Structure-based rationale for selectivity in the asymmetric allylic alkylation of cycloalkenyl esters employing the Trost 'Standard Ligand' (TSL): Isolation, analysis and alkylation of the monomeric form of the cationic η3-cyclohexenyl complex [(η3-c-C6H 9)Pd(TSL)]+

Article abstract of DOI:10.1021/ja8099757

The solution-phase structures of the monomeric forms of the cationic Pd-η³-allyl and Pd-η³-cyclohexenyl complexes [Pd(R,R)-1(η³-C₃H₅)]⁺ (7 ⁺) and [Pd(R,R)-1(η³-C₆H₉)] ⁺ (8⁺) bearing the trans-cyclohexylenediamine-based Trost 'Standard Ligand' (R,R)-1 have been elucidated by NMR, isotopic labeling and computation. In both complexes, (R,R)-1 is found to adopt a C1-symmetric conformation, leading to a concave shape in the 13-membered chelate in which one amide group in the chiral scaffold projects its NH unit out of the concave surface in close vicinity to one allyl terminus. The adjacent amide has a reversed orientation and projects its carbonyl group out of the concave face in the vicinity of the opposite allyl terminus. Stoichiometric and catalytic asymmetric alkylations of [8⁺][X^-] by MCHE₂ (E = ester, M = 'escort' counterion, X = Pd allyl counterion) show the same selectivities and trends as have been reported for in situ-generated catalysts, and a new model for the enantioselectivity has been explored computationally. Three factors are found to govern the regioselectivity (pro-S vs pro-R) of attack of nucleophiles on the η³-C₆H₉ ring in 8⁺ and thus the ee of the alkylation product: (i) a pro-R torquoselective bias is induced by steric interaction of the η³-C₆H₉ moiety with one phenyl ring of the ligand; (ii) pro-S delivery of the nucleophile can be facilitated by hydrogen-bonding with the concave orientated amide N-H; and (iii) pro-R delivery of the nucleophile can be facilitated by escort ion (M) binding to the concave orientated amide carbonyl. The latter two opposing interactions lead to the selectivity of the alkylation being sensitive to the identities of X^- and M⁺. The generation of 8⁺ from cyclohexenyl ester substrate has also been explored computationally. The concave orientated amide N-H is able to activate the leaving group of the allylic ester by hydrogen bonding to its carbonyl group. However, this interaction is only feasible for the (S)-enantiomer of substrate, leading to the prediction of a powerful kinetic resolution (k_S ? k_R), as is found experimentally. This new model involving two regiochemically distinct (NH) and (CO) locations for nucleofuge or nucleophile binding, may prove of broad utility for the interpretation of the selectivity in asymmetric allylic alkylation reactions catalyzed by Pd complexes of (R,R)-1 and related ligands.

Full text of DOI:10.1021/ja8099757

Published on Web 05/12/2009
Structure-Based Rationale for Selectivity in the Asymmetric
Allylic Alkylation of Cycloalkenyl Esters Employing the Trost
‘Standard Ligand’ (TSL): Isolation, Analysis and Alkylation of
the Monomeric form of the Cationic η3-Cyclohexenyl Complex
[(η3-c-C6H9)Pd(TSL)]+
Craig P. Butts,† Emane Filali,† Guy C. Lloyd-Jones,*,† Per-Ola Norrby,*,‡
David A. Sale,† and York Schramm†
School of Chemistry, UniVersity of Bristol, Cantock’s Close, Bristol, BS8 1TS, U.K., and
Department of Chemistry, UniVersity of Gothenburg, Kemigården 4,
SE-412 96 Go¨teborg, Sweden
Received December 29, 2008; E-mail: guy.lloyd-jones@bris.ac.uk; pon@chem.gu.se
Abstract: The solution-phase structures of the monomeric forms of the cationic Pd-η3-allyl and Pd-η3-
cyclohexenyl complexes [Pd(R,R)-1(η3-C3H5)]+ (7+) and [Pd(R,R)-1(η3-C6H9)]+ (8+) bearing the trans-
cyclohexylenediamine-based Trost ‘Standard Ligand’ (R,R)-1 have been elucidated by NMR, isotopic labeling
and computation. In both complexes, (R,R)-1 is found to adopt a C1-symmetric conformation, leading to a
concave shape in the 13-membered chelate in which one amide group in the chiral scaffold projects its NH
unit out of the concave surface in close vicinity to one allyl terminus. The adjacent amide has a reversed
orientation and projects its carbonyl group out of the concave face in the vicinity of the opposite allyl terminus.
Stoichiometric and catalytic asymmetric alkylations of [8+][X-] by MCHE2 (E ) ester, M ) ‘escort’ counterion,
X ) Pd allyl counterion) show the same selectivities and trends as have been reported for in situ-generated
catalysts, and a new model for the enantioselectivity has been explored computationally. Three factors are
found to govern the regioselectivity (pro-S vs pro-R) of attack of nucleophiles on the η3-C6H9 ring in 8+ and
thus the ee of the alkylation product: (i) a pro-R torquoselective bias is induced by steric interaction of the
η3-C6H9 moiety with one phenyl ring of the ligand; (ii) pro-S delivery of the nucleophile can be facilitated by
hydrogen-bonding with the concave orientated amide N-H; and (iii) pro-R delivery of the nucleophile can
be facilitated by escort ion (M) binding to the concave orientated amide carbonyl. The latter two opposing
interactions lead to the selectivity of the alkylation being sensitive to the identities of X- and M+. The
generation of 8+ from cyclohexenyl ester substrate has also been explored computationally. The concave
orientated amide N-H is able to activate the leaving group of the allylic ester by hydrogen bonding to its
carbonyl group. However, this interaction is only feasible for the (S)-enantiomer of substrate, leading to
the prediction of a powerful kinetic resolution (kS . kR), as is found experimentally. This new model involving
two regiochemically distinct (NH) and (CO) locations for nucleofuge or nucleophile binding, may prove of
broad utility for the interpretation of the selectivity in asymmetric allylic alkylation reactions catalyzed by Pd
complexes of (R,R)-1 and related ligands.
proven difficult to control with other systems,2 and the trans-
diaminocyclohexane-based ‘Standard Ligand’ (1) has, in most
cases, been found to provide very high selectivities, under
carefully optimized conditions.4
1. Introduction
Since its inception in 1992, the Trost Modular Ligand series
(“TML”, Chart 1)1 has been applied to an extraordinary range
of asymmetric allylic alkylation reactions2 and is among the
small group of chiral ligands that offer genuine utility for
asymmetric transition-metal catalyzed C-C bond-formation.3
It has been especially successful with allylic substrates that have
The generally accepted mechanism for Pd-catalyzed allylic
alkylation can be abbreviated5 to a cycle involving two key
steps: oxidative addition of a low valent Pd species, “L2Pd(0)”,
to the allylic electrophile and nucleophilic attack on the resulting
Pd(II) allyl cation, to liberate the product and regenerate the
Pd(0) complex. For reactions involving allylic esters (typically
carboxylates and carbonates) and stabilized nucleophiles (typi-
cally stabilized enolates, such as malonates), the overall
stereochemical pathway is one of retention, arising from 2-fold
inversion.6 Within this framework, there are opportunities for
asymmetric induction at both stages of the reaction.2,4 Thus,
for the oxidative addition event, asymmetric catalysis can
† University of Bristol.
‡ University of Gothenburg.
(1) Trost, B. M.; Van Vranken, D. L. Angew. Chem. Int Ed. 1992, 31,
228.
(2) Recent reviews. (a) Lu, Z.; Ma, S. Angew Chem., Int. Ed. 2008, 47,
258. (b) Trost, B. M.; Lee, C. Catalytic Asymmetric Synthesis, 2nd
ed; Ojima; I., Ed.; Wiley-VCH: New York, 2000; pp 593-649. (c)
Pfaltz; A.; Lautens, M. ComprehensiVe Asymmetric Catalysis; Jacob-
sen, E. N., Pfaltz, A., Yamamoto, H., Eds.; Springer: Heidelberg, 1999;
pp 833-886.
9
10.1021/ja8099757 CCC: $40.75  2009 American Chemical Society
J. AM. CHEM. SOC. 2009, 131, 9945–9957 9945

