Communication
warding 87% ee (entry 2). This strongly suggested that the
moderate de in the case of 26 is not due to the substitution
pattern of the phenyl ring but due to the structure of the
Grignard reagent 14. That there are boundaries at the substitu-
tion pattern on the aromatic ring became clear in a parallel
study. When the methyl protecting groups in 13 were replaced
by tert-butyldimethylsilyl groups leading to enone 28, this sub-
strate did not react under standard 1,2-addition conditions
(entry 3). Also 29 was prepared as the precursor for (R,R,R)-a-
tocopherol 1, but the product of the addition of 14, 30, was
almost racemic (entry 4).
de in the subsequent 1,2-addition, probably because the steric
bulk was not directly positioned at the phosphorus center.
Therefore we had to conclude that 73% de was the maximum
achievable stereoselectivity at present.
With the most optimal ligand L9, we produced the key
chiral tertiary alcohol 26 in 73% de and 93% yield. Straightfor-
ward debromination of 26 with tBuLi at À788C for 0.5 h afford-
ed 31 in 91% yield (Scheme 7).[42] Reduction of the double
These studies forced a re-evaluation of the applied chiral
ligand, rev-Josiphos L1. In the development of the catalytic
asymmetric 1,2-addition of Grignard reagents we had already
experienced that L1 was unique in its chiral induction. Both
closely related ligands such as the parent Josiphos, and unre-
lated ligands such as BINAP, and phosphoramidites performed
badly. These studies were carried out with isobutylmagnesium
bromide as the nucleophile.
We therefore studied various chiral ligands, including the
commercial ligands L2–L4 and L8 and L9 and the ligands L5–
L7 prepared for this purpose,[39,40] in combination with
Grignard reagent 14. Most ligands performed less well com-
pared to rev-Josiphos L1 (entries 5–10). Josiphos-type ligand
L8, being an exception, afforded a virtually identical de as L1
(entry 11). Ligand L8 bears a sterically demanding di-tert-butyl
phosphine group in combination with the Josiphos-like ar-
rangement of a dialkyl phosphine on the ethyl branch and
a diaryl phosphine on the ferrocene ring. In our experience,
the enantioselectivity of the copper-catalyzed 1,2-addition
profits from rev-Josiphos type ligands, that is, a dialkyl phos-
phine on the ferrocene and a diaryl phosphine ethyl branch.
Following this idea, we studied commercial ligand L9. To our
delight, a significant improvement of the diastereoselectivity
to 73% was observed in the 1,2-addition of Grignard reagent
14 to 13 (entry 12). So, the three rev-Josiphos-type ligands L7,
L1, and L8, gave us a clear hint to increase the de in the 1,2-
addition of 14 to 13, by increasing the steric bulk at the ferro-
cene phosphorus substituent of rev-Josiphos type ligands. This
lured us into an attempt to prepare several new rev-Josiphos
type ligands with sterically very hindered phosphorus substitu-
ents on position 2 of the ferrocenyl ring (Figure 2). A consider-
Scheme 7. Synthesis of (R,R,R)-g-tocopherol. a) 14, CuBr·SMe2, L9, TBME,
À788C, overnight, 93%; b) tBuLi, Et2O, À788C, 0.5 h, 91%; c) flavin catalyst,
O2, N2H4·H2O, EtOH, r.t., overnight; 90%; d) i. Ce(NH4)2(NO3)6, THF/H2O, 08C,
0.5 h; ii. Na2S2O4, acetone/H2O, r.t., 0.5 h; iii. p-TSA, toluene, 608C, 5 min,
72% over three steps. p-TSA=p-toluenesulfonic acid.
bond in 31 turned out to be a showcase for flavine-catalyzed
diimide reduction.[43] Heterogeneous transition metal catalysts,
for example, Pd/C,[44] Pt/C, PtO2,[45] and Pd/C/NaOAc,[10c] in com-
bination with H2 invariably provided the hydrogenolysis prod-
uct and also the recently disclosed diimide reduction with
FeCl3 as the catalyst suffered from hydrogenolysis.[46] Flavine-
catalyzed double bond reduction of 31 afforded 32 as the only
product in 90% yield.[47] To prepare for the ring-closing step,
32 was oxidized to the corresponding quinone by treatment
with cerium(IV) ammonium nitrate, followed by subsequent re-
duction to afford hydroquinone 11. Finally, acid-catalyzed and
oxygen-induced cyclization of 11 provided the desired (R,R,R)-
g-tocopherol 3 in 72% yield over three steps.[22b,c] The synthet-
ic material was identical in all aspects (1H and 13C NMR, mass
analysis) with the reported data.[48]
Figure 2. Designed rev-Josiphos type Ligands.
In summary, we developed an efficient synthesis of (R,R,R)-g-
tocopherol based on copper-catalyzed asymmetric 1,2-addi-
tion. Starting from commercially available 2,3-dimethyl hydro-
quinone and phytol, (R,R,R)-g-tocopherol was prepared in 12
steps (longest linear sequence), 36% overall yield and 73% de
at the C2 chiral center. The synthesis is not misplaced in the
current collection of catalytic asymmetric approaches to the
tocopherols, as the route is straightforward, in particular in its
introduction of chirality at C2, and its use of readily available
able effort was invested in the preparation of L1-type ligands
with R=EtMe2C, Et3C, iPrMe2C, and adamantyl according to lit-
erature procedures for related ligands.[41] However, the cou-
pling of the R2PCl reagent with the ortho-lithiated ferrocene in-
variably failed, probably due to this (desired) steric hindrance.
Ligand L7 was accessible by this method but afforded a low
Chem. Eur. J. 2014, 20, 14250 – 14255
14253
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim