 for enantioselective hydroformylations
constitute another important, growing area of development.
The understanding of fundamental aspects of catalytic
hydroformylation trails well behind the discovery of new and
more effective catalysts.1,6 Basic issues, such as the identities of
the regioselectivity-determining and turnover-limiting steps, the
Herein we report our initial findings of a large-scale
computational project aimed at modelling propene hydro-
formylation as catalysed by Rh complexes containing the
† Based on the presentation given at Dalton Discussion No. 4, 10–13th
January 2002, Kloster Banz, Germany.
DOI: 10.1039/b108649a
J. Chem. Soc., Dalton Trans., 2002, 729–742
This journal is © The Royal Society of Chemistry 2002
729

View Online
xantphos13,14 ligand. One goal of this project is simply to
explore whether computing a representative collection of
intermediates, transition states, and their conformers for a
realistic model of a multi-step catalytic process is reasonable. In
this context, one criterion for feasability is the use of modern,
easily accessible computational methods and hardware. A sec-
ond goal is to establish the accuracy of the computational
models via comparison with pertinent empirical data. Catalysts
based on the xantphos ligand were chosen for this initial study
because these catalysts exhibit high selectivity for linear alde-
hydes, good activities, and a relatively rich set of empirical
data in the open literature. The ultimate goal of this project is
the complete characterization of the free energy surface for
hydroformylation via computations. From such studies we hope
to sharpen our understanding of fundamental mechan-
istic issues. By coupling these computational models with a
nascent experimental program exploring the kinetic and
spectroscopic characterization of the catalytic process, we
hope to develop enough constraints so as to resolve some of the
basic mechanistic uncertainties mentioned above. This first
report focuses on the structures and energetics of conform-
ational and geometrical isomers of HRh(xantphos)(CO)2 (1),
HRh(xantphos)(CO) (2), HRh(xantphos)(CO)(propene) (3),
(propyl)Rh(xantphos)(CO) (4), (propyl)Rh(xantphos)(CO)2
(5), and (butanoyl)Rh(xantphos)(CO) (6) complexes and the
transition states connecting these complexes and their various
isomers.
phosphorus.30 All other atoms used the 6-31G basis set.31 All
reported intermediates are true minima, as determined by the
absence of any negative eigenvalue in the Hessian matrix in the
vibrational frequency analysis of the potential energy surface.
The located transition-state structures have a single negative
eigenvalue (imaginary frequency) on the Hessian matrix as
determined by using the synchronous-guided quasi-Newton
(QST2 and QST3) methods as incorporated in Gaussian 98. All
computations were performed on a 64-cpu cluster of Pentium
III processors running under the LINUX operating system.
Results
A general working mechanism for the hydroformylation of
propene as catalyzed by a Rh complex of xantphos is provided
in Fig. 1. The catalytic cycle comprises two main branches: one
for the production of the linear product, butanal, and one for
the branched product, 2-methylpropanal. The species 1, 2, and
3 are common to both branches, with different isomers of 4
being produced at the branching point. A fundamental ques-
tion, that will be only partly addressed in this paper, concerns
whether the regioselectivity of propene hydroformylation is
fixed upon alkene association (2
(3 4), or at the subsequent CO insertion step (4
Although our working mechanism implies that the transform-
ation 3 4 is irreversible and, hence, fixes the regioselectivity
3), at the branching point
5).
we emphasize that this is a simple convenience to be tested
through computation, and ultimately, experiment.
Computational methods
Structures and thermodynamics of key intermediates
All calculations were performed using the two-layered
ONIOM15–18 method as implemented in the Gaussian 98 pro-
gram.19 The calculations have been done wholly at the ab initio
quantum mechanical level without the support of a molecular
mechanics (MM) method. As a result, our calculations can be
denoted as (High : Medium), where High is the level of theory
on the core layer and Medium is the level on the whole system
in generally accepted ONIOM formalism. All the species
involved in the bond formation and bond breaking during the
catalytic cycle of hydroformylation are designated as “core-
layer” including two monodentate phosphine (PH3) ligands as
replacements for the bidentate xantphos ligand where the
linked carbon atoms of xantphos are substituted with hydro-
gens. The idea is to reduce the computational costs while
retaining the fundamental properties of the chelating xantphos
ligand which is believed to control the regio- and stereo-
selectivity of the catalytic hydroformylation process.
All intermediates and the transition states have fully been
optimized at (B3LYP/LANL2DZ : HF/LANL2MB) levels of
theory where density functional (DFT),20,21 with Becke’s three-
parameter functional (B3)22 and Lee, Yang, and Parr (LYP)23
correlation functional and Hartree–Fock theories have been
utilized. For the core-layer, Hay and Wadt’s LANL2DZ basis
sets have been used with B3LYP and for the outer-layer, which
includes the whole system, medium-level LANL2MB basis sets
have been used and treated with HF theory. The LANL2DZ
has a double-ζ quality basis set for the valence and penultimate
shells, with effective core potentials at rhodium24 and phos-
phorus,25 and a Dunning/Huzinaga full double-ζ basis26 on C,
H, N and O. The outer-layer, which includes the whole system,
has been treated with HF theory incorporation with LAN-
L2MB basis sets which contain the same effective core poten-
tials with a minimal basis set on the valence and penultimate
shells of rhodium and phosphorus, plus minimal STO-3G basis
sets27 on C, H, and O. Computations have also been performed
with the ‘Stuttgart’ basis sets and ECPs. The ECPs replaced 28
core electrons on rhodium and 10 on phosphorus. The Stuttgart
valence basis sets have the following contraction schemes:
(311111/22111/411) for rhodium28 and (31/31) for phos-
phorus.29 One d function with an exponent of 0.34 was added to
An aspect commonly bedevilling computations involving
realistic models concerns the existence of many isomers and
conformers. Although the xantphos ligand appears rather
simple and rigid, we find a variety of conformational and
geometric isomers as illustrated in Fig. 2. Three conformer
classes of the xantphos ligand are populated; in the isomer with
approximate C2 symmetry the best-fit plane of the xantphos
ligand lies nearly parallel with the P–Rh–P plane of the
coordinated ligand. Two conformers with approximate Cs
symmetry are found; they differ in the sense by which the non-
planar xanthene array of fused cycles ‘domes’ or bends about
the middle ring. One direction of bending places the O atom of
the xanthene ring closer (proximal) to the Rh than the other
(distal) direction.
Geometric isomers, also, abound. The xantphos ligand has a
sufficiently large bite angle that trans arrangements in four
coordinate complexes (e.g., 2 and 4) cannot be precluded.
