The mechanism for the rhodium-catalyzed decarbonylation of aldehydes: A combined experimental and theoretical study

Article abstract of DOI:10.1021/ja710270j

The mechanism for the rhodium-catalyzed decarbonylation of aldehydes was investigated by experimental techniques (Hammett studies and kinetic isotope effects) and extended by a computational study (DFT calculations). For both benzaldehyde and phenyl acetaldehyde derivatives, linear Hammett plots were obtained with positive slopes of +0.79 and +0.43, respectively, which indicate a buildup of negative charge in the selectivity-determining step. The kinetic isotope effects were similar for these substrates (1.73 and 1.77 for benzaldehyde and phenyl acetaldehyde, respectively), indicating that similar mechanisms are operating. A DFT (B3LYP) study of the catalytic cycle indicated a rapid oxidative addition into the C(O)-H bond followed by a rate-limiting extrusion of CO and reductive elimination. The theoretical kinetic isotope effects based on this mechanism were in excellent agreement with the experimental values for both substrates, but only when migratory extrusion of CO was selected as the rate-determining step.

Full text of DOI:10.1021/ja710270j

Published on Web 02/28/2008
The Mechanism for the Rhodium-Catalyzed Decarbonylation
of Aldehydes: A Combined Experimental and Theoretical
Study
Peter Fristrup,† Michael Kreis,† Anders Palmelund,† Per-Ola Norrby,‡ and
Robert Madsen*,†
Center for Sustainable and Green Chemistry, Department of Chemistry, Building 201, Technical
UniVersity of Denmark, DK-2800 Lyngby, Denmark, and Department of Chemistry, UniVersity of
Gothenburg, Kemigården 4, SE-412 96 Göteborg, Sweden
Received November 13, 2007; E-mail: rm@kemi.dtu.dk
Abstract: The mechanism for the rhodium-catalyzed decarbonylation of aldehydes was investigated by
experimental techniques (Hammett studies and kinetic isotope effects) and extended by a computational
study (DFT calculations). For both benzaldehyde and phenyl acetaldehyde derivatives, linear Hammett
plots were obtained with positive slopes of +0.79 and +0.43, respectively, which indicate a buildup of
negative charge in the selectivity-determining step. The kinetic isotope effects were similar for these
substrates (1.73 and 1.77 for benzaldehyde and phenyl acetaldehyde, respectively), indicating that similar
mechanisms are operating. A DFT (B3LYP) study of the catalytic cycle indicated a rapid oxidative addition
into the C(O)-H bond followed by a rate-limiting extrusion of CO and reductive elimination. The theoretical
kinetic isotope effects based on this mechanism were in excellent agreement with the experimental values
for both substrates, but only when migratory extrusion of CO was selected as the rate-determining step.
Introduction
couple of years later an increase in temperature allowed
substoichiometric amounts of the metal catalyst to be used.5
Reactions involving transfer of carbon monoxide via transition
metals (carbonylation,1 hydroformylation,2 and decarbonylation)
belong to the very heart of organometallic chemistry, but to
our knowledge, only the two former have been studied by
computational methodsspresumably owing to their established
industrial importance. As part of a continued effort toward the
development of more sustainable chemical transformations and,
in particular, the utilization of bioderived starting materials in
organic synthesis, we have turned our attention toward the
rhodium-mediated decarbonylation.3 In its stoichiometric form,
the reaction was discovered by Tsuji and Ohno in 1965,4 and a
In the more than 40 years that have passed since the discovery
of the reaction, there has been an ongoing interest in the
development of the methodology6 as well as its application in
synthesis.7,8 Significant progress toward a more efficient de-
carbonylation methodology was achieved by Doughty and
Pignolet in 1978 through the introduction of chelating bispho-
sphines as ligands for rhodium.9 The favored catalyst was found
to be [Rh(dppp)2]Cl, and this advancement was followed by a
(5) Ohno, K.; Tsuji, J. J. Am. Chem. Soc. 1968, 90, 99–107.
(6) (a) Fessard, T. C.; Andrews, S. P.; Motoyoshi, H.; Carreira, E. M.
Angew. Chem., Int. Ed. 2007, 46, 9331–9334. (b) Lee, H. W.; Chan,
A. S. C.; Kwong, F. Y. Chem. Commun. 2007, 2633–2635. (c) Beck,
C. M.; Rathmill, S. E.; Park, Y. J.; Chen, J.; Crabtree, R. H.; Liable-
Sands, L. M.; Rheingold, A. L. Organometallics 1999, 18, 5311–5317.
(d) O’Connor, J. M.; Ma, J. J. Org. Chem. 1992, 57, 5075–5077. (e)
Baldwin, J. E.; Barden, T. C.; Pugh, R. L.; Widdison, W. C. J. Org.
Chem. 1987, 52, 3303–3307.
† Technical University of Denmark.
‡ University of Gothenburg.
(1) Monsanto process (carbonylation of methanol): (a) Cheong, M.;
Ziegler, T. Organometallics 2005, 24, 3053–3058. (b) Feliz, M.;
Freixa, Z.; van Leeuwen, P. W. N. M.; Bo, C. Organometallics 2005,
24, 5718–5723. (c) Daura-Oller, E.; Poblet, J. M.; Bo, C. Dalton Trans.
2003, 92–98. (d) Kinnunen, T.; Laasonen, K. J. Organomet. Chem.
2003, 665, 150–155. (e) Cavallo, L.; Solà, M. J. Am. Chem. Soc. 2001,
123, 12294–12302. (f) Ivanova, E. A.; Gisdakis, P.; Nasluzov, V. A.;
Rubailo, A. I.; Rösch, N. Organometallics 2001, 20, 1161–1174.
(2) Hydroformylation: (a) Sparta, M.; Børve, K. J.; Jensen, V. R. J. Am.
Chem. Soc. 2007, 129, 8487–8499. (b) Gleich, D.; Hutter, J.
Chem.sEur. J. 2004, 10, 2435–2444. (c) Landis, C. R.; Uddin, J.
J. Chem. Soc., Dalton Trans. 2002, 729–742. (d) Carbó, J. J.; Maseras,
F.; Bo, C.; van Leeuwen, P. W. N. M. J. Am. Chem. Soc. 2001, 123,
7630–7637. (e) Alagona, G.; Ghio, C.; Lazzaroni, R.; Settambolo, R.
Organometallics 2001, 20, 5394–5404. (f) Decker, S. A.; Cundari,
T. R. J. Organomet. Chem. 2001, 635, 132–141. (g) Schmid, R.;
Herrman, W. A.; Frenking, G. Organometallics 1997, 16, 701–708.
(3) For a recent review on rhodium-catalyzed C-C bond cleavage
reactions, see: Necas, D.; Kotora, M. Curr. Org. Chem. 2007, 11,
1566–1591.
(7) For recent applications in total synthesis, see: (a) Padwa, A.; Zhang,
H. J. Org. Chem. 2007, 72, 2570–2582. (b) Malerich, J. P.; Maimone,
T. J.; Elliott, G. I.; Trauner, D. J. Am. Chem. Soc. 2005, 127, 6276–
6283. (c) Harmata, M.; Wacharasindhu, S. Org. Lett. 2005, 7, 2563–
2565. (d) Weatherhead, G. S.; Cortez, G. A.; Schrock, R. R.; Hoveyda,
A. H. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 5805–5809. (e) Banwell,
M. G.; Coster, M. J.; Edwards, A. J.; Karunaratne, O. P.; Smith, J. A.;
Welling, L. L.; Willis, A. L. Aust. J. Chem. 2003, 56, 585–595. (f)
Allin, S. M.; James, S. L.; Elsegood, M. R. J.; Martin, W. P. J. Org.
Chem. 2002, 67, 9464–9467. (g) Kato, T.; Hoshikawa, M.; Yaguchi,
Y.; Izumi, K.; Uotsu, Y.; Sakai, K. Tetrahedron 2002, 58, 9213–9222.
(h) Boeckman, R. K., Jr; Zhang, J.; Reeder, M. R. Org. Lett. 2002, 4,
3891–3894.
(8) For applications in carbohydrate chemistry, see: (a) Monrad, R. N.;
Madsen, R. J. Org. Chem. 2007, 72, 9782–9785. (b) Andrews, M. A.;
Gould, G. L.; Klaeren, S. A. J. Org. Chem. 1989, 54, 5257–5264.
(9) Doughty, D. H.; Pignolet, L. H. J. Am. Chem. Soc. 1978, 100, 7083–
7085.
(4) Tsuji, J.; Ohno, K. Tetrahedron Lett. 1965, 3969–3971.
9
5206 J. AM. CHEM. SOC. 2008, 130, 5206–5215
10.1021/ja710270j CCC: $40.75
2008 American Chemical Society

Rhodium-Catalyzed Decarbonylation of Aldehydes
A R T I C L E S
Scheme 1. Competition Studies for the Decarbonylation of
Benzaldehyde versus p-Substituted Benzaldehydes
mechanistic study, which resulted in a proposal comprising five
elementary steps: (i) coordination of the aldehyde, (ii) oxidative
addition of the aldehydic C-H bond to form a rhodium-acyl
complex, (iii) migratory extrusion of carbon monoxide, (iv)
reductive elimination of the product, and (v) dissociation of the
extruded carbon monoxide.10
Recently, we have developed a more efficient decarbonylation
methodology, which relies on the in situ formation of the active
catalyst from the commercially available RhCl3•3H2O and dppp
ligand in diglyme as a solvent.11 This reaction showed good
functional group compatibility and, in addition, could be used
as part of a tandem Oppenauer oxidation-decarbonylation
sequence. In an attempt to discover more efficient catalysts for
the decarbonylation, we have undertaken a combined experi-
mental and theoretical study of the existing Rh-dppp system.
We believe that a good understanding of the existing system is
a necessary prerequisite for performing knowledge-based design
of a new catalyst system. During the last decades, the develop-
ment of computers and algorithms for molecular modeling has
allowed the computational chemist to investigate systems of
increasing molecular complexity with reasonable accuracy. Here,
we incorporate computational chemistry as an integrated part
of a mechanistic study directed toward the elucidation of the
mechanism for the rhodium-catalyzed decarbonylation.
the beginning and at intervals, and the disappearance of both
benzaldehyde and the para-substituted aldehyde under inves-
tigation was quantified by GC relative to the internal standard.
Assuming that the two substrates follow the same mechanism,
the reaction order in each component will be the same for both
substrates in a competition study. If the reaction is first-order
in substrate, the relative reactivity can then be obtained as the
slope of the line when ln(c0/c) for one substrate is plotted against
the same values for benzaldehyde. As can be seen in Figure 1,
the good linear fit validates the reaction order assumptions.
The kinetic data resulted in seven straight lines all with R2
values above 0.99 (Figure 1). With the slopes of the seven
curves, which equal kX/kH, the Hammett plot could now be
constructed using σ values from the literature (Figure 2).15 The
slope of the line was positive, showing that the reaction is
accelerated by electron-withdrawing aldehyde substituents. The
observed relative reactivities were in accordance with the
“standard” σ values, thus eliminating the possibility of a radical
mechanism12a which has been shown for the catalytic decar-
bonylation with transition metal porphyrin complexes.16
The relatively small positive slope of +0.79 indicates that a
small negative charge is buildup in the transition state (TS) of
the selectivity-determining step. This in itself is unsurprising
since both oxidative addition and migratory extrusion lead to
structures with a carbon-metal bond, thus giving complexes
with more negative charge in either the benzylic position or on
the aromatic ring. One of the possible methods to differentiate
between these two possibilities is a Hammett study of substituted
phenyl acetaldehydes. If the selectivity-determining step is the
oxidative addition, then there should be no conjugation between
the para-substituent of the arene and the reacting site, resulting
in a low slope for the Hammett plot. If the selectivity-
determining step, on the other hand, is the migratory extrusion
of carbon monoxide, then the reacting site would be a benzylic
position which is in conjugation with the para-substituent.
