now published in J. Am. Chem. Soc.

, 127, 10699-10706 (2005)

Can the π−facial selectivity of solvation be predicted by atomistic simulation?

Roberto BerardiA, Gianfranco CainelliB*, Paola GallettiB, Daria GiacominiB, Andrea GualandiB, Luca

MuccioliA, and Claudio ZannoniA*.

Dipartimento di Chimica Fisica e Inorganica and INSTM, Università di Bologna, Viale Risorgimento 4,

40136 Bologna, Italy, and Dipartimento di Chimica “G. Ciamician”, Università di Bologna, Via Selmi

2, 40126 Bologna, Italy.

*[email protected], *[email protected]

ABSTRACT: This work is concerned with the rationalization and prediction of solvent and temperature

effects in nucleophilic addition to α-chiral carbonyl compounds leading to facial diastereoselectivity.

We study, using Molecular Dynamics simulations, the facial solvation of (R)-2-phenyl-propionaldehyde

in n-pentane and n-octane at a number of temperatures and compare it with experimental selectivity data

for the nBuLi addition leading to syn and anti (2R)-2-phenyl-3-heptanol, which give non-linear Eyring

plots with the presence of inversion temperatures. We have found from simulations that the facial

solvation changes with temperature and alkane. Moreover, by introducing a suitable molecular chirality

index we have been able to predict break temperatures (TCI) for the two solvents within less than 20

degrees of the inversion temperatures experimentally observed in the diastereoselective nBuLi addition.

We believe this could lead to a viable approach for predicting inversion temperatures and other subtle

solvent effects in a number of stereoselective reactions.

1
INTRODUCTION

It is well known that a number of chemical reactions can be affected by the medium in which the

reaction takes place and that solvents can change both equilibrium constants and reaction rates 1. The

effects of solvent on stereoselectivity have already been well described by organic chemists, and in

particular several examples of solvent dependent face-selectivity have been discussed 2. However, a

proper rationalization of these effects, and even more the possibility of predicting them from just

knowledge of the compounds involved, is far from being in sight. To date, the two most common

qualitative models employed to rationalize the reaction products of nucleophilic addition to α-chiral

carbonyls are based on two arguments that actually disregard the effect of the solvent, assuming that the

attack comes from the less sterically hindered direction, and that every aldehyde is at the same, fixed

conformation. These are the Cram’s model 3, predicting for this aldehyde a unique conformation in

which the double bond is flanked by the two smallest groups (H, CH3) attached to the chiral center, and

the most successful Felkin–Ahn model 4,5, which instead predicts in this simple case the double bond to

be flanked by the medium size group (CH3) and lying between the two bulkiest groups (CH3, Ph).

At a very simple level, the stereoselectivity is related, through the modified Eyring equation 6, to the

difference ΔΔG≠ on the free activation energies for the syn and anti reaction pathways leading to the two

diastereoisomers,

lnS(T) = ln(k/k') = -ΔΔG≠(T) /RT = -ΔΔH(T)≠/RT + ΔΔS(T)≠/R, (1)

where S = [anti]/ [syn] is the stereoselectivity and k and k' are the overall rate constants for the synthesis

of the two stereoisomers. Many experimental results show a non-linear temperature dependence for the

selectivity, and a plot of lnS vs. 1/T often consists of two straight lines with different slopes

(corresponding to different activation parameters), intersecting at a temperature called the inversion

diastereoselectivity of α-chiral imines and carbonyl compounds, a case for which some of us have

reported several interesting temperature dependence of the diastereomeric excess (de%) of the products

2
in polar and nonpolar solvents. For the model reaction involving addition of BuLi to 2-phenyl-

propionaldehyde (scheme 1), the effect has been studied over a wide range of temperatures for various

common organic solvents (n-pentane, n-hexane, cyclopentane, n-octane, n-decane, n-dodecane,

cyclohexane, THF, tBuOMe, 1,4-dioxane) 8. While the ethers show a linear dependence of de% on the

temperature, all the alkanes present a break point at a certain temperature Tinv. In particular, for linear

alkanes it has been observed that: (i) the anti product is favored at all temperatures; (ii) the selectivity

decreases with the increase of temperature; (iii) the longer the solvent chain, the lower the diasterofacial

selectivity; (iv) the inversion temperature correlates with the melting point of the alkane.

