Reevaluation of the mechanism of the Baylis-Hillman reaction: Implications for asymmetric catalysis

Article abstract of DOI:10.1002/anie.200462462

(Chemical Equation Presented) One step beyond: Proton transfer (step 3, see scheme), not C-C bond formation (step 2) as previously thought, is the rate-limiting step (RLS) in the initial stage of the Baylis-Hillman reaction, which involves the amine-catal

Full text of DOI:10.1002/anie.200462462

Communications
Baylis–Hillman Reaction
Reevaluation of the Mechanism of the Baylis–
Hillman Reaction: Implications for Asymmetric
Catalysis**
Varinder K. Aggarwal,* Sarah Y. Fulford, and
Guy C. Lloyd-Jones*
[
1]
The Baylis–Hillman reaction (BHR, see Scheme 1) involves
[
2]
the amine- (or phosphine-) catalyzed addition of an
aldehyde to an activated alkene, such as an acrylate, 1, to
[
3]
generate an allylic alcohol of type 6. The commonly
[
4]
accepted mechanism for this process involves reversible
conjugate addition of the nucleophilic amine catalyst to the
activated alkene to generate an enolate (step 1), nucleophilic
attack of the enolate on the aldehyde to generate a second
zwitterionic intermediate (step 2), and then elimination
(step 3) to generate the product and liberate the amine
catalyst (Scheme 1). The reaction shows pseudobimolecular
kinetics, and the rate-limiting step (RLS) of the reaction has
been determined as step 2 on the basis that no primary kinetic
[
4a]
isotope effect (KIE) was observed.
Protic solvents are
[
5]
known to accelerate the BHR, and it has been proposed that
this acceleration occurs through activation of the aldehyde by
hydrogen bonding to thereby promote step 2 (see intermedi-
ate A, Scheme 1). However, hydrogen bonding to the
aldehyde would have to compete with the enolate, which is
a much better hydrogen-bond acceptor. Indeed, this more
thermodynamically favorable interaction will stabilize the
enolate and render it less reactive and so should slow down
Scheme 1. Proposed mechanism of the Baylis–Hillman reaction with
potential for autocatalysis through A or B.
[*] Prof. V. K. Aggarwal, S. Y. Fulford, Prof. G. C. Lloyd-Jones
School of Chemistry
Bristol University
Cantock’s Close, Bristol BS81TS (UK)
Fax: (+44)117-929-8611
E-mail: v.aggarwal@bristol.ac.uk
guy.lloyd-jones@bristol.ac.uk
[**] We thank Bristol University and the EPSRC for financial support
Supporting information for this article is available on the WWW
under http://www.angewandte.org or from the author.
1
706
ꢀ 2005 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
DOI: 10.1002/anie.200462462
Angew. Chem. Int. Ed. 2005, 44, 1706 –1708