Although trans complexation at five-coordinate, trigonal
bipyramidal complexes (e.g., 1 and 3) engenders too much steric
crowding to persist, both the diequatorial (ee) and axial-
equatorial (ae) dispositions of the xantphos P atoms give rise to
low energy, true minima on the energy surface. Even further
elaboration of isomers in the trigonal bipyramidal structures
can be achieved by different positioning of, for example, the
hydride, CO, and propene ligands of 3
Descriptions of the structures of complexes 1–6 are pre-
sented below. Relative energies, enthalpies, entropies, and free
energies of stable minima at 298 K as computed in this study
are provided in Table 1. Our reference point for all relative
energies is the most stable HRh(xantphos)(CO) isomer, 2a. In
the interest of brevity, we focus on the qualitative descriptions
of the structures and relative energies for these complexes. Fine
structural details of various isomers are tabulated but not
discussed.
HRh(xantphos)(CO)2 complexes (1). In situ spectroscopic
studies of hydroformylation as catalyzed by bisphosphines
similar to xantphos invariably reveal isomers of 1 as the only
detectable species in solution.14 We have located six minima
730
J. Chem. Soc., Dalton Trans., 2002, 729–742

View Online
Fig. 1 A general mechanism for catalytic hydroformylation.
corresponding to isomers of 1 (see Fig. 3 and Table 2). Three of
HRh(xantphos)(CO) complexes (2). Loss of CO from 1 yields
the isomers (1a, 1b, and 1d) correspond to axial-equatorial
placements of the P atoms and the remaining isomers (1c, 1e,
1f ) have diequatorial dispositions. All isomers have the H
ligand in an axial position. Of the axial-equatorial isomers, the
xantphos fragment of 1d demonstrates a pseudo-C2 symmetry
whereas 1a and 1b exhibit pseudo-Cs symmetry of the xantphos
fragment. For the diequatorial isomers, the xantphos fragment
displays pseudo-Cs symmetry. The primary difference between
1b and 1e concerns the proximity of the H to the xanthene
rings.
At this point in our search we find that the most stable isomer
of five-coordinate HRh(xantphos)(CO)2, 1a, has an axial-
equatorial disposition of P atoms. In terms of estimated free
energies at 298 K, the most stable axial-equatorial isomer is
about 1.3 kcal molϪ1 more stable than the diequatorial isomer
of lowest energy, 1c. In contrast, experimental measurements of
1JRh–H coupling constants at 298 K in benzene suggest that the
diequatorial isomers predominate by a ratio of 7 : 3 ee : ae.14
Given the uncertainties presented by both experiment and
computation, perhaps the main conclusion to be drawn at this
point is that diequatorial and axial-equatorial isomers of 1
are close in energy. From the computational side, a clear
issue concerns finding all relevant conformers. For example,
we have not yet located a diequatorial isomer of 1 that has
a pseudo-C2 symmetric conformation of the xantphos frag-
ment. Given that 1a has this conformation, the small dis-
crepancy with experiment may reflect the absence of the most
representative conformer of the diequatorial isomer from our
current tabulation.
the intermediate 2. Of the six isomers of 2 that we have found
(see Fig. 4 and Table 3), the three cis isomers (2a–2c) are lower in
energy than the three trans isomers (2d–2f ) by several kcal
molϪ1. Despite the rather small ideal bite angle for the xantphos
ligand (111Њ), trans configurations of the ligand have low
enough energies that they could play a role in catalysis. The
various cis isomers differ in the conformation of the xantphos
fragment but are nearly isoenergetic.
The formation of 2 from 1 is thermodynamically uphill by
ca. 15 kcal molϪ1 in free energy under standard conditions.
Correspondingly, one cannot expect 2 to be spectroscopically
observable under catalytically relevant conditions. From com-
putation, the bond dissociation enthalpy for loss of the first CO
from 1 is estimated to be on the order of 26 kcal molϪ1. These
values are a few kcal molϪ1 higher than those found by others
using PH3 model complexes.8–10
HRh(xantphos)(CO)(propene) complexes (3). Addition of
propene to 2 generates a large array of possible isomers and
conformers of complex 3. We have located a total of 16 minima
for 3 (see Table 4 and Fig. 5). All but one of these minima, the
structure 3o, have the propene located cis with respect to the
Rh–H bond. As a result, there are at least 15 minima that could
be direct precursors to the crucial migratory insertion step in
which the first C–H bond is formed. All structures of 3 conform
well to trigonal bipyramidal geometry with the H always
occupying an axial site. With the exception of 3o, the propene is
always coordinated such that the C᎐C bond vector lies in the
᎐
equatorial plane.9,10
J. Chem. Soc., Dalton Trans., 2002, 729–742
731

View Online
Fig. 2 Classes of isomers and conformers of intermediates.
Multiple examples of both axial-equatorial (3a–3h) and die-
methyl group of coordinated propene most remote to the
quatorial (3i–3p) coordination modes are found. This complex-
ity arises from the possibility of four different orientations of
the propene ligand with respect to the Rh–H axis and from
different conformations of the xantphos fragment. The most
stable isomer of 3 with diequatorial disposition of the xantphos
chelate, 3m, is virtually isoenergetic with the axial-equatorial
isomer. Thus the xantphos ligand is not computed to strongly
favour either chelate mode for isomers of 1 or 3.
sterically bulky groups. Rotation of propene to place the methyl
group in the other three positions results in higher energies by
about 1 kcal molϪ1 increments in the order 3c, 3d, and 3e.
The lone isomer with propene in an axial position, 3o, cannot
undergo direct insertion into the Rh–H bond without prior
isomerization. The relatively high energy of 3o indicates that
significant pooling of catalyst into this species will not occur.