A R T I C L E S
Butts et al.
Chart 1. Trost Modular Ligand Series (‘TML’) and ‘Standard
Ligand’ 1, with ‘Wall’ and ‘Flap’ Rings Indicated
Figure 1. Pd(η3-c-C6H9) complexes of the P,X-ligands 312 and 413 designed
for asymmetric alkylation of cycloalkenyl esters, illustrating the two control
elements: exo/endo biasing and activation trans to P.
simple cycloalkenyl esters with malonate and phthalimide
nucleophiles, Scheme 1. Such reactions in principle10 proceed
via a single Pd-η3-cycloalkenyl cationic intermediate (2), thus
while kinetic resolution of cycloalkenyl ester may attend the
process,11 the enantio-determining step is the nucleophilic attack
on the ‘meso’ Pd-η3-cycloalkenyl moiety (2) to generate the
alkylation product. The product enantiomer ratio is thus
determined by the regioselectivity of attack at the pro-R versus
pro-S allylic terminus, the formal meso symmetry10a of the Pd-
η3-cycloalkenyl moiety (2) being broken by the chirality of the
appended ligand (1).
Scheme 1. Trost and Bunt’s Asymmetric Allylic Alkylation of
Racemic Cycloalkenyl Esters (X ) Acetate or Methyl Carbonate)
Using Catalytic Pd(R,R-1) to Induce Highly pro-S Selective Attack
on ‘Meso’-2 by Malonate or Phthalimide Nucleophiles (Nu)9
It later emerged that other ligands could also exert a high
degree of stereocontrol in these previously troublesome reactions
involving ‘slim’ cyclic allylic substrates.2 However, in contrast
to P,P-based ligand 1, the use of a mixed donor ‘P,X’ ligand
system (where P is a phosphorus-based donor and X is a
heteratom such as N or S) was required to achieve high
enantioselectivities (>95% ee). Prime examples are the cyman-
trene-based phosphinoaryl oxazolines (3) developed by Helm-
chen,12 and the phosphinite sulfides (4), developed by Evans,13
illustrated for alkylation of Pd(η3-c-C6H9) in Figure 1.
In both cases (312 and 413), through the systematic study of
structure-selectivity relationships, single crystal X-ray analysis and
NMR spectroscopy of intermediates, two synergistic effects were
confirmed as essential components in the overall performance and
design of the ligand. First, steric interactions between the ligand
and the cycloalkenyl moiety must strongly disfavor one rotameric
isomer (endo) over the other (exo) in the [(P,X)Pd(η3-cycloalk-
enyl)]+ intermediate. Second, with the η3-cycloalkenyl unit orien-
tated in the exo isomer with respect to the ligand framework, and
thus the pro-R vs pro-S allylic termini orientated in the square plane
with respect to the P and X donors, one allylic terminus is more
activated than the other to nucleophilic attack, due to its trans
relationship with the P. This combination of η3-cycloalkenyl
orientation with terminus-specific activation can then induce the
proceed by for example enantioface selection in prochiral
substrates, kinetic resolution, or desymmetrization. In the
nucleophilic attack on the Pd-allyl species, selectivity can be
induced for example by allyl enantioface selection, allyl
regioselection (pro-R vs pro-S), or nucleophile enantioface
selection. What distinguishes the TML series from the vast range
of other ligands developed for asymmetric allylic alkylation2 is
that it has proven utility3,4 in all of these regimes, including
use of O- and N-based nucleophiles4 as well as ‘harder’
nucleophiles such as enolates7 and lithiated methylpyridines.8
As a consequence, it has enjoyed widespread application in
synthesis.3,4
A major development in the field came in 1994, when Trost
and Bunt9 demonstrated that unprecedented levels of enanti-
oselectivity (93-99% ee) could be achieved in the reaction of
(8) Trost, B. M.; Thaisrivongs, D. A. J. Am. Chem. Soc. 2008, 130, 14092.
(9) Trost, B. M.; Bunt, R. C. J. Am. Chem. Soc. 1994, 116, 4089.
(10) (a) Trost, B. M.; Bunt, R. C. J. Am. Chem. Soc. 1996, 118, 235. (b)
Lloyd-Jones, G. C.; Stephen, S. C. Chem.sEur. J. 1998, 4, 2539. (c)
Fristrup, P.; Jensen, T.; Hoppe, J.; Norrby, P.-O. Chem.sEur. J. 2006,
12, 5352, and references therein.
(3) At the time of submission (Dec 2008), Chemical Abstracts listed over
205 publications reporting on the efficient application of “Pd(1)” in
asymmetric catalysis.
(4) For leading references, see: Trost, B. M.; Crawley, M. L. Chem. ReV.
(11) (a) Dominguez, B.; Hodnett, N. S.; Lloyd-Jones, G. C. Angew. Chem.,
Int. Ed. 2001, 40, 4289. (b) Lloyd-Jones, G. C.; Stephen, S. C. Chem.
Commun. 1998, 2321. (c) Gais, H.-J.; Eichelmann, H.; Spalthoff, N.;
Gerhards, F.; Frank, M.; Raabe, G. Tetrahedron: Asymmetry 1998, 9,
235. (d) Frank, M.; Gais, H.-J. Tetrahedron: Asymmetry 1998, 9, 3353.
(e) Gais, H. J.; Spalthoff, N.; Jagusch, T.; Frank, M.; Raabe, G.
Tetrahedron Lett. 2000, 41, 3809.
2003, 103, 2921.
(5) Evans, L. A.; Fey, N.; Harvey, J. N.; Hose, D.; Lloyd-Jones, G. C.;
Murray, P.; Orpen, A G.; Osborn, R.; Owen-Smith, G. J. J.; Purdie,
M. J. Am. Chem. Soc. 2008, 130, 14471, and references therein.
(6) (a) Trost, B. M.; Weber, L.; Strege, P. E.; Fullerton, T. J.; Dietsche,
T. J. J. Am. Chem. Soc. 1978, 100, 3416. (b) Mackenzie, P. B.; Whelan,
J.; Bosnich, B. J. Am. Chem. Soc. 1985, 107, 2046. (c) Hayashi, T.;
Yamamoto, A.; Hagihara, T. J. Org. Chem. 1986, 51, 723.
(7) Trost, B. M.; Schroeder, G. M. Chem.-Eur. J. 2005, 11, 174.
(12) Kudis, S.; Helmchen, G. Angew. Chem., Int. Ed. 1998, 37, 3047.
(13) Evans, D. A.; Campos, K. R.; Tedrow, J. S.; Michael, F. E.; Gagne,
M. R. J. Am. Chem. Soc. 2000, 122, 7905.
9
9946 J. AM. CHEM. SOC. VOL. 131, NO. 29, 2009

Asymmetric Induction by the Trost Standard Ligand
A R T I C L E S
Chart 2. Cationic Monomeric Palladium η3-Allyl (7) and
η3-Cyclohexenyl (8) Complexes of the Standard Ligand (R,R)-1
Generated in situ from Oligomer, then Studied by NMR
Figure 2. Wall-and-flap cartoon model20 for asymmetric alkylation
involving Pd complexes of (R,R)-1, with the asymmetric alkylation of
cyclohexenyl acetate (5) as an example.
of nucleofuge and entry of nucleophile in one front and one
rear quadrant of the allyl plane.
requisite regioselectivity, to generate the alkylation product in high
ee.12,13 In light of this analysis, it is tempting to ascribe an
analogous mode of selectivity by ligand 1, whereby it acts as a
P,O-bidentate ligand, via coordination of one of the amide carbonyl
groups. However, results from the systematic removal of various
ligand components,10b,14 the isolation of a catalytically active but
nonenantioselective binuclear bis-P,O-Pd-allyl complex of 1,15 and
preliminary 31P NMR spectroscopic studies,16 strongly support the
active and selective species to be P,P-, not P,O-coordinated Pd
complexes.
This ‘cartoon model’17 has broad application in that it can
be used to rationalize the outcome from nearly all of the
optimized reactions catalyzed by Pd(1) and related ligands in
the TML series.20 As an example, the reaction of rac-
cyclohexenyl acetate (()-(5) with malonate nucleophile
(MCHE2, E ) CO2Me or CO2Bn, M ) ‘escort’ cation), is
correctly predicted to proceed with powerful kinetic resolution
of 5 (kS/R g 45 with R,R-1),11 and selective nucleophilic attack
by MCHE2 at the pro-S terminus (f(S)-6).9
Despite the undeniable utility of the cartoon model,20 detailed
information about the intermediates involved in the reactions
would allow a structural basis for rationalization of the selectiv-
ity, and facilitate further development and application of the
ligand class. However, the complexity of the system, vide infra,
demands that such a study begins with a simple example
comprising solely the basic components necessary for the
reaction. The reaction of cyclohexenyl acetate (5) with MCHE2
to generate (S)-6 (Figure 2) fulfils this requirement.9 Moreover,
it also provides the challenge that any new model must a priori
predict that the pro-S selectivity is M+-dependent,9 a fact not
easily rationalized by the current cartoon model,20 with the ee
of (S)-6 rising from 0-98% ee in the order: Li+ < Na+ < K+ <
Cs+ < R4N+ (R ) n-Bu, n-Hex.).
In contrast to the detailed structural and spectroscopic data
obtained in the study of the mechanism of selectivity of most
of the other classes of ligand for asymmetric Pd-catalyzed
allylation, the analysis of the asymmetric induction by 1, in all
of its numerous successful applications,3,4 has been made
extremely challenging due to a paucity of pertinent structural
data.16,17 Thus, despite extensive but largely unrewarding efforts
to structurally characterize the intermediates involved in catalysis
by Pd(1),10b,15-18 rationalization has, by necessity, been achieved
by way of empirical correlation. In 1999 Trost and Toste17
replaced a first-generation mnemonic that simply predicted
torquoselectivity on the basis of ligand configuration,19 with a
second-generation three-dimensional model based coordination
of 1 to generate a chiral pocket around the η2-alkene/η3-allyl
coordination plane.17,20 The time-averaged structure of coordi-
nated ligand 1 is represented by a C2-symmetric folded surface,20
Figure 2, in which selectivity is based on a steric deactivation
mechanism. The four phenyl rings play key roles by adopting
pseudoaxial (‘wall’) and pseudoequatorial (‘flap’) orientations
(see Chart 1), with the ‘walls’ acting to selectively impede egress
2
Herein we report on a combined computational and H/13C
label-facilitated NMR analysis of Pd((R,R)-1) complexes bearing
η3-allyl and η3-cyclohexenyl moieties (7+ and 8+, Chart 2).The
structural details are then allied with computational and
experimental analyses of the generation of the η3-cyclohexenyl
complex 8+ from rac-cyclohexenyl acetate (()-(5), and the
pro-S selective addition of MCHE2, to develop a model for
the catalytic process (()-5f(S)-6, Figure 2. We also show that
the model may be of utility in the rationalization and prediction
of the much broader range of asymmetric allylic alkylation
reactions catalyzed by Pd(R,R)-1.3,4
(14) Trost, B. M.; Breit, B.; Organ, M. G. Tetrahedron Lett. 1994, 35,
5817.
(15) Butts, C. P.; Crosby, J.; Lloyd-Jones, G. C.; Stephen, S. C. Chem.
Commun. 1999, 1707.
(16) (a) Fairlamb, I. J. S.; Lloyd-Jones, G. C. Chem. Commun. 2000, 2447.
(b) Lloyd-Jones, G. C.; Stephen, S. C.; Fairlamb, I. J. S.; Martorell,
A.; Dominguez, B.; Tomlin, P. M.; Murray, M.; Fernandez, J. M.;
Jeffery, J. C.; Riis-Johannessen, T.; Guerziz, T. Pure. Appl. Chem.
2004, 76, 589.
2. Results and Discussion
To date, structural details for the key intermediates in the
asymmetric alkylation reaction mediated by Pd(1), namely the
η3-allylic complexes, have remained elusive.10,16,17,21 Indeed,
1H NMR spectra of even apparently simple Pd allyl complexes
of 1 are reported as uninterpretable17 due to, inter alia, overlap
with signals from numerous other species.16 We earlier reported
(17) Trost, B. M.; Toste, F. D. J. Am. Chem. Soc. 1999, 121, 4545 and
footnotes 15 and 18 and associated text therein.
(18) Amatore, C.; Jutand, A.; Mensah, L.; Ricard, L. J. Organomet. Chem.
2007, 692, 1457.
(19) Trost, B. M. Acc. Chem. Res. 1996, 29, 355.
(20) Trost, B. M.; Machacek, M. R.; Aponick, A. Acc. Chem. Res. 2006,
39, 747.
9
J. AM. CHEM. SOC. VOL. 131, NO. 29, 2009 9947