These facts are not easy either to rationalize intuitively or to be predicted by using quantum modelling 9,

not only in the traditional “one molecule in vacuum approach”, but also using fixed conformation

continuum or discrete models with a few explicit solvent molecules to account for the solvent effect
10,11
. Indeed on one hand important differences in inversion temperatures are found in different linear

alkanes with essentially the same dielectric constant and, on the other, the good conformational

sampling in an equilibrium condensed phase that would be necessary to undertake the task is beyond the

current possibilities of quantum chemical calculations.

Some examples have been reported for enantio- or diastereoselective reactions, where different reaction

solvents are able to drive the diastereofacial selectivity 13; the solvent plays a fundamental role as only

small differences in activation enthalpies of the two diastereomeric transition states are expected. The

presence of a Tinv would thus depend on dynamic solvation effects and represents the interconversion

temperature of two different solvation clusters which act as two different supramolecules with different

stereoselectivities. The dynamic solvent effects have been confirmed by temperature dependent studies
13
of the C NMR, CD and UV spectra of some solvated aldehydes in the absence of any reaction,

revealing the presence of peculiar temperatures TCD, TUV, and TNMR, whose values are identical and

match with experimentally found Tinv 14, thus supporting its solvent-dependent nature.

3
To confirm this interpretation and establish its microscopical origin as well as to try to develop a

predictive approach, we have decided to study the solvation of (R)-2-phenyl propionaldehyde in two

solvents at first glance very similar: n-pentane and n-octane. The present paper is divided into two parts.

In the first we investigate the experimental temperature-dependent diastereofacial selectivity in the

addition reaction of n-BuLi to (R)-2-phenyl propionaldehyde: while data for the racemic aldehyde were

already available 8, here we have obtained experimental data for the (R) enantiomer in n-pentane and in

n-octane in a wide range of temperatures. In the second part we study, with the help of the molecular

dynamics (MD) technique15, 16, the temperature dependence of diastereosolvation in n-pentane and n-

octane and we introduce some tools for following the change in solvation and, to some extent, to predict

the inversion temperature.

EXPERIMENTAL PART

Optically pure (R)-2-phenyl propionaldehyde was prepared accordingly to the procedure reported in the
17
literature by means of oxidation of (R)-2-phenyl-propanol with the Dess-Martin periodinane (see

scheme 1).

Ph Ph
periodinane
OH Me
Me
1 O

n-pentane or Ph Ph
n-octane Bu Bu
1 + nBuLi Me + Me
Temperature OH OH
1a 1b
anti syn

Scheme 1. The one step synthesis followed to prepare the chiral aldehyde 1
and the nucleophilic addition reaction conditions.

In all experiments the addition reaction was performed by introducing nBuLi into aldehyde 1 diluted in
anhydrous n-pentane or n-octane at constant temperature (see supporting information). The temperature
value was varied over a range of approx. 120 K. The reactions proceeded smoothly to give anti (1a) and
syn (1b) alcohols (scheme 1). The diastereomeric ratio [anti]/[syn] was determined in each experiment
by GC analysis of the corresponding trimethylsilylethers and results are reported as Supporting
Information. The data for ln (anti/syn) versus 1/T were analyzed by a linear least-squares fitting to the
Eyring equation and for each data set a residual analysis was applied to evaluate the number of linear

4
trends and to ascertain the presence of an inversion temperature. More specifically, the inversion
temperature has been first estimated by fitting the complete {lnSi , 1/Ti } data set (ordered with
increasing values of 1/T) with a single regression line. The data pair corresponding to the largest
absolute value residual has been used to partition the data set into left and right subsets, with the
dividing point assigned to the left subset. The two groups of points have then been separately fitted with
a least squares straight line. The intersection of these regression curves gives the estimated inversion
temperature, while the differences ΔΔH≠ and ΔΔS≠ were obtained from the slopes and the intercepts of
these linear plots (Table 1). The thermodynamic parameters are quite different in the two solvents. In
particular in pentane, where little values of ΔΔH≠ resulted, the differential activation entropy represents
the main contribution to stereoselectivity.