Binding propene to 2 to generate various isomers of 3 is
weakly exothermic, at best. At the level of theory which we have
employed, the most exothermic reaction releases just 0.9 kcal
Among the axial-equatorial isomers of 3, the preferred orien-
tation of the propene ligand (see 3a) is that which places the
732
J. Chem. Soc., Dalton Trans., 2002, 729–742

View Online
Table 1 Thermodynamic valuesa for intermediates along the reaction pathway
∆HЊ/kcal molϪ1
∆GЊ/kcal molϪ1
∆SЊ/cal molϪ1 KϪ1
∆E/kcal molϪ1
1a
1b
1c
1d
1e
1f
Ϫ26.15
Ϫ22.63
Ϫ25.07
—
Ϫ24.55
—
Ϫ15.09
Ϫ11.75
Ϫ13.72
—
Ϫ13.39
—
Ϫ37.09
Ϫ36.48
Ϫ38.07
—
Ϫ37.42
—
Ϫ27.69
Ϫ24.04
Ϫ26.65
Ϫ24.72
Ϫ26.19
Ϫ22.03
2a
2b
2c
2d
2e
2f
0.00
0.55
1.77
2.07
7.32
5.16
0.00
0.51
2.13
3.41
8.02
6.75
0.00
0.14
Ϫ1.20
Ϫ4.47
Ϫ2.36
Ϫ5.33
0.00
0.59
1.71
2.03
7.27
5.19
3a
3b
3c
3d
3e
3f
3g
3h
3i
3j
3k
3l
3m
3n
3o
3p
Ϫ0.91
—
—
1.42
2.13
2.92
—
2.50
—
0.15
1.54
0.32
Ϫ0.59
0.64
5.90
0.18
13.82
—
—
16.49
17.39
18.19
—
16.94
—
15.11
16.49
14.97
13.29
14.76
19.74
14.83
Ϫ49.40
—
—
Ϫ50.54
Ϫ51.17
Ϫ51.19
—
Ϫ48.44
—
Ϫ50.20
Ϫ50.13
Ϫ49.14
Ϫ46.55
Ϫ47.34
Ϫ46.43
Ϫ49.13
Ϫ2.56
Ϫ0.72
Ϫ1.53
Ϫ0.30
0.40
1.14
2.86
0.92
Ϫ0.40
Ϫ1.57
Ϫ0.33
Ϫ1.35
Ϫ2.31
Ϫ1.22
4.17
Ϫ1.49
4a
4b
4c
4d
4e
4f
4g
4h
4i
Ϫ4.19
Ϫ5.04
Ϫ6.17
Ϫ2.43
0.55
Ϫ4.14
Ϫ1.63
Ϫ2.54
Ϫ3.96
4.18
8.23
7.76
6.48
11.82
13.28
9.02
12.23
10.37
9.43
Ϫ41.70
Ϫ42.97
Ϫ42.45
Ϫ47.80
Ϫ42.69
Ϫ44.14
Ϫ46.52
Ϫ43.32
Ϫ44.93
Ϫ48.58
Ϫ50.09
Ϫ50.07
Ϫ7.86
Ϫ8.76
Ϫ9.85
Ϫ6.19
Ϫ3.05
Ϫ7.89
Ϫ5.29
Ϫ6.11
Ϫ7.45
0.57
4j
4k
4l
18.66
15.11
18.59
0.17
3.66
Ϫ3.34
0.26
5a
5b
5c
5d
5e
5f
5g
5h
5i
Ϫ30.13
Ϫ30.65
Ϫ24.61
Ϫ28.05
Ϫ28.46
Ϫ30.94
Ϫ28.30
Ϫ27.52
Ϫ28.12
Ϫ5.18
Ϫ5.69
2.08
Ϫ4.11
Ϫ4.10
Ϫ7.44
Ϫ4.16
Ϫ2.22
Ϫ1.94
Ϫ83.69
Ϫ83.73
Ϫ89.54
Ϫ80.28
Ϫ81.69
Ϫ78.81
Ϫ80.96
Ϫ84.85
Ϫ87.82
Ϫ35.49
Ϫ35.97
Ϫ29.46
Ϫ33.49
Ϫ33.83
Ϫ36.35
Ϫ33.54
Ϫ32.67
Ϫ33.32
6a
6b
6c
6d
6e
6f
Ϫ37.79
Ϫ38.55
Ϫ38.09
Ϫ39.22
Ϫ38.61
Ϫ38.69
Ϫ14.61
Ϫ16.05
Ϫ13.93
Ϫ14.04
Ϫ15.42
Ϫ15.36
Ϫ77.75
Ϫ75.47
Ϫ81.06
Ϫ84.42
Ϫ77.77
Ϫ78.26
Ϫ44.21
Ϫ45.04
Ϫ44.63
Ϫ45.52
Ϫ45.02
Ϫ44.93
a All energies are relative to 2a at 298 K.
Table 2 Geometric parameters computed for 1
Rh–H/Å
Rh–C1/Å
Rh–C2/Å
Rh–P1/Å
Rh–P2/Å
Bite angle/Њ
1a
1b
1c
1d
1e
1f
1.586
1.587
1.609
1.588
1.613
1.609
1.905
1.898
1.935
1.897
1.936
1.941
1.906
1.912
1.892
1.914
1.891
1.877
2.516
2.549
2.482
2.532
2.485
2.533
2.555
2.605
2.482
2.573
2.485
2.532
103.2
103.6
105.3
108.6
107.0
105.3
J. Chem. Soc., Dalton Trans., 2002, 729–742
733

View Online
Fig. 3 Structures computed for isomers and conformers of 1.
Fig. 4 Structures computed for isomers and conformers of 2.
molϪ1. Given the unfavourable entropy of binding propene, the
interactions that would disfavour coordination of alkene.
overall binding constant of propene to 2 is quite unfavourable.
Hence, spectroscopically observable quantities of 3 should
never accumulate. For comparison, the recent work of Decker
and Cundari10 reported exothermic (11–12 kcal molϪ1) binding
of ethene to analogous model complexes using PH3 as the
phosphine ligand at the B3LYP level. However, and perhaps
significantly for this work, the reaction was computed to be
even more strongly exothermic (ca. 22 kcal molϪ1) at the
CCSD(T)/B3LYP level. It is not surprising that the more real-
istic xantphos ligand model would create additional steric
However, the discrepancy of nearly 20 kcal molϪ1 between our
computations and those of Decker and Cundari raises
serious concerns that the B3LYP model may be seriously
underestimating the strength of the Rh–propene interaction.
(Propyl)Rh(xantphos)(CO) complexes (4). Migratory inser-
tion of the alkene into the Rh–H bond transforms 3 into 4. This
process lends an additional degree of complexity to the overall
analysis because both i-propyl (4h–l) and n-propyl (4a–j) groups
can be produced. These structures are depicted in Fig. 6 and the
734
J. Chem. Soc., Dalton Trans., 2002, 729–742

View Online
geometrical parameters are summarized in Table 5. Due to
the four coordinate nature of 4, the isomers may be further
broken down into cis (4a,b,h–j) and trans (4c–g,k,l) models of
diphosphine coordination.
range of 25 kcal molϪ1 are computed for formation of both the
n-propyl (5a–f ) and i-propyl isomers (5g–i) from 4 (see Table 6
and Fig. 7). Many of the energetic and structural features seen
for the two previous five coordinate complexes, 1 and 3,
reappear in isomers of 5. Namely, both diequatorial and axial-
equatorial coordination modes are seen and are computed to
have similar energies (for the i-propyl complexes the axial-
equatorial mode has not yet been thoroughly searched). The
n-propyl isomers are slightly more stable than the i-propyl
isomers by the increment (ca. 3 kcal molϪ1) seen for 4.