The reactions were carried out under the same conditions as
for the benzaldehydes, but because the reactions were too fast,
the amount of catalyst had to be reduced to 1.5 mol % (Scheme
2). Again, preliminary studies showed that the reactions of the
phenyl acetaldehydes proceeded in quantitative yield with no
byproduct formation.
Results and Discussion
Experimental Study. The experimental part of the mechanistic
study will incorporate Hammett studies in line with earlier
work,12 where the observed relative reactivity of different para-
substituted substrates is used to characterize the selectivity-
determining step. Furthermore, the method can be used to
differentiate between different mechanistic proposals based on
the quality of the fit against different sets of σ values (e.g.,
“normal”, “cationic”, or “radical”), which could then serve as
a starting point for a computational exploration of the reaction.
In addition to the Hammett study, deuterated substrates have
been synthesized, allowing for a determination of the kinetic
isotope effect, which can be directly compared to a calculated
value derived from a theoretical catalytic cycle. In the current
study, we have selected a set of para-substituted benzaldehydes
and envisioned that their relative reactivity could be determined
by means of a series of competition experiments.13 In these
experiments, two substrates were decarbonylated simultaneously,
and the relative disappearance of the starting material could be
quantified using an internal standard (Scheme 1).
Seven different para-substituted benzaldehydes that had been
found to give good yields with negligible formation of byproduct
were included in this study.11 The decarbonylations were
performed in a 0.1 M diglyme solution, which we have shown
earlier to be the favored solvent for the reaction.11 Under these
conditions, oxidative addition into the aryl-chloride bond of
4-chlorobenzaldehyde does not occur.14 Samples were taken at
The kinetic data resulted in seven straight lines all with R2
values above 0.99 (Figure 3). As for the benzaldehydes, electron-
withdrawing substituents accelerate the reaction. The resulting
Hammett plot is shown in Figure 4.
(10) Doughty, D. H.; Anderson, M. P.; Casalnuovo, A. L.; McGuiggan,
M. F.; Tso, C. C.; Wang, H. H.; Pignolet, L. H. AdV. Chem. Ser. 1982,
196, 65–83.
The linear regression gave a slope of +0.43 but a poorer fit
than for the benzaldehydes, with a R2 of 0.70. While the relative
reactivities for the substrates containing non-halide substituents
still fall on a reasonable line, the halide-containing phenyl
(11) Kreis, M.; Palmelund, A.; Bunch, L.; Madsen, R. AdV. Synth. Catal.
2006, 348, 2148–2154.
(12) (a) Keinicke, L.; Fristrup, P.; Norrby, P.-O.; Madsen, R. J. Am. Chem.
Soc. 2005, 127, 15756–15761. (b) Fristrup, P.; Le Quement, S.; Tanner,
D.; Norrby, P.-O. Organometallics 2004, 23, 6160–6165.
(13) (a) Fristrup, P.; Jensen, G. H.; Andersen, M. L. N.; Tanner, D.; Norrby,
P.-O. J. Organomet. Chem. 2006, 691, 2182–2198. (b) Fristrup, P.;
Tanner, D.; Norrby, P.-O. Chirality 2003, 15, 360–368.
(14) A Hammett study by Doughty et al. failed, due to the deactivation of
4-chlorobenzaldehyde in neat solution; see ref 10.
(15) Hansch, C.; Leo, A.; Taft, R. W. Chem. ReV. 1991, 91, 165–195.
(16) (a) Belani, R. M.; James, B. R.; Dolphin, D.; Rettig, S. J. Can. J. Chem.
1988, 66, 2072–2078. (b) Domazetis, G.; Tarpey, B.; Dolphin, D.;
James, B. R. J. Chem. Soc., Chem. Commun. 1980, 939–940.
9
J. AM. CHEM. SOC. VOL. 130, NO. 15, 2008 5207

A R T I C L E S
Fristrup et al.
Figure 1. Kinetic data for the decarbonylation of para-substituted benzaldehydes.
zaldehydes resulted in a slope of 1.69 with a resonance
contribution of 43%. Thus, the substituent influence is weaker
for phenyl acetaldehyde, but it is still clear that electrons are
shifted toward the aryl ring in the selectivity-determining TS.
The lower F value for phenyl acetaldehyde is in line with what
is expected for introducing a methylene between the aryl ring
and the reaction center.18 Together with the good correlations
in Figure 1 and Figure 3, this is an indication that all substrates
follow the same reaction mechanism.
We also tested the relative reactivity of benzaldehyde versus
that of phenyl acetaldehyde. Somewhat surprisingly, in a direct
competition, phenyl acetaldehyde reacted approximately 20
times faster than benzaldehyde, whereas kinetic measurements
in separate experiments only gave a factor of 6 in relative
reactivity. Thus, the substrate selection includes additional
factors beyond the preference in the rate-limiting step, but the
difference between these two ratios only corresponds to an
energy difference of 3–4 kJ/mol. To see if there is a change in
the rate-determining step between the two substrates and to get
a better idea whether the oxidative addition or the migratory
extrusion is the rate-determining step, we decided to determine
the deuterium isotope effect of benzaldehyde and phenyl
acetaldehyde. Using 1.5 mol % of RhCl3•3H2O and 3 mol %
of dppp in a comparative kinetic study of benzaldehyde-d0 or
benzaldehyde-d1, respectively, versus p-tolualdehyde, a deute-
rium isotope effect kH/kD of 1.77 was observed for benzaldehyde.
For the comparative kinetic study of phenylacetaldehyde-d0 or
phenylacetaldehyde-d1, respectively, versus 2-(4-methoxyphe-
nyl)acetaldehyde, a deuterium isotope effect kH/kD of 1.73 was
observed (for further details, see Supporting Information). The
similarity of the two kinetic isotope effects indicates that the
rate-determining step is the same for both benzaldehyde and
phenyl acetaldehyde. The absolute value of the deuterium
isotope effect of 1.7–1.8 clearly indicates that the C-H bond
has been broken in going from the resting state to the selectivity-
determining step. This does not reveal the exact nature of the
transition state but is again a strong indication that the C-H
bond is intact in the resting state.
Figure 2. Hammett plot for the decarbonylation of para-substituted
benzaldehydes.
Scheme 2. Competition Studies for the Decarbonylation of Phenyl
Acetaldehydes versus p-Substituted Phenyl Acetaldehydes
acetaldehydes (F, Cl, and CF3) reacted faster than would be
expected from their σ value. The same tendency was also
observed for the benzaldehydes, but it is much less pronounced
in that case. To rationalize this, there are two possibilities: either
the halide-substituted aldehydes react via a different mechanism
or the fit to a single empirical σp value is simply not sufficiently
accurate. To check the latter possibility, we used the R and F
values introduced by Swain and Lupton,17 which split the σ
values into a resonance and an inductive contribution, thereby
giving a two-parameter fit. The correlation is, as expected, very
much improved, yielding a R2 ) 0.93 and a slope of 1.0 with
36% resonance contribution. Similar calculations for the ben-
Computational Study. To address the ambiguities observed
in the experimental study, the investigation was extended with
(17) Swain, C. G.; Lupton, E. C., Jr J. Am. Chem. Soc. 1968, 90, 4328–
(18) Smith, M. B.; March, J. March’s AdVanced Organic Chemistry, 5th
4337.
ed.; Wiley: New York, 2001.
9
5208 J. AM. CHEM. SOC. VOL. 130, NO. 15, 2008

Rhodium-Catalyzed Decarbonylation of Aldehydes
A R T I C L E S
Figure 3. Kinetic data for the decarbonylation of para-substituted phenyl acetaldehydes.
Figure 4. Hammett plot for the decarbonylation of para-substituted phenyl
acetaldehydes.
Figure 5. Tentative catalytic cycle for the rhodium-catalyzed decarbony-
lation of aldehydes.
a computational study. Although recent advances in computing
power allow a treatment of the full experimental system, it is
often useful to start by investigating a small model system where
different mechanistic proposals can be compared within a
reasonable time frame. In the present work, the obvious model
system consists of a simplified model ligand where the phenyl
groups have been removed and replaced by hydrogen. The
electronic properties of this model ligand are expected to be
reasonably close to those of the real ligand, whereas the reduced
size does not offer a possibility for van der Waals and/or
electrostatic interactions between the phenyl groups in the
substrate and those in the catalyst. However, for the initial
evaluation of different mechanistic possibilities, the system
should be adequate.
On the basis of the many mechanistic similarities with the
rhodium-catalyzed carbonylation and hydroformylation, we
suggest a catalytic cycle based on (i) coordination of aldehyde,
(ii) oxidative addition, (iii) migratory extrusion, and (iv)
reductive elimination as illustrated in Figure 5.
In the original work by Doughty and Pignolet, a catalytic
cycle was suggested which contained two dppp molecules per
rhodium through the entire cycle. However, calculations per-
formed with the model ligand (Ph ) H) indicate that coordina-
tion of PhCHO to Rh(dppp)2+ is highly unfavorable, which were
further substantiated by a series of energy minimizations with
different fixed distances between RhI and the benzaldehyde
oxygen atom (distance is varied from 2.7 to 2.0 Å in steps of
0.1 Å, which resulted in an energy increase of about 100 kJ/
mol; plot has been included as Supporting Information).
Although the model phosphine may not describe the steric
interactions accurately, the model is sufficient in this case, where
introduction of the additional phenyl groups only leads to an
increase of the steric repulsion between the approaching
substrate and the RhI complex. Instead, we suggest Rh(dpp-
+
p)(CO)2 as a starting point for the catalytic reaction, thus
indicating an exchange of CO for PhCHO as the first elementary
step. A significant amount of the added catalyst may very well
be tied up as Rh(dppp)2+, CO-bridged dimers, or other low-
energy, unreactive resting states, but in the present work, we
have focused our attention on the complexes which can be part
of the catalytic cycle.
In the coordination chemistry of RhI, a square-planar geom-
etry is often preferred, which is also the result achieved when
the cationic Rh(model-dppp)(CO)2 complex is optimized using
B3LYP/LACVP*. Coordination of the aldehyde takes place via
the oxygen atom and gives rise to another square-planar complex
which is higher in Gibbs free energy (26 kJ/mol).
Since oxidative addition results in the formation of an
additional ligand, there exists the possibility for the generation
of three different geometrical isomers of the five-coordinate,
octahedral RhIII intermediate. The three different isomers can
have either hydride, CO, or the acyl moiety in the apical position
9
J. AM. CHEM. SOC. VOL. 130, NO. 15, 2008 5209

A R T I C L E S
Fristrup et al.
Figure 6. The three possible geometrical isomers of the five-coordinated
RhIII intermediate (Gibbs free energies, 298 K).
Figure 9. Structure of the transition state for migratory extrusion of carbon
monoxide.
Figure 7. The structure of the transition state for oxidative addition.
Figure 10. Transition state for reductive elimination of benzene.
will therefore postpone the evaluation of all possible isomers
of the transition state for migratory extrusion until the full ligand
is included. However, we have located one of the possible
transition states (Ph equatorial, H axial, Figure 9) and determined
the length of the breaking C-C bond to be 1.838 Å.
After migratory extrusion, the three components originating
from the aldehyde (i.e., the “R group”, hydride, and CO) take
up three coordination sites, but for reductive elimination to take
place, the hydride and the phenyl must be in a cis relationship.
Again, we anticipate small differences in energy between the
isomers with either hydride or phenyl in the apical position and
postpone the final evaluation of these isomers. However,
employing the model system, we found a transition state for
reductive elimination (Figure 10), and the length of the forming
C-H bond is 1.625 Å. The elimination of benzene regenerates
the Rh(dppp)(CO)2 complex, which is ready to participate in
the next catalytic cycle.
