R. S. Paton, D. J. Dixon et al.
triple-z valence basis set and LANL2TZ(f) ECP plus triple-
z valence basis combination was used for single-point
energy calculations.[23] An ultrafine (99,590) grid containing
99 radial shells and 590 angular points per shell was used for
the numerical integration in the calculation of single-point
energetics.[24] Stationary points were verified as minima
(zero imaginary frequencies) or transition structures (a
single imaginary frequency) with harmonic frequency calcu-
lations at the same level of theory as optimizations, and
Gibbs free energies were computed at 298 K after shifting
very low frequencies below 50 cmÀ1 to a value of 50 cmÀ1 to
prevent spuriously large vibrational entropic terms. Due to
the insensitivity of conversion and enantioselectivity with re-
spect to the solvent observed in our experiments, optimiza-
tions were performed in the gas phase; however, the effects
of solvation have been included through single-point energy
calculations with a CPCM model by using UAKS radii.[25]
Where applicable, all possible efforts were made to find the
lowest-energy conformations by performing geometry opti-
mizations from different starting points. Natural bond orbi-
tal (NBO) calculations were performed with NBO 3.1[26] and
Wiberg bond orders computed.[27]
Figure 9. Top: Optimized M062X/6-31G(d) structures of lowest-energy
À
competing TSs for C C bond formation of 2a catalyzed by a simplified
version of 1b (H atoms are omitted for clarity). ZPE-inclusive M062X/6-
31G(d) relative energies, with M062X/6-311+GACTHNUTRGNE(NUG d,p) in parentheses,
To gain a qualitative and quantitative understanding of
the observed stereoinduction we modeled competing stereo-
shown in kcalmolÀ1. Bottom: Wiberg bond orders and distances (ꢅ) for
key interactions in the TSs and for a simplified model.
À
determining C C bond-forming transition structures (TSs)
by using quantum chemical calculations. The nature and ge-
ometry of this TS were established by performing calcula-
tions on a model system: a syn-carbocupration TS in which
observed stereoselectivity of deuterium incorporation in
Scheme 1. Pre-catalyst coordination to CuI occurs through
the quinuclidine nitrogen atom, leaving the urea group to
hydrogen bond with one of the oxygen atoms. Coordination
of the urea oxygen to Cu is disfavored by ca. 13 kcalmolÀ1,
and other H-bonding modes were calculated; however, the
pre-catalyst scaffold evidently favours the mode shown in
Figure 9. The optimized TS geometries for enolate Re and
Si face attack show the former is favored by 1.3 kcalmolÀ1,
in accord with experiment in which this is the major enan-
tiomer in 92% ee (the M062X/6-31G(d) computed ee from
competing Boltzmann factors is 82%).
One marked structural difference between the diastereo-
meric TSs that could account for the facial selectivity is the
lack of any ester O–Cu interaction in the disfavored TS. The
metal plays a dual role in the Conia-ene TS: to coordinate
strongly to the charged b-dicarbonyl enolate while also acti-
vating the h2-coordinated alkyne. It is the former of these
two roles that differs between the Re and Si TS. Relative to
a trimethylamine: a CuI model system lacking any hydro-
gen-bonding contributions, the effect of the two H-bonding
interactions formed between the urea group and one of the
oxygen atoms is to weaken the strength of the correspond-
ing O–Cu interaction. This is evident from bond distances
and calculated bond orders (BOs) in Figure 9, in which rela-
tive to a model TS, the ketone O–Cu interaction lengthens
by 0.13 ꢅ and loses 0.05 in terms of BO in the major TS,
and the ester O–Cu interaction effectively disappears in the
minor TS. The strength of hydrogen-bonding interactions is
very similar for the two TSs, as judged by O–H distances
and computed BOs, and serves to compensate for the loss of
À
the two C O bonds are aligned was computed to be ener-
getically preferred over isomeric Cu enolates and alternative
mechanisms involving hydrogen-transfer to the terminal
alkyne from an enol nucleophile (Figure S1). Models were
also considered in which the Cu counterion (chloride) was
still present in the carbocylization step; however, the com-
À
puted activation free-energy barrier of C C bond formation
is around 10 kcalmolÀ1 higher than for a Cu/amine complex
(see Figure S2 in the Supportig Information) and thus these
other models were discounted. We thus focussed on the
computational description of the carbocyclization step
shown in Figure 9. In our model, a simplified structure lack-
ing the olefin substituent on the quinuclidine was used to
describe the pre-catalyst 1b. This modelling assumption is
justified by the similar ee (with opposite sign) obtained
when using pseudoenantiomeric pre-catalyst 1d, differing
only in the positioning of the vinyl substituent. For substrate
2a and model pre-catalyst 1b, we performed a number of
optimizations starting from different geometries of the pre-
catalyst with respect to the cyclic core of the TS. This search
process yielded eight unique conformations (four enolate Re
face attack, four enolate Si face) from which the lowest-
energy and thus most important diastereomeric pair, making
up 87% of the conformer distribution at 298 K, are shown
below in Figure 9.
The low-energy TSs are both characterized by a six-mem-
bered cyclic TS in which the alkyne and enolate are both co-
ordinated to CuI. The syn-carbometallation across the
alkyne of these models is consistent with the experimentally
&
8
&
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 0000, 00, 0 – 0
ÝÝ
These are not the final page numbers!
Sladojevich, Filippo