Thermodynamically, the transformation 3
4 is slightly
favourable (of the order of 1–6 kcal molϪ1), both in terms of
enthalpy and free energy. As points of comparison, Decker
and Cundari10 report a 1–3 kcal molϪ1 energy lowering upon
insertion of ethene into a Rh–H bond and Morokuma and
coworkers9 report a potential energy increase of 4.2 kcal molϪ1
.
The most stable isomer, 4c, is the n-propyl with trans xantphos
coordination and the lowest energy i-propyl complex, 4i, has a
cis-disposition of the xantphos ligand. Formation of the
n-propyl isomers is more favourable than formation of
the i-propyl isomers by an increment of ca. 3 kcal molϪ1. How-
ever all isomers of 4 demonstrate strong distortions from ideal
square planar coordination. Presumably this distortion reflects
the inability of the xantphos ligand to achieve either 90 or 180Њ
bite angles.
(Butanoyl)Rh(xantphos)(CO) complexes (6). Like the
migratory insertion of the C᎐C double bond into a Rh–H
᎐
bond, the computed insertion of CO into the Rh–propyl bond
of 5 to yield butanoyl isomers (6) is mildly exothermic (ca. 7
kcal molϪ1) with similar values for the free energy. Values
obtained by Decker and Cundari10 for the analogous PH3
model system average to about 9 kcal molϪ1. Six true minima of
6 (see Table 7 and Fig. 8) have been optimised and analyzed
thermodynamically. Four of the minima are n-butanoyl isomers
(6a–d) and two are the i-butanoyl isomers (6e,f ). Although the
sampling of the i-butanoyl isomers is rather limited at this
point, it appears that the thermodynamic bias in favour of the
n-propyl isomers computed for 5 largely disappears upon con-
version to 6. Presumably this reflects release of steric repulsions
as the alkyl group is placed farther from the crowded metal
center.
(Propyl)Rh(xantphos)(CO)2complexes (5). Based on the exo-
thermicity of the transformation 2
1, one expects that the
addition of CO to 4 should produce 5 with the release of
significant heat. Such is the case as reaction enthalpies in the
Table 3 Computed geometric parameters for 2
At this point we have computed 28 additional true minima
corresponding to isomers of 7 and 8. However, discussion of
these intermediates and their free energies lies beyond the scope
of this presentation.
Rh–H/Å
Rh–C/Å
Rh–P1/Å
Rh–P2/Å
Bite angle/Њ
2a
2b
2c
2d
2e
2f
1.594
1.591
1.589
1.651
1.618
1.655
1.839
1.835
1.835
1.894
1.907
1.887
2.514
2.508
2.503
2.347
2.356
2.349
2.497
2.519
2.539
2.347
2.356
2.349
106.3
104.8
108.7
146.1
138.1
144.1
Transition states and free energies of activation
Having established the relative thermodynamics of various
potential reaction intermediates and their isomers and con-
Table 4 Geometric parameters computed for 3
Isomer
Rh–H/Å
Rh–CO/Å
Rh–C1/Å
Rh–C2/Å
Rh–P1/Å
Rh–P2/Å
C᎐C/Å
᎐
Bite angle/Њ
3a
3b
3c
3d
3e
3f
3g
3h
3i
3j
3k
3l
3m
3n
3o
3p
1.569
1.565
1.568
1.568
1.566
1.569
1.565
1.569
1.605
1.615
1.603
1.608
1.601
1.611
1.575
1.617
1.891
1.888
1.896
1.888
1.897
1.893
1.892
1.889
1.934
1.926
1.929
1.927
1.926
1.926
1.863
1.924
2.216
2.232
2.209
2.222
2.196
2.208
2.213
2.232
2.206
2.216
2.204
2.204
2.207
2.207
2.373
2.211
2.229
2.226
2.239
2.258
2.259
2.264
2.215
2.235
2.245
2.242
2.236
2.245
2.238
2.234
2.434
2.236
2.518
2.526
2.553
2.265
2.582
2.588
2.637
2.600
2.469
2.449
2.476
2.464
2.458
2.467
2.514
2.461
2.549
2.579
2.514
2.535
2.518
2.517
2.569
2.555
2.459
2.469
2.466
2.484
2.464
2.485
2.500
2.463
1.428
1.426
1.427
1.424
1.427
1.424
1.428
1.422
1.427
1.425
1.429
1.427
1.428
1.428
1.389
1.427
102.0
105.7
101.7
101.0
103.9
106.2
104.2
106.7
119.0
117.9
105.7
105.7
107.5
106.7
104.3
111.95
Table 5 Geometric parameters computed for isomers of 4
Rh–C(R)/Å
Rh–CO/Å
Rh–P1/Å
Rh–P2/Å
Bite angle/Њ
4a
4b
4c
4d
4e
4f
4g
4h
4i
2.121
2.131
2.119
2.181
2.133
2.154
2.138
2.145
2.141
2.187
2.166
2.154
1.836
1.831
1.830
1.878
1.895
1.893
1.898
1.837
1.833
1.884
1.895
1.898
2.538
2.571
2.576
2.374
2.394
2.385
2.373
2.549
2.562
2.384
2.403
2.382
2.554
2.542
2.491
2.374
2.412
2.388
2.387
2.563
2.530
2.409
2.387
2.414
103.1
102.9
104.2
144.9
144.5
135.2
134.1
102.8
102.2
142.3
133.3
139.1
4j
4k
4l
J. Chem. Soc., Dalton Trans., 2002, 729–742
735

View Online
Fig. 5 Structures computed for 3.
formers, we turn to the barriers associated within isomers and
by computing the transition states that connect the
the barriers for reaction to form subsequent intermediates
(Table 8).
proximal conformation of 1a with the distal conformation of
1d. As expected the conformational exchange has a small free
energy of activation, 4.1 kcal molϪ1 (relative to 1a in Table 1),
and a low frequency imaginary mode of Ϫ34cmϪ1. Axial-
Isomerization and interchange of complexes 1 and 2
equatorial exchange, 1a
1c is computed to occur with a
barrier in free energy of 12.5 kcal molϪ1, which is just slightly
higher than the value of 10.5 kcal molϪ1 reported by
It is generally expected that conformational equilibria will have
low barriers to exchange. We have tested this expectation
736
J. Chem. Soc., Dalton Trans., 2002, 729–742

View Online
Table 6 Geometric parameters computed for isomers of 5
Rh–CPr/Å
Rh–CCO1/Å
Rh–CCO2/Å
Rh–P1/Å
Rh–P2/Å
Bite angle/Њ
5a
5b
5c
5d
5e
5f
5g
5h
5i
2.176
2.179
2.196
2.187
2.177
2.180
2.192
2.197
2.202
1.926
1.926
1.925
1.906
1.903
1.903
1.912
1.989
1.928
1.885
1.888
1.871
1.902
1.906
1.909
1.894
1.879
1.883
2.547
2.527
2.693
2.635
2.617
2.642
2.672
2.581
2.547
2.548
2.536
2.549
2.544
2.526
2.546
2.558
2.581
2.547
103.6
106.4
99.7
101.0
102.2
99.9
101.0
101.8
104.5
Fig. 6 Computed structures for isomers of 4.