Since the dppp ligand occupies two coordination sites and
the fragments from the aldehyde a maximum of three (Ph, H,
and CO), there will be one additional site available during the
entire catalytic cycle (denoted “L” in Figure 5). A priori the
identity of this spectator ligand L is unknown, but good
candidates would be the ubiquitous carbon monoxide formed
during the reaction or the chloride counterion present in the
metal precursor (RhCl3•3H2O). Up until now, we have limited
ourselves to CO for simplicity, but we have also investigated
the energetics of the entire reaction using chloride, water, or a
phosphine as “spectator ligand”. To allow a reasonable com-
parison between the cationic and neutral systems, we have used
the PB-SCRF solvation model with parameters suitable for THF.
Figure 8. The four possible geometrical isomers of the octahedral RhIII
complex (Gibbs free energies, 298 K).
(Figure 6), and since an apical acyl moiety is strongly favored
(Figure 6, left), this indicates that the transition state leading to
this isomer should also be the most favorable.
The transition state for oxidative addition was found to have
a C-H distance of 1.625 Å (Figure 7). This type of three-center
TS is ubiquitous in organometallic chemistry.
The next step in the reaction sequence is migratory extrusion
of carbon monoxide, which also leads to the occupation of the
sixth and last coordination site in the octahedral RhIII complex.
Also here, we found it necessary to evaluate several geometrical
isomers, and again, we began with the inspection of the
octahedral complex formed after migratory extrusion. With two
identical carbon monoxide ligands, the number of isomers was
reduced to four (Figure 8). The two favored isomers both have
one of the ligands with the strongest trans-influence (H or Ph)
trans to a ligand with weak trans-influence (CO). The structure
with both H and Ph trans to phosphines (medium trans-influence)
is less favored, and the structure with both ligands with strong
trans-influence opposite each other is strongly disfavored.19
With such small differences in energy, the influence of the
phenyl groups on the ligand may become important, and we
(19) Appleton, T. G.; Clark, H. C.; Manzer, L. F. Coord. Chem. ReV. 1973,
10, 335–422.
9
5210 J. AM. CHEM. SOC. VOL. 130, NO. 15, 2008

Rhodium-Catalyzed Decarbonylation of Aldehydes
A R T I C L E S
Figure 11. Energy profiles (enthalpy) for four different spectator ligands using the model diphosphine ligand obtained with the solvation model describing
THF.
Figure 12. Two X-ray structures of Rh(dppp) complexes illustrating the structure of the five-coordinated RhIII intermediate formed after oxidative addition.
Left: Complex with two chlorides and the benzoyl moiety in the apical position. Right: Complex with two iodides and an acetyl moiety in the apical
position.
The structures were fully optimized using this solvation model,
and the resulting solution phase energies are shown in Figure
11. It is clear that the differences between three of the spectator
ligands are insignificant, but the solution phase energy for the
water-ligated pathway is constantly ∼50 kJ/mol higher than the
others. As a consequence, it is therefore difficult to determine
the one most likely to be present under experimental conditions,
but two of them deserves special notion in this respect. Since
chloride is present on rhodium even before formation of the
catalytically active species, it may very well stay attached to
the metal during the reduction of the RhIII precatalyst to RhI.
In addition, chloride is the only anionic ligand, which should
be favorable for coordination to the cationic metal center. In
the current study, however, we choose to favor carbon monoxide
as a spectator ligand since the abundance of carbon monoxide
increases dramatically during the course of the reaction. From
these calculations of relative energies, a phosphine ligand also
seems plausible; however, these calculations were performed
with a model phosphine (PH3), whereas the real ligand is
significantly larger than both chloride and CO and, as a
consequence, sterically disfavored.
geometry and, as a consequence, renders the approach of an
incoming substrate even more difficult.
Computational Study with Full Experimental System:
Benzaldehyde. Substitution of the simple hydrogen atoms in the
model ligand with phenyl groups dramatically increases the
number of possible conformations. However, relevant analogous
complexes can be found in the Cambridge Crystallographic
Database. The database contains two acyl-rhodiumIIIdppp
complexes, depicted in Figure 12 (CCDB codes BZPPRH10
and YIZDIU02). The dppp ligand forms a six-membered chelate,
where one side of the ring is blocked by two phenyl groups,
whereas the other two phenyl substituents are equatorial, leaving
one phase of the metal open. In both structures, the apical acyl
ligand is placed on the open face of the catalyst.
With these structures as starting points, we have investigated
the decarbonylation of benzaldehyde using DFT calculations
(B3LYP functional,20 LACVP* basis set21) in the gas phase.
At this level of theory, the decarbonylation of benzaldehyde is
exergonic (∆G298K ) –8 kJ/mol). This is attributable to the
entropically favored liberation of carbon monoxide, and at even
higher temperatures (experimental conditions, T ) 440 K), the
reaction is calculated to be exergonic by 27 kJ/mol. For the
As suggested originally by Doughty and Pignolet, the
additional coordination could also be occupied by another dppp
ligand,9,10 which would dissociate one of the phosphines in the
octahedral RhIII intermediate, but as mentioned previously, the
approach of benzaldehyde to Rh(dppp)2+ is strongly disfavored.
If the full experimental dppp ligand is included, the combined
steric demand forces the complex to adopt a non-square-planar
(20) (a) Becke, A. D. J. Chem. Phys. 1993, 98, 5648–5652. (b) Becke,
A. D. J. Chem. Phys. 1993, 98, 1372–1377. (c) Lee, C.; Yang, W.;
Parr, R. G. Phys. ReV. B 1988, 37, 785–789.
(21) LACVP* uses the 6-31G* basis set for all light elements and the
Hay-Wadt ECP and basis set for rhodium; see: Hay, P. J.; Wadt,
W. R. J. Chem. Phys. 1985, 82, 299–310.
9
J. AM. CHEM. SOC. VOL. 130, NO. 15, 2008 5211

A R T I C L E S
Fristrup et al.
