Günter Helmchen*, Henning Steinhagen, Michael Reggelin‡
and Steffen Kudis
Organisch-Chemisches Institut, Universität
Heidelberg, D-69120 Heidelberg, FRG
‡Organisch-Chemisches
Institut, Universität Frankfurt D-60439 Frankfurt/Main, FRG
Abstract: Phosphinooxazolines were found to be highly effective ligands in Pd catalyzed asymmetric substitutions of allylic compounds. Malonates, amines and nitronates were employed as nucleophiles. Mechanistic aspects were studied by NMR and x-ray crystal structure analysis. Particular emphasis was placed on the identification of Pd-bound intermediates in order to gain a clear understanding of the enantioselective step in the catalytic cycle.
Pd catalyzed asymmetric C-C and C-N bond forming substitutions
at allylic compounds are being employed by many research groups
(ref. 1). These reactions essentially involve oxidative addition
of a Pd° fragment to a chiral, racemic allylic derivative
to yield a p-allyl complex (Scheme
1) that with a nucleophile furnishes a chiral product. It was
logical to try to achieve enantioselectivity in this preparatively
useful reaction with the help of chiral ligands L*. Remarkably,
the C2-symmetric chelate diphosphines giving excellent
results in hydrogenations, i.e., CHIRAPHOS, BINAP etc.,
gave disappointing results in allylic substitutions, particularly
so with cyclic allylic substrates. Only in the early 1990s, it
was demonstrated with bisoxazolines (ref. 2)
and new types of diphosphines (ref. 3)
that high enantioselectivity
is possible with proper combinations of a substrate and a C2-symmetric
ligand.
Scheme 1
In our own work we launched an attempt to develop Pd complexes
of non-C2-symmetric ligands for allylic substitutions.
Our approach was inspired by work of Faller, who had realized
remarkably high enantioselectivity in stoichiometric substitutions
at cyclopentadienyl-molybdenum complexes. Fallers success is due
to the use of ligands (CO, NO) with different electronic rather
than steric properties (ref. 4). In catalysis this approach would
involve the use of a chiral chelate ligand with two electronically
distinct donor centers. This was apparently first probed by Caesarotti
with the ligand PROLOPHOS with two slightly, by bonding to O or
N, differentiated P atoms (ref. 5). A fairly low level of enantioselectivity
was obtained. We felt that a more pronounced difference in electronic
as well as steric properties was required and, therefore, chose
the combination of a hard, N, and a soft, P, S or Se, donor. Realization
of this proposal relied on the proven usefulness of the oxazoline
moiety. Aryl groups were preferred as substituents at P because
triarylphosphines are normally stable to air. These considerations
led to the development of phosphinooxazoline (PHOX) ligands 1
(ref. 6).
The same concept was independently pursued by the groups of Pfaltz (ref. 7) , Williams (ref. 8) and, with a different type of P-N chelate ligands (QUINAP), J.M. Brown (ref. 9a).
Oxazolines are available from amino alcohols which can be prepared
from the chiral pool of natural amino acids (Scheme 2). There
are many established routes from amino alcohols to oxazolines
(ref. 10). Usually, one step procedures are employed
(ref. 11,12,13).
However, better yields are often achieved with a three step procedure
involving formation of an N-acyl amino alcohol, activation of
the OH group and ring closure with base (ref. 14).
Scheme 2
Introduction of phosphorus is described in Scheme 3 (ref. 14). Nucleophilic substitution of fluorine with a diarylphosphide proceeds with 70-90 % yield. In the case of stereogenic phosphorus, with, e.g., Ph and 1-naphthyl or 2-biphenylyl substituents, ca. 7:3 mixtures of diastereomers are formed which can be easily separated by flash chromatography or crystallization. Electrophilic phosphorus and also sulfur and selenium compounds can be reacted with the Grignard compound obtained from the corresponding bromo derivative and activated magnesium. Yields with halophosphines are only 30-50 %, but the P-diastereomers are formed with selectivity of 85:15. Configurations at phosphorus were determined by crystal structure analysis.
Scheme 3
Malonates are the nucleophiles most often used in allylic substitutions (ref. 6,7,8). In addition, a variety of other nucleophiles were investigated: amines (ref. 15), N-acylamides (ref. 15), nitro compounds (ref. 16), and sulfinates (ref. 17). These nucleophiles are less reactive than malonates; however, enantioselectivities are very similar and the steric course is grosso modo independent of the nucleophile.