T>Tinv T<Tinv

Solvent Tinv (K) ΔΔH≠ ΔΔS≠ ΔΔH≠ ΔΔS≠

(kcal mol-1) (cal mol-1K-1) (kcal mol-1) (cal mol-1K-1)

Pentane 204 -0.33 ± 0.06 +1.2 ± 0.2 +0.46 ± 0.14 +5.2 ± 0.7

Octane 288 -1.0 ± 0.1 -1.0 ± 0.4 -0.58 ± 0.04 +0.6 ± 0.2

Table 1. Differential activation parameters ΔΔH≠, ΔΔS≠ and inversion temperature Tinv for nBuLi addition to aldehyde 1

in n-pentane and n-octane as obtained from linear regression analysis of the experimental data.

The Eyring plots in figure 1 show the inverse temperature dependence of lnS to 1 in n-pentane and n-

octane. Notwithstanding the little variation of the diastereomeric ratio [anti]/[syn] with 1/T, the plots

show for n-pentane an inversion temperature (Tinv) at 204 K, suggesting the presence of two different

kinds of solute-solvent clusters: one above and one below Tinv. A similar behavior, although less

evident, is observed using n-octane as the solvent, with the anti diastereoisomer being favored at all

temperatures.

5
Figure 1. Eyring plots for the diastereomeric ratio obtained in the addition of nBuLi to aldehyde 1 in pure n-pentane (left)
and n-octane (right). The two linear regression curves, obtained as described in the text, and the inversion temperature are
shown together with the experimental results (circles).

COMPUTER SIMULATIONS

Our aim here is to investigate the existence of differences at molecular level in the solvation of the

aldehyde in the two alkanes studied experimentally, via extensive molecular dynamics simulations at
15, 16
atomistic level . In practice, we have explicitly simulated (R)-2-phenyl-propionaldehyde in n-

pentane and n-octane, examining the distribution of solvent molecules around the solute and the

properties of the chiral solute in this explicit medium. To do this, we have described the n-alkane
18
solvent with the NERD united-atoms Force Field (FF) which was specifically developed for
19
hydrocarbons, while the solute has been modeled by suitably modifying the OPLS-AMBER FF by

adding specifically derived point charges and by changing a torsional term. As the description of

aldehyde torsion with respect to the bond CHO-C* is of fundamental importance in determining the

conformations of the solute and the reaction products, we have performed first a preliminary test of the

standard AMBER FF, by a comparison with quantum mechanical calculations at Hartree-Fock 6-31G*
20
level for the torsional potential. This has shown that the standard FF gives a description of the φ1

torsion that is even qualitatively incorrect (figure 2) and, consequently, we have replaced the relevant

AMBER FF parameters with new ones using the following procedure.

6
 Ph
H Me H Ph
Me R O O O
O
Ph H Me Ph H Me
H H H

A B C

Figure 2. Definition of the torsional angles φi (left) and a plot of the torsional potential for the rotation around the CHO-C*

bond (right) of (R)-2-phenyl-propionaldehyde as obtained from various levels of theory: original AMBER FF (black, long

dashed), HF/6-31G* (black, short dashed), B3LYP/cc-pVDZ (red, thick line) used for parametrizing the new force field

torsional term (blue dots, calculated in vacuum). The Newman projections of the minimal energy conformers are also shown.

The molecular geometry was optimized with the B3LYP density functional (DFT) using Dunning

correlation consistent polarized valence double zeta (cc-pVDZ) basis set20, and electrostatic-potential

fitting atomic charges were also calculated and added to the force field. At the same level of theory, and

for 24 conformations corresponding to uniformly spaced torsional angles in the range –180o <φ1<180o,

the molecular structure was then relaxed while keeping fixed φ1 at the desired value, thus obtaining a

reference torsional energy profile (red curve in figure 2 and Table B in the Supporting Information).

This overall DFT potential contains non-bonded contributions terms which in the AMBER type FF are

given separately, and that have thus to be accounted for to avoid counting them twice in the simulation.

In practice, the contribution Unb(φ1) of the non-bonded terms to the FF has been calculated from the

logarithm of the “non-bonded” torsional angular distribution Pnb(φ1) obtained from an MD simulation of

an isolated molecule at 1000 K, in which the FF torsional potential was fixed to zero (Unb=-kBT lnPnb). B

The rationale of the method is that, at this high temperature, the conformations corresponding to all
7
angles are populated, and consequently a good conformational sampling can be achieved. The

difference between the DFT and the nonbonded torsional potential has then been fitted with a truncated

Fourier series expansion (six cosine and three sine terms) that has been used as the new FF torsional
21
potential (see e.g. ref. ). We have verified the new force field torsional potential by two 600 ns

simulations in vacuum at the temperatures of 270 and 320 K; the agreement with the original DFT

potential is good (root mean square deviation: 0.14 kcal/mol) and the discrepancies are due to the

limited number of Fourier series terms used in the force field for torsion φ1.