Angewandte
Chemie
[
8]
the reaction, albeit with an accumulation of the hydrogen-
bonded enolate. Interestingly, in the absence of proton donors
the reaction shows autocatalysis, presumably because the
product can act as a hydrogen-bond donor and promote the
kinetics of the reaction were readily simulated by the use of
two simple models, one (see A, Scheme 1) that follows the
conventional mechanism in which the product catalyzes the
addition of the enolate 3 to the aldehyde 4, and the second
(see B, Scheme 1) in which the product catalyzes a rate-
limiting breakdown of the zwitterionic intermediate 5, for
example, by proton transfer.
[
6]
reaction. These issues made us consider an alternative
hypotheses for the role of the protic solvent, including the
possibility that it could accelerate formal proton transfer from
the a-keto methine to the alkoxide in the zwitterion 5 to
facilitate liberation of the product and regeneration of the
catalyst through elimination (see intermediate B, Scheme 1).
We now show that step 3 is the RLS in the initial phase of the
reaction and that once the concentration of product has built
up, step 2 becomes the RLS. The consequences of these
findings with respect to asymmetric catalysis are also dis-
cussed.
The two mechanistic scenarios (A and B) can, in principle,
be distinguished on the basis that upon employing a-
2
deuterated acrylate (2-[ H ]-1; = d -1),
a primary KIE
1
1
should be completely absent in A but evident in B prior to
the autocatalyzed breakdown of zwitterion 5 becoming more
efficient than its generation. As the latter condition might
only be fulfilled very early in the reaction, the comparison of
absolute rates over a number of half-lives, as performed by
Isaacs and co-workers with deuterated (> 99%) acrylonitri-
First, we examined the profile of the reaction of ethyl
acrylate with benzaldehyde catalyzed by quinuclidine in
which the onset of significant autocatalysis was readily
[
4a]
le,
is unlikely to be informative. In fact, a competition
experiment between d-1 and 1 would clearly identify the RLS
because we would expect a primary KIE to increase the mole
fraction of d-1 in the acrylate (xd-1) if step 3 is rate-limiting and
a secondary KIE to decrease xd-1 if step 2 is rate-limiting. We
thus analyzed the effect of C(2)-deuteration by H NMR
spectroscopic analysis of the BHR of approximately equimo-
lar mixtures of d-1 and 1 (xd-1 = 0.505 Æ 0.005, Figure 1). As
evident from Figure 1b, in the early stages of reaction (up to
[
6]
apparent between 0 and 20% conversion. Study of the
1
reaction mixture by H NMR spectroscopy (600 MHz)
[
9]
showed that there were no species (< 1%) other than the
substrates, catalyst, and product throughout the entire
reaction. Thus all equilibria prior to thermodynamically
favorable liberation of 6 must lie predominantly on the side
1
[
4b]
of the starting materials (see Scheme 2).
ꢀ
20% conversion), the mole fraction of d-1 increases
substantially. As reaction proceeds further into the phase
where autocatalysis is by far the dominant process, the ratio
[
9]
then stabilizes (xd-1max = 0.55). This shows that in the early
phase (< 20% conversion), step 3 is rate-limiting (primary
KIE evident). By using model B as a starting point, we were
able to simulate the kinetics of the competition experi-
[
8,9]
ments;
lines) and observed (circles) kinetics in Figure 1, graphs a
for example, see the comparison of the predicted
Scheme 2. Proton-transfer mechanism.
(
and c. For satisfactory simulation, the model required
incorporation of a substantial primary KIE (k /k = 5 Æ 2)
H
D
The rate acceleration was confirmed to arise from
autocatalysis owing to the protic nature of the product 6 by
the absence of such an induction period in control reactions
for the noncatalyzed step 3. The fact that the simplification of
k /k = 1 for autocatalysis of step 3 allows a satisfactory
H
D
simulation suggests that the autocatalysis causes a change in
[
7]
[9]
with catalytic quantities of 1) the product or 2) MeOH. The
the RLS from step 3 to step 2 early in the reaction.
2
Figure 1. a) Evolution of the Baylis–Hillman reaction of PhCHO (4, 4.0m) with [C(2)- H0.51]-ethyl acrylate (1/d-1, 4.3m) catalyzed by quinuclidine
(
2, 1.0m) at room temperature to give a mixture of O-d-6 and 6. Lines through data points are kinetic simulations based on models A and B with
[
7]
refinement for contraction (see Supporting Information for full details). b) Relationship between conversion (based on PhCHO) and the mole
fraction d-1 (xd-1 =[dÀ1]/[1+d-1]). c) Simulated variation of the concentrations of 1 and d-1 ([1] and [d-1]) in the first 40% of the reaction by
employing model B with k /k =4.8 for step 3 (non-autocatalyzed).
H
D
Angew. Chem. Int. Ed. 2005, 44, 1706 –1708
www.angewandte.org
ꢀ 2005 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
1707