Rationalizing the steric course of the nucleophilic substitution is difficult because there are two diastereomeric -allyl complexes, designated exo- and endo-isomer here (4x, 4n in Scheme 4). The products can be formed via four pathways and the preferred product can arise by reaction at the allylic C trans to P of the exo or cis to P at the endo isomer. For the decision between these possibilities, a postulate of Bosnich (ref. 18) was helpful. Based on the assumption of an early transition state, this postulate states that the more abundant isomer is the more reactive one (ref. 19). The more abundant is generally the exo isomer (see below). In conjunction with the known configuration of the products of allylic substitutions it is deduced that the nucleophile preferentially attacks the carbon trans to P (ref. 20).
There is so far no direct proof for this mechanistic proposal.
There is, however, support by circumstantial evidence from 13C
NMR shifts, NMR studies on the interconversion of exo and
endo diastereomers, and x-ray crystal structures (ref. 20).
The crystal structures, for examples see Figure 1, allowed
us to understand why exo isomers are more stable than endo
isomers. There are several important general observations: (i)
The "inner" chelate cycle PdNCCCP is non-planar. (ii)
A consequence of non-planarity is conformational nonequivalence
of the substituents at P, one is axially, the other equatorially
arranged. Aryl ring planes are nearly perpendicular to each other,
with the axial group pointing its edge, the equatorial its face
to the metal. (iii) The substituent of the oxazoline ring
occupies an axial position in a way that only the equatorial H
can interact with the allylic moiety. The dominating interaction
is the one with the equatorial aryl group at P. Minimization of
this interaction is the reason for preference of the exo
over the endo diastereomer.
Scheme 4
The analysis of x-ray structures of p-allyl
complexes was helpful for an understanding of equilibria of p-allyl
complexes and structural aspects important for selectivity. In
addition, the reaction course was studied by modern 2D NMR spectroscopic
methods in order to gain a more detailed insight into the mechanism
of enantiodiscrimination in the catalytic reaction (ref. 21).
 |  |
Figure 1. Front view (left) and side view (right) of the
x-ray crystal structure of complex 4x (R = iPr,
R' = Ph)
Initially the stoichiometric reaction between the p-allyl complexes 4 (R = iPr, R' = Ph; 10:1 mixture of 4x and 4n) and sodium dimethyl malonate as a nucleophile was examined. A sample containing the p-allyl complexes 4 and the nucleophile was prepared at low temperature and warmed up inside the NMR probehead to rt. The progress of the reaction was then monitored by 31P NMR spectroscopy (Figure 2). During the course of the reaction 4x and 4n are always in rapid equilibrium. As first new species a compound with a singlet at d = 11.18 ppm appears, whose concentration reaches a maximum already after 90 s before it is consumed whithin a few minutes. This compound is the Pd0 alkene complex 5a. Consumption of 5a is accompanied by the appearance of a new species with a characteristic AB spin system (dA korr = 9.41, dB korr = 12.91, 2JP,P = 128 Hz) which is the main component after a reaction time of ca. 1 h. We assign structure 7 to this long-lived intermediate. The precise geometry of this complex could not yet be determined.
Only traces of the metal-free product (S)-6, which is formed from 7, were detected in the reaction mixture when the reaction was stopped at a conversion of approximately 50 %. The concentration of (S)-6 only increased when conversion exceeded 50 %. Therefore, complex 7 is a stable by-product of the stoichiometric reaction. For the determination of the constitution of the mechanistically meaningful transient species 5a, the reaction was carried out with 13C3-labeled NaDMM at -20 to -30°C and stopped after 2 min by cooling to -78°C. A sample prepared in this way contained ca. 75 % of the phosphorus in the form of the Pd0 alkene complex 5a and was stable for several weeks at 78 °C in an inert atmosphere.
Figure 2. 31P{1H} NMR-Spectra recorded
at various reaction times (-60°C). A: equilibrating
p-allyl complexes 4x and 4n;
B, C, D: reaction mixtures 370 s, 500 s
and 1 h after addition of the sodium dimethyl malonate.
Assignment of the resonances of all NMR-active nuclei was possible
by use of 13C3-labeled malonate with a large
set of 2D NMR experiments (1H,1H-COSY, -TOCSY,
NOESY, ROESY, 13C,1H-HSQC, -HMBC,
-HMQC-TOCSY and 31P,1H-HMBC). By quantitative
analysis of NOE- and ROE data information on distances of H nuclei
could be obtained, which is only in accordance with conformer
5a and not with 5b (Scheme 5). For example, characteristic
is the NOE between 2-H and 17-H which is altered during the transformation
of 4x to 5a. The value of this NOE contact in 5a
corresponds to a reduction of the distance between 2-H and 17-H
by 1.4 Å, as compared with the distance in 4x
determined by x-ray diffraction analysis.
Scheme 5
Statements concerning the mechanism of the allylic substitution require the following plausible assumptions: (a) The attack of malonate at 4x under formation of 5a proceeds via a "least motion" reaction path, i.e., a rotation of 30° from 4x to 5a is the main process, (b) rotation of the alkene fragment relative to the N-Pd-P-plane in complex 5a is sufficiently slow so that equilibration between 5a and 5b is slow compared with the rate of their formation.
