Combining Q2MM modeling and kinetic studies for refinement of the osmium-catalyzed asymmetric dihydroxylation (AD) mnemonic

Article abstract of DOI:10.1016/j.jorganchem.2005.11.009

The interactions between the substrate and the ligand in the Sharpless AD reaction have been examined in detail, using a combination of substrate competition experiments and molecular modeling of transition states. There is a good agreement between comput

Full text of DOI:10.1016/j.jorganchem.2005.11.009

Journal of Organometallic Chemistry 691 (2006) 2182–2198
www.elsevier.com/locate/jorganchem
Combining Q2MM modeling and kinetic studies for refinement
of the osmium-catalyzed asymmetric dihydroxylation (AD) mnemonic
Peter Fristrup, Gitte Holm Jensen, Marie Louise Nygaard Andersen,
*
*
David Tanner , Per-Ola Norrby
Department of Chemistry, Technical University of Denmark, Building 201 Kemitorvet, DK- 2800 Kgs. Lyngby, Denmark
Received 29 September 2005; received in revised form 7 November 2005; accepted 7 November 2005
Available online 13 December 2005
Abstract
The interactions between the substrate and the ligand in the Sharpless AD reaction have been examined in detail, using a combination
of substrate competition experiments and molecular modeling of transition states. There is a good agreement between computational and
experimental results, in particular for the stereoselectivity of the reaction. The influence of each moiety in the second-generation ligand
(DHQD)2PHAL on the rate and selectivity of the reaction has been elucidated in detail.
Ó 2005 Elsevier B.V. All rights reserved.
Keywords: Osmium; Asymmetric synthesis; Dihydroxylation; Molecular modeling; Kinetics
1. Introduction
a benchmark for research in the burgeoning field of cata-
lytic asymmetric synthesis, and has found extensive use in
target-oriented synthesis [4]. An empirical mnemonic
device was introduced (Fig. 1, left) [5] and later refined
(Fig. 1, middle) [6] for prediction of absolute stereochemis-
try, and a number of mechanistic investigations have
appeared, including kinetic [6,7], computational [8–11],
and isotope effect [12] studies. We recently published an
updated version of the AD mnemonic (Fig. 1, right) [13]
and in the current paper we describe how a combination
of Q2MM modeling [14] and kinetic measurements can
be used to further refine our understanding of the factors
underlying the selectivity in the AD reaction.
The revised mnemonic by Sharpless and coworkers
(Fig. 1, middle) [6], based upon kinetic [6] and modeling
[9] studies, importantly identified that an area of the cata-
lytic complex could provide stabilization for the substrate,
and thus rationalized the phenomenon of ligand acceler-
ated catalysis (LAC) [15] that had been observed in the
AD reaction. Our work on substrate selectivities extended
this concept and suggested that two separate areas provide
independent stabilization in two different regions of the
ligand [13]. The first of these is identical to the one
In 1980, Hentges and Sharpless published a seminal
paper describing asymmetric induction in the reaction of
osmium tetroxide with alkenes [1]. The process required a
stoichiometric amount of osmium, and relied on the
‘‘pseudoenantiomeric’’ ligands dihydroquinine acetate
and dihydroquinidine acetate, respectively, for induction
of chirality in the diol products (25–90% ee). This relatively
humble beginning was followed by a major breakthrough
in 1988, when the reaction was made catalytic by Sharpless
and coworkers by use of N-methyl-morpholine-N-oxide as
a reoxidant [2]. Rapid development of the process then fol-
lowed, resulting in new ligands for acceleration of the reac-
tion and experimental protocols which are operationally
simple, provide excellent chemical yields and enantioselec-
tivities for a wide range of 1,2-diols at low catalyst load-
ings, and equally easy access to both enantiomers of a
given chiral diol [3]. The Sharpless osmium-catalyzed
asymmetric dihydroxylation (AD) reaction thus serves as
*
Corresponding author. Tel.: +45 4525 2123; fax: +45 4593 3968.
E-mail address: pon@kemi.dtu.dk (P.-O. Norrby).
0022-328X/$ - see front matter Ó 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.jorganchem.2005.11.009

P. Fristrup et al. / Journal of Organometallic Chemistry 691 (2006) 2182–2198
2183
Fig. 1. Left: The original mnemonic device proposed by Sharpless and coworkers [5]. Middle: The revised version of the mnemonic device proposed by
Sharpless and coworkers [6]. Right: The modified mnemonic device recently proposed by Norrby and coworkers [13].
identified by Sharpless and coworkers, caused by interac-
tions between alkene substituents and the aromatic linker
of the AD-ligand [9], whereas the second corresponds to
the interactions with the quinoline moieties first postulated
by Corey et al. [16] (Fig. 2).
For simple systems, it has been shown that fair predic-
tions of experimental selectivity can be obtained by man-
ual selection of a few transition state structures followed
by computational optimization to the transition states
and comparison of the energies thus calculated [10]. How-
ever, to obtain quantitative predictions for flexible sys-
tems, it is necessary to sample the space of all possible
transition states, either by regular conformational search-
ing or molecular dynamics [11,13], followed by Boltzmann
population analysis. The number of conformations to be
sampled calls for a relatively simple and fast computa-
tional method. At the present level of computer technol-
ogy, this requires the use of molecular mechanics force
fields, either in conjunction with a DFT representation
for the reaction center (QM/MM), or by employing force
fields designed to represent the transition states accu-
rately. In the current work, we have chosen the latter
approach, utilizing TS force fields based on our own
Q2MM methodology [14], which has previously been
shown to give highly accurate results for the AD reaction
[11].
2. Experimental results
Recently, we have published a limited study of selectiv-
ity and reactivity of two groups of tri-substituted alkenes
[13]. This study led to the development of a modified mne-
monic device, which is able to predict the absolute config-
uration of the diol product (Fig. 1, right).
Fig. 2. Illustration of the two attractive pockets described by Sharpless (left) and Corey (right). The TS structures of 2a (left) and 4b (right) have been
oriented in accordance with the Sharpless mnemonic device.

2184
P. Fristrup et al. / Journal of Organometallic Chemistry 691 (2006) 2182–2198
Here we wish to expand the study to include two new
thus orient the large substituent into the NE corner, where
it can interact favorably with the PHAL linker, the Sharp-
less pocket [6,9]. Similarly, the large substituents of the b
alkenes go into the SW corner where they can be stabilized
by the ligand quinolines, the orientation suggested by
Corey and Noe [16]. Finally, the substituents of the c alk-
enes must go into the NW corner, where the Sharpless
mnemonic predicted a minor steric interference [5,6],
whereas our modified mnemonic simply predicts an
absence of stabilizing interactions [13]. With the longest
chains (alkenes 3 and 4), we expect that the flexibility of
the substituent will allow it to orient itself into any proxi-
mal stabilizing area of the ligand, and thus the difference
between substituents should be lower than for the more
rigid alkenes 1 and 2.
The relative rates of reactivity within each class of alk-
enes are determined by pair-wise competition experiments,
where the rate of disappearance of the alkenes is followed
by GC, as described previously [13]. Since the reaction is
first order in each alkene, and of equal order for all other
reagents, the relative rate of any two alkenes in competition
is trivially obtained as the slope in a plot of ln([alkene]0/
[alkene]) for one alkene against the other. The plots
obtained for all four groups of alkenes reacted with stoichi-
ometric OsO4 in toluene are shown in Figs. 4–7. Within
each group of isomeric alkenes, one was selected as refer-
ence compound either since it had a reactivity in between
that of the other two alkenes or simply due to ease of
separation (GC).
groups of tri-substituted alkenes, as well as a wider set of
reaction conditions. All alkenes included in the current
study are shown in Fig. 3.
The alkenes from the previous study have been included
in the discussion to present a complete picture of the selec-
tivity-determining interactions in the AD reaction. The
tri-substituted alkenes are ideal probes in this particular
reaction, since the large steric crowding in the so-called
SE corner of the mnemonic device (Fig. 1) only allows
the orientation where the vinylic hydrogen of the substrate
is pointing towards the SE corner. Alkenes denoted a will
Ph
Ph
Previously
published
Ph
1a
2a
3a
4a
1b
2b
3b
4b
1c
2c
Ph
Ph
This study
Ph
Ph
Ph
Ph
Previously
published
3c
4c
Ph
Ph
This study
Ph
For the first two groups of alkenes the differences in
reactivity within the group was rather large. This called
for the use of the alkene with an intermediate reactivity
Fig. 3. The four sets of alkenes used to investigate the selectivity-
controlling features in the AD-reaction.
Kinetic dihydroxylation group 1 (toluene)
4
3.5
3
1b vs. 1a
1c vs. 1a
y = 3.75x
1b
R2 = 0.998
2.5
2
1a
y = x
1.5
1
0.5
0
1c
y = 0.081x
R2 = 0.969
0
0.5
1
1.5
2
2.5
ln([olefin 1a]/[olefin 1a 0])
Fig. 4. Relative rates of reaction within group 1. 1b and 1c were both measured against 1a and for this reason 1a has the relative rate 1 (y = x).