The torsional potential obtained, which is in excellent agreement with recently published calculations
22
and free jet millimeter-wave absorption spectroscopy experimental data , shows three minima

corresponding to the preferred conformers. The most stable one, labeled A in figure 2, is the one with

the carbonyl oxygen eclipsing the methyl group, while the conformer corresponding to the eclipsed

hydrogen (B) is found to be less stable of about 1.4 kcal/mol and the conformer with the eclipsed phenyl

(conformer C) has an energy 1.7 kcal/mol higher than conformer A.

We have performed MD simulations of samples composed of one solute: a (R)-2-phenyl-

propionaldehyde molecule, and 50 surrounding solvent molecules using ORAC, a multiple time-step
23
code developed by Procacci et al. . This relatively small sample, contained in a periodic cubic box

kept at atmospheric pressure, is sufficient to describe well the first solvation sphere and partially the

second, as we have verified with additional simulations performed on larger samples containing 250

molecules of pentane at the temperature of 270 K and of octane at the temperature of 320 K (see figure

6). The temperature ranges, 170-270 K for n-pentane and 220-320 K for n-octane, have been chosen

since they are sufficiently large to enclose the experimental inversion temperatures. The systems
24
investigated required around 5 ns to equilibrate: a reasonable time for a system composed of
25
relatively simple molecules moving in low viscosity isotropic solvents . Following the equilibration

stage we have performed very long production runs of about 300 nanoseconds to get sufficiently good

statistics. We have monitored the torsional angle φ1 as function of time and verified that under these

8
conditions the sampling during the MD production runs is adequate to acquire statistically significant

conformational population probabilities.

In view of assessing the molecular origin of the experimental diastereoselectivity, we have started

comparing the conformational distributions around the CHO-C* bond (torsional angle φ1) as computed

from the MD trajectories for the pentane and octane solvents at the highest and lowest temperatures

studied with the distributions calculated from a simulation of the aldehyde in vacuum (figure 3). We see

that in both solvents the conformations of type A are more probable than in vacuum (i.e. are stabilized

by the alkane); the conformations of type C and B are correspondingly destabilized, while the order of

the conformational energies is preserved (A<B<C). The low temperature distributions (black curves)

show that almost uniquely the conformation of type A is present in these thermodynamic conditions.

Figure 3: Normalized conformational distributions P(φ1) in pentane (left) and octane (right) at the lowest (black) and highest

(red) MD simulation temperatures, compared with the distribution obtained form a 600 ns MD NVT simulation of the

aldehyde in vacuum at the highest temperature (B, blue).

To measure more quantitatively both solvent and temperature effects on the molecular structure, we

have considered the mean squared angular fluctuations of the acyl, methyl and phenyl groups (figure 4),

that have been computed as the average <(Δφi)2 > at coarse grained intervals Δt of one picosecond for

each temperature. Not surprisingly, the mean squared fluctuations increase linearly with temperature

and their slopes provide an estimate of a local “torsional diffusion coefficient”. However an evident

jump in the curves is observed when changing solvent from pentane to octane (at 220 K): this is again
9
consistent with a change in the local mobility due to a modification in solvation, with octane allowing

larger conformational fluctuations of the substrate than pentane.

Figure 4: Mean squared angular fluctuations <(Δφi)2 > in pentane (left, 170-270 K) and octane (right, 220-320 K) solutions,

calculated for rotations of the acyl (φ1, squares), methyl (φ2, circles) and phenyl (φ3, boxes) groups.

Having verified that solvent effects are present, we have examined if our data support the existence of

an inversion temperature, in particular trying to identify some indicators connected to changes in

solvation with temperature that can on one hand be obtained from the MD simulations, and on the other

be related to the diastereoselectivity.

We start by examining the solvent structuring around the carbonyl through one of the many atom-atom

radial distribution functions that can be defined. In particular, we calculate the distribution g0(r) of the

distances between the carbonyl carbon taken at the center and each of the united atoms belonging to

solvent molecules (see figure 5), sampled on a spherical region:

1 (2)
g 0 (r ) = δ (r − rij ) ,
4π r 2 ρ ij

where rij is the carbon-solvent distance and ρ=N/V is the number density.