A R T I C L E S
Butts et al.
Chart 3. Enantiomerically Pure, d5-Phenyl Bearing Isotopologues
of Ligand (R,R)-1, Prepared for NMR Studies of 7+ and 8+
Scheme 2. Synthesis of P-Chiral d5-ortho-Diphenylphosphino
Benzoic Acids (RP)-d5-9 and (SP)-d5-9a
that the effect of temperature and concentration on both the net
optical rotatory power and the 31P{1H} NMR spectra of solutions
of η3-allyl complexes (7+) clearly indicates that the system has
a high propensity for reversible aggregation or oligomeriza-
tion;16,22 the η3-cyclohexenyl complex (8+), the key intermediate
involved in the conversion of cyclohexenyl acetate (5) to (S)-
6, Figure 2, behaves analogously. We also proposed that
catalytic turnover via these oligomeric complexes can substan-
tially reduce the selectivity in asymmetric allylic alkylation
reactions involving Pd(1), thus accounting for the inverse
dependence of selectivity and catalyst loading that has been
observed in such reactions.16 The goal of the study reported
herein was to study the structure and reactivity of the monomeric
Pd allyl complexes of 1, and three issues had therefore to be
addressed: first, the complexes must be generated free from
signal overlap by oligomeric and aggregated complexes, second,
it needed to be confirmed that the species studied were indeed
the monomeric complexes (7+) and (8+), and third a number of
deuterated forms of ligand (R,R)-1 were required to facilitate
the detailed NMR analysis of complexes 7+ and 8+, vide infra.
Synthesis of 2H-Labeled Ligand (R,R)-1. In addition to a
perdeuterated η3-cyclohexenyl system, vide infra, four enan-
tiomerically pure deuterated forms of 1 were preparedsd10-1a,
d20-1b, d10-1c, and d10-1d, Chart 3. The synthesis of each
required preparation of the appropriate d5- or d10-ortho-
diphenylphosphino benzoic acid (9) to combine with known
procedures for amide coupling with (R,R)-cyclohexane di-
amine.23 By adaptation of the procedure of Rauchfuss for the
synthesis of unlabeled 9 from PPh3,24 the perdeuterodiphenyl
derivative d10-9 was readily prepared from d15-PPh3 and then
coupled with the requisite amines23,25 to generate isotopically
desymmetrized ligand d10-1a and the symmetrical ligand d20-
1b.
a Conditions: (i) 100 equiv of H2NCH2CH2NH2;30 (ii) 8 equiv of KOH
in THF;31 (iii) BuLi, THF, -78 °C, AcOH.
The preparation of the benzoic acids (RP)-d5-9 and (SP)-d5-9,
P-chiral through isotopic substitution, proved significantly more
challenging. All attempts to adapt known methods for the
asymmetric synthesis of P-chiral phosphines using stereospecific
displacements of chiral auxiliaries26 failed. We thus developed
a procedure involving resolution, Scheme 2. Key to the strategy
was the use of a p-Br substituent on the nondeuterated phenyl
ring as a ‘traceless resolving handle’, so as to overcome the
inherent low-grade chirality arising solely from isotopic sub-
stitution. Using a tert-butyl ester to protect the carboxylic acid,
we assembled (()-d5-10 from (Et2N)2PCl, via a repeated
sequence of substituting of Cl with Ar, using ArMgX reagents,
then Et2N with Cl, using anhydrous HCl in ether.27 With (()-
d5-10 in hand, resolution was achieved by employing the method
of Ibers and Otsuka,28 for which the 1-naphthyl analogue 11a29
proved efficient for diastereomer separation via column
chromatography.
Both diastereomers of 11a were microcrystalline or amor-
phous, so P-configurations were assigned by X-ray diffraction
studies on a single crystal of the much more crystalline benzo
complex 11b·Et2O. This same crystal (ca. 3 µg) was then
dissolved in CD2Cl2 and analyzed by 31P{1H} NMR to allow
assignment of the 31P chemical shifts of (RC,RP)-11b/(RC,SP)-
11b. After decomplexation of the separated diastereomers of
11a, the absolute configurations of the resulting (RP)-d5-10 and
(SP)-d5-10 were assigned by 31P{1H} NMR by recomplexation
(21) Molecular mechanics structures of several allyl complexes of Pd-1
and analogues have been calculated: Helena Hagelin, “Palladium-
Catalyzed Aromatic Coupling and Allylic SubstitutionsAn Experi-
mental and Theoretical Study” Ph D. thesis, Royal Inst. of Technology,
Stockholm, Sweden, 1999.
(22) A detailed analysis of the anion, solvent, temperature and concentra-
tion-dependencies of the monomer-oligomer populations will be
reported in due course.
(26) For a recent review, see. (a) Johansson, M. J.; Kann, N. C. Mini-
ReViews In Organic Chemistry 2004, 1, 233.
(27) Markl, G.; Amrhein, J.; Stoiber, T.; Striebl, U.; Kreitmeier, P.
Tetrahedron 2002, 58, 2551.
(28) Tani, K.; Brown, L. D.; Ahmed, J.; Ibers, J. A.; Yokota, M.; Nakamura,
A.; Otsuka, S. J. Am. Chem. Soc. 1977, 99, 7876.
(23) Trost, B. M.; Van Vranken, D. L.; Bingel, C. J. Am. Chem. Soc. 1992,
114, 9327.
(29) Alcock, N. W.; Brown, J. M.; Hulmes, D. I. Tetrahedron Asym. 1993,
4, 743.
(24) Hoots, J. E.; Rauchfuss, T. B.; Wrobleski, T. B. Inorg. Syntheses 1982,
21, 175.
(30) Dunina, V. V.; Kuz’mina, L. G.; Rubina, M. Y.; Grishin, Y. K.; Veits,
Y. A.; Kazakova, E. I. Tetrahedron Asym. 1999, 10, 1483.
(31) Filali, E.; Lloyd-Jones, G. C.; Sale, D. A. Synlett 2009, 205.
(25) Correia, J. D. G.; Domingos, A.; Santos, I. Eur. J. Inorg. Chem. 2000,
1523.
9
9948 J. AM. CHEM. SOC. VOL. 131, NO. 29, 2009

Asymmetric Induction by the Trost Standard Ligand
A R T I C L E S
Scheme 3. Synthesis of Oligomeric Cationic Pd(1)-Allyl BAr′F
Complexes
Figure 3. Use of isotopically desymmetrized ligand (R,R)-d10-1a to
distinguish monomeric (n ) 1) from oligomeric (n > 1) forms of Pd
complexes (d10-7a+BAr′F-)n and (d10-8a+BAr′F-)n via 31P31PCOSY.
to generate 11b. Finally, tert-butyl ester hydrolysis31 then
protodebromination gave samples of (RP)-d5-9 and (SP)-d5-9
(g95% ee) which were coupled with (R,R)-cyclohexanedi-
amine23 to give (RCRCRPRP)-d10-1c and (RCRCSPSP)-d10-1d in
high diastereoisotopomeric purity.
ated from the isotopically desymmetrized ligand (R,R)-d10-1a,
Scheme 2. Key to the analysis is that any monomeric complexes
(n ) 1) in which ligand 1 is P,P-chelating can only be generated
as d10-isotopologues (one ArP(C6D5)2 and one ArP(C6H5)2
coordinate to the Pd). In contrast, any species generated by
ligand 1 acting as a bis-monodentate P-ligand, to generate
oligomeric rings or chains, must generate a statistical 1:2:1 ratio
of d0/d10/d20 isotopologues at the Pd centers in the oligomer,
Figure 3. The 31P nucleus in the P(C6D5)2 moiety experiences a
ca. -0.5 ppm net isotope shift upfield of that in the P(C6H5)2
moiety, and consequently the isotopologues are readily distin-
Preparation of Pd-Allyl Complexes 7+ and 8+. In earlier work
we found that coordinating counterions increase the equilibrium
population of higher-order species in the simple allyl complex
7+. Key to our present investigation has been the very low
interactivity of the [B((3,5-(CF3)2)C6H3)4]- anion (“BAr′F”)32
with Pd-allyl cations,5 which has allowed us to prepare mono-
meric complexes, relatively free of oligomer.16 As shown in
Scheme 3, reaction of the readily prepared allylic palladium
chloride complexes [(η3-C3H5)PdCl]2 (12)33 and [(η3-c-C6H9)-
PdCl]2 (13)33 with NaBAr′F affords the cationic complexes [(η3-
C3H5)Pd(MeCN)2][BAr′F] (14+[BAr′F-]) and [(η3-c-C6H9)Pd-
(MeCN)2][BAr′F] (15+[BAr′F-]) as white amorphous solids.These
stable precursors react cleanly with ligand (R,R)-1 in CH2Cl2
to afford the oligomeric complexes (7+BAr′F-)n and
(8+BAr′F-)n as glassy solids, in high yield (>96%). By
preparation of d9-cyclohex-2-enol from d10-cyclohexene, we also
generated, via d9-13 and d9-15, the perdeuterated η3-cyclohex-
enyl isomer (d9-8+BAr′F-)n, and using ligands d10-1a, d20-1b,
d10-1c, and d10-1d (Chart 3) the isotopologous complexes (d10-
7a)n, (d10-8a+BAr′F-)n, (d20-8b+BAr′F-)n, (d10-8c+BAr′F-)n, and
(d10-8d+BAr′F-)n.
2
guished by analysis of JPP patterns in the 31P31PCOSY (see
Supporting Information).
Monomeric η3-Allylic Complex 7+BAr′F-. 31P{1H} NMR
analysis of the concentration-dependence of the monomer-
oligomer distribution in (7+BAr′F-)n, showed that below a
threshold [Pd]TOTAL concentration of 23 mM in CD2Cl2, the
solution became sufficiently free of oligomers16 (<5%), to allow
detailed analysis of the monomeric complex 7+ by 1H, 13C, and
31P NMR. On complexation to the (η3-C3H5)Pd cation the two
phosphine groups in ligand 1 become inequivalent and an ‘AB’
spin system (2PdJPP) will be generated in the 31P{1H} NMR
spectrum. If the ligand maintains a time-average C2-symmetry,
then the two 180° related orientations of the allyl group
(formally ‘rotamers’) results in a degenerate pair of complexes,
and while interconversion will result in exchange of PA and PB
nuclei, a single complex should be observed. However, the
31P{1H} NMR spectrum contains two distinct monomeric species
(exo-7+ and endo-7+, both with JPP ) 35 Hz), present in equal
ratio (K ) 1.0), showing that, at the NMR time scale, the
conformation of 1 in the 13-membered chelate does not have
C2-symmetry. A combination of 1H TOCSY, 1H,31P-HMBC, and
1H,13C-HMQC allowed a complete assignment of the δ- and
J-values in the P,P-Pd-η3-allyl units in the two isomers, Figure
4, and showed that there is no significant difference in allyl
structure between isomers and no significant difference in
Oligomeric complexes (7+BAr′F-)n and (8+BAr′F-)n dissolve
in d8-THF and CD2Cl2 to give solutions containing both the
monomeric (n ) 1) and oligomeric (n > 1) forms. These were
unambiguously distinguished via analysis of complexes gener-
(32) (a) Brookhart, M.; Grant, B.; Volpe, A. F. Organometallics 1992, 11,
3920. (b) BAr′F engages in ‘very modest but not zero ion pairing’,
see. (c) Nama, D.; Butti, P.; Pregosin, P. S. Organometallics 2007,
26, 4942.
(33) (a) Smidt, J.; Hafner, W. Angew. Chem. 1959, 71, 284. (b) Imaizumi,
S.; Matsuhisa, T.; Senda, Y. J. Organomet. Chem. 1985, 280, 441.
(c) Auburn, P. R.; Mackenzie, P. B.; Bosnich, B. J. Am. Chem. Soc.
1985, 107, 2033.
9
J. AM. CHEM. SOC. VOL. 131, NO. 29, 2009 9949