P. Fristrup et al. / Journal of Organometallic Chemistry 691 (2006) 2182–2198
2185
Kinetic dihydroxylation group 2 (toluene)
4.5
4
2a
2b
y = x
y = 6.74x
y = 0.71x
2c
R2 = 0.971
R2 = 0.997
3.5
3
2a vs. 2b
2c vs. 2b
2.5
2
1.5
1
0.5
0
0
1
2
3
4
5
6
ln([olefin 2b]/[olefin 2b 0])
Fig. 5. The relative rates of reaction within group 2.
Kinetic dihydroxylation group 3 (toluene)
4.5
4
3b
3a vs. 3c
3b vs. 3c
y = 2.60x
R2 = 0.995
3a
3.5
3
3c
y = 2.37x
R2 = 0.998
y = x
2.5
2
1.5
1
0.5
0
0
0.5
1
1.5
2
2.5
3
3.5
ln([olefin 3c]/[olefin 3c 0])
Fig. 6. The results of the kinetic dihydroxylations with the group 3 alkenes.
to be used a reference compound (i.e., alkene 1b in group
1 and alkene 2b in group 2). Within groups 3 and 4 the
alkenes could not all be separated on GC, which made
it necessary to use the alkenes 3c and 4a as reference com-
pounds. Within these groups the differences in reactivity
were rather small, so the accuracy in the determination
of relative reactivity was unaffected by the choice of refer-
ence compound.
Compound 1b is clearly the fastest reacting alkene, doc-
umenting the importance of the stabilizing interactions in
the SW corner of the Sharpless mnemonic (Fig. 1).
Substrate 1a is the second fastest reacting alkene,

2186
P. Fristrup et al. / Journal of Organometallic Chemistry 691 (2006) 2182–2198
Kinetic dihydroxylation group 4 (toluene)
3.5
3
4a
4c vs. 4a
4b vs. 4a
y = x
4b
2.5
2
y = 0.87x
R2 = 0.999
1.5
1
4c
y = 0.28x
R2 = 0.998
0.5
0
0
0.5
1
1.5
2
2.5
3
3.5
4
ln([olefin 4a]/[olefin 4a 0])
Fig. 7. The results of the kinetic dihydroxylation of the alkenes 4.
approximately four times slower than 1b. The fact that it is
more than 10 times faster than the slowest reacting alkene
(1c) shows the NE corner also interacts favorably with the
phenyl group. The slowest reacting alkene (1c) points the
phenyl group towards the NW corner of the Sharpless
mnemonic, which does not afford stabilization. The results
from group 1 reveal a difference in reaction rate of more
than a factor of 40 between the slowest and the fastest
alkene. The large difference can be attributed to the fact
that the phenyl group is connected directly to the double
bond, and the conjugation does not allow the phenyl group
to rotate freely to relieve strain or steric repulsion inside the
pocket created by the ligand. The results from group 1 are
in accordance with the original Sharpless mnemonic [6],
which singles out the SW corner as especially attractive
towards aromatic groups.
The group 2 alkenes were investigated in a similar man-
ner and the results are shown in Fig. 5. Within this group
2a is by far the fastest reacting alkene. Clearly this means
that the NE corner is far better in stabilizing a benzyl
group than any other corner. Alkene 2b reacts about 40%
faster than 2c, which shows that the SW corner is only mar-
ginally better in stabilizing a benzyl group than the NW
corner. These results are in sharp contrast to the predic-
tions from the Sharpless revised mnemonic (Fig. 1, middle)
which predicts the SW corner to be more favorable than
NE. Our own revised mnemonic (Fig. 1, right) does not
attempt to differentiate between the SW and NE corners,
and the results from alkenes 1 and 2 clearly show that no
such differentiation is possible with such a simple device.
The results for the group 3 alkenes are shown in Fig. 6.
In this case two parts of the osmium–ligand complex, cor-
responding to the NE and SW corners of the mnemonic
device, are equally good in stabilizing the aromatic part
of the alkene. Again, this is in contrast to the revised
mnemonic device suggested by Sharpless, but in good
agreement with the mnemonic device suggested by us pre-
viously (Fig. 1, right). The difference in reactivity is much
smaller for this substrate group than for the previous
group. This can be attributed to the fact that the phenyl
groups in alkenes 3 and 4 are connected to the double bond
with a flexible tether, allowing it to reach through space
towards any stabilizing interactions available, with less
dependence on the position of the origin of the tether.
The results for the kinetic dihydroxylation of group 4
are shown in Fig. 7. The same explanations as the ones
advanced for group 3 can also be applied here. Substrates
4a and 4b show very similar reactivity. The slowest reacting
alkene is 4c, which has the large group in the NW corner.
This alkene can obtain stabilizing interactions only by
‘‘wrapping’’ around to the other areas of the ligand.
Having established the relative rates of reactivity for the
one-phase system in toluene, we turned our attention to the
more synthetically useful two-phase t-BuOH:H2O solvent
system. Our initial results using alkenes from group 1 were
unsuccessful due to too large differences in reactivity
between the substrates. When the difference in reactivity
approaches a factor of 10 the changes in the concentration
of the slowest reacting alkene become very small, thus lead-
ing to large uncertainties in the determination. Instead, we
chose to select competing pairs from the entire set of 12
alkenes, based solely upon relative reactivity. Using alkene
2a as reference compound the relative reactivities of 1a, 1b
and 3a could be established (Fig. 8).

P. Fristrup et al. / Journal of Organometallic Chemistry 691 (2006) 2182–2198
2187
Kinetic dihydroxylation (2a as reference)
2.5
2
1a vs. 2a
1b vs. 2a
3a vs. 2a
1b
2a
y = 4.70x
R2 = 0.980
y = x
1a
1.5
1
3a
y = 0.56x
R2 = 0.990
y = 1.49x
R2 = 0.997
0.5
0
0
0.5
1
1.5
2
2.5
3
ln([olefin 2a]/[olefin 2a 0])
Fig. 8. Kinetic plots for alkenes 1a, 1b and 3a, which were all determined relative to alkene 2a.
The reactivity of the substantially slower-reacting c alk-
enes from groups 1–3 was determined relative to alkene 2c,
and it was also possible to determine the relative reactivity
of alkene 2b in the same group (Fig. 9).
For the group 3 alkenes it was possible to determine the
relative reactivity of all three alkenes within the group, and
in addition alkene 4b could also be related to 3b (Fig. 10).
The remaining determinations of relative reactivities
have been collected in Fig. 11.
The results of the kinetic dihydroxylations are sum-
marized in Fig. 12. A dotted arrow indicates the differ-
ence in reactivity was too large to be determined. A
full line arrow indicates the difference in reactivity could
be determined. The reactivity of each alkene compared
Kinetic dihydroxylation (2c as reference)
2.5
2
2b
y = 1.29x
R2 = 0.997
3c
y = 3.56x
R2 = 0.981
1.5
1
1c vs. 2c
2b vs. 2c
3c vs. 2c
1c
y = x
2c
y = 0.54x
R2 = 0.972
0.5
0
0
0.2
0.4
0.6
0.8
ln([olefin 2c]/[olefin 2c 0])
1
1.2
1.4
1.6
Fig. 9. Kinetic plots for alkenes 1c, 2b and 3c, which were all determined relative to alkene 2c.