Assumption (b) is supported by experiments in which the reaction was carried out in a temperature gradient (-78 °C to rt) and monitored by 1H NMR spectroscopy. In this experiment broadening of the resonances of complex 5a, which would be expected for a dynamic exchange process, was not found. Accepting assumption (b), the configuration of the more reactive allyl complex 4x is conserved in the Pd0 alkene complex 5a. Considering the known absolute configuration of the product, an attack of the nucleophile trans to phosphorus at the exo p-allyl complex 4x can be derived. This result is in accordance with earlier interpretations (ref. 9,20a), but is here based on a precisely characterized Pd0 alkene product complex in the Pd-complex catalyzed allylic substitution.
The front view in Figure 1 shows quite clearly that the chiral
ligand mainly provides interactions at its wings. It appears likely
that allylic systems with big substituents, such as phenyl, should
display high exo-endo-ratios and enantioselectivity, but
narrow systems, with small substituents or cyclic compounds, might
give low selectivity. This is exactly what was found. In Scheme
6 substrates are ordered according to their "broadness"
and it is quite remarkable how closely the ee values parallel
the steric extension (the isopropyl case is taken from ref. 7).
The importance of this parameter is further underlined by NMR
data of the corresponding p-complexes:
ratios of 1.8:1, 4:1, and 9:1 for the cyclohexenyl, the 1,3-dimethyl-
and the 1,3-diphenylallyl derivative (CDCl3 solution),
respectively. The cause of enantioselectivity though is a kinetic
phenomenon, i.e., a function of differing reaction rates
at the allylic termini in exo and endo complexes.
Recent results indicate a significant difference of relative reaction
rates in acyclic and cyclic substrates with respect to exo
and endo isomers.
Scheme 6
The rather clear relationship between the size of the -allyl moiety
and enantioselectivity was very satisfactory, as it was in excellent
agreement with our mechanistic assumptions. However, the production
of racemic product from the cyclic substrate (cf. Scheme 6) was
somewhat unsatisfactory from a preparative point of view. Clearly,
a ligand was required that would reach into the narrow area directly
above or below the allylic sp2 centers. As such ligands
the biphenylyl derivatives A-C (Scheme 3) were conceived
(ref. 22). We were able to obtain
a high resolution x-ray crystal structure of the complex [Pd(3-C6H9)
(A)]SbF6
derived from ligand A (R = i-Pr), and indeed, in
the crystal conformer a of the
cyclohexenyl -allyl Pd complex is found in which the phenyl of
the 2-biphenylyl group is located directly above the allylic moiety
as described in Scheme 7.
Scheme 7
|
|
Despite the fact that conformer a is favored in the solid state enantioselectivities resulting with the ligand A (cf. Scheme 3) were not satisfactory. Distinct dependence on ring size of the substrate and, to a certain extent, on reaction conditions are apparent from the data given in Scheme 8. In order to enhance electronic effects, the ligand B with electron withdrawing CF3 groups was prepared. With this ligand improved enantioselectivity was obtained with methylene chloride as solvent.
Nevertheless, results were still not satisfactory. A hint towards
improvement was gained by an NMR analysis of the complex [Pd(h
3-C6H9)(A)]SbF6
which indicated the existence of several conformers in solution,
including the unfavorable conformer b
with the crucial phenyl group rotated away from the allylic moiety.
In order to destabilize conformers of this type, the cymantrene-based
ligand D (Scheme 7) was conceived and could be prepared
in a reasonably straightforward way (ref. 23). This ligand induces
excellent catalytic activity and displays long shelf life. Conformers
analogous to are apparently destabilized by interaction with
the manganese tricarbonyl group. High enantioselectivity with
this new ligand was indeed obtained.
Scheme 8
Method a: 1.5 eq LiCH(COOCH3)2, dioxane,
rt.
Method b: 2.5 eq CH2(COOCH3)2,
BSA method, methylene chloride, 0 °C.
Method c: 1.5 eq NaCH(COOCH3)2, dimethylformamide,
-50 °C to 0 °C.
Prior to the development of the new ligand D high enantioselectivities with cyclic substrates were achieved with salts of the easily available b-phosphinocarboxylic acid E as chiral ligand (ref. 24): enantiomeric excess of 85, 98 and >99 % ee for the 5-, 6- and 7-membered ring derivatives, respectively [with LiCH(COO-t-Bu)2 as nucleophile]. Products with S-configuration are formed with E.
Acknowledgements. This work was supported by the Deutsche Forschungsgemeinschaft (SFB 247) and the Fonds der Chemischen Industrie.