Communications
The efficient autocatalysis of step 3 by a simple hydroxyl
[
1] A. B. Baylis, M. E. D. Hillman, Offenlegungsschrift 2155113, U.S.
Patent 3,743,669, 1972, [Chem. Abstr. 1972, 77, 34174q].
2] K. Morita, Z. Suzuki, H. Hirose, Bull. Chem. Soc. Jpn. 1968, 41, 2815.
3] For reviews, see: a) D. Basavaiah, A. J. Rao, T. Satyanarayana,
Chem. Rev. 2003, 103, 811–892; b) P. Langer, Angew. Chem. 2000,
112, 3177–3180; Angew. Chem. Int. Ed. 2000, 39, 3049 – 3052;
c) E. Ciganek, Org. React. 1997, 51, 201–350; d) D. Basavaiah,
P. D. Rao, R. S. Hyma, Tetrahedron 1996, 52, 8001–8062; e) S. E.
Drewes, G. H. P. Roos, Tetrahedron 1988, 44, 4653–4670.
4] a) J. S. Hill, N. S. Isaacs, J. Phys. Org. Chem. 1990, 3, 285 – 288;
b) L. S. Santos, C. H. Pavam, W. P. Almeida, F. Coelho, M. N.
Eberlin, Angew. Chem. 2004, 116, 4430 – 4433; Angew. Chem.
Int. Ed. 2004, 43, 4330 – 4333.
moiety is easily accommodated by the model outlined in
Scheme 2 which involves a six-membered proton transfer
from ROH to the alkoxide with concomitant deprotonation
of the a-methine to give an enolate, followed by elimination
[
[
[
10]
of the amine (E1cB). The intramolecular four-membered
direct proton transfer is presumably disfavored owing to
strain induced in attaining the appropriate eclipsed confor-
mation of the C(O)ÀC(H) bond and the transfer angle
[
between O-H-C which is far from optimal. Moreover, the
approximately 908 transfer angle between O-H-C in the
transition state is expected to limit the primary KIE to
roughly 2.3, a much lower value than that observed in the non-
[5] a) F. Ameer, S. E. Drewes, S. Freese, P. T. Kaye, Synth. Commun.
988, 18, 495 – 500; b) I. E. Markꢀ, P. R. Giles, N. T. Hindley,
[
11]
1
autocatalytic phase of the reaction.
Tetrahedron 1997, 53, 1015 – 1024; c) Y. M. A. Yamada, S.
Ikegami, Tetrahedron Lett. 2000, 41, 2165 – 2169.
In summary, these studies have shown that in the absence
of added protic species, the initial stage of the BHR involves
rate-limiting proton transfer (step 3). As the product concen-
tration builds, proton transfer becomes increasingly efficient
and the RLS step is then step 2, as in the conventional model.
This finding has considerable implications for asymmetric
catalysis of the BHR. The successful catalysts to date (> 80%
ee) have hydrogen-bond donors appended at some point to
[
[
6] V. K. Aggarwal, I. Emme, S. Y. Fulford, J. Org. Chem. 2003, 68,
692 – 700.
7] The reactions display a noticeable contraction in volume as they
proceed. However, the nonlinear relationship between the
conversion and the concentrations of all components is not the
origin of the induction period. See Supporting Information for a
full discussion.
[
12]
[8] The kinetics were simulated using MacKinetics (Leipold Asso-
ciates, USA); see Supporting Information for full discussion.
[
the nucleophile. It is quite likely that all four diastereomers
of the intermediate alkoxide are formed, but only one has the
hydrogen-bond donor suitably positioned to allow fast proton
9] The magnitude of the secondary KIE on any individual step in
the reaction of 1/d-1 would be small—typically 1.15 ꢁ k /k ꢁ
H
D
[
13]
transfer. The other diastereomers revert back to starting
materials and eventually the reaction filters through the
pathway that leads to fast elimination (Scheme 3). The low
0.87—but the net effect across equilibria may be larger and for
steps 1 and 2 is expected to favor the reaction of d-1 (i.e. k
/k
<
H
D
2
3
1) owing to a change in hybridization at C(2) from sp in 1 to sp
in the zwitterion 5. However, control experiments (see Support-
ing Information) revealed a slow background exchange of H and
D between 1/d-O-6 and d-1/6 catalyzed by quinuclidine and
presumably through the enolate 3/d-3. This observation com-
promises any meaningful analysis of the decrease in xd-1 observed
in the later phase of the reaction when significant 6/d-O-6 has
accumulated.
[
10] The question remains as to what mediates step 3 in the absence
of significant quantities of product. The possibility that a second
molecule of quinuclidine acts as a base was eliminated by a study
of the effect of doubling the catalyst loading which caused only
an approximately 1.75-fold increase in rate in the crucial early
stages of reaction—exactly as would be predicted on the basis of
first-order dependency on each of the three reaction components
when the dilution of aldehyde and alkene caused by increased
catalyst loading is taken into account. We therefore suggest that
traces of protic species, for example, water, enol, etc., may well
be sufficient to initiate reaction. A hemiacetal anion intermedi-
ate (derived from 5 and 4) has been proposed by McQuade and
co-workers to effect proton transfer. See Reference [14].
Scheme 3. Likely origin of enantioselection in the Baylis–Hillman
1
–4
reaction. D are diastereomers of the alkoxide adduct.
success rate in the design of chiral catalysts for the BHR could
be because the focus has been on controlling the stereo-
chemistry of the CÀC bond in the RLS (step 2) of the
reaction. On the basis of the above study, we now believe that
the positioning of suitable hydrogen-bond donors for selec-
tive proton transfer of one of the alkoxide diastereoisomers,
and not the others, is likely to more successful. The alkoxide
diastereomer that undergoes the fast, selective proton-trans-
fer reaction may also be the diastereomer that is preferen-
tially formed, but this is not a prerequisite. As a caveat, the
use of an aprotic solvent may be crucial for attaining high
enantioselectivity, and enantiomeric excesses could be
[11] R. A. More OꢁFerrall, J. Chem. Soc. B 1970, 785.
[12] a) Y. Iwabuchi, M. Nakatani, N. Yokoyama, S. Hatakeyama, J.
Am. Chem. Soc. 1999, 121, 10219 – 10220; b) K.-S. Yang, W.-D.
Lee, J.-F. Pan, K. Chen, J. Org. Chem. 2003, 68, 915 – 919; c) M.
Shi, L.-H. Chen, Chem. Commun. 2003, 1310 – 1311; d) N. T.
McDougal, S. E. Schaus, J. Am. Chem. Soc. 2003, 125, 12094 –
12095; e) J. E. Imbriglio, M. M. Vasbinder, S. J. Miller, Org. Lett.
2003, 5, 3741 – 3743.
[
14]
decreased by competitive nonselective autocatalysis.
[13] Such a model to account for asymmetric induction has been
suggested by Hatakeyama, see Reference [12a].
[
14] Note added in proof: McQuade and co-workers have independ-
ently found a substantial kinetic isotope effect during the initial
phase of the Baylis–Hillman reaction (< 10% conversion). See:
K. E. Price, S. J. Broadwater, H. M. Jung, D. T. McQuade, Org.
Lett. 2005, 147 – 150.
Received: October 29, 2004
Published online: February 3, 2005
Keywords: asymmetric catalysis · autocatalysis · kinetics ·
.
proton transfer · reaction mechanisms
1
708
ꢀ 2005 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
Angew. Chem. Int. Ed. 2005, 44, 1706 –1708