Figure 13. Left: Structure of the transition state for oxidative addition with the reaction taking place in a manner which delivers the acyl moiety to the
“open” face of the catalyst. Right: Energy profile for the dissociative pathway.
phenyl acetaldehyde, the decarbonylation becomes more favor-
able, with calculated Gibbs free energies of 43 and 64 kJ/mol,
respectively, at 298 and 440 K. When the selected computational
method was applied to an energy minimization of the two X-ray
structures depicted in Figure 12, the overall correspondence is
good, but the bond lengths to Rh are slightly overestimated in
the calculations. Overlaying the metals and directly attached
atoms gave an rms deviation of 0.06 Å for BZPPRH10 and
0.16 Å for YIZDIU02, in the latter case mostly due to the overly
long Rh-I bonds.
Our starting point and energetic reference was the cationic
+
square-planar RhI(dppp)(CO)2 complex. Replacement of one
CO by PhCHO leads to a free energy increase of 39 kJ/mol.
The dissociative pathway involves expulsion of a CO molecule,
leading to the tricoordinated Rh(dppp)CO+ complex, which has
a higher Gibbs free energy (298 K, 92 kJ/mol), both compared
Figure 14. The four isomers for which the migratory extrusion was
investigated by DFT calculations.
+
to the migrating CO moiety can be safely ignored as the barrier
was ∼50 kJ/mol higher than for the TS with lowest energy.
Backward energy minimizations from the TS revealed the
existence of a “preorganized” five-coordinate complex with a
free coordination site cis to the acyl moiety. For the lowest
energy TS, this complex was characterized and found to have
a relative Gibbs free energy of 179 kJ/mol (Figure 15, left).
The transition state for migratory extrusion has a free energy
of 221 kJ/mol for the most favored isomer (Figure 15, center,
and Figure 16). This TS is very similar to the ones discussed
by Ziegler22 and Cavallo,23 and in our system, the key distances
are C(ipso)-C(O) ) 1.928 Å, Rh-C(O) ) 1.919 Å, and
Rh-C(ipso) ) 2.254 Å. The TS was verified by a full frequency
calculation which resulted in a single imaginary frequency of
270 cm-1. After migratory extrusion, an octahedral RhIII
intermediate is formed with a free energy of 170 kJ/mol.
Subsequently, reductive elimination of benzene goes through
another three-center TS (Figure 15, right, and Figure 16). Here
the barrier is relatively low, only 57 kJ/mol, suggesting that
reductive elimination takes place readily after the octahedral
RhIII intermediate has been formed. The relative Gibbs free
energy of this TS is 227 kJ/mol when compared to the
RhI(dppp)(CO)2 complex, and the key distances are C-H )
to the RhI(dppp)(CO)2 complex. Attempts to identify an
associative pathway using the model phosphine were unsuc-
cessful since an approach of benzaldehyde leads to an energy
increase as high as 300 kJ/mol without resulting in dissociation
of a CO molecule. Benzaldehyde strongly prefers coordination
via the oxygen atom, but oxidative addition of the C-H bond
requires formation of a prereactive complex with Rh-H
coordination (relative Gibbs energy ) 114 kJ/mol). From this
complex, oxidative addition goes through a three-center transi-
tion state with a relative Gibbs free energy of 158 kJ/mol (Figure
13).
The nature of the TS was verified by a full frequency
calculation which resulted in a single imaginary frequency of
639 cm-1. After oxidative addition, a five-coordinated inter-
mediate is formed, which has a relative Gibbs free energy of
124 kJ/mol.
The subsequent step in the catalytic cycle is migratory
extrusion of carbon monoxide. Of the several possible TSs, only
the one leading to a trans relationship between H and Ph could
be discarded based on the model studies. The results for the
remaining TSs are depicted in Figure 14.
The hydride already present on rhodium prefers a position
trans to either phosphorus or spectator carbon monoxide, which
can be fulfilled in three of the TSs and results in almost similar
energies. These reaction pathways may all be operating under
experimental conditions. The TS with hydride positioned trans
(22) Margl, P.; Ziegler, T.; Blöchl, P. E. J. Am. Chem. Soc. 1996, 118,
5412–5419.
(23) Cavallo, L.; Solà, M. J. Am. Chem. Soc. 2001, 123, 12294–12302.
9
5212 J. AM. CHEM. SOC. VOL. 130, NO. 15, 2008

Rhodium-Catalyzed Decarbonylation of Aldehydes
A R T I C L E S
Figure 15. Left: The prereactive complex for the favored isomer of the transition state for migratory extrusion. Center: The favored isomer for the transition
state for migratory extrusion. Right: The transition state for reductive elimination of benzene resulting in regeneration of the catalytically active Rh(dppp)(CO)2
complex.
Figure 16. Overview of the catalytic cycle for decarbonylation of benzaldehyde. For clarity, only the isomer having the lowest energy has been included.
1.609 Å, Rh-H ) 1.630 Å, and Rh-C ) 2.224 Å. The single
imaginary frequency was found to be 943 cm-1. Two additional
isomers of this TS were found within 1 kJ/mol, and the reaction
may therefore very well proceed through several reaction
pathways simultaneously. In addition, it can be expected that
different substrates could potentially be decarbonylated through
different isomeric reaction pathways.
Computational Study with Full Experimental System:
Phenyl Acetaldehyde. Although many similarities could be
expected between the benzaldehydes and the phenyl acetalde-
hydes, a thorough computational investigation was still required
in order to locate all relevant intermediates for this more flexible
substrate. Interestingly, the overall transformation is considerably
more favorable for this substrate since the products (toluene
and CO) are stabilized by 43 kJ/mol in Gibbs free energy (298
K) compared to the phenyl acetaldehyde starting material (Figure
17).
For this substrate, the initial displacement of CO from the
cationic square-planar RhIdppp(CO)2 is less facile than that for
benzaldehyde, resulting in an increase in Gibbs free energy of
57 kJ/mol for the coordination of the aldehyde instead of CO.
Again, we expect a dissociative mechanism to be operating.
We have not been able to locate a prereactive complex with
coordination to the hydrogen atom of the aldehyde functionality
as was the case for benzaldehyde. A structure like this is not a
stable minimum, but the reaction may still proceed through such
a geometry.
It was advantageous to employ the small model phosphine
in the initial calculations, but also here the full dppp ligand must
be included to achieve a full understanding of the mechanism
and energetics of the reaction. In the following discussion, we
have limited ourselves to the results obtained using the catalyst
with the full dppp ligand.
Oxidative addition takes place through a three-center TS with
a relative Gibbs free energy of 152 kJ/mol and a C-H distance
of 1.548 Å (Figure 18, left). The two other key bond lengths
are 1.681 Å (Rh-H) and 2.109 Å (Rh-C). The identity of the
TS was confirmed by a full frequency analysis, which yielded
a single imaginary frequency of 629 cm-1. After oxidative
addition has taken place, a five-coordinate acyl-RhIII complex
is formed with a relative Gibbs free energy of 135 kJ/mol. Also
here, the free apical site can be occupied by uptake of an
additional CO molecule resulting in a decrease in enthalpy (35
kJ/mol), but a concomitant increase in Gibbs free energy of 20
kJ/mol indicates that this complex is not important under
catalytic conditions.
This complex can, in turn, undergo migratory extrusion
through a TS with a relative Gibbs free energy of 213 kJ/mol,
9
J. AM. CHEM. SOC. VOL. 130, NO. 15, 2008 5213

A R T I C L E S
Fristrup et al.
Figure 17. Energy profile for the decarbonylation of phenyl acetaldehyde.
Figure 18. Illustration of the three TSs in the decarbonylation of phenyl acetaldehyde. Left: Oxidative addition into the aldehydic C-H bond. Center:
Migratory extrusion of CO to form an octahedral RhIII intermediate. Right: Reductive elimination of toluene, which returns the active RhI(dppp)(CO)2
complex.
which is 7 kJ/mol lower than that for benzaldehyde (Figure 18,
center). Initially, a series of restricted minimizations were carried
out with a fixed C-C distance, which revealed that the lowest
energy conformation had the two reacting components in an
equatorial position. The key distances for the TS geometry are
C-C ) 1.955 Å, Rh-C(O) ) 1.931 Å, Rh-C(benzylic) )
2.579 Å, and the nature of the TS was confirmed by the
The calculated energy profiles can rationalize this difference
in reactivity, as the barrier for the rate-determining migratory
extrusion is 20 kJ/mol lower for phenyl acetaldehyde. In
comparison with the relative reactivities determined experimen-
tally, this difference is obviously too large, but we are satisfied
with having captured the correct trend in reactivities and note
that an error of about 10 kJ/mol is not surprising when one
considers the size of the calculated systems.
existence of a single imaginary frequency of 311 cm-1
.
After migratory extrusion, an octahedral RhIII complex is
formed (relative energy ) 137 kJ/mol) from which toluene can
be formed by reductive elimination (Figure 18, right). For this
process, the activation energy is merely 49 kJ/mol, which results
in an early TS with a C-H distance of 1.649 Å. The Rh-H
distance is 1.621 Å, whereas the distance from Rh to the
benzylic carbon is 2.395 Å. The imaginary frequency was 961
cm-1 for this TS.
Although the calculated energy profiles for benzaldehyde and
phenyl acetaldehyde are rather similar, there are also differences
which deserve attention. As mentioned earlier, in a direct
competition experiment, phenyl acetaldehyde reacted about 20
times faster than benzaldehyde, which is in contrast to the results
obtained when the two aldehydes are reacted in separate flasks.
In this case, there is also a marked difference is reactivity, which
results in about 6 times faster conversion of the phenyl
acetaldehyde.
For phenyl acetaldehyde, the five-coordinate intermediate is
stabilized by 17 kJ/mol compared to the TS for oxidative
addition, whereas this intermediate is only stabilized by 6 kJ/
mol when benzaldehyde is the substrate. The observed experi-
mental differences between reactions of both substrates, together
or in separate flasks, thus give an indication of a catalytic
reaction which is not completely governed by Curtin-Hammett
conditions. In conclusion, the observed differences in reactivity
between benzaldehyde and phenyl acetaldehyde can be ex-
plained by the calculated energy profiles, and in addition, the
experimental results can be used to pinpoint inaccuracies in the
calculated energies.
Deuterium Kinetic Isotope Effect. For benzaldehyde, the
experimental kinetic isotope effect (KIE) was determined to be
1.73 (kH/kD), whereas for phenyl acetaldehyde, the value was
1.77. The close resemblance for the two different substrates
suggests that the same elementary step is rate-determining for
9
5214 J. AM. CHEM. SOC. VOL. 130, NO. 15, 2008

Rhodium-Catalyzed Decarbonylation of Aldehydes
A R T I C L E S
both substrates. A KIE of this magnitude would suggest that
the bond to C-H(D) is at least partially broken in the rate-
determining step. In the oxidative addition, the C-H(D) bond
of the aldehyde is converted to a Rh-H(D) bond, which has a
significantly lower stretching frequency (C-H(D) ∼ 2900
(2100) cm-1, Rh-H(D) ∼ 2100 (1500) cm-1).24 Thus, any of
the three types of transition states could be candidates for the
rate-limiting step based on the observed KIE. The calculated
energy surfaces strongly suggest that either the migratory
extrusion or the reductive elimination can be rate-limiting, but
for completeness, we have determined theoretical isotope effects
for all three types of transition states. The ground state is either
the isolated reactants or the catalyst-aldehyde complex. Both
options have been tested, but since the two sets of results were
virtuallyidentical,onlythecalculationsusingthecatalyst-aldehyde
complex are shown here. As a first crude approximation, we
have examined the difference in zero-point energies, which
results in a calculated KIE for benzaldehyde of 4.00 at 298 K
when oxidative addition is considered to be rate-determining.
