to the exclusive formation of enyne 5 (Table 1, entry 1, 84%,
E/Z = 58:42). On repeating the reaction in the presence of
arylboronic acid, the desired reactivity was restored, with 3d
and 3e leading to a complementary 9:1 mixture of THF
products 4d and 4e respectively (Table 1, entries 2 and 3). In
each case, the major diastereomer was formed with retention
of configuration of the propargylic stereocenter, as evidenced
by NOE enhancements between H2 and H5 in 4e (arising
from carbonate 3e) which were absent in 4d.[8] To investigate
whether the boronic acid was acting as a Brønsted or Lewis
acid, we conducted a screen of other additives. Whilst
trimethyl borate delivered 4d with high stereoselectivity
(Table 1, entry 4), its ability to prevent enyne formation was
reduced, suggesting a purely Lewis acidic role to be unlikely.
Boric acid itself was also able to mediate THF formation,
albeit with reduced selectivity (Table 1, entry 5). In contrast,
PPTS proved most effective in terms of yield and selectivity
(Table 1, entry 6); reducing the loading of PPTS and lowering
the reaction temperature (to 508C) led to highly diastereo-
selective cyclizations for both 3d and 3e (Table 1, entries 7
and 8).
With these encouraging results in hand, we turned our
attention to the nature of the propargylic leaving group. It was
clear that if a cyclic carbonate is used, the hydroxy groups of
its diol precursor need not be differentiated; however, its
synthesis is likely restricted to seven-membered rings, which
in turn limits the ring size of the product oxacycle. We
therefore decided to investigate an acyclic carbonate leaving
group, with the internal nucleophile now an alcohol rather
than the alkoxide we presumed to be generated in situ from
the cyclic carbonate. We selected acyclic carbonate 6a
(Table 1) as a test substrate for this chemistry, where Noyori
transfer hydrogenation[10] was used to install the propargylic
stereocenter (98% ee).[8] To our surprise, no reaction was
observed under the optimized conditions developed for the
cyclic carbonates (i.e. PPTS as additive; Table 1, entry 9);
however, reaction in the presence of boric acid gave the
tetrahydrofuran 4a in quantitative yield but a disappointing
20% ee (Table 1, entry 10).
We hypothesized that the surprising difference in reac-
tivity and stereoselectivity between the cyclic and acyclic
carbonates might arise from the more reactive nature of the
alkoxide generated from the cyclic carbonates (compared to
the alcohol in 6a). It has been shown that bidentate phosphine
ligands can reduce the extent of stereochemical erosion in
propargylic cross-coupling reactions,[6a,11] and we were
pleased to find that the combination of [Pd(dba)2] or Pd-
(OAc)2 with a range of these ligands led to consistently high
stereoselectivity (ꢀ 96% ee; Table 1, entries 11–14). Despite
this success, these reactions suffered from the formation of
varying amounts of enol ether 7, a byproduct which likely
forms through competitive attack of the alcohol on the central
carbon atom of the allenylpalladium(II) intermediate. For
reasons which remain unclear, the production of enol ether 7
could be minimized by using dppf as ligand (Table 1,
entries 13 and 14) and eliminated entirely in the presence of
boric acid (Table 1, entry 15).
Scheme 2. Proposed cyclization mechanisms.
SN2’ oxidative addition[6c,12] leads, following loss of CO2 and
protonation, to the allenylpalladium intermediate 8, which
can form the major furan product syn-4 through an anti-SN2’-
type reductive elimination (path A). It is well-known that
allenyl–palladium complexes can undergo syn-facial 1,3-
migrations to give complexes of type 9,[13] in this case
potentially stabilized by the proximal nucleophile. Syn
reductive elimination from this species would generate the
minor isomer anti-4. From either intermediate 8 or 9, a
competing hydride elimination, possibly mediated by the
alkoxide, could give enyne 5. For the acyclic carbonates, the
high stereoselectivity observed when bidentate phosphine
ligands are used may be explained by the intermediacy of the
cationic h3-allenylpalladium(II) complex 10a,[14] which not
only reduces the propensity for 1,3-migration,[6a,11] but also
renders the attack of the alcohol stereospecific (path B).
However, this intermediate also accounts for the formation of
enol ether 7, which arises from competitive attack of the
alcohol on the central allene carbon of 10a (path C);[14]
indeed, enol ether 7 has been prepared in this manner.[15] In
fact, the formation of propargylic substitution products from
attack of heteroatoms on the terminal allene carbon of
complexes such as 10 is rarely observed.[16] This outcome is
likely due to the relative positioning of the palladium and
nucleophile in the allenylpalladium intermediate, and the
resulting trajectory of nucleophilic attack: In the majority of
reported cases,[14,15] the palladium is positioned on the allene
carbon proximal to the nucleophile, and the trajectory of
attack is perpendicular to this alkene (path D, complex 10b),
whereas in our system, the palladium is located on the distal
allene carbon, and central carbon attack takes place via an
orthogonal trajectory (path C). We speculate that the role of
the acidic additives in preventing the formation of 5 or 7 may
be to “soften” the alkoxide nucleophile (for example through
formation of a borate complex, as proposed by Trost et al. in
related allylic etherifications).[17]
With optimized sets of reaction conditions established, a
selection of cyclic and acyclic carbonates were prepared to
explore the scope of the reaction for asymmetric oxacycle
synthesis (Table 2).[8,18] Pleasingly, the monosubstituted cyclic
Possible mechanisms for the two cyclizations are illus-
trated in Scheme 2. For the cyclic carbonate syn-3, initial anti-
Angew. Chem. Int. Ed. 2011, 50, 11506 –11510
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Daniels, David S. B.