2188
P. Fristrup et al. / Journal of Organometallic Chemistry 691 (2006) 2182–2198
Kinetic dihydroxylation (3b as reference)
1.4
1.2
1
3a vs. 3b
3c vs. 3b
4b vs. 3b
4c
y = 0.96x
R2 = 0.992
3b
y = x
0.8
0.6
0.4
0.2
0
3a
y = 0.51x
R2 = 0.995
y = 0.092x
R2 = 0.977
3c
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
ln([olefin 3b]/[olefin 3b 0])
Fig. 10. Kinetic plots for alkenes 3a, 3c and 4c, which were all determined relative to alkene 3b.
Kinetic dihydroxylation
4
3.5
3
1a vs. 2b
y = 6.05x
R2 = 0.991
4a vs. 4b
3c vs. 4c
2.5
2
1.5
1
y = 0.95x
R2 = 0.990
y = 1.09x
R2 = 0.993
0.5
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
ln([olefin]/[olefin])
Fig. 11. The remaining kinetic dihydroxylations, which enable us to determine the relative reactivity of all 12 alkenes. Note that the competitive
dihydroxylations collected in this plot have been performed using different reference compounds. Thus, the relative reactivity of the compounds within this
plot cannot be assessed directly.
to the slowest reacting alkene 1c is shown in a square
next to each alkene, whereas the experimentally deter-
mined relative reactivity from each successful pair-wise
competition is shown in an ellipse on the corresponding
arrow.
Examining Fig. 12 in more detail one realizes that the
largest difference in reactivity determined successfully was
10.8:1 measured between 3b and 3c. Multiplying with the
reactivity of 3c one arrives at a reactivity of 71.5 for 3b rel-
ative to the reference compound 1c. The accuracy of this

P. Fristrup et al. / Journal of Organometallic Chemistry 691 (2006) 2182–2198
2189
in t-BuOH:H2O and the resulting ee was measured by chi-
ral GC. In Table 1 is summarized experimental reactivity
determined in toluene along with reactivity and selectivity
(ee) in the two-phase t-BuOH:H2O. It is interesting to note
that, despite the fact that trisubstituted alkenes generally
are considered favored substrates for the AD reaction, sev-
eral of the substrates investigated here display moderate or
even low enantioselectivity.
121.3
1
14.5
1a
1b
1c
6.05
4.70
1.78
1.86
1.86
25.8
2a
2b
2c
1.29
10.8
2.40
1.49
3.56
3. Computational methods
73.5
Force field modeling of transition states has been per-
formed using the previously described Q2MM force field
[14] for the AD reaction [11a], updated [13] to work with
MacroModel v8.0 [17]. The results are not identical to
the published force field [11a], but relative energies gener-
1.97
6.62
38.4 3a
3b
3c
1.04
1.05
ally deviate only by 1–2 kJ/molꢀ1
. Conformational
1.09
77.1
4a
4b
4c
6.95
searches consisted of systematic Monte Carlo searches
[18] of selected torsions as described previously [13] fol-
lowed by a re-optimization in vacuo with a tight energy
cutoff (20 kJ/molꢀ1). The ensemble of conformations thus
generated was used as a starting point for a mixed
Monte-Carlo [19]/low mode [20] search which was per-
formed for 100,000 steps. In the Monte Carlo part of the
search the full set of torsions was now included. Finally,
another reoptimization was performed either in vacuo or
with the built-in solvation model for H2O. These two sets
of calculations represent two extremes. The experimental
results were expected to lie in between these two sets, with
the toluene results closer to gas phase, and the bi-phasic
results (where the reaction probably occurs in the t-
BuOH-phase) closer to the results obtained with the con-
tinuum model for water. Separate conformational searches
were performed for each of the four possible approach vec-
tors of the alkene [6,13], corresponding to the four possible
orientations in the mnemonic device. When referencing
individual ensembles, they are designated by which corner
of the mnemonic the alkene hydrogen is pointing to. Thus,
for all alkenes, from the mnemonic we would expect the
energetically best ensemble to be the one designated SE.
The free alkenes were subjected to systematic Monte Carlo
searches [18] only, since these have few enough degrees of
freedom so that it is possible to guarantee a complete cov-
erage of the conformational space using a systematic
search.
70.7
Fig. 12. Overview of the experimentally determined relative reactivities in
the two-phase t-BuOH:H2O system. dotted arrow indicates the
difference in reactivity was too large to be determined. A full arrow
indicates a successful determination, with the relative reactivity shown in
an ellipse.
A
determination can be assessed by following the path
3c ! 2c ! 2b ! 1a ! 2a ! 3a arriving at 3b now having
a relative reactivity of 75.6 – a relative error of ca. 5%,
which is acceptable considering the fact that six consecutive
determinations of relative reactivity were used. A priori we
cannot judge which of the two determinations is the most
accurate one – a single determination of a relatively high
reactivity (with an uncertain accuracy), or the use of six
consecutive determinations each with relatively good accu-
racy. Consequently, in Fig. 12 we have chosen to show an
average value of 73.5 for the reactivity of alkene 3b.
Along with the kinetic dihydroxylations the resulting
diols were synthesized using catalytic amounts of osmium
Table 1
Experimentally determined reactivities and enantioselectivities for the
dihydroxylation of all 12 alkenes
Alkene
krel (experimental)a
ee (%) exp
Toluene
t-BuOH:H2O
Ensemble energies for each conformational search are
obtained by Boltzmann summation over all conformations
according to Eq. (1), and thus include conformational
entropy, but not vibrational or solvation contributions,
unless included explicitly in each individual conformational
energy. Note that Eq. (1) can also be used to combine sev-
eral different ensembles (like those for all alkene orienta-
tions) into one global ensemble
1a
1b
1c
2a
2b
2c
3a
3b
3c
4a
4b
4c
a
12.1
45.3
1
9.3
1.40
1
2.37
2.58
1
3.53
3.06
1
14.4
119.5
1
13.8
1.29
1
5.8
11.3
1
11.3
10.4
1
96.4
99.8
72.6
97.6
28.0
21.3
95.6
89.0
49.4
96.7
94.2
46.4
!
X
i
E ¼ ꢀRT ln
eꢀE =RT
.
ð1Þ
i
Reactivities (krel) relative to alkene c in each group of isomers.

2190
P. Fristrup et al. / Journal of Organometallic Chemistry 691 (2006) 2182–2198
Enantioselectivities are calculated by combining all
ture in the AD reaction [3]. Looking at the results in
Table 2, two features are striking. Firstly, the reactivity
span is much wider for alkenes 1 and 2 than for 3 and 4.
We attribute this mainly to a difference in the reactivity
of the respective reference alkenes c, which have tethers
long enough to allow the alkene to fold the phenyl group
into attractive areas of the ligand, irrespective of the actual
attachment point of the tether. Secondly, the benzyl-substi-
tuted alkenes 2 are unique in that there is a large difference
between the reactivity of 2a and 2b. Since this difference
can also be seen in the experimental data (Fig. 5), we
decided to look in more detail at the most favorable paths
for dihydroxylation of 2b.
The most surprising feature of the transition states for
2b is that the orientation which was expected to be favored
based on the mnemonic device, SE (which has the benzyl
group oriented into the SW corner), is found to be isoener-
getic with the NE orientation, where the benzyl group
points into the non-interacting NW corner and a methyl
group is positioned into the sterically crowded SE corner.
The reason for this can be found in the exact orientation
of the benzyl group in the pocket created by the quinolines.
This pocket is well suited for a phenyl group directly
attached to the reacting alkene, but the intervening methy-
lene in 2b requires that the direction of the phenyl group of
2b differs by ca 70° from that of the phenyl group in 1b.
The fairly narrow cleft defined by the two quinolines can-
not accommodate this drastic change in orientation. When
the ligand is forced to accept a benzyl group in this cleft, it
undergoes a conformational change which allows it to
interact favorably with the benzyl group, but at the cost
of a high conformational energy (Fig. 13, left).
paths contributing to one enantiomer into one ensemble
using Eq. (1), and are reported as relative activation barri-
ers. The enantiomeric ratio (er) and ee can then be calcu-
lated using Eqs. (2) and (3). The calculation of ee is
defined to yield a positive number if the major enantiomer
is obtained from the expected SE orientation of the alkene
in the mnemonic device, that is, the energy for the ensemble
containing the SE (and NW) conformations is lowest in
energy
SWþNEꢀESEþNWÞ=RT
er ¼ eðE
;
ð2Þ
ð3Þ
er ꢀ 1
ee ¼
.
er þ 1
Relative rates must be obtained isodesmically. As in our
previous work [13], the competition is seen as a pseudo-
equilibrium where the alkene in one TS exchanges with free
alkene to form an alternative TS. The relative activation
energy can then be obtained from Eq. (4), where E in all
cases corresponds to total ensemble energies obtained from
all contributing conformations for one type of TS or
alkene. We define the reference alkene as alkene 1, so that
Eq. (4) always yields a positive number for any alkene that
reacts faster than the reference (alkene c in each group, or
1c globally). We note that when the two products in the
competition are more different than simple stereoisomers
(i.e., for all comparisons except between b and c within
each group), the isodesmic comparison is not isoparamet-
ric, which means that systematic errors present in all molec-
ular mechanics treatments do not necessarily cancel. Thus,
we will also be testing the importance of the systematic
errors in our force field for an isodesmic comparison
The best possible conformation of 2b can instead be
found in the NW orientation, with the benzyl group point-
ing towards the NE corner (Fig. 13, right). In this orienta-
tion, the benzyl group can be stabilized by the PHAL
linker, which provides a much more open, L-shaped pocket
[9], but at the cost of having to point a methyl group into the
crowded SE corner, rationalizing the low reactivity of 2b.
Interestingly, the SW orientation, with the benzyl group
pointing into the sterically crowded SE corner (Fig. 14),
was found to be only 3 kJ/mol higher in energy than the
best conformation, the NW orientation shown in Fig. 13.
On analyzing the structure, we see that the auxiliary alka-
loid unit has undergone a drastic change of orientation to
achieve some stabilizing interactions with the benzyl substi-
tuent at only a moderate steric cost. Since this orientation
leads to the opposite enantiomer compare to the favored
NW orientation, a very low ee is expected for 2b, in good
agreement with the experimental data (Table 1).
¼
DDE ¼ ETS1 ꢀ ETS2 ꢀ Ealkene1 þ Ealkene2
.
ð4Þ
4. Computational results
The calculated relative rates of dihydroxylation of all
alkenes relative to alkene c within each group are shown
in Table 2.
In all cases the a and b alkenes react faster than the c
alkenes due to the stabilizing interactions with the ligand
environment. This stabilization also accounts for the corre-
lation between selectivity and rate, which is a general fea-
Table 2
Calculated values for the differences in activation energy within each
isomeric group of alkenes (Eq. (4))
Group
DDEà calc (kJ/mol)
In vacuo
H2O
For the c alkenes neither of the four possible orienta-
tions of the alkene within the osmium–ligand complex
are especially favored, but as expected from previous work,
we can see that the long tethers of alkenes 3c and 4c indeed
allow the alkene to fold and put the phenyl group in close
proximity to various parts of the ligand. The folded con-
formers suffer a penalty of a few kJ/mol due to gauche
a
b
a
b
1
2
3
4
12.1
12.2
4.3
12.3
2.9
3.7
5.6
10.5
10.8
3.2
9.0
2.9
2.9
2.5
4.6
5.1
The c alkene within each group has been used as reference.