However, at elevated temperatures, the possibility of populating
the excited vibrational states must be included, which is possible
using the reduced partition function introduced by Bigeleisen
and Mayer,25 and used for similar systems by, for example, the
groups of Houk26 and Singleton.27
The calculation of frequencies and thermodynamic properties
for all intermediates and transition states allows the three
possibilities to be tested, by simply comparing the calculated
KIE at 440 K with the experimentally determined value. With
the Rh(dppp)(CO)PhCHO complex as resting state and the
assumption that oxidative addition is the rate-determining step,
the KIE can be calculated to be 2.85 (kH/kD) at 440 K. This
value is somewhat higher than the experimental value, and
therefore, another scenario was considered where the migratory
extrusion is the rate-determining step with the Rh(dppp)(CO)Ph-
CHO complex as a resting state. This resulted in a calculated
KIE of 1.80 at 440 K, which is in excellent agreement with the
experimental value of 1.73. On the other hand, if reductive
elimination is the rate-determining step, the calculated KIE is
2.81, which is almost identical to the value calculated when
oxidative addition is rate-determining, and still significantly
higher than the experimental value. In those cases where the
bond to H(D) is either broken or formed in the rate-determining
step, the calculated KIE is too high, whereas for the migratory
extrusion where the “probe atom” (i.e., H or D) is not directly
involved in the rate-determining elementary step, the calculated
KIE is very close to the experimental value. In a similar manner,
the KIE for phenyl acetaldehyde was calculated to be 1.79 for
the scenario where the resting state is Rh(dppp)CO(PhCH2CHO)
and the migratory extrusion of CO the rate-determining step.
This is in excellent agreement with the experimental value of
1.77, and also here oxidative addition and reductive elimination
as the rate-determining steps both lead to higher calculated KIEs
of 2.81 and 2.60, respectively.
(dppp) was investigated by experimental and theoretical meth-
ods. All available data indicate that the mechanism consists of
oxidative addition, migratory extrusion, and reductive elimina-
tion with migratory extrusion as the rate-determining step. The
overall good agreement between experiment and theory il-
lustrates that modern theoretical methods have matured to a point
where the results can be used to predict and to rationalize
experimental results. This opens up the possibility for a number
of new applications, such as in silico ligand screening and virtual
design of new synthesis routes.
Computational Details
Visualization and comparison of structures were performed in
Maestro version 7.5.106.28 For phenyl acetaldehyde, an initial,
simple molecular mechanics conformational search was carried out
which resulted in a single conformation with the plane of the phenyl
ring and the plane of the CdO moiety arranged perpendicular to
each other. DFT calculations were performed with the B3LYP
functional20 employing the LACVP* basis set, which uses the
Hay-Wadt effective core potential (ECP)21 and basis set for
rhodium, and the 6-31G* basis set for all other atoms. Solvation
was simulated using a polarized continuum model (PB-SCRF) with
parameters suitable for THF,29 which was deemed sufficiently
similar to the ethereal diglyme used in the experimental investiga-
tions.30 In the gas phase, intermediates and transition states were
characterized by a full, analytic frequency calculation, which
resulted in only positive frequencies for intermediates and exactly
one imaginary frequency for the transition states. In all cases, did
a visualization of this imaginary frequency correspond to the
expected normal mode for the elementary step under investigation.
A theoretical estimation of the kinetic isotope effect was
performed using the available Hessian from the full frequency
calculation by simply changing the isotope of the aldehydic H atom.
To allow comparison to the experimental value, the Gibbs free
energies were calculated for a temperature of 440 K.
Acknowledgment. Financial support from the Lundbeck
Foundation is gratefully acknowledged. The Center for Sustain-
able and Green Chemistry is supported by the Danish National
Research Foundation. We thank Dr. Fedor Zhuravlev and Dr.
Mårten Ahlquist for fruitful discussions.
Supporting Information Available: Detailed experimental
procedures, additional kinetic plots along with XYZ coordinates,
SCF energies, Gibbs free energies (298 and 440 K) for all
structures with the full experimental ligand. This material is
available free of charge via the Internet at http://pubs.acs.org.
JA710270J
(26) Houk, K. N.; Gustafson, S. M.; Black, K. A. J. Am. Chem. Soc. 1992,
114, 8565–8572.
(27) Singleton, D. A.; Wang, Z. J. Am. Chem. Soc. 2005, 127, 6679–6685.
(28) For updated versions, see: http://www.schrodinger.com.
(29) (a) Marten, B.; Kim, K.; Cortis, C.; Friesner, R. A.; Murphy, R. B.;
Ringnalda, M. N.; Sitkoff, D.; Honig, B. J. Phys. Chem. 1996, 100,
11775–11788. (b) Tannor, D. J.; Marten, B.; Murphy, R.; Friesner,
R. A.; Sitkoff, D.; Nicholls, A.; Ringnalda, M.; Goddard, W. A., III;
Honig, B. J. Am. Chem. Soc. 1994, 116, 11875–11882.
(30) The parameters for THF were used since these have been determined
with significantly better accuracy than the corresponding parameters
for diglyme. Calculations on the small model system in Figure 11 (L
) CO) using parameters for diglyme (ꢀ ) 7.0, F ) 0.94) gave results
which were within 1 kJ/mol of the ones obtained with the parameters
for THF.
Conclusion
The mechanism for the rhodium-catalyzed decarbonylation
of aldehydes in the presence of a bidentate phosphine ligand
(24) Kaesz, H. D.; Saillant, R. B. Chem. ReV. 1972, 72, 231–281.
(25) Bigeleisen, J.; Mayer, M. G. J. Chem. Phys. 1947, 15, 261–267.
9
J. AM. CHEM. SOC. VOL. 130, NO. 15, 2008 5215