A R T I C L E S
Butts et al.
Figure 4. Color-coding showing the static P,P-Pd-η3-allyl unit during interconversion (k ) 2.6((0.3) s-1, CD2Cl2, 25 °C) of monomeric exo-7+ and
endo-7+ with selected NMR data (δ/ppm, J/Hz) and NOE contacts.
bonding/hybridization between the allyl termini within each
isomer. Using 2D EXSY techniques (1H1H, 31P31P, and 13C13C)
the two isomers (exo-7+ and endo-7+) were found to interconvert
via a pathway that preserves the entire set of stereochemical
relationships between the nuclei within the P,P-Pd-η3-allyl unit,
see color coding in Figure 4. This striking feature rules out all
of the well-known isomerization paths of Pd-allyls (η1-η3
interconversion, ligand dissociation, Berry pseudorotation)34 as
these would result in syn/anti exchange of allyl protons and/or
cis/trans exchange between allyl termini and the P-donors. The
isomerization process must thus involve a conformational change
of the ligand 1, which does not actively involve the P,P-Pd-
η3-allyl unit.
The origins of this conformational change were elucidated
by 1D-1H NOE experiments which allowed identification of
contacts that were distinct and complementary between the two
isomers. Thus in exo-7, strong correlations were present between
one syn allyl proton (red) and one of the two amide protons,
and between both anti allyl protons (blue and green) and
aromatic protons. In contrast, correlations in endo-7 were present
between one anti allyl proton (green) at the opposite allyl
terminus and one of the two amide protons, and between both
syn allyl protons (magenta and red) and unspecified aromatic
protons. In both isomers, there is a ca. 1.5 ppm difference in
chemical shifts between the two amide protons and in both cases
the NOE contacts to the allyl protons are from the higher field
NH protons (ca. 5.8 ppm), which shift to a lower field
environment (ca. 7.3 ppm) on isomer interconversion. The only
explanation that accounts for all of the above NMR data is that
the ligand conformation results in a shallow curved surface being
formed around the Pd-allyl moiety, with one amide NH (ca.
5.8 ppm) on the concave surface, and the other on the convex
surface; isomerization then involves inversion of the conforma-
tion such that the concave surface becomes the convex, and
vice versa.
Monomeric η3-Cyclohexenyl Complex 8+BAr′F-. Analogous
31P{1H} NMR studies conducted with the η3-cyclohexenyl
complex (8+BAr′F-)n revealed two distinct differences to
(7+BAr′F-)n. First, there was a far higher propensity for
oligomerization,22 resulting in much lower concentration samples
being required to obtain oligomer-free solutions of 8+ (4.2 mM
in CD2Cl2, or 2.0 mM in d8-THF). Second, instead of there being
2Pd
two isomers present (cf. exo/endo-7+), a single AB system (
J
PP
) 34 Hz) was apparent in the 31P{1H} NMR spectrum, with no
evidence for resolution into two species, even on cooling to
-50 °C.
Detailed analysis of the H NMR spectrum of [8]+[BAr′F]-
1
proved challenging. There was substantial signal overlap (there
are 40 aryl CH protons, including BAr′F, in the range δH
6.6-7.7 ppm and 14 aliphatic protons in the range δH 1.2-2.2
ppm), as well as rapid T1,2 relaxation, which rendered 1D
TOCSY, 2D HSQC, and HBMC experiments uninformative and
thwarted extraction of H,H coupling constants. Nonetheless, 2D
COSY, TOCSY, and NOESY analyses, assisted by 2H-labeled
complexes (Scheme 3), allowed assignment of the chemical
shifts of the cyclohexenyl moiety, the ligand scaffold and a
partial assignment of the aromatic protons (see Supporting
(34) See for example: (a) Vrieze, K. In Dynamic Nuclear Magnetic
Resonance Spectroscopy; Jackman, L. M., Cotton, F. A., Eds.;
Academic Press: New York, 1975. (b) Albinati, A.; Kunz, R. W.;
Ammann, C. J.; Pregosin, P. S. Organometallics 1991, 10, 1800. (c)
Hansson, S.; Norrby, P.-O.; Sjo¨gren, M.; Åkermark, B.; Cucciolito,
M. E.; Giordano, F.; Vitagliano, A. Organometallics 1993, 12, 4940.
(d) Breutel, C.; Pregosin, P. S.; Salzmann, R.; Togni, A. J. Am. Chem.
1
Information for full details). As with 7+, the H NOESY and
¨
Soc. 1994, 116, 4067. (e) Gogoll, A.; Ornebro, J.; Grennberg, H.;
amide chemical shift data was indicative of a curved ligand
surface. The central allylic proton (red in Figure 5) has an NOE
contact to the C(H)N unit in the cyclohexane scaffold (ring C),
the chemical shift of the adjacent amide proton (7.09 ppm)
indicates that this methine proton (C(H)N) is on the concave
surface of the complexed ligand. Three of the four protons on
methylene groups adjacent to the allyl termini in the η3-
Ba¨ckvall, J.-E. J. Am. Chem. Soc. 1994, 116, 3631. (f) Pregosin, P. S.;
Salzmann, R.; Togni, A. Organometallics 1995, 14, 842. (g) Malet,
R.; Moreno-Man˜as, M.; Parella, T.; Pleixats, R. J. Org. Chem. 1996,
61, 758. (h) Ramdeehul, S.; Barloy, L.; Osborn, J. A.; De Cian, A.;
Fischer, J. Organometallics 1996, 15, 5442. (i) Boog-Wick, K.;
Pregosin, P. S.; Trabesinger, G. Organometallics 1998, 17, 3254. (j)
Gogoll, A.; Grennberg, H.; Axe´n, A. Organometallics 1998, 17, 5248.
(k) Solin, N.; Szabo´, K. J. Organometallics 2001, 20, 5464.
9
9950 J. AM. CHEM. SOC. VOL. 131, NO. 29, 2009

Asymmetric Induction by the Trost Standard Ligand
A R T I C L E S
Figure 5. Key NOE contacts for exo-8+, d9-8+, (d20)-8b+, (d10)-8c+, and (d10)-8d+ where gray indicates a deuterated moiety and associated absence of NOE.
Note that all ortho protons experience time-average NOE contact. See Supporting Information for 2D NOESY data.
cyclohexenyl rings display NOE contacts (blue and green in
Figure 5) to aryl rings.
hybrid DFT methods, like B3LYP.35 Such calculations are time-
consuming, requiring days for each optimized geometry with
complexes of the current size. To circumvent this problem, fast
molecular mechanics (MM) methods can be utilized for the
conformational searches, and only the most promising structures
are then subjected to more rigorous DFT investigations. For
this approach to be successful, the MM method must be reliable;
use of “off the shelf” unvalidated MM is not recommended for
metal complexes. In the current work, we do all conformational
searches with a modified MM method that has previously been
found to have good accuracy for (η3-allyl)Pd complexes.36 Final
energies come from DFT optimizations in solvent, as described
in the Computational Methods section. Structures 7+ and 8+
yielded very similar results, except that for 7+, there was very
little difference between endo and exo structures in terms of
the number of conformations found or their energies, whereas
for 8+, endo structures were significantly fewer and higher in
energy. The conformational ensemble for 8+ up to 18 kJ mol-1
above the global minimum is depicted in Figure 6. The
corresponding ensembles for 7+ can be visualized as taking the
exo ensemble for 8+ and inserting either an exo- or endo-η3-
allyl moiety in place of the exo-η3-cyclohexenyl. It can be seen
that all low energy conformations are remarkably similar, folding
the backbone cyclohexane into close proximity with the Pd-
allyl moiety. Only two conformations in this energy range
display a fold with the backbone endo to the η3-cyclohexenyl
unit. The only other major difference between the different folds
is in the orientation of one of the amide units. For several
conformations, the amide N-H is pointing into the concave
embrasure, close to one allyl terminus (yellow hydrogen in
Figure 6), whereas for the others, only carbonyl oxygens are
pointing inward toward the Pd-allyl moiety. It can also be seen
that for the endo conformations, the amide units are far from
the reactive allyl termini. Other, more minor differences arise
from the orientations of the phenyl groups combined with minor
rotation around the Pd-P bonds. Interestingly, one side of the
η3-cyclohexenyl moiety interacts mainly with a backbone
The aryl rings exhibiting NOE contacts with the η3-cyclo-
hexenyl ring were identified as simple Ph groups (rings B, D,
E, or G) by the absence of these NOE contacts in the
perdeuterophenyl complex (d20)-8b+. The use of the d9-8
complex in which the η3-cyclohexenyl ring is perdeuterated
confirmed that the NOE contacts with the Ph rings are from
methylene groups in the η3-cyclohexenyl ring, rather than
methylenes in the cyclohexane ligand scaffold (ring C). The
triangulating set of contacts (red, blue, and green in Figure 5)
indicate that the orientation of the η3-cyclohexenyl ring is
analogous to exo-7+, identifying the single species observed for
[8]+[BAr’F]- as exo-8+. Since the two isomers of the simple
allyl complexes exo/endo-7+ are isoenergetic, there must be a
substantial destabilization in endo-8+ caused by the allyl unit
being substituted at one or both anti positions. Computational
studies, vide infra, identified this destabilization as arising from
steric strain being induced in endo-8+ by clash of the ligand
scaffold (ring C) with the two methylene units at the allyl termini
in the η3-cyclohexenyl ring.
The orientation of the phenyl rings in the Ph2P units was
1
explored by way of comparison of the 2D H NOESY data of
unlabeled complex 8+ with that obtained for diastereoisotopo-
meric complexes (d10)-8c+ and (d10)-8d+, Figure 5. With (RP,RP)-
based ligand 1c, the NOE between the methylene group at the
pro-R allyl terminus in (d10)-8c+ was absent, clearly identifying
the NOE (green protons) in exo-8+ as arising from phenyl ring
B bisecting the methylene group. Analogously, with (SP,SP)-
based ligand 1d, the NOE between the pseudo-axial proton on
the methylene group at the pro-S allyl terminus in (d10)-8d+
was absent, identifying this NOE (blue protons) in exo-8+ as
arising from proximity to phenyl ring D.
Computational Investigation of Complexes 7+ and 8+. The
structures of the Pd complexes 7+ and 8+ pose nontrivial
computational problems. Ligand 1 is large and flexible, neces-
sitating a thorough and unbiased search of several thousand
plausible conformations. On the other hand, transition metal
complexes in general require advanced quantum chemical
methods in order to obtain structures and energies with an
accuracy that is sufficient to allow meaningful comparisons to
solution NMR data. Today, the methods of choice are usually
(35) (a) Lee, C.; Yang, W.; Parr, R. G. Phys. ReV. B 1988, 37, 785. (b)
Becke, A. D. J. Chem. Phys. 1993, 98, 5648. (c) Stephens, P. J.;
Devlin, F. J.; Chabalowski, C. F.; Frisch, M. J. J. Phys. Chem. 1994,
98, 11623.
(36) Hagelin, H.; Åkermark, B.; Norrby, P.-O. Organometallics 1999, 18,
2884.
9
J. AM. CHEM. SOC. VOL. 131, NO. 29, 2009 9951

A R T I C L E S
Butts et al.
Figure 6. Two views of the MM conformational ensemble of 8+, up to 18 kJ mol-1. C-H hydrogens are hidden, the η3-cyclohexenyl and amide moieties
are shown as tubes, and Pd in CPK style. Left hand image ‘front view’; right-hand image ‘side view’ of 8+, see arrow in ‘front view’. In the ‘side view’ the
favored conformer has an amide hydrogen (yellow) close to the pro-S allyl terminus.
Figure 7. DFT-optimized lowest energy conformers for exo-7+, endo-7+ and exo-8+. Key NOE contacts color coded as in Figures 4 and 5. Allyl termini
are highlighted in pale orange. All heteroatoms (O, N, Pd, P) are colored dark gray.
benzene, not the pendant phenyl groups, in most low energy
conformers. Attempts to enforce ligand C2-symmetry or any
geometry where the four pendant phenyl groups are close
enough to the η3-allyl moiety to give a structure with any
resemblance to that in Figure 2, results in steric energies more
than 100 kJ mol-1 above the global minimum. Earlier rational-
izations of selectivities with ligand 1 have predominantly been
built on the assumption that it is the pendant phenyl groups
that interact with the η3-allyl moiety or the nucleophile.20 The
structures determined in the current work indicate that additional
control mechanisms are active.
catalytic reaction, and specifically to the selectivity for delivery
of malonate to the pro-S carbon of the η3-cyclohexenyl ring to
generate (S)-6, we reacted malonate nucleophile (CHE2) with
monomeric [8+][BAr′F] in THF and measured the ee of the
alkylkation product 6 by chiral HPLC,38 ‘ee obs’, Table 1. These
stoichiometric reactions were conducted with a range of escort
ions (M+) to the malonate nucleophile and gave selectivities
that perfectly matched the trend reported for the catalytic
reaction4,9 (Figure 2). Thus with Li+ (Table 1, entry 1) near-
racemic cyclohexenyl malonate 6 was obtained and the selectiv-
ity increased progressively as M+ was changed from Li to Na,
K and Cs (Table 1, entries 1, 2, 7, and 8). With Bu4N+ as escort
ion, (S)-6 was obtained in very high ee (entry 9).
Stoichiometric reaction of NaCHE2 or Bu4NCHE2 with
oligomeric ([8+][BAr′F])n in which ca. 93% of (R,R)-1 is
complexed in a binuclear nonchelate mode, gave 6 with low
selectivity (entries 4 and 10), the sense of induction actually
being slightly reversed with Na as escort ion, in both cases
confirming that catalytic flux via the monomer (8+) is essential
for attaining high selectivity.
The lowest energy conformations of both 7+ and 8+ were
optimized using DFT, as described in the Computational
Methods section, validating the gross structural features and
energy ranking of the MM conformations. The most populated
conformations were used to map expected NOE contacts.37
Comparison of these maps to the NOESY data showed excellent
agreement (Figure 7) reinforcing the assignment of the experi-
mental structures.
Alkylation of Complex exo-8+. To ensure that our structural
studies on the isolated complex exo-8+ were pertinent to the
(38) The use of dibenzyl malonate, rather than dimethyl malonate, facilitates
reliable chiral HPLC separation of the enantiomers of 6 (E ) CO2Bn),
see: (a) Trost, B. M.; Radinov, R. J. Am. Chem. Soc. 1997, 119, 5962.
(37) Hagelin, H.; Åkermark, B.; Norrby, P.-O. Chem.sEur. J. 1999, 5,
902.
9
9952 J. AM. CHEM. SOC. VOL. 131, NO. 29, 2009