P. Fristrup et al. / Journal of Organometallic Chemistry 691 (2006) 2182–2198
2191
Fig. 13. Left: The best SE orientation of 2b. Right: The best TS structure for 2b, the NW orientation, with the benzyl group stabilized by the PHAL
moiety in the NE corner.
interactions, but also some gain due to conformational
entropy. In other words, several slightly stabilized confor-
mations are possible for the c alkenes, whereas only a
few highly stabilized conformations can be achieved in
the ligand pockets of the SW and NE corners.
As a final point on the results in Table 2, we can see that
utilization of a computational solvation model leads to a
small but systematic decrease in rate differences. This is
entirely in line with expectations, since the surface area
available for stabilization by the solvent is largest for alk-
enes c, where the large group is mostly pointing out into
the solvent.
Turning now to enantioselectivities, the calculated
results are shown as relative activation barriers in Table
3. Numbers are shown as positive for all cases where the
calculated major enantiomer corresponds to what would
be predicted from the mnemonic devices (in all cases orien-
tation SE). We first note that the solvation model has little
effect on calculated enantioselectivities, as expected for the
AD reaction where ee values measured in water and tolu-
ene usually differ by only a few percent [3]. The target accu-
racy of the Q2MM method is around 2 kJ/mol [11a];
differences lower than this are considered to be within the
uncertainty of the method. We can immediately see that
only two substrates, 1a and 1b, are expected to yield very
Fig. 14. The SW orientation of 2b, with the benzyl group in the SE corner,
crowded by the PHAL moiety.
high enantioselectivities. For the remaining substrates,

2192
P. Fristrup et al. / Journal of Organometallic Chemistry 691 (2006) 2182–2198
Table 3
Table 4
Calculated values for the differences in activation energies for the
diastereomeric transition states leading to opposite enantiomers
The experimental reactivities and selectivities converted to differences in
activation energy
Group
DDEà calc (kJ/mol)
Group
DDEà exp (kJ/mol)
In vacuo
H2O
Toluene t-BuOH:H2O ee
a
b
c
a
b
c
Isomer:
a
b
a
b
a
b
c
1
2
3
4
13.1
4.8
7.4
7.6
14.2
3.1
7.4
7.2
3.3
13.0
6.9
8.3
8.9
11.9
5.7
7.2
5.2
2.0
0.9
2.2
0.6
1
2
3
4
a
6.2
5.5
2.1
3.1
9.4
0.8
2.3
2.8
6.6
6.5
4.4
6.0
11.9
0.6
6.0
5.8
9.9
10.9
9.4
>17a
1.4
7.0
8.7
4.6
1.1
2.7
2.5
ꢀ1.8
1.5
ꢀ0.8
10.1
Only the major enantiomer was observed and consequently the ee was
assumed to be at least 99.8%.
where the longer tethers allow stabilization of several
orientations, the difference between competing paths is
smaller, and largely arises from the steric crowding of sin-
gle methyl groups in the SE corner for orientations leading
to the minor enantiomer. For the alkenes c, we can even see
that the absolute stereochemistry of the major enantiomer
is uncertain. In terms of orientation in the mnemonic
device, this is due to the fact that there is a competition
between on the one hand orientation SE, which experiences
neither steric penalty nor attractive interactions, and on the
other hand orientations SW and NE, where an attractive
interaction can be obtained, but at the cost of a steric clash
between a methyl group and the PHAL linker in the SE
corner. For alkenes 2c and 4c, the prediction of absolute
stereochemistry is even dependent on whether a solvent
model is used or not, but since all of the corresponding
energies are within 2 kJ/mol, the predictions must all be
considered to be inconclusive in view of the expected uncer-
tainty of our method.
energy. For example, an error of 2 kJ/mol is approximately
equal to the difference between 0% and 30% ee at room
temperature, but also to the difference between 98% and
99% ee. To investigate the correlation between experimen-
tal and computational results, it is usually more appropri-
ate to convert the former to energies, and then compare
those to the computational data. The data from Table 1
have therefore been converted to relative activation ener-
gies for the two competing process, Table 4.
The experimental differences in activation energy
obtained in toluene were plotted against the theoretical val-
ues obtained in vacuo (Fig. 15).
Each data point corresponds to one of the eight a or b
substrates – the c alkenes are left out since they are used
as reference compounds within each group and thus
assigned a relative reactivity of 1. The correlation is good,
with only two values deviating more than 2 kJ/mol from
the regression line. Interestingly, we see no significant dif-
ference in performance for the a and b series, despite the
fact that the isodesmic comparison is fully isoparametric
only for the b series. It therefore seems that the systematic
errors in bending and torsion functions that are not fully
compensated in the a vs. c comparison give an insignificant
contribution to the error in the current study. However, if
bond types change (as in a comparison of 1a vs. 2a), the
molecular mechanics comparison is no longer valid, since
5. Comparison between theory and experiment
Computational results are initially obtained as energies,
which can later be converted to experimental observables
for direct comparison to experimental numbers. However,
properties like reaction rates are exponentially dependent
on the energies, with the result that an error in the calcu-
lated property is a non-linear function of the underlying
Calculated vs. Experimental
18
16
14
12
10
8
y = 1.65x
R2 = 0.77
1b
1a
2a
4b
3b
6
3a
4a
4
2b
2
0
0
2
4
6
8
10
E‡ (exp.)
Fig. 15. Activation energies relative to the c alkenes, computational results in vacuo vs. experimentally determined values in toluene.

P. Fristrup et al. / Journal of Organometallic Chemistry 691 (2006) 2182–2198
2193
the unknown systematic errors connected to bond type
changes and varying degrees of conjugation are significant.
The slope in Fig. 15 is significantly higher than 1, indi-
cating that energy differences are systematically overesti-
mated by the force field. We can see two obvious sources
for this systematic error. It is a known deficiency of the
Q2MM methodology that, even though distortions perpen-
dicular to the reaction coordinate are handled well by the
force field, changes along the reaction coordinate cannot
be correctly represented [14]. In real systems, the system
can relax along the reaction coordinate in response to steric
and electronic changes, whereas in the current Q2MM
implementation, the reaction coordinate is in effect frozen.
This naturally leads to a steep energy increase in the model,
where the real system might have the possibility for a soft
relaxation. The second possibility is that the non-bonded
interactions responsible for the steric attraction, and there-
fore for the difference between the various orientations,
have less influence in solvent. An indication of the relative
importance of these two factors can be obtained from a
comparison with the results obtained using a computa-
tional solvation model. We decided to make this compari-
son for the experimental data obtained in the two-phase
t-BuOH:H2O system which are plotted against the calcu-
lated values in vacuo in Fig. 16 and against data from
the solvated calculations in Fig. 17.
With the experimental data from the two-phase system,
the correlation is significantly worse than for the toluene
data, but interestingly the slope is now close to 1. Further-
more, the slope is smaller for the calculations employing
the solvation model, indicating that a part of the systematic
error in Fig. 15 is due to an overestimation of the non-
bonded interactions. However, the larger errors in the
two-phase system preclude a full assignment of the source
of the systematic error. We also note that the experimental
system, which contains a high concentration of salts, may
not be well represented by either the gas phase or the water
model calculations. To conclude the comparison of relative
rates, for the reaction in non-polar solvent, we obtain a
good correlation between calculated and experimental
Calculated vs. Experimental
16
14
12
10
8
y = 1.16x
R2 = 0.47
1a
1b
2a
4a
3b
6
4b
4
3a
2
2b
0
0
2
4
6
8
10
12
14
E‡ (exp.)
Fig. 16. The calculated differences in activation energies (in vacuo) relative to the c compounds against the experimentally determined values in the two-
phase t-BuOH:H2O system.
Calculated vs. Experimental
12
y = 0.93x
2a
R2 = 0.27
1a
10
1b
8
6
3b
4
2b
4a
3a
4b
2
0
0
2
4
6
8
10
12
14
E‡ (exp.)
Fig. 17. The calculated differences in activation energies (in water) relative to the c compounds against the experimentally determined values in the two-
phase t-BuOH:H2O system.

2194
P. Fristrup et al. / Journal of Organometallic Chemistry 691 (2006) 2182–2198
data. For reactions in polar solvent, the errors are still
small in an absolute sense, but too large to allow a truly
predictive model for relative reactivity.
are plotted against the data calculated in vacuo in Fig. 18,
and against the data calculated in solvent in Fig. 19.
The agreement is very good, with most of the compu-
tational predictions falling within the target accuracy,
2 kJ/mol. The agreement is slightly worse in the calcula-
tions employing the solvation model, despite the fact that
the experimental data were obtained in a polar solvent,
indicating that in this case, the solvation introduces a
slightly higher level of random noise in the calculations,
and does not correct for any significant systematic errors.
The largest error obtained is for the benzyl-substituted
alkene 2a. The absolute value of the error is still fairly
small, ca 4 kJ/mol for the in vacuo calculations and
6 kJ/mol employing the water model, and might therefore
just represent the extreme of a random distribution of
errors. However, we also wish to point out that the deriva-
tion of the original Q2MM force field [11a] included quan-
tum chemical data for reactions of phenyl- and alkyl-
substituted alkenes, but no benzyl-substituted alkenes. It
is therefore plausible that torsional parameters for the
bond connecting the benzyl group to the reacting alkene,
which should be subject to hyperconjugation effects unique
For the determination of ee, the error in terms of energy
is expected to be smaller since the comparison is now
between diastereomeric structures, and furthermore, the
systematic errors which may affect the results when using
molecular mechanics calculations for comparing isomers
should cancel completely when comparing diastereomers.
We therefore plot the calculated versus the experimental
results without performing a regression. The dotted line
instead symbolises perfect agreement (slope = 1). Since
the absolute stereochemistry has not yet been determined
for the previously unknown products in the current study,
the energy derived from the experimental data is assumed
to be positive, that is, that all alkenes give the major product
predicted from the mnemonic device. For some of the prod-
ucts displaying very low ee, this assumption may be false,
but will not have a major impact on the following discus-
sion. The enantioselectivities in the current study have been
determined for the products obtained from the two-phase t-
BuOH:H2O system. The experimentally derived energies
Calculated vs. Experimental
1a
18
16
14
12
10
8
1b
3a
4b
4a
3b
6
2b
2a
4
1c
2
3c
4c
0
2c
-2
-4
0
2
4
6
8
10
12
14
16
18
ΔΔE‡ (exp.)
Fig. 18. Calculated enantioselectivities in vacuo plotted against the experimentally determined enantioselectivities.
Calculated vs. Experimental
18
16
1a
14
1b
12
4a
10
8
3a
4b
3b
2a
2b
6
4
3c
4c
1c
2
2c
0
0
2
4
6
8
10
12
14
16
18
ΔΔE‡ (exp.)
Fig. 19. Calculated enantioselectivities using the water model plotted against the experimentally determined enantioselectivities.

P. Fristrup et al. / Journal of Organometallic Chemistry 691 (2006) 2182–2198
2195
to the benzyl moiety, have been improperly implemented in
the current force field, leading to incorrect orientations of
the benzyl group in the chiral pockets of the ligand.
unique in that all substituents in the current study can be
stabilized in this pocket (alkenes a).
The auxiliary quinoline moiety (on the alkaloid unit
which does not coordinate to Os) also has two important
functions. Firstly, it stabilizes the back-side of the reacting
alkene, providing a energy bonus for reaction involving the
equatorial oxygen pointing into the pocket. This interac-
tion can truly be described as a p–p-interaction, since the
electrons of the p-system of the quinoline presumably sta-
bilize the r*-orbital of the forming C–O bond. The interac-
tion is also electrostatic in nature; population analysis of
quantum chemical transition states has revealed significant
positive charge on the alkene carbons in the transition
state. Secondly, both quinolines form a cleft which can sta-
bilize certain alkene substituents in the SW corner of the
mnemonic device. It is obvious that this stabilization is
strongly dependent on shape, since the benzyl moiety could
not be accommodated (alkene 2b), whereas all other alk-
enes in the b series experience stabilization in this cleft.
The quinoline on the primary alkaloid unit participates in
some of this stabilization, and also provides significant
dipole stabilization of the complex by interaction between
the MeO-substituent and the axial oxo-group.
In view of the generally excellent agreement between the
calculated and experimental selectivities reported here and
previously [11a], we have analyzed the interactions in sev-
eral of the most important conformations of the transition
states in the current study. Our findings are summarized in
Fig. 20, where the best transition state for the dihydroxyla-
tion of 1b has been overlaid on our revised mnemonic
device. To summarize, the strongest selectivity-determining
interactions come from the PHAL linker, which is respon-
sible for the steric crowding of the point in space corre-
sponding to the SE corner of the mnemonic device. It is
important to note that this interaction has been present
in all successful ligands for the AD reaction. Even the ace-
tate used in the first report by Hentges and Sharpless [1]
prefers a conformation where the carbonyl carbon is virtu-
ally superimposable on one of the PHAL nitrogen atoms.
Another important contribution of the PHAL is to pro-
vide attractive stabilization to substituents pointing into
the NE corner of the mnemonic device. In conjunction with
the backside of the quinuclidine coordinating to Os, the
PHAL moiety forms an L-shaped pocket which is able to
accommodate a wide range of substituent shapes. It is
6. Summary
The revised mnemonic [13] for the AD reaction has been
validated by a broad substrate selectivity study, and the
limitations of this qualitative prediction tool have been
pointed out. Several of the substrates investigated here
should yield good enantioselectivities according to the
mnemonic, but in fact give only low to moderate selectiv-
ity. We have verified that our Q2MM force field [11a] for
the AD-reaction provides excellent results for prediction
of enantioselectivities, and fair correlation with relative
reactivities of different substrates if the comparison is per-
formed isodesmically. The latter point is surprising in view
of the known deficiencies in empirical force fields for com-
parisons of substrates with different bonding patterns. The
influence of each moiety of the PHAL ligand on the rate
and selectivity of the AD-reaction has been elucidated.
7. Experimental
7.1. General
EtOAc and hexane for chromatography were distilled
under nitrogen. THF was distilled from Na/benzophenone
under nitrogen. Purification of crude products was done by
˚
flash column chromatography on Matrex 60 A silica gel.
All moisture-sensitive reactions were performed in flame-
dried glassware under argon. Optical rotations were mea-
sured on a Perkin–Elmer 241 polarimeter at 589 nm
(CH2Cl2). H and 13C NMR spectra were obtained using
a Varian Mercury300 operating at 300 and 75 MHz,
respectively. MS (electron impact) was performed on VG
Masslab Trio-2. GC-HRMS was performed on a Waters
1
Fig. 20. The global minimum structure of alkene 1b with the phenyl group
stabilized in the south-west corner has been overlaid on the revised
mnemonic device. The most important stabilizing interactions have been
mapped.

2196
P. Fristrup et al. / Journal of Organometallic Chemistry 691 (2006) 2182–2198
Micromass GCT equipped with an Agilent DB-5MS
column.
ate vinyl bromide and 2 mol% of the Pd(PPh3)4 catalyst
were used.
2-Methyl-6-phenyl-2-hexene (4a) was isolated as a clear
oil in 46% yield (1.86 mmol) after chromatography. 1H
NMR (CDCl3) 1.64 (s, 3H, CH3), 1.65–1.75 (m, 2H,
CH2), 1.75 (d, 1.2 Hz, 3H, CH3), 2.06 (q, 7.3 Hz, 2H,
CH2), 2.65 (t, 7.8 Hz, 2H, CH2), 4.55 (m, 1H, @C–H),
7.15–7.35 (m, 5H, aryl-H). 13C NMR (CDCl3) 17.7 (1C,
CH3), 25.7 (1C, CH3), 27.7 (1C, CH2), 31.6 (1C, CH2),
35.5 (1C, CH2), 124.3, 125.5, 128.2, 128.4 (6C, aromatic
carbons), 131.7 (1C, @C), 142.7 (1C, @C). m/z (EI): 175
(5%, [M + 1]+), 174 (38%, M+), 118 (20%, [Ph–CH2–
CH@CH]+), 117 (25%, [Ph–CH2–CH@CÅ]+), 105 (42%,
[Ph–CH2–CH2]+), 104 (100%, [Ph–CH2–CÅ]+), 92 (52%,
[Ph–CH3]+), 91 (69%, [Ph–CH2]+), 69 (36%, [(CH3)2C@
CH–CH2]+), 55 (42%, [(CH3)2C@CH]+), 41 (63%,
[(CH3)C@CH2]+). Exact mass calculated for C13H18:
174.1409, found 174.1411.
(E)-3-Methyl-6-phenyl-2-hexene (4b) was isolated as a
clear oil in 60% yield (2.41 mmol) after chromatography.
1H NMR (CDCl3) 1.60 (d, 7.5 Hz, 3H, CH3), 1.61 (s, 3H,
CH3), 1.68–1.80 (m, 2H, CH2), 2.09 (t, 7.7 Hz, 2H, CH2),
2.61 (t, 7.8 Hz, 2H, CH2), 5.24 (m, 1H, @C–H), 7.15–
7.35 (m, 5H, aryl-H).13C NMR (CDCl3) 13.3 (1C, CH3),
15.8 (1C, CH3), 29.9 (1C, CH2), 35.8 (1C, CH2), 39.5
(1C, CH2), 118.8, 125.8, 128.4, 128.6 (6C, aromatic car-
bons), 135.7 (1C, @C), 143.0 (1C, @C). m/z (EI): 174
(5%, M+), 117 (2%, [Ph–CH2–CH@CÅ]+), 105 (15%, [Ph–
CH2–CH2]+), 104 (100%, [Ph–CH2–CÅ]+), 92 (7%, [Ph–
CH3]+), 91 (17%, [Ph–CH2]+), 55 (15%, [(CH3)2C@CH]+),
41 (17%, [(CH3)C@CH2]+). Exact mass calculated for
C13H18: 174.1409, found 174.1411.
7.2. Synthesis
Alkenes 1a–c were synthesized using an adapted Suzuki-
Miyaura cross-coupling procedure catalyzed by palla-
dium(0) [21–23] as reported previously [13]. The gem-
substituted alkene 2a was synthesized by a Wittig reaction
between acetone and the appropriate phosphonium salt
and isolated as a clear oil after chromatography in 16%
1
yield [24]. H NMR (CDCl3) 1.73 (s, 3H, CH3), 1.75 (s,
3H, CH3), 3.35 (d, 7.5 Hz, 2H, Ph-CH2), 5.34 (m, 1H,
@C–H), 7.15–7.35 (m, 5H, aryl-H). 13C NMR (CDCl3)
18.1 (1C, CH3), 26.0 (1C, CH3), 34.7 (1C, CH2), 123.5,
125.9, 128.5, 128.6 (6C, aromatic carbons), 132.7 (1C,
@C–H), 142.1(1C, @C). m/z (EI): 147 (6%, [M + 1]+),
146 (51%, M+), 131 (100%, [M ꢀ CH3]+), 91 (50%,
C6H5CHþ). Exact mass calculated for C11H14: 146.1096,
2
found 146.1099.
The alkenes 2b and 2c were synthesized from the vinyl
bromides as follows: THF (100 mL) was cooled to
ꢀ78 °C. The vinyl bromide was added followed by slow
addition of two equivalents of t-BuLi. After 1 h, 1.1 equiv-
alents of CuCN was added. Finally, 1.1 equivalents of ben-
zyl bromide was added slowly and the reaction mixture was
kept at ꢀ40 °C while being monitored by TLC
(Rf(BnBr) = 0.20, Rf(vinyl bromide) = 0.28 and Rf(alk-
ene) = 0.48 all in hexane). The reaction was quenched with
sat. aq. NH4Cl followed by standard workup. (E)-2-
Methyl-1-phenyl-2-butene (2b, 10.6 mmol scale) was iso-
1
lated as a clear oil in 44% yield (4.3 mmol, 684 mg) H
NMR (CDCl3) 1.56 (s, 3H, CH3), 1.63 (d, 6.6 Hz, 3H,
CH3), 3.30 (s, 2H, Ph-CH2), 5.32 (m, 1H, @C–H), 7.15–
7.35 (m, 5H, aryl-H). 13C NMR (CDCl3) 13.5 (1C, CH3),
15.5 (1C, CH3), 46.2 (1C, CH2), 120.4, 125.8, 128.1, 128.8
(6C, aromatic carbons), 135.1 (1C, @C), 140.5 (1C, @C).
m/z (EI): 147 (6%, [M + 1]+), 146 (50%, M+), 131 (100%,
(Z)-3-Methyl-6-phenyl-2-hexene (4c) was isolated as a
clear oil in 47% yield (1.90 mmol) after chromatography.
1H NMR (CDCl3) 1.64 (d, 6.9 Hz, 3H, CH3), 1.70 (s, 3H,
CH3), 1.70–1.80 (m, 2H, CH2), 2.09 (t, 7.8 Hz, 2H, CH2),
2.61 (t, 7.8 Hz, 2H, CH2), 5.24 (m, 1H, @C–H), 7.15–
7.35 (m, 5H, aryl-H). 13C NMR (CDCl3) 13.6 (1C, CH3),
23.6 (1C, CH3), 29.8 (1C, CH2), 31.3 (1C, CH2), 36.0
(1C, CH2), 119.5, 125.9, 128.5, 128.6 (6C, aromatic car-
bons), 136.0 (1C, @C), 142.9 (1C, @C). m/z (EI): 174
(7%, M+), 117 (2%, [Ph–CH2–CH@CÅ]+), 105 (17%, [Ph–
CH2–CH2]+), 104 (100%, [Ph–CH2–CÅ]+), 92 (8%, [Ph–
CH3]+), 91 (19%, [Ph–CH2]+), 55 (12%, [(CH3)2C@CH]+),
41 (12%, [(CH3)C@CH2]+). Exact mass calculated for
C13H18: 174.1409, found 174.1411.
[M ꢀ CH3]+), 91 (96%, C6H5CHþ). Exact mass calculated
2
for C11H14: 146.1096, found 146.1098.
(Z)-2-Methyl-1-phenyl-2-butene (2c, 22.7 mmol scale)
was isolated as a clear oil in 68% yield (14.0 mmol,
1
2.247 g). H NMR (CDCl3), 1.62 (s, 3H, CH3) 1.72 (d,
6.9 Hz, 3H, CH3), 3.38 (s, 2H, Ph-CH2), 5.40 (m, 6.9 Hz,
1H, @C–H),7.35–7.15 (m, 5H, aryl-H) 13C NMR (CDCl3)
13.7 (1C, CH3), 23.3 (1C, CH3), 37.5 (1C, CH2), 120.3,
125.8, 128.3, 128.5 (6C, aromatic carbons), 134.7 (1C,
@C), 140.3 (1C, @C). m/z (EI): 147 (5%, [M + 1]+), 146
(48%, M+), 131 (100%, [M ꢀ CH3]+), 91 (94%,
8. General dihydroxylation procedure
C6H5CHþ). Exact mass calculated for C11H14: 146.1096,
The dihydroxylation and characterization of the diols
5a–c and 7a–c have been reported earlier [13]. The synthesis
of the diols 6a–c and 8a–c was performed in a similar man-
ner. In order to allow determination of enantiomeric purity
by GC the diols 6–8 were derivatized as follows. To ꢁ1 mg
of diol in a small vial was added one drop of pyridine and
two drops of acetic anhydride. The vial was closed and
heated to 100 °C for two hours. The resulting mixture of
2
found 146.1101.
Alkenes 3a–c were synthesized using an one-pot hydrob-
oration – Suzuki-Miyaura cross-coupling procedure [13].
The fourth group of alkenes (4a–c) were synthesized in a
similar manner using allylbenzene instead of styrene in
the first hydroboration step. In all three reactions 4.4 mmol
allylbenzene, 4.4 mmol 9-BBN, 4.0 mmol of the appropri-

P. Fristrup et al. / Journal of Organometallic Chemistry 691 (2006) 2182–2198
2197
acetate and excess reagent was diluted with ꢁ1 mL of
EtOAc and analyzed by chiral-GC. The racemic diols were
synthesized using quinuclidine as ligand for osmium and
derivatized in a similar manner to validate the method.
found: C, 74.66 H, 9.68. Calc. for C13H20O2: C, 74.96; H,
9.68%. The racemate was isolated as a pale yellow oil in
quantitative yield.
(2R,3R)-3-Methyl-6-phenyl-2-hexane diol (8b) was iso-
lated as a colorless solid in quantitative yield; m.p. 80–
82 °C. 1H NMR (CDCl3) 1.09 (s, 3H, CH3), 1.12 (d,
6.3 Hz, 3H, CH3), 1.45–1.60 (m, 2H), 1.67–1.80 (m, 2H),
2.08 (br s, 2H, 2 OH), 2.63 (t, 7.5 Hz, 2H, CH2), 3.64 (q,
6.3 Hz, 1H, CH), 7.15–7.35 (m, 5H, aryl-H). 13C NMR
(CDCl3) 17.5 (1C, CH3), 20.6 (1C, CH3), 25.2 (1C, CH2),
36.4 (1C, CH2), 38.7 (1C, CH2), 72.8 (1C, C–OH), 74.9
(1C, C–OH), 125.7, 128.2, 128.3, 142.2 (6C, aromatic car-
1
The racemates had identical H and 13C, but in some cases
the physical appearance was different (see below).
(2R)-3-Methyl-1-phenyl-2-butane diol (6a) was isolated
as colorless crystals in quantitative yield; m.p. 72–74 °C.
1H NMR (CDCl3) 1.24 (s, 3H, CH3), 1.27 (s, 3H, CH3),
2.22 (br s, 2H), 2.56 (dd, 13.5 Hz, 10.8 Hz, 1H), 2.88 (dd,
13.5 Hz, 2.1 Hz, 1H), 3.59 (dd, 10.8 Hz, 2.1 Hz, 1H),
7.20–7.37 (m, 5H, aryl-H). 13C NMR (CDCl3) 23.7 (1C,
CH3), 26.4 (1C, CH3), 38.3 (1C, CH2), 72.6 (1C, C–OH),
79.1 (1C, C–OH), 126.4, 128.6, 129.2, 138.9 (6C, aromatic
20
bons). ½aꢂD ¼ 1:3 (c = 2.7, CH2Cl2). Enantiomeric excess
was determined to 94.2% by analysis of the acetate. Anal.
found: C, 74.70 H, 9.67. Calc. for C13H20O2: C, 74.96; H,
9.68%. The racemate was isolated as a colorless solid in
88% yield; m.p. 85–87 °C.
20
carbons). ½aꢂD ¼ 56:0 (c = 1.5, CH2Cl2). Enantiomeric
excess was measured to 97.6% by analysis of the acetate.
Anal. found: C, 73.34; H, 9.08. Calc. for C11H16O2: C,
73.30; H, 8.95%. The racemate was isolated as a yellow,
amorphous solid in quantitative yield; m.p. 60–61 °C.
(2R,3R)-2-Methyl-1-phenyl-2-butane diol (6b) was iso-
lated as a colorless solid in quantitative yield; m.p. 57–
59 °C. 1H NMR (CDCl3) 1.09 (s, 3H, CH3), 1.22 (d,
6.3 Hz, 3H, CH3), 1.82 (br s, 1H, OH), 2.18 (br s, 1H,
OH), 2.80 (s, 2H, CH2), 3.67 (q, 6.3 Hz, 1H, CH), 7.20–
7.40 (m, 5H, aryl-H). 13C NMR (CDCl3) 17.4 (1C, CH3),
21.0 (1C, CH3), 45.0 (1C, CH2), 72.3 (1C, C–OH), 74.8
(1C, C–OH), 126.5, 128.2, 130.5, 136.9 (6C, aromatic car-
(2R,3S)-3-Methyl-6-phenyl-2-hexane diol (8c) was iso-
1
lated as a pale yellow oil in 99% yield. H NMR (CDCl3)
1.13 (d, 6.3 Hz, 3H, CH3), 1.15 (s, 3H, CH3), 1.35–1.50
(m, 1H), 1.55–1.88 (m, 3H), 2.05 (br s, 2H, 2 OH), 2.55–
2.73 (m, 2H, CH2), 3.62 (q, 6.3 Hz, 1H, CH), 7.15–7.35
(m, 5H, aryl-H). 13C NMR (CDCl3) 17.4 (1C, CH3), 23.5
(1C, CH3), 25.3 (1C, CH2), 35.5 (1C, CH2), 36.5 (1C,
CH2), 74.1 (1C, C–OH), 74.6 (1C, C–OH), 125.7, 128.2,
20
128.3, 142.3 (6C, aromatic carbons). ½aꢂD ¼ ꢀ0:5 (c = 0.6,
CH2Cl2). Enantiomeric excess was determined to 46.4%
by analysis of the acetate. Anal. found: C, 74.72 H, 9.45.
Calc. for C13H20O2: C, 74.96; H, 9.68%. The racemate
was isolated as a pale yellow oil in quantitative yield.
20
bons). ½aꢂD ¼ 4:5 (c = 2.8, CH2Cl2). Enantiomeric excess
was measured to 28.0% by analysis of the acetate. Anal.
found: C, 73.53; H, 9.00. Calc. for C11H16O2: C, 73.30;
H, 8.95%. The racemate was isolated as a colorless solid
in 90% yield; m.p. 57–59 °C.
9. Kinetic measurements in toluene
(2R,3S)-2-Methyl-1-phenyl-2-butane diol (6c) was iso-
lated as a colorless solid in 91% yield; m.p. 73–75 °C.1H
NMR (CDCl3) 1.09 Hz (s, 3H, CH3), 1.26 (d, 6.3 Hz, 3H,
CH3), 1.70 (br s, 1H, OH), 2.21 (br s, 1H, OH), 2.62 (d,
13.5 Hz, 1H, CH2), 3.00 (d, 13.5 Hz. 1H, CH2), 3.74 (q,
6.3 Hz, 1H, CH), 7.20–7.40 (m, 5H, aryl-H). 13C NMR
(CDCl3) 17.2 (1C, CH3), 23.4 (1C, CH3), 41.3 (1C, CH2),
72.8 (1C, C–OH), 74.5 (1C, C–OH), 126.5, 128.3, 130.7,
The competitive stoichiometric dihydroxylations were
performed in 2 mL toluene. The two alkenes were added
giving a concentration of each of about 1 mg/mL. This
allowed direct analysis of the reaction mixture by GC.
Two equivalents of (DHQD)2PHAL with respect to the
total amount of alkene were added. n-Decane or n-tride-
cane in a concentration of ca 1 mg/mL was used as internal
standard. When the difference in reactivity was large (>5)
the accuracy in the determination of alkene concentration
was improved by using less of the most reactive alkene.
After each GC measurement 2 drops of a 0.1 M solution
of OsO4 in toluene were added (about 10 lL). The reaction
mixture and the OsO4-solution were both kept at 0 °C.
20
137.0 (6C, aromatic carbons). ½aꢂD ¼ ꢀ8:9 (c = 1.9,
CH2Cl2). Enantiomeric excess was determined to 21.3%
by analysis of the acetate. Anal. found: C, 73.00; H, 8.80.
Calc. for C11H16O2: C, 73.30; H, 8.95%. The racemate
was isolated as a colorless solid in quantitative yield;
m.p. 63–64 °C.
(3R)-2-Methyl-6-phenyl-2-hexene (8a) was isolated as a
1
pale yellow oil in 90% yield. H NMR (CDCl3) 1.14 (s,
10. Kinetic measurements in t-BuOH:H2O (1:1)
3H, CH3), 1.19 (s, 3H, CH3), 1.30–1.58 (m, 2H), 1.61–
1.77 (m, 1H), 1.86–2.04 (m, 1H), 2.28 (br s, 2H, 2 OH),
2.58–2.75 Hz (m, 2H, CH2), 3.39 (dd, 10.2 Hz, 2.4 Hz,
1H, CH), 7.15–7.35 (m, 5H, aryl-H). 13C NMR (CDCl3)
23.2 (1C, CH3), 26.5 (1C, CH3), 28.5 (1C, CH2), 31.2
(1C, CH2), 35.8 (1C, CH2), 73.1 (1C, C–OH), 78.3 (1C,
C–OH), 125.7, 128.2, 128.3, 142.2 (6C, aromatic carbons).
The reaction was performed in 3 mL t-BuOH:H2O (1:1)
containing ca 0.03 mmol of each alkene. In addition three
equivalents of K2CO3, 10 mol% (DHQD)2PHAL and 0.5
equivalents of naphthalene (internal standard) relative to
the total amount of alkene were added. The resulting sus-
pension was stirred until all solids had dissolved. Then stir-
ring was stopped and the phases were allowed time to
separate. 10 lL of the organic layer was withdrawn and
20
½aꢂD ¼ 28:3, (c = 4.2, CH2Cl2). Enantiomeric excess was
determined to 96.7% by analysis of the acetate. Anal.