Castellanos-Páez et al.32 for axial-equatorial exchange in the
complex HRh(BDPP)(CO)2 (BDPP = (2S,4S)-bis(diphenyl-
phosphino)pentane).
Isomers of 2 undergo exchange and conformational transi-
tions with little energetic barrier. For example, the transition
state which connects the cis isomer 2a with the trans isomer 2e
J. Chem. Soc., Dalton Trans., 2002, 729–742
737

View Online
Table 7 Geometric parameters computed for isomers and conformers of 6
Structure
Rh–H/Å
Rh–C/Å
Rh–P1/Å
Rh–P2/Å
Bite angle/Њ
6a
6b
6c
6d
6e
6f
2.011
1.992
2.015
2.055
1.995
2.069
1.844
1.849
1.844
1.915
1.848
1.915
2.627
2.583
2.592
2.381
2.595
2.387
2.558
2.550
2.569
2.382
2.557
2.390
103
104
103
132
104
132
Fig. 7 Computed structures for 5.
lies just 5.5 kcal molϪ1 higher in free energy than 2a. Similarly,
the proximal distal conformational conversion which
converts 2a into 2b traverses a free energy barrier of just 2.3
kcal molϪ1
exchange measurements in solution at 40 ЊC. This important
reference point is one of the few good opportunities to directly
compare relevant experimental and empirical data. Based on
this assessment it appears that the ONIOM model performs
reasonably for a gas phase simulation of a solution process.
.
Dissociation of CO from 1a to form 2a is a barrierless pro-
cess. The lowest energy trajectory for CO loss involves signifi-
cant bending of the Rh–C–O angle as the CO dissociates. This
type of trajectory has been identified previously in Krogh-
Jespersen et al.’s33 computations of CO dissociation from Ir
complexes. Because we find a barrierless process, the kinetics of
CO dissociation can be estimated from the relative free energies
of 1a and 2a. The free energy difference of 15.1 kcal molϪ1
based on gas phase computations at 298 K compares reason-
ably with experimental values of 20 kcal molϪ1 14 for CO dis-
sociation in closely related complexes derived from 13CO
Formation and fluxional processes of complex 3
Association of propene with 2 yields the trigonal bipyramidal
isomers of 3. Recall that the thermodynamic free energies of
these association reactions cluster around 0 kcal molϪ1. We have
located six transition states for these processes. Four of the
transition states take isomer 2a to the isomers 3a, 3c, 3d, and 3e.
The other two connect the isomer 3j with 2b and 2d. Whereas
association of CO has no computed barrier, our computations
738
J. Chem. Soc., Dalton Trans., 2002, 729–742

View Online
Table 8 Transition state energies computed for various processes relative to 2a
∆HЊa
∆GЊa
∆SЊb
∆Ea
Imaginary frequency i/cmϪ1
1
1a
1a
1 isomerization
1c
1d
Ϫ14.09
Ϫ23.09
Ϫ2.64
Ϫ10.98
Ϫ38.41
Ϫ40.61
Ϫ14.32
Ϫ23.93
Ϫ410
Ϫ34
2
2a
2a
2 isomerization
2e
2b
5.03
1.65
5.47
2.33
Ϫ1.49
Ϫ2.26
5.87
2.23
Ϫ196
Ϫ58
3
3l
3c
3 isomerization
3l
9.67
21.13
22.96
34.73
Ϫ44.56
Ϫ45.60
8.65
20.83
Ϫ29
Ϫ385
3n
2
3 transition states
2b
2a
2a
2a
2a
2d
3g
3a
3d
3e
3c
3j
4.77
5.63
5.94
5.50
6.42
6.42
17.09
18.89
19.30
18.74
18.84
18.84
Ϫ40.58
Ϫ41.33
Ϫ44.47
Ϫ44.83
Ϫ44.42
Ϫ41.65
4.02
4.78
5.13
4.71
5.61
5.61
Ϫ82
Ϫ71
Ϫ73
Ϫ71
Ϫ69
Ϫ68
3
4 transition states
3n
3a
3j
4e
4f
17.48
12.07
13.69
31.64
26.53
27.81
Ϫ47.50
Ϫ48.51
Ϫ47.37
16.98
11.85
13.23
Ϫ764
Ϫ719
Ϫ645
4a
3
4i transition states
3a
3n
3a
4k
4k
4h
20.81
15.32
18.92
34.89
29.25
33.73
Ϫ47.23
Ϫ46.72
Ϫ49.66
20.60
14.77
18.74
Ϫ946
Ϫ766
Ϫ988
a kcal molϪ1 at 298 K. b cal molϪ1 KϪ1
.
Fig. 8 Computed structures for 6
consistently yield significant activation free energies for propene
association that span a narrow range of 17.1–19.3 kcal molϪ1. It
appears that the combination of basis sets and ONIOM layers
do not suffice to give realistic energies for the interaction of the
catalyst with propene, either in the kinetic or thermodynamic
sense. Unfortunately we have been unable to find published
transition states and energies for alkene association in simpler
model systems to compare with our results. Thus, this crucial
step in the overall catalytic process will require significantly
more attention in future work.
We have examined two fluxional processes at the trigonal
bipyramidal isomers of 3: rotation of the coordinated propene
and interconversion of axial-equatorial and diequatorial xant-
phos coordination modes. Rotation of the propene group at the
diequatorial isomer 3l (transformation 3l
3l in Table 8)
incurs a free energy barrier of 8 kcal molϪ1 (relative to 3l in
J. Chem. Soc., Dalton Trans., 2002, 729–742 739

View Online
Table 1) at 298 K. Essentially this is the energy for rotation of
the C–C bond vector perpendicular to the equatorial coordin-
ation plane. Conversion of the diequatorial isomer 3p to an
axial-equatorial similar to 3b must cross a free energy barrier of
about 20 kcal molϪ1. This is a rather high barrier for this
exchange process. It remains to be seen if this computation is
generally representative of diequatorial/axial-equatorial
exchange processes at 3.
bite angle favour normal aldehydes? What is the catalyst resting
state? What is the selectivity-determining step of 1-alkene
hydroformylation? How could the catalyst be modified to
increase activity? What should be the catalyst rate law and why?
How do isotopic substitutions affect the reaction rates and
why?
Important empirical observations concerning xantphos-
based hydroformylation catalysts are reproduced by compu-
tations. It is clear from the computations that most of the
catalyst will pool in the form of 1. This species, which is the
only observed species under catalytic conditions according
to NMR and IR experiments, is computed to populate both
axial-equatorial and diequatorial coordination modes of the
xantphos ligand.
Formation of complex 4
Particularly in the recent work of Carbó et al., attention has
been focussed on the transformation 3
4 under the hypo-
thesis that this is the step in which the reaction regioselectivity is
fixed. Let us first consider the absolute magnitude of the acti-
vation free energies that we compute for insertion of propene
into Rh–H bonds. Relative to complex 2a, the free energies of
the seven transition states that we have located so far, four to
give n-propyl complexes and three to give i-propyl complexes,
range from 26.5 to 34.9 kcal molϪ1. Given that 2a is neither the
computed nor the experimentally observed resting state of the
catalyst and given that hydroformylation with these ligands
proceeds with turnover frequencies on the order of 100 hϪ1, the
computed insertion transition state free energies are far too
high. As mentioned previously, the weak exothermicity com-
puted for binding propene is a major contributor to the large
discrepancy between experiment and computation.
Experiment also suggests significant population of both
isomer classes, although the experimental estimates (7 : 3 ee :
ae) differ somewhat from computation (1 : 9 ee : ae). The
computed free energies of CO dissociation (15 kcal molϪ1
)
are somewhat lower than experimental estimates (20 kcal
molϪ1) based on isotopic exchange measurements but are not
unreasonable. Computation of the insertion barriers to form
n-propyl isomers vs. i-propyl isomers agree with the n : i
aldehyde ratio observed in experimental studies, but this value
must be considered fortuitous at this stage due to the egregious
errors that are seen in estimates of the overall hydroformylation
activity.
Even at the early stages (3a
4f ) of the hydroformylation
reaction, such high free energy barriers (ca. 41 kcal molϪ1) are
computed for taking the catalyst from its resting state (1a) to
product that any insights into the kinetics of the catalytic pro-
cess are precluded. It appears that a major culprit is a dramatic-
ally underestimated enthalpy of association of propene with 2.
However, it is not clear if other factors also contribute signifi-
cantly, such as intrinsically high barriers for alkene insertion,
etc. With such uncertainties in the magnitudes of important
thermodynamic and kinetic free energies, it is impossible to
meaningfully evaluate which step is selectivity-determining or
which step is turnover-limiting. Such evaluations require high
confidence in the relative transition state free energies of steps
with different molecularities; that level of confidence is not
justified.
Although the sampling of transition states for the trans-
formation of 3 to 4 is still sparse, some interesting trends are
seen. The lowest energy pathway (3a
4f ) for the formation
of the n-propyl isomers is initiated from an axial-equatorial
isomer and yields a trans square planar complex. In contrast,
the lowest energy pathway (3n
4k) yielding an isopropyl
isomer converts di-equatorial five-coordinate complex
a
into a trans square planar complex. If one assumes that
the ratio of these two lowest energy pathways determines
the n/i ratio of the product aldehyde, one obtains the ratio
42 : 1 at 80 ЊC. This number, fortuitously, coincides remark-
ably well with the experimental ratio of 52 : 1 observed for
1-octene hydroformylation with the xantphos-ligated catalyst at
80 ЊC.14
Why are the estimated activation free energies for catalysis so
inconsistent with experiment? At this point it is not certain, but
our previous experience with modelling Rh(diphosphine)-
catalyzed enamide hydrogenation did demonstrate the need to
have high quality basis sets on Rh and P. Another limitation
concerns the B3LYP method. As shown by Decker and
Cundari,10 the energy of propene binding to the catalyst in
simple model systems changes dramatically when the energies
of B3LYP-optimized geometries are evaluated with CCSD(T)
methods. To test the influence of basis sets, computational
methods, and basis set superposition errors on the relative ener-
gies of key intermediates we have performed the computations
summarized in Tables 9 and 10.
Discussion
Computations play an increasingly important role in under-
standing the details of catalytic reaction mechanisms. During
the past decade the speed and accuracy of modern quantum
mechanical (QM) codes as well as the computational power
available from clusters based on commodity PC’s has grown
rapidly. Our first goal for this project is to explore whether
attempting realistic QM simulations of the hydroformylation
catalytic cycle is reasonable. Based on the sheer magnitude of
intermediates, transition states, and their conformers that
have been located during the initial stages of this project it is
clear that a reasonable exploration is within our reach, even
with rather modest resources. Although we have not achieved
the level of conformational sampling that could be achieved by
dynamics methods, these early results show that a large variety
of geometric isomers and conformers can be located using
non-dynamics methods and, more importantly, must be con-
sidered for realistic modelling. This work also amply demon-
strates that computational estimates of free energies, a most
useful quantity for comparing experimental observations with
computational predictions, can be achieved in a reasonable time
frame.
The data in Table 9 summarize the effect of computational
method and basis set on the thermodynamics of 1
2 and 2
3. The changes in energy for the core ONIOM and total
ONIOM energies are consistently close to one another. For the
dissociation of CO from 1 to yield 2 the B3LYP and CCSD(T)
computations yield similar results, but the BP86 functional and
the MP2 method give CO dissociation energies that are much
higher than experimental estimates. For the association of
propene and 2 to yield 3, the B3LYP method yields larger
exothermicities as the basis set becomes more sophisticated.
Thus, consistency with experiment should improve with better
basis sets. Again the BP86 method leads to tighter ligand bind-
ing than B3LYP. Interestingly, the CCSD(T) computations with
LANL2DZ basis functions give a strongly exothermic reaction
(ca. Ϫ14 kcal molϪ1) but most of this disappears when basis set
superposition error (BSSE) corrections are estimated with
Catalytic hydroformylation is an important industrial pro-
cess that is poorly understood in terms of the detailed reaction
mechanisms. A second goal of this work is to gain insights into
the reaction mechanism that are not easily obtained from
experiment alone. Relevant questions include: Why does a wide
740
J. Chem. Soc., Dalton Trans., 2002, 729–742

View Online
Table 9 Calculated ∆E(1
2) (kcal molϪ1) and ∆E(2
3) (kcal molϪ1) at different levels of theory and with different basis sets
Method
∆Eopt
∆Ecore
1
2
ONIOM(B3LYP/LANL2DZ : HF/LANL2MB)a
ONIOM(BP86/LANL2DZ : HF/LANL2MB)a
ONIOM(CCSD(T)/LANL2DZ : HF/LANL2MB)c
ONIOM(MP2/LANL2DZ : HF/LANL2MB)c
ϩ27.69
ϩ35.9
ϩ27.2
—
ϩ26.6
ϩ34.6
26.1
ϩ33.0
2
3
ONIOM(B3LYP/LANL2DZ : HF/LANL2MB)a
ONIOM(BP86/LANL2DZ : HF/LANL2MB)a
B3LYP/(Stuttgart and 6-31G)b, c
Ϫ2.56
Ϫ10.1
Ϫ6.9
7.7
Ϫ13.9
Ϫ4.6
Ϫ11.9
—
—
Ϫ15.8
Ϫ26.1
B3LYP/(Stuttgart and LANL2DZ)b, d
ONIOM(CCSD(T)/LANL2DZ : HF/LANL2MB)e
ONIOM(MP2/LANL2DZ : HF/LANL2MB)e
a Full ONIOM optimisation. b Full B3LYP optimisation. c Rh and P atoms use the Stuttgart basis sets and ECP’s as described in the Methods
section; the 6-31G basis set is used on all other atoms. d Rh and P atoms use the Stuttgart basis sets and ECP’s as described in the Methods section;
the LANL2DZ basis set is used on all other atoms. e Single point calculations using ONIOM(B3LYP/LANL2DZ : HF/LANL2MB) optimised
geometries.