Asymmetric Induction by the Trost Standard Ligand
A R T I C L E S
Table 1. Enantiomeric Excesses (ee’s) in Stoichiometrica
Alkylation of Monomericb [8+] with Excess [MCHE2] (M ) Li, Na,
K, Cs, Bu4N; E ) CO2Bn) to Give (S)-6, Compared to the
Selectivity Predicted by DFT Calculations
abstracting acetate ion from Pd-(allyl)OAc complexes,5 in this
case continually regenerating [8+][BAr′F]. Consistent with this
conclusion, for reactions with the Bu4N+ escort-ion, the nature
of X (BAr′F, OTf, Cl, or AcO) does not impact significantly
on the selectivity (entries 9, 11, and 12).
Earlier rationales for selectivity in the alkylation of cycloalk-
enyl esters (Figure 2) were based on steric deactivation of the
nucleophile by phenyl rings.20 As is evident from the NMR
and computational data outlined above, in no case can any of
these rings reach past the allyl moiety to influence the incoming
nucleophile. Any influence of these substituents on the reaction
selectivity would be expected to come from induced torquese-
lectivity41 caused by specific interactions with the η3-cyclohex-
enyl moiety. The DFT-optimized structure for exo-8+ shows
the closest phenyl ring to the η3-cyclohexenyl ring to be ring B
in Figure 5, which gives rise to the NOE contacts highlighted
in green in Figure 7. This interaction causes a slight rotation of
η3-cyclohexenyl ring (anticlockwise when viewed with the η3-
C6H9 moiety in front of the Pd) such that one allyl terminus is
located slightly above and the other slightly below the formal
square plane around Pd, see central structure in Figure 8. On
its own, this induced torqueselectivity41 will favor nucleophilic
attack of malonate anion (CHE2) at the pro-R terminus as this
leads to continued anticlockwise rotation of the η3-cyclohexenyl
ring during formation of the new C-C bond in the nascent η2-
bound product (R)-6. However, this (R)-selectivity is opposite
to that observed under the catalytic9,42 and stoichiometric
conditions (Figure 2 and Table 1), where (S)-selectivity ranges
from near-racemic to near-perfect, depending on the identity
of the ‘escort-ion’ to the nucleophile (M+) and, in some cases
the identity of X-, Table 1.
entry
M+
X-
ee obsc
ee calcdd
1
Li
BAr′F
BAr′F
BAr′F
BAr′F
OTf
3
-16
67
-
2
Na
44 (49)a
85
3e
4f
5
Nae
Na
-16f
-
Na
Na
K
Cs
Bu4N
Bu4N
Bu4N
Bu4N
83 (83)a
93 (87)a
76
-
6
7
8
9
Cl
-
BAr′F
BAr′F
BAr′F
BAr′F
OTf
84
>99
>99
-
89
98g (95)a
6f
10f
11
12
90 (94)a
92 (95)a
-
-
Cl
a Values in parentheses are for catalytic alkylations employing 2.5
b
mol % [8+][X] (1.0 mM) and (S)-cyclohexenyl acetate. [8+][BAr′F],
1.0 mM, >95% monomeric. For [8+][OTf] and [8+][Cl] (prepared in situ
from 13/15+[OTf-] and (R,R)-1) some oligomer is present. c ee of (S)-6
by chiral HPLC on AD-H column. d ee of (S)-6 calculated from
DFT-derived ∆E between TS for addition of MCHE2 to exo-8+. e 20%
f
Bu4N, 80% Na. Stoichiometric reaction employing 27 mM [8+][BAr′F]
(7% monomeric). g Analogous reactions conducted after pre-equili-
bration of 1 mM [8+][BAr′F] with 100 mM MeOL (L ) H or D) in
THF (1H NMR analysis shows full exchange NH to NL in [8+][BAr′F])
gave identical enantioselectivity, within experimental error (four runs
with L ) H, four runs with L ) D).
To investigate the underlying causes for the above selectivity
trend, we located approximate transition states for addition of
malonates with a range of escort ions to exo-8+. As discussed
in the Computational Methods section, a continuum solvation
model is both necessary and sufficient for accurate location of
these very polar transition states. From the studies above, we
considered the common computational practice of excluding
the counterion to exo-8+ from the calculations, but the malonate
escort ion M+ must be included since one of the goals of the
Addition of catalytic (20 mol %) Bu4NCHE2 to the reaction
of NaCHE2 with [8+][BAr′F] raised the selectivity for (S)-6
substantially (compare entries 2 and 3), suggesting at least a
ca. 3-fold selectivity for attack by Bu4NCHE2,39 which can be
regenerated via equilibrium: [Bu4N][BAr′F] + NaCHE2
f
Bu4NCHE2 + NaBAr′F.5 With the concept that greater dis-
sociation of the nucleophile from the escort-ion (M+) is a key
component for facilitating a monomer-specific actiVated path-
way that increases pro-S selectivity, we compared the effect of
X-, the counterion to the Pd complex (X ) BAr′F, OTf, Cl),
on reactions involving Na+ and Bu4N+ escort ions. Consistent
with this concept of dissociation, increasing the strength of the
interaction between Na+ and X-, so as to increase the dissocia-
tion of the Na+ from the CHE2 anion, gave a substantial rise in
selectivity (compare entries 2, 5, and 6), despite the increase in
oligomer concentration due to the change from X ) BAr′F to
OTf and Cl. Moreover, with the Bu4N+ escort-ion, which
interacts less strongly, there was little effect (compare entries
9, 11, and 12).39
(40) To avoid complications arising from the memory effects that are known
to attend these reactions under some conditions (see ref 10), we
analysed the enantiomeric excess of 6 after ca. 10 turnovers, at which
point the conversion of cyclohexenyl acetate (5) is e50% after which
turnover rate drops precipitously. In this manner, the powerful kinetic
resolution for this process leads to selective study of the outcome of
generation and attack on [8+][OAc-] via the matched manifold
involving (S)-5. For reactions involving Bu4N as escort ion, subsequent
slow turnover via the mismatched manifold proceeded without drop
in selectivity. In contrast, and as expected,10 with reactions involving
Na+ as the escort ion, the selectivity dropped after the onset of turnover
via the mismatched manifold.
(41) Induced rotation of the allyl moiety can have a major influence on
the relative reactivity of the two allyl termini: Oslob, J. D.; Åkermark,
B.; Helquist, P.; Norrby, P.-O. Organometallics 1997, 16, 3015.
(42) For cyclopentenyl substrates the same trend in M+ is observed, except
that reaction with LiCHE2 gives rise to the opposite enantiomer of
product in moderate selectivity (ca. 40% ee R; see ref 19). Due to the
propensity of Pd η3-cyclopentenyl species to undergo ꢀ-H elimination
to generate cyclopentadiene (see: (a) Fairlamb, I. J. S.; Lloyd-Jones,
G. C.; Vyskocˇil, S.; Kocˇovský, P. Chem. Eur. J. 2002, 8, 4443.) We
have only studied the η3-cyclohexenyl complexes. However, the
mechanistic rationale presented herein is fully consistent with the
results obtained with η3-cyclopentenyl substrates under catalytic
conditions. The difference between the two systems is simply that
the ‘crossing point’ where the net selectivity for pro-R attack by CHE2
begins to dominate over pro-S attack is at M ) Li for η3-cyclohexenyl,
but between M ) Li/Na for η3-cyclopentenyl.
When [8+][X] was used as a catalyst for the reaction of
cyclohexenyl acetate (5),40 the selectivities (ee values in
parentheses in entries 2, 5, 6, 9, 11, and 12 in Table 1) were
very similar to those obtained in the stoichiometric reactions.
For the reactions involving NaCHE2 (entries 2, 5 and 6) this
initially appears to contradict the above conclusions, as under
catalytic conditions X should be an acetate ion. However, we
have previously shown that catalytic NaBAr′F (which will be
generated from NaCHE2 + [8+][BAr′F]) is highly efficient at
(39) Madec, M.; Prestat, G.; Martini, E.; Fristrup, P.; Poli, G.; Norrby,
P.-O. Org. Lett. 2005, 7, 995.
9
J. AM. CHEM. SOC. VOL. 131, NO. 29, 2009 9953

A R T I C L E S
Butts et al.
Figure 8. (Center) DFT-optimized structure of exo-8+, with allyl termini highlighted in pale orange. The induced (ring B) pro-R (anticlockwise) torquoselectivity
is indicated with arrow. (Outer Images) DFT-optimized transition states (TSs) for pro-R and pro-S addition of NaCHE2. The attacking central carbon of
malonate is highlighted in pale orange, resulting torquoselection with curved arrow, and the key enolate-amide hydrogen bond with a bold arrow.
current study is to reproduce the selectivity-trend in Table 1.
We utilized the global minimum energy conformation of exo-
8+ together with a malonate chelating with both carbonyl
oxygens to the escort ion (M+).39,43 This still left three possible
staggered approaches of the malonate to each allyl terminus.
Very interestingly, with most escort ions, one TS had a
substantially lower energy than any alternative approach,
illustrated for the Na+ malonate in Figure 8.44 In this pro-S TS,
leading to the experimentally observed (S)-6, we find a strong
hydrogen bond between one malonate enolate-oxygen and the
amide hydrogen (highlighted with arrow in Figure 8). Simul-
taneously, the escort ion gains some stability from an interaction
with the proximal backbone benzene moiety. Likewise, for some
pro-R TSs, the escort ion could gain stability by a long-range
dipole-ion interaction with the amide carbonyl pointing into the
concave face (Figure 6) in close proximity to the allyl terminus.
In calculations without the continuum solvation model (gas
phase), which are not discussed further, this interaction becomes
a strongly exaggerated coordination.
shown in Table 1, entry 1. Such a good agreement is surely
fortuitous; due to the problem of intersecting continuum cavities
at the TS, we see apparently random energy variations of a few
kJ mol-1 between virtually identical structures. However, the
trend itself is not sensitive to these small variations, and it
matches well the experimental trend: compare ee obs and ee
calcd values for entries 1, 2, 7, 8, and 9, in Table 1. We note
that H-bond mediated delivery of the nuclophile requires that
it possess a strongly negative hydrogen bond acceptor, in a 1,3-
relationship to the reactive (nucleophilic) site. This feature,
present in for example carboxylates, 1,3-dicarbonyls, carbon-
ates,46 or phthalimides, is found in nearly all known cases where
ligand (R,R)-1 leads to high selectivity for attack at the pro-S
allyl terminus, or corresponding position in analogous cyclic
intermediates.20 We also note that the energy difference between
endo-8+ and exo-8+ is not a requirement for high selectivity
with cycloalkenyl substrates. In the endo isomer, the cycloalk-
enyl ring will block delivery of the nucleophile to the pro-R
allyl terminus via the amide hydrogen-bond route and therefore
the reaction will be funneled through the exo isomer, provided
that the conformational exchange is fast relative to nucleophilic
attack.45 Moreoever, in 2-substituted cycloalkenyl substrates,
which will generate η3-intermediates analogous to 8+ but with
a substituent at the central allyl carbon, the amide-H-bond
mediated delivery of the nucleophile will encounter steric strain
in both the exo and endo isomers, while for open-chain allylic
substrates, H-bond mediated delivery of the nuclophile is not
strained for either isomer. Both factors may help to rationalize
the lower selectivities that are obtained with these types of
substrates.20
For weakly coordinating escort ions like Cs+, the hydrogen
bond interaction dominated, favoring the pro-S TS by ca. 22 kJ
mol-1 in THF (18 kJ mol-1 in CH2Cl2).45 As discussed in the
Computational Methods section, the magnitude of this selectivity
is exaggerated due to the missing vibrational entropy, which in
reality must favor the looser pro-R TS; the experimental
selectivity (ee obs, Table 1) corresponds to a free energy
difference of ca. 8 kJ mol-1, although this value itself is
compromised by some competing low-selectivity reaction
occurring via the oligomer. Using more strongly coordinating
escort ions will attenuate the hydrogen bond and at the same
time increase the importance of the amide carbonyl interaction,
so that for Li+, the calculated difference between the two paths
is less than 1 kJ mol-1, in perfect agreement with the result
It was recently shown that Pd complexes of (R,R-1) and
analogues retain high selectivity in the catalyzed reaction of
cycloalkenyl and simple linear allyl esters with “hard” nucleo-
philes, such as lithium or tin enolates,7 and lithium methylpy-
ridine trifluoroborates.8 Under such strongly basic conditions,
ligand N-H deprotonation would lead to enhanced interaction
of the carbonyl with the escort ion M+ (Li, Sn), thus augmenting
selectivity Via this pathway.47
Generation of Complex exo-8+ from Cyclohexenyl Esters.
Ligand 1 is known for a dual mode of reactivity; not only will
it induce a high selectivity in nucleophilic attack on cycloalkenyl
substrates, but it also achieves powerful kinetic resolution of
chiral substrates and is able to efficiently desymmetrize meso
(43) (a) Norrby, P.-O.; Mader, M. M.; Vitale, M.; Prestat, G.; Poli, G.
Organometallics 2003, 22, 1849. (b) Svensen, N.; Fristrup, P.; Tanner,
D.; Norrby, P.-O. AdV. Synth. Catal. 2007, 349, 2631. (c) Fristrup,
P.; Ahlquist, M.; Tanner, D.; Norrby, P.-O. J. Phys. Chem. A 2008,
112, 12862.
(44) Due to the stabilization provided by the continuum model, it was also
possible to find completely “open approaches” without interaction with
any ligand moiety. These open approaches were insignificant for pro-S
attack where these had a higher energy than the amide H-bonded
transition states. For pro-R attack, these open approaches were more
prevalent and it was necessary in all cases to locate both types of
transition states and compare their energies.
(45) Higher enantioselectivity is obtained in CH2Cl2 versus THF under
catalytic conditions (see ref 9). This may arise from a combination of
factors, including less oligomerization of 8+ and lower MCHE2
concentration in CH2Cl2, the latter allowing more efficient equilibration
of oligomeric 8+/endo monomeric 8+ with exo-8+ relative to nucleo-
philic attack.
(46) For the use of [HCO3]- as nucleophile for the asymmetric ‘hydrolysis’
of cycloalkenyl esters to the corresponding alcohols, via decarboxy-
lation of an intermediate cycloalkenyl hydrogen carbonate ester, see.
(a) Lu¨ssem, B. J.; Gais, H.-J. J. Am. Chem. Soc. 2003, 125, 6066.
9
9954 J. AM. CHEM. SOC. VOL. 131, NO. 29, 2009

Asymmetric Induction by the Trost Standard Ligand
A R T I C L E S
substrates.Thus, in the presence of Pd(R,R)-1, (S)-cyclohexenyl
acetate 5 reacts up to 2 orders of magnitude faster than the (R)-
enantiomer,11 and in the meso-forms of 3,6-X2-cyclohexene (X
) carboxylate leaving groups such as ester, carbonate, N-tosyl
carbamate,), it is the X group on the (S)-configured allylic carbon
that is expelled with very high selectivity.20 The carboxylate
ionization is the microscopic reverse of nucleophilic attack on
the (η3-allyl)Pd complex; in fact, the reaction is reversible5,48
and Pd(R,R)-1 catalyzed substitution of, e.g., trichloroethyl
carbonate by benzoate proceeds with very high selectivity.20
The same type of computational approach can be applied here
as was used for the nucleophilic attack. One major difference
is that the overall system is net neutral and ionogenic: the two
neutral fragments Pd(R,R-1) and (R)-5/(S)-5 react to give exo-
8+ and an acetate anion, making the use of the continuum model
absolutely essential.37
Looking first at the neutral preionization η2-allyl ester
complexes, we can see that already in the ground state, the
acetate carbonyl of (S)-5 accepts a hydrogen bond from the
amide hydrogen on the concave side of 1, whereas no corre-
sponding stabilization is available for (R)-5. The hydrogen bond
stabilizes the [η2-(S)-5]Pd(1) complex by 14 kJ mol-1 relative
to the [η2-(R)-5]Pd(1) diastereomer. Upon ionization, this energy
difference increases due to the increased negative charge on
the acetate, until at the TS, Figure 9, we calculate an energy
difference of 38 kJ mol-1 between the isomers. As was the case
with nucleophilic attack, vide supra, this energy difference is
exaggerated by our neglect of entropic contributions; the
experimental selectivity corresponds to a free energy difference
that is about half of what we calculate. meso-3,6-Diacetoxycy-
clohexene gives results that are entirely consistent with those
from monoacetate 5. Upon generation of the neutral preioniza-
tion η2-allyl diester complex, only one of the two enantiotopic
acetate groups can engage in hydrogen bonding to the amide;
this acetate will be selectively ionized, due to the stabilization
of the leaving acetate anion through hydrogen bonding.
Figure 9. ‘Side view’ of DFT-optimized TS structures leading to exo-8+
via ionization of mismatched cyclohexenyl acetate (R)-5 (upper) and matched
cyclohexenyl acetate (S)-5 (lower). The approximate motions of the acetate
groups leading to ionization are indicated with arrows.
lonate (S)-6 in up to 98% ee, depending on the identity of M+
and, in some cases, X– (ee obs, Table 1) in excellent agreement
with Pd(R,R)-1 catalyzed reactions involving cyclohexenyl
acetate (5).9
P,P-Coordination of ligand 1 in 7+ and 8+ was then explored
computationally, using a modified molecular mechanics method
to sift out favored conformers of the very flexible 13-membered
chelate, which were then refined by DFT. The outcome was
very similar for 7+ and 8+, in both cases revealing the
13-membered chelate ring to be puckered such that the backbone
cyclohexane is placed in close proximity to the Pd-allyl
moiety.49 There was an excellent correlation between the DFT
derived structures and the experimental NOESY data (Figure
7), cross-validating both approaches. The proximity of the amide
groups to the Pd center also accounts for the known instability14,50
of the Pd(0) complexes of 1.
3. Summary
By exploiting the concentration-dependency and reversibility
of the oligomerization, in concert with the exceptionally low
interactivity of the B[((3,5-(CF3)2)C6H3)4]- anion (“BAr′F”)33
with Pd-allyl cations,5 we have prepared monomeric cationic
complexes [(η3-C3H5)Pd(R,R)-1]+ (7+, two isomers: exo/endo)
and [(η3-c-C6H9)Pd(R,R)-1]+(8+, single isomer: exo) that are
sufficiently free of oligomer to permit their detailed study by
1H, 13C, and 31P NMR spectroscopy. Employing deuterium
labeling (Chart 3, Schemes 2 and 3), the complexes were shown
to exist as monomeric chelates (Figure 3) with structures that
could be partially eluciated by a study of NOE contacts (Figures
4 and 5). Monomeric complex exo-8+ was shown to react with
malonate nucleophile (MCHE2) to give the cyclohexenyl ma-
Nucleophilic attack on exo-8+ by malonate (MCHE2) was
then studied by DFT. For the hypothetical addition of a free
malonate anion (no M+), an H-bonding interaction between
the enolate oxygen of the malonate and the amide NH on the
concave surface of Pd-coordinated (R,R)-1 was found to guide
the enolate carbon to the proximal (pro-S) terminus of the η3-
c-C6H9 unit with perfect selectivity. Introduction of metal
counterions resulted in 1,3-dicarbonyl chelation of M+ by the
malonate and a resulting attenuation of the H-bonding interaction
of the enolate with the amide. Simultaneously, there were
(47) Ligand carbonyl coordination of the Li escort ion in lithium 2-py-
ridylmethyl trifluoroborate (LiHMDS + methyl pyridine + BF3) would
augment the inherent (R)-torquoselectivity for attack in exo-8.
However, with (R,R)-1, pro-S selectivity still dominates (70 pro-S/30
pro-R). Nonetheless, in agreement with the preceding discussion noting
the possible interaction of the escort ion with aromatic rings,
replacement of the cyclohexane scaffold in (R,R)-1 with a 9,10-
dihydroanthracene moiety results in highly selective nucleophilic attack
at the pro-R carbon.
(48) (a) Amatore, C.; Jutand, A.; Mensah, L.; Meyer, G.; Fiaud, J.-C.;
Legros, J.-Y. Eur. J. Org. Chem. 2006, 1185. (b) Amatore, C.; Gamez,
S.; Jutand, A.; Meyer, G.; Moreno-Man˜as, M.; Morral, L.; Pleixats,
R. Chem.sEur. J. 2000, 6, 3372.
(49) A similar coordination geometry and conformation for 1 to that
observed for allylic Pd complexes 7 and 8, is found in the complex
Pt(R,R-1)Cl2 where the Cl centers hydrogen bond to the amide NH
groups: (a) Burger, S.; Therrien, B.; Su¨ss-Fink, G. Acta Crystallogr.
2004, E60, m1163.
(50) Pd(0) complexes of ligand 1, and its naphtho analogue, are unstable
in solution, losing both amide hydrogens to form two covalent Pd-N
bonds, thus generating a neutral, P,P,N,N-tetracoordinate mononuclear
species. This bright yellow complex is catalytically inactive for
allylation reactions and has been characterized by X-ray crystal-
lography (see footnote 11 in ref 18).
9
J. AM. CHEM. SOC. VOL. 131, NO. 29, 2009 9955

A R T I C L E S
Butts et al.
increased interactions between the escort ion M+ and the
carbonyl of the amide group proximal to the pro-R terminus.
Overall, as M+ became more oxophilic (Cs+fLi+) this led to
a net trend of decreasing attack at the pro-S terminus and
increasing attack at the pro-R terminus to the point where with
Li+ the (S)-pathway was virtually isoenergetic with the (R)-
pathway; the trend in selectivity (ee calcd, Table 1) matching
that observed experimentally (ee obs) under both stoichiometric
and catalytic conditions.
Analogous studies on the second key part of the catalytic
cycle, generation of exo-8+ from cyclohexenyl acetate (5),
revealed a fundamental difference in ionization mode between
enantiomers of 5, Figure 9. For the matched substrate (S)-5, a
hydrogen bond from the amide proton on the concave surface
to the carbonyl oxygen of the acetate preorients the allyl unit
for ionization. This interaction is absent in the less stable (14
kJ mol-1) mismatched pairing with (R)-5 and while not
deactivated per se, this process is kinetically noncompetitive
with ionization of the matched substrate (S)-5 (kS . kR).11
In summary, we have identified that hydrogen-bond interac-
tions of one N-H unit in Pd-coordinated (R,R)-1 can substan-
tially accelerate both ionization and nucleophilic attack. This
hydrogen-bond directed delivery of the nucleophile has prece-
dent in the elegant design of chiral ferrocene ligands of Hayashi
and Ito,51 although the hydrogen bonding unit appears to be
substantially more localized and orientated in Pd(R,R-1) com-
plexes. The interaction provides not only a new rationale for
observed selectivities20,52 but also a design principle for further
development of the reaction class.51 For attack involving harder
nucleophiles7,8,20 with tightly ion-paired escort-ions (M+) such
as Li+, an alternative selective pathway is available via favorable
interaction of M+ with the carbonyl of the other amide unit.
We note that the selectivity mechanism, outlined in cartoon form
in Figure 10, differs from most commonly advanced rationaliza-
tions in stereoselective synthesis; the high selectivity is obtained
by selective favoring of one pathway, not by disfavoring of all
undesired paths. Thus, the current reaction is an example of
Ligand Accelerated Catalysis,53 and even a rather special case
within this class since the acceleration is strongly dependent
on the ligand conformation. Only in a chelate will the structure
display a concave face with regioselectively placed activating
amide groups, thus allowing the desired pathway to out-compete
background turnover with poor selectivity by the oligomeric
catalysts in the reaction milieu.
Figure 10. Cartoon model for kinetic resolution (upper graphic) and
asymmetric induction (lower-graphic) in the Pd(R,R)-1 catalyzed reaction
of cyclohexenyl esters with nucleophiles. pro-R versus pro-S selectivity in
nucleophilic attack depends identity of escort ion M+, counterion X-, and
availability of a hydrogen bond acceptor, in a 1,3-relationship to the
nucleophilic site (e.g., malonate, phathalimide, carboxylate, carbonate, etc.).
VNMR 500 spectrometer fitted with Varian 500 H(C/X) PFG
1
or Varian 500 Auto X BM PFG probes. 2D H1H NOESY
spectra were generated by employing standard Varian Chempack
experiments. Varian Chempack experiments were modified to
conduct multinuclear analyses where necessary. See Supporting
Information for full details, analysis, and assignments.
Computational Methods. Using in-house parametrization
methods,54 we have earlier developed MM3* force fields55
specifically for (η3-allyl)Pd complexes,36 based on a combination
of DFT and experimental data. The force field has been slightly
modified to accommodate version updates of MacroModel; the
modified parameter subset used in the current work is available
as Supporting Information. This force field, when utilized for
structures similar to those in the training set, has a structural
accuracy that rivals any other available method (including
DFT),36 and delivers conformational energies that from previous
experience are accurate to within ca. 5-6 kJ/mol, slightly worse
than the accuracy that can be obtained for small, purely organic
molecules.56 When further refined by B3LYP optimization, the
conformational energy accuracy is usually improved to 1-2 kJ
mol-1, sufficient for a reliable comparison to experimental data.
We want to identify all distinct conformations present in at least
5% (the approximate detection limit in the NMR experiments,
vide supra), which translates to an energy range of ca. 8 kJ
mol-1 above the global minimum at ambient temperature. Using
a wide safety margin of 10 kJ mol-1, this means that signifi-
cantly different conformations within 18 kJ mol-1 from the
global minimum in the MM conformational searches should be
subjected to DFT validation.
NMR Methods. Samples were sealed in NMR tubes (178 mm
long, 5 mm diameter) fitted with a Young’s valve, under an
inert (N2) atmosphere, using standard Schlenk-line techniques.
CD2Cl2 was freshly distilled from preactivated 3 Å molecular
sieves under an atmosphere of N2(g) prior to use. THF was
freshly distilled from Na-benzophenone ketyl under an atmo-
sphere of N2(g) prior to use. [7]+[BAr′F]- (23.5 mg, 13.8 × 10-3
mmol) and [8]+[BAr′F]- (4.2 mg, 2.4 × 10-3 mmol) were
dissolved in 0.6 cm3 CD2Cl2 and the resulting 23.0 mM
([7]+[BAr′F]-) or 4.0 mM ([8]+[BAr′F]-) samples checked for
the presence of >95% monomeric chelate by 31P{1H} NMR prior
to detailed analysis. NMR spectra were recorded on a Varian
Molecular modeling is basically a gas phase method, and for
comparison with data determined in solvent, it is necessary to
(54) (a) Norrby, P.-O.; Liljefors, T. J. Comput. Chem. 1998, 19, 1146. (b)
Norrby, P.-O.; Brandt, P. Coord. Chem. ReV. 2001, 212, 79. (c) Norrby,
P.-O. In Computational Organometallic Chemistry; Cundari, T. Ed.;
Marcel Dekker: New York, 2001; pp 7-37.
(55) MM3* is a modification of the original MM3(89) force field: (a)
Allinger, N. L.; Yuh, Y. H.; Lii, J.-H. J. Am. Chem. Soc. 1989, 111,
8551. (b) For some additions in the MM3(94) force field, see: Allinger,
N. L.; Zhou, X.; Bergsma, J. J. Mol. Struct. (Theochem) 1994, 312,
69.
(51) Hayashi, T.; Yamamoto, A.; Ito, Y.; Nishioka, E.; Miura, H.; Yanagi,
K. J. Am. Chem. Soc. 1989, 111, 6301.
(52) For an example involving malononitriles in asymmetric decarboxylative
cycloaddition of electron-deficient olefins see: (a) Wang, C.; Tunge,
J. A. J. Am. Chem. Soc. 2008, 130, 8118.
(53) Berrisford, D. J.; Bolm, C.; Sharpless, K. B. Angew. Chem., Int. Ed.
Engl. 1995, 34, 1059.
9
9956 J. AM. CHEM. SOC. VOL. 131, NO. 29, 2009

Asymmetric Induction by the Trost Standard Ligand
A R T I C L E S
augment the modeling method by some type of solvent
representation. For DFT and other quantum chemical techniques,
reliable continuum methods are available (vide infra), but we
know of no current solvation method that allows application to
metal complexes in MM. The best method available to us was
to simply screen electrostatic interactions by increasing the
dielectric constant to 9, as a very rough simulation of dichlo-
romethane or THF solvation, previously validated in similar
applications.57 Conformational searches employed the pseudo-
systematic Monte Carlo Search58 implemented in Macro-
Model,59 with further fine-grained refinement using Low Mode
Searching.60
All energies reported herein come from DFT optimization
using the B3LYP functional35 in conjunction with the LACVP*
basis set.61 Solvent effects were represented by the PBF
continuum model62 with default parameters for THF (solvent
) tetrahydrofuran) and also CH2Cl2 (solvent ) dichlo-
romethane). As expected, the calculations in the two solvents
gave virtually identical structures, and relative energies that
agreed to within a few kJ mol-1. All DFT structures and energies
are available as Supporting Information. The B3LYP results
agreed with the ordering of the MM conformations, even though
individual relative energies were shifted by a few kJ mol-1, in
good agreement with previous usage of the force field.
We also studied reactions where 8+ was either the product
or reactant, that is, ionization of cyclohexenyl acetate 5 by
Pd(R,R-1) leading to 8+ and acetate anion, or nucleophilic attack
of dimethyl malonate anion with or without alkali metal
counterion on 8+ leading to 6 bound to Pd(R,R-1). Considerable
previous experience has shown us that the combination of
double-ꢁ basis set with a continuum solvation model is both
necessary and sufficient for an accurate reproduction of (η3-
allyl)Pd reactivity.37,43 In particular, it has been shown that in
gas phase calculations, attack of a negative nucleophile on
cationic (η3-allyl)Pd complexes is a monotonous downhill
process; there is no transition state.37 Thus, continuum solvation
is a requirement for location of the relevant stationary points.
In the current project, as well as previously,43 we have noted
that the reaction coordinate has a long, flat region around the
TS, sometimes with small bumps that are artifacts from the
cavity model used for continuum solvation. In our experience,
regular transition state searches frequently converge to a minor
bump or flat region that may be quite far from the actual TS.
On the other hand, the reaction coordinate is closely correlated
with the C-Nu distance; a relaxed scan over this coordinate
produces a relatively smooth curve without the sudden energy
jumps (“snap effects”) indicative of a discontinuity in the
calculated reaction coordinate. The absence of discontinuities
could also be verified by visual inspection and animation of
the converged points along the path. Thus, choosing the highest
point on a relaxed coordinate scan allows a reliable location of
transition structures, validated by inspection of the PES. All
“transition states” have been located this way. However, this
procedure has a certain drawback. We are currently unable to
calculate reliable frequencies for this size of complex in solvent.
Therefore, we cannot calculate the vibrational contribution to
the free energy, in particular the entropy component. Since
higher energy, wider stationary points tend to also have higher
entropies, we neglect a contribution that will tend to compensate
some (but not all) of the calculated energy differences. Thus,
we see that all large energy differences will be overestimated,
sometimes severely. We are therefore careful to draw conclu-
sions only from calculated trends, not absolute energy differences.
Acknowledgment. We thank AstraZeneca, EPSRC, University
of Bristol, and the Federal Educational Department of the State of
Aargau, Switzerland, for generous funding, Johnson Matthey for
donation of PdCl2, and Chirotech for donation of both enantiomers
of the ‘Standard Ligand’ (1). The computations were performed
on C3SE computing resources in Gothenburg. G.C.L.J. is a Royal
Society Wolfson Research Merit Awardee, and P.O.N. is supported
by the Swedish Research Council. Dr. H. Hagelin-Weaver (ne´e
Hagelin), Dr. I. J. S. Fairlamb, Dr. A. Martorell, Dr. P. M. Tomlin,
and Mr. Antony Meadowcroft are thanked for preliminary studies.
Supporting Information Available: Details for preparation of
Pd complexes 7+, 8+, d9-8+, and 8a-d+, together with NMR
analyses and full computational details. This material is available
free of charge via the Internet at http://pubs.acs.org.
JA8099757
(57) Norrby, P.-O.; Åkermark, B.; Hæffner, F.; Hansson, S.; Blomberg,
M. J. Am. Chem. Soc. 1993, 115, 4859.
(58) Goodman, J. M.; Still, W. C. J. Comput. Chem. 1991, 12, 1110.
(59) Maestro 8.5, MacroModel 9.5, and Jaguar 7.5, Schrodinger, LLC, New
York, NY, 2008. For documentation and current versions of the
programs see www.schrodinger.com.
(56) An error of ca 2 kJ mol-1 corresponds to a factor of 2 in the calculated
populations at ambient temperature, for both equilibria and rates.
Comparisons to experimental data for typical organic molecules
indicate that this accuracy is achievable with well-parameterized
molecular mechanics, and improved to slightly below 2 kJ mol-1 using
B3LYP with a double-ꢁ basis set: (a) Liljefors, T.; Gundertofte, K.;
Norrby, P.-O.; Pettersson, I. In Computational Medicinal Chemistry
for Drug DiscoVery; Bultinck, P., Tollenaere, J. P., De Winter, H.,
Langenaeker, W., Eds.; Marcel Dekker: New York, 2004; p 1. (b)
Gundertofte, K.; Liljefors, T.; Norrby, P.-O.; Pettersson, I. J. Comput.
Chem. 1996, 17, 429. (c) Pettersson, I.; Liljefors, T. In ReViews in
Computational Chemistry; Lipkowitz, K. B., Boyd, D. B., Eds.; VCH:
New York, 1996; Vol. 9, p 167.
(60) Kolossvary, I.; Guida, W. C. J. Am. Chem. Soc. 1996, 118, 5011.
(61) The LACVP* basis set uses 6-31G* for the lighter elements. Heavy
metals (Pd, Cs, K) use the Hay-Wadt small-core ECP with ac-
companying basis set: Hay, P. J.; Wadt, W. R. J. Chem. Phys. 1985,
82, 299.
(62) (a) Tannor, D. J; Marten, B.; Murphy, R.; Friesner, R. A.; Sitkoff,
D.; Nicholls, A.; Ringnalda, M.; Goddard, W. A., III; Honig, B. J. Am.
Chem. Soc. 1994, 116, 11875. (b) Marten, B.; Kim, K.; Cortis, C.;
Friesner, R. A.; Murphy, R. B.; Ringnalda, M. N.; Sitkoff, D.; Honig,
B. J. Phys. Chem. 1996, 100, 11775.
9
J. AM. CHEM. SOC. VOL. 131, NO. 29, 2009 9957