2198
P. Fristrup et al. / Journal of Organometallic Chemistry 691 (2006) 2182–2198
(d) D.W. Nelson, K.B. Sharpless, A. Gypser, P.T. Ho, H.C. Kolb,
T. Kondo, H.-L. Kwong, D.V. McGrath, A.E. Rubin, P.-O. Norrby,
K.P. Gable, J. Am. Chem. Soc. 119 (1997) 1840.
diluted by addition of 20 lL t-BuOH. The catalyst was
added (5 mol% K2OsO2(OH)4), and the next sample was
withdrawn after 10 min of stirring, again allowing time
for phase separation. Each withdrawal of sample from
the reaction mixture was followed by the addition of
10 mol% K3Fe(CN)6, and this cycle was repeated 10 times.
All the samples were diluted with t-BuOH as mentioned
above and analyzed by GC immediately or stored in a free-
zer until analysis could be performed.
´
[8] (a) A.K. Rappe, W.A. Goddard III, J. Am. Chem. Soc. 104 (1982)
3287;
(b) K.A. Jørgensen, R. Hoffmann, J. Am. Chem. Soc. 108 (1986) 1867;
(c) K.A. Jørgensen, Tetrahedron Lett. 31 (1990) 6417;
(d) P.-O. Norrby, H.C. Kolb, K.B. Sharpless, Organometallics 13
(1994) 344;
(e) A. Veldkamp, G. Frenking, J. Am. Chem. Soc. 116 (1994) 4937;
(f) P.-O. Norrby, H. Becker, K.B. Sharpless, J. Am. Chem. Soc. 118
(1996) 35;
(g) U. Pidun, C. Boehme, G. Frenking, Angew. Chem., Int. Ed. Engl.
35 (1996) 2817;
Acknowledgements
´
(h) S. Dapprich, G. Ujaque, F. Maseras, A. Lledos, D.G. Musaev, K.
We thank Jørgen Øgaard Madsen for performing the
MS analysis and for assistance in the interpretation of
the fragmentation patterns, and DTU for financial
support.
Morukuma, J. Am. Chem. Soc. 118 (1996) 11660;
(i) M. Torrent, L. Deng, M. Duran, M. Sola, T. Ziegler, Organomet-
allics 16 (1997) 13.
[9] P.-O. Norrby, H.C. Kolb, K.B. Sharpless, J. Am. Chem. Soc. 116
(1994) 8470.
´
[10] G. Ujaque, F. Maseras, A. Lledos, J. Am. Chem. Soc. 121 (1998) 1317.
References
[11] (a) P.-O. Norrby, T. Rasmussen, J. Haller, T. Strassner, K.N. Houk,
J. Am. Chem. Soc. 121 (1999) 10186;
´
(b) N. Moitessier, B. Maigret, F. Chretien, Y. Chapleur, Eur. J. Org.
Chem. (2000) 995;
[1] S.G. Hentges, K.B. Sharpless, J. Am. Chem. Soc. 102 (1980) 4263.
[2] E.N. Jacobsen, I. Marko´, W.S. Mungall, G. Schro¨der, K.B. Sharp-
less, J. Am. Chem. Soc. 110 (1988) 1968.
[3] (a) B.B. Lohray, Tetrahedron: Asymmetry 3 (1992) 1317;
(b) H.C. Kolb, K.B. Sharpless, in: M. Beller, C. Bolm (Eds.),
Transition Metals for Organic Synthesis, vol. 2, VCH, Weinheim,
1998, pp. 219–242;
(c) N. Moitessier, C. Henry, C. Len, Y. Chapleur, J. Org. Chem. 67
(2002) 7275;
(d) N. Moitessier, C. Henry, C. Len, D. Postel, Y. Chapleur, J.
Carbohyd. Chem. 22 (2003) 25;
(e) G. Drudis-Sole, G. Ujaque, F. Maseras, Chem. Eur. J. 11 (2005)
1017.
´
(c) J.S. Svendsen, I. Marko, in: E.N. Jacobsen, A. Pfaltz, H.
[12] (a) E.J. Corey, M.C. Noe, M.J. Grogan, Tetrahedron Lett. 37 (1996)
4899;
Yamamoto (Eds.), Comprehensive Asymmetric Catalysis, Springer
Verlag, Berlin, 1999, pp. 713–790;
(b) A.J. Delmonte, J. Haller, K.N. Houk, K.B. Sharpless, D.A.
Singleton, T. Strassner, A.A. Thomas, J. Am. Chem. Soc. 119 (1997)
9907.
(d) R.A. Johnson, K.B. Sharpless, in: I. Ojima (Ed.), Catalytic
Asymmetric Synthesis, second ed., VCH, New York, 2000, pp. 357–
398;
[13] P. Fristrup, D. Tanner, P.-O. Norrby, Chirality 15 (2003) 360.
[14] (a) P.-O. Norrby, J. Mol. Struct. (Theochem.) 506 (2000) 9;
(b) P.-O. Norrby, F. Jensen, Theor. Chem. Acc. 109 (2003) 1.
[15] D.J. Berrisford, C. Bolm, K.B. Sharpless, Angew. Chem., Int. Ed.
Engl. 34 (1995) 1059.
(e) See also: K.A. Jørgensen, in: M. Beller, C. Bolm (Eds.), Transition
Metals for Organic Synthesis, vol. 2, VCH, Weinheim, 1998, pp. 157–
172.
[4] For selected examples, see: (a) E.M. Carreira, J. Du Bois, J. Am.
Chem. Soc. 116 (1994) 10825;
[16] (a) E.J. Corey, M.C. Noe, S. Sarshar, Tetrahedron Lett. 35 (1994)
2861;
(b) Z.-M. Wang, H.C. Kolb, K.B. Sharpless, J. Org. Chem. 59 (1994)
5104;
(b) E.J. Corey, M.C. Noe, M.J. Grogan, Tetrahedron Lett 35 (1994)
6427;
(c) E.J. Corey, A. Guzman-Perez, M.C. Noe, J. Am. Chem. Soc. 116
(1994) 12109;
(c) E.J. Corey, M.C. Noe, J. Am. Chem. Soc. 118 (1996) 11038;
(d) E.J. Corey, J. Zhang, Org. Lett. 3 (2001) 3211.
[17] MacroModel v. 8.0 from Schrodinger Inc.: F. Mohamadi, N.G.J.
Richards, W.C. Guida, R. Liskamp, M. Lipton, C. Caulfield, G.
Chang, T. Hendrickson, W.C. Still, J. Comput. Chem. 11 (1990) 440,
For current versions, see: http://www.schrodinger.com.
[18] J.M. Goodman, W.C. Still, J. Comput. Chem. 12 (1991) 1110.
[19] G. Chang, W.C. Guida, W.C. Still, J. Am. Chem. Soc. 112 (1990)
1419.
(d) K.C. Nicolaou, C.N.C. Boddy, H. Li, A.E. Koumbis, R. Hughes,
S. Natarajan, N.F. Jain, J.M. Ramanjulu, S. Bra¨se, M.E. Solomon,
Chem. Eur. J. 5 (1999) 2602;
(e) J.S. Lee, P.L. Fuchs, Org. Lett. 5 (2003) 2247.
[5] (a) K.B. Sharpless, W. Amberg, M. Beller, H. Chen, J. Hartung, Y.
Kawanami, D. Lubben, E. Manoury, Y. Ogino, T. Shibata, T.
¨
Ukita, J. Org. Chem. 56 (1991) 4585;
(b) K.B. Sharpless, W. Amberg, Y.L. Bennani, G.A. Crispino, J.
Hartung, K.-S. Jeong, H.-L. Kwong, K. Morikawa, Z.-M. Wang, D.
Xu, X.-L. Zhang, J. Org. Chem. 57 (1992) 2768.
[6] H.C. Kolb, P.G. Andersson, K.B. Sharpless, J. Am. Chem. Soc. 116
(1994) 1278.
[20] I. Kolossvary, W.C. Guida, J. Am. Chem. Soc. 118 (1996) 5011.
[21] N. Miyaura, T. Ishiyama, M. Ishikawa, A. Suzuki, Tetrahedron Lett.
52 (1986) 6369.
[22] N. Miyaura, T. Ishiyama, H. Sasaki, M. Ishikawa, M. Satoh, A.
Suzuki, J. Am. Chem. Soc. 111 (1989) 314.
[7] (a) T. Go¨bel, K.B. Sharpless, Angew. Chem., Int. Ed. Engl. 32 (1993)
1329;
[23] S.R. Chemler, D. Trauner, S.J. Danishefsky, Angew. Chem. Int. Ed.
40 (2001) 4544.
(b) P.-O. Norrby, K.P. Gable, J. Chem. Soc., Perkin Trans. II (1996)
171;
[24] Y.S. Angelis, M. Orfanopoulos, J. Org. Chem. 62 (1997) 6083.
(c) E.J. Corey, M.C. Noe, J. Am. Chem. Soc. 118 (1996) 319;