Table 10 ∆E (kcal molϪ1) and ∆E (kcal molϪ1) corrected for basis set
superposition errors (BSSE) for 2
(CO)HRh(PH3)2 ϩ propene)
methods required for future progress in realistic modelling
3 ((CO)HRh(PH3)2(propene)
of hydroformylation. Overall, this work provides the most
comprehensive sketch to date of the structural diversity
present in realistic models of hydroformylation catalysts. As
such, this work will provide a valuable foundation for
subsequent studies to be performed at more sophisticated levels
of computation.
Method
∆E
Ϫ4.1
Ϫ11.2
Ϫ26.0
Ϫ16.0
∆Ecorr(BSSE)
B3LYP/LANL2DZa
BP86/LANL2DZ//B3LYP/LANL2DZ
MP2/LANL2DZ//B3LYP/LANL2DZ
CCSD(T)/LANL2DZ//B3LYP/LANL2DZ
Ϫ1.4
Ϫ8.7
Ϫ9.0
ϩ1.2
Are we at the stage where computations can help us to under-
stand turnover-limiting and selectivity-determining steps for
highly active catalysts? Realistic modelling requires, firstly, the
ability to search large numbers of conformers and isomers all
along the catalytic cycle of model catalysts that reflect the steric
and electronic characteristics of real catalysts. As these initial
results indicate, computations with a realistic scope are feasible.
Obtaining reaction energetics with sufficient accuracy to make
kinetic distinctions, such as the nature of the turnover-limiting
and selectivity-determining steps, presents a far greater chal-
lenge. Catalysts that are highly active necessarily are restricted
to a small range of activation free energies. For multi-step reac-
tions such as hydroformylation this means that high accuracies
are required to determine whether, for example, insertion of
propene into a Rh–H bond will be reversible or irreversible
under catalytic conditions. At the level of computation utilized
in this work, the necessary accuracy of the thermodynamic and
kinetic activation parameters has not been achieved. In the case
of hydroformylation, another limitation is the paucity of good
empirical data that might be used as constraints for some of the
reaction steps.
This is not to say that useful, realistic models are out of
reach. In recent studies of the mechanism of asymmetric
hydrogenation we have shown that strikingly good agreement
between experimental and computational energetics can be
achieved. Although the work reported herein falls short of our
goals, good progress has been made in much of the difficult
work (searching conformational and isomer space, obtaining
good structures for intermediates and transition states). Fur-
thermore, there are obvious, feasible approaches to improving
the computations. Therefore we conclude that a significant first
step has been made to understand catalytic hydroformylation
reactions. Our future efforts will focus on both improving
computational accuracy and expanding the pool of empirical
kinetic data relevant to catalytic hydroformylation.
a Fully optimised.
counterpoise calculations (see Table 10). Such results under-
score the inadequacy of the LANL2DZ basis set functions. It is
commonly observed that basis set superposition errors become
less significant as the quality of the basis sets improve.34 Thus
the next phase of these computations will focus on the appli-
cation of higher level basis sets for the core part of the ONIOM
computation. As always, more levels of sophistication may be
needed such as including solvation effects and the use of larger
core fragments in the ONIOM model.
An alternative explanation for the unrealistically high acti-
vation energies derived from computations is that the computed
mechanism is wrong. For example, it is possible that propene
binding to the catalyst is preceded by dissociation of one arm
of the xantphos chelate from 1, rather than dissociation of CO.
Due to the strain in the xantphos chelate this could provide a
lower energy route to creating a sterically uncongested vacant
coordination site to bind propene. To test this hypothesis we
have performed preliminary computations along this pathway.
Our initial results reveal that addition of propene to 1 along
this path is still thermodynamically uphill and that higher than
realistic activation energies will necessarily result. Thus, it
appears that future studies should focus on the computational
methods and basis sets used, rather than the exploration of
alternate pathways.
Conclusions
The data presented herein represent substantial progress toward
comprehensive modelling of the catalytic hydroformylation
reaction. Relative to previous computational studies,8–12 this
work demonstrates clearly that consideration of multiple
isomers and conformers is essential to realistic modelling. For
example, these results conclusively dispel the assumption12 that
only diequatorial isomers of the critical intermediate 3 contrib-
ute to product formation as the axial-equatorial isomers are
essentially isoenergetic. From the computational perspective,
this work establishes useful guidelines for the basis sets and
Acknowledgements
Support for this work was provided by the National Science
Foundation. The authors wish to thank Mr Chris Pawela for his
expert assistance with the computer hardware used in this
project.
J. Chem. Soc., Dalton Trans., 2002, 729–742
741

View Online
17 R. D. J. Froese and K. Morokuma, in Encyclopedia of
Computational Chemistry, ed. P. v. R. Schleyer, Wiley, New York,
1998.
18 S. Dapprich, I. Komaromi, K. S. Byun, K. Morokuma and
M. J. Frisch, THEOCHEM, 1999, 462, 1.
References
1 For an excellent overview of relevant hydroformylation chemistry
see: Rhodium Catalyzed Hydroformylation, ed. P. W. N. M. van
Leeuwen and C. Claver, Kluwer-CMC, Dordrecht, The
Netherlands, 2000 and references cited therein.
2 C. P. Casey and G. T. Whiteker, Isr. J. Chem., 1990, 30, 299;
P. W. N. M. N. van Leeuwen, P. C. J. Kamer and J. N. H. Reek,
Pure Appl. Chem., 1999, 71, 1443; T. J. Devon, G. W. Phillips,
T. A. Puckette, J. L. Stavinoha and J. J. Vanderbilt, (to Texas
Eastman), US Pat., 4,694,109, 1987.
3 G. Stanley, in Catalysis of Organic Reactions, ed. M. G. Scaros and
M. L. Prunier, Marcel Dekker, Inc., New York, 1995, p. 363;
U. Nettekoven, P. C. J. Kamer, M. Widhalm and P. W. N. M. van
Leeuwen, Organometallics, 2000, 19, 4596.
4 J. E. Babin and G. T. Whiteker (to Union Carbide), WO 93/03839,
US Pat., 911518, 1992; G. J. H. Buisman, M. E. Martin, E. J. Vos,
A. Klootwijik, P. C. J. Kamer and P. W. N. M. van Leeuwen,
Tetrahedron: Asymmetry, 1995, 6, 719.
5 K. Nozaki, N. Sakai, T. Nanno, T. Higashijima, S. Mano,
T. Horiuchi and H. Takaya, J. Am. Chem. Soc., 1997, 119, 4413.
6 C. P. Casey, E. L. Paulsen, E. W. Beuttenmueller, B. R. Proft,
L. M. Petrovich, B. A. Matter and D. R. Powell, J. Am. Chem. Soc.,
1997, 119, 11817.
7 W. R. Rocha and W. B. de Almeida, Int. J. Quantum Chem., 2000, 78,
42.
8 R. Schmid, W. A. Herrmann and G. Frenking, Organometallics,
1997, 16, 701.
9 T. Matsubara, N. Koga, Y. Ding, K. G. Musaev and K. Morokuma,
Organometallics, 1997, 16, 1065.
10 S. A. Decker and T. R. Cundari, Organometallics, 2001, 20, 2827.
11 D. Gleich, R. Schmid and W. A. Herrmann, Organometallics, 1998,
17, 2141; D. Gleich, R. Schmid and W. A. Herrmann,
Organometallics, 1998, 17, 4828; D. Gleich and W. A. Herrmann,
Organometallics, 1999, 18, 4354.
12 J. J. Carbó, F. Maseras, C. Bo and P. W. N. M. van Leeuwen, J. Am.
Chem. Soc., 2001, 123, 7630.
13 M. Kranenburg, Y. E. M. van der Burgt, P. C. J. Kamer and
P. W. N. M. van Leeuwen, Organometallics, 1995, 14, 3081.
14 P. W. N. M. van Leeuwen, C. P. Casey and G. T. Whiteker, in
Rhodium Catalyzed Hydroformylation, ed. P. W. N. M. van Leeuwen
and C. Claver, Kluwer-CMC, Dordrecht, The Netherlands, 2000,
ch. 4; L. A. van der Veen, P. C. J. Kamer and P. W. N. M.
van Leeuwen, Organometallics, 1999, 18, 4765.
19 M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria,
M. A. Robb, J. R. Cheeseman, V. G. Zakrzewski, J. A. Montgomery
Jr., R. E. Stratmann, J. C. Burant, S. Dapprich, J. M. Millam,
A. D. Daniels, K. N. Kudin, M. C. Strain, O. Farkas, J. Tomasi,
V. Barone, M. Cossi, R. Cammi, B. Mennucci, C. Pomelli,
C. Adamo, S. Clifford, J. Ochterski, G. A. Petersson, P. Y. Ayala,
Q. Cui, K. Morokuma, D. K. Malick, A. D. Rabuck,
K. Raghavachari, J. B. Foresman, J. Cioslowski, J. V. Ortiz,
B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi,
R. Gomperts, R. L. Martin, D. J. Fox, T. Keith, M. A. Al-Laham,
C. Y. Peng, A. Nanayakkara, C. Gonzalez, M. Challacombe,
P. M. W. Gill, B. Johnson, W. Chen, M. W. Wong, J. L. Andres,
C. Gonzalez, M. Head-Gordon, E. S. Replogle and J. A. Pople,
Gaussian 98, Revision A.7, Gaussian, Inc., Pittsburgh, PA, 1998.
20 R. G. Parr and W. Wang, Density-Functional Theory of Atoms and
Molecules, Oxford University Press, New York, 1994.
21 R. Neumann, R. H. Nobes and N. C. Handy, Mol. Phys., 1996, 87,
1.
22 A. D. Becke, Phys. Rev. A, 1988, 38, 3098; A. D. Becke., J. Chem.
Phys., 1993, 98, 5648.
23 C. Lee, W. Yang and R. G. Parr, Phys. Rev. B, 1988, 37, 785.
24 P. J. Hay and W. R. Wadt., J. Chem. Phys., 1985, 82, 299.
25 P. J. Hay and W. R. Wadt, J. Chem. Phys., 1985, 82, 285.
26 T. H. Dunning and P. J. Hay, in Modern Theoretical Chemistry, ed.
H. F. Schaefer III, Plenum, New York, 1976, vol. 3, p. 1.
27 W. J. Hehre, R. F. Stewart and J. A. Pople, J. Chem. Phys., 1969, 51,
2657.
28 D. Andrae, U. Haeussermann, M. Dolg, H. Stoll and H. Preuss,
Theor. Chim. Acta, 1990, 77, 123–141.
29 A. Bergner, M. Dolg, W. Kuenchle, H. Stoll and H. Preuss, Mol.
Phys., 1993, 80, 1431–1441.
30 S. Huzinaga, Gaussian Basis Sets for Molecular Calculations,
Elsevier, Amsterdam, 1984.
31 R. Ditchfield, W. J. Hehre and J. A. Pople, J. Chem. Phys., 1971, 54,
724–728; W. J. Hehre, R. Ditchfield and J. A. Pople, J. Chem. Phys.,
1972, 56, 2257–2261; P. C. Hariharan and J. A. Pople, Theor. Chim.
Acta, 1973, 28, 213–222; P. C. Hariharan and J. A. Pople, Mol. Phys.,
1974, 27, 209–214.
32 A. Castellanos-Páez, S. Castillón, C. Claver, P. W. N. M. van
Leeuwen and W. G. J. de Lange, Organometallics, 1998, 17, 2543.
33 F. Abu-Hasanayn, K. Krogh-Jespersen and A. S. Goldman, J. Am.
Chem. Soc., 1994, 116, 5979.
15 M. Svensson, S. Humbel, R. D. Froese, T. Matsubara, S. Sieber and
K. Morokuma, J. Phys. Chem., 1996, 100, 19357.
16 S. Humbel, S. Sieber and K. Morokuma, J. Chem. Phys., 1996, 105,
1959.
34 K. A. Peterson and T. H. Dunning, J. Chem. Phys., 1995, 102, 2032.
742
J. Chem. Soc., Dalton Trans., 2002, 729–742