The results for g0(r) are plotted in figure 5. We see a local structuring of the solvent, more pronounced

in pentane than in octane, where the first and second peak of g0(r) are actually quite similar; as already

mentioned, we notice that the samples sizes we have used are sufficient to allow for the formation of

10
two solvation spheres. Test runs on larger samples with N=250 solvent molecules show that no further

relevant structuring occurs as the bulk limiting value g0(r)=1 is obtained.

Figure 5: The radial distribution function g0(r) around the carbonylic carbon and the average number of CHn centers N(r) in

pentane and octane calculated from the highest and lowest temperature MD simulation trajectories. We also show the radial

distribution function calculated from simulations of a larger sample with 250 solvent molecules (blue dotted lines).

Regarding the effect of temperature, we notice that an increase of temperature of 100 K, from well

below to well above the experimentally observed inversion temperature, broadens the peaks of g0(r), but

is not sufficient to modify the extent of the local, short range order. In addition, for both solvents the

shape of the radial distribution changes very little in the simulated temperature interval.

Figure 6: Snapshot of the united atom pentane molecules partially or totally contained in the first solvation sphere around

the carbonyl carbon (pentane, 180 K) with the definition of positive and negative faces of the carbonyl. Solvent atoms have

been rendered 5 times smaller than the actual size to increase visibility.

11
 Since the standard radial distribution depends only on distance it cannot be used to discriminate

between atoms above or below the aldehyde plane. Thus, in order to try to quantify the facial selectivity

of solvation, we have modified this indicator by defining a specific parameter s = (n ⋅ rij ) / | n ⋅ rij | , where

n is the normal to the plane identified by the three atoms of the acyl group. This plane is defined so that

s is positive on the face of the π carbonyl bond leading under nucleophilic attack to an anti product, and

negative on the other face (see figure 6).

We have then distinguished the two contributions to the total radial distribution function according

the sign of s and computed the density distribution g0(sr) around the two faces:

1
g 0 (s r ) = sδ (r − rij ) . (3)
2π r 2 ρ ij

These g0(sr) curves (figure 7) show clearly that the solvation is strongly facial selective, with the

positive side being on average more populated than the negative one at every temperature studied. It is

also possible to introduce a solvation radius for each face of the π bond, defined as the position of the

first minimum of the distribution function: the ratio of these distances is solvent dependent and changes

with temperature, as shown in figure 8. We find that the solvation radius measured for the positive face

(empty and filled red squares) is always smaller than one measured for the negative face (empty and

filled blue circles), in spite of the former region being more populated: we see that the solvation is

strongly face-selective and the positive face is surrounded by a more structured solvent cage. All radii

plotted in figure 8 show some discontinuity in their temperature dependence: in principle this might be

related to the experimental inversion temperatures, but we have found these data not sensitive enough to

be used for a recovery of the experimental result. The number of united atoms centers N(sr) for the two

faces does not change significantly with temperature (figures 5 and 7), and the positive face is always

more solvated: this is reflected in the experimental outcome that the anti product is favored at all

temperatures.

12
Figure 7: The face-dependent radial distribution function g0(sr) around the carbonylic carbon and the corresponding number

N(sr) of CHn groups contained in each hemispherical region.

Figure 8: Radii of the first solvation shell for the carbonyl carbon of the aldehyde in pentane (empty symbols) and octane

(filled symbols), as estimated from the g0(sr) distribution: face independent (black triangles), positive face (red squares), and

negative face values (blue circles).

We are still left with problem of finding an observable that is sensitive enough to changes in solute

solvation as a function of temperature. Since it has been demonstrated experimentally by CD studies of

the solvated aldehyde [14] that the inversion temperatures are directly connected to dynamic solute-

solvent clustering effects, and since it is reasonable to suppose that the diasteroselectivity might depend

on the degree of chirality of the chiral reactant, we have examined as a candidate indicator of selectivity

change the average chirality of (R)-2-phenyl propionaldehyde, as a function of temperature and solvent

nature. There are of course many quantitative estimators of molecular chirality, based on different
13
 26, 27, 28
operational definitions of chirality or on the choice of different molecular observables (for a

recent review see 29 ). Here we are not concerned with spectroscopic properties like CD itself or optical

rotation, but rather with an indicator of the effective (conformationally averaged) chirality in molecular

shape as relevant to solute-solvent interactions. Thus we have chosen to calculate a chiral index that

has already been successfully used in tackling problems connected with effects of solute chirality,

particularly in theoretical studies of helical twisting power, i.e. of the effectiveness of a chiral solute in
30, 31
generating a twisted (cholesteric) phase when dissolved in a nematic solvent . This molecular

chirality index can be written, as described in detail in 30 as:

G0 S = 4!/(3 N 4 ) ∑ Pˆijkl {mi m j mk ml ⎡⎣(rij × rkl ) ⋅ ril ⎤⎦ (rij ⋅ r jk )(r jk ⋅ rkl ) / ⎡⎣ (rij rjk rkl ) 2 ril ⎤⎦} ,
{ } (4)
{ijkl }

where rij =ri-rj is an interatomic vector and ri, mi the position and mass in atomic units of the i-th atom.

The summation is performed over all different sets of four atoms {ijkl} out of the N atoms in the

molecule, while P̂ijkl {Aijk} is an operator generating a sum over all the permutations of i, j, k, l . The

index G0S assigns to each observed molecular conformation a measure of chirality obtained as a mass-

weighted degree of asymmetry of the atomic coordinates, with a sign that can be used for a right/left

classification 26,27. We report in figure 9 the logarithm of the conformationally averaged value of <G0S>

obtained from simulation trajectories. We see that the conformationally averaged solute chirality

decreases with T, and that it is systematically higher in pentane than in octane (possibly an effect of

temperature). In figure 9 we also see a clear evidence of two regions with distinct chiral index vs

temperature slopes. From these results a break temperature (named TCI) can be determined with the

same algorithm described in the experimental section. It is gratifying to see that the break temperatures

obtained (223 K in pentane and 274 K in octane) are close to the inversion temperatures derived from

the Eyring plots of the experimental diastereoselectivity (see figure1). On the other hand there is no

reason to believe that only solvation effects directly connected to chirality are the only enthalpic effects

that can contribute to the reaction outcome and this is probably the origin of the differences between the

slopes of the <G0S> and selectivity curves. On the whole our findings definitely supports the hypothesis

14
that the inversion temperature for the stereoselective reaction studied is largely solvent-driven and that

the chiral index <G0S> is a useful monitor of these solvation changes.

Figure 9: The logarithm of the average molecular chiral index <G0S> for (R)-2-phenyl-propionaldehyde dissolved in pentane

(left) and octane (right) as obtained from the MD simulations (circles) together with the regression lines and the chiral break

temperatures TCI.

CONCLUSIONS

The molecular level description of solvation processes and of their effects on the details of chemical

reactions is an important topic that still has a large margin for development, both from the point of view

of understanding and rationalization and, even more, from that of predicting temperature dependent

effects like inversion temperatures 7,8. Here we have approached the specific case of stereoselectivity in
12
a nucleophilic addition, with the aim of verifying the hypothesis that the observed solvent and

temperature effects on facial selectivity are closely related to the nature of solvation in the local

microenvironment surrounding solute molecules.

We have performed stereoselectivity experiments for (R)-2-phenyl-propionaldehyde in two quite similar

solvents: pentane and octane at a number of temperatures and found (i) that they show an inversion

temperature and (ii) that these temperatures are quite different (204 and 288K). We have then performed

15
extensive molecular dynamics computer simulations for the two aldehyde-solvent systems and

measured the facial solvation of the carbonyl bond. We have found that the close presence of a chiral

center determines different average solvent densities on the two faces of the π-bond and that the density

depends on the solvent used; from this point of view, pentane and octane show remarkable differences,

which are likely to affect the diastereoselectivity of a nucleophilic reaction if we think at the solvent

density as a measure of accessibility of a face. The higher diastereoselectivity reported in pentane can

be rationalized as an higher accessibility of the face which leads to the anti product. This concept is not

sufficient in itself to explain the presence of an inversion temperature, and the quantification from our

simulations of the amount of solvent in the neighborhood of a side of the carbonylic bond has proven to

be not sufficiently sensitive for the purpose. We have then focused on a solute physical property, the
30,31
conformationally averaged molecular chiral index <G0S>, that we have determined from the

simulations. We have found that ln<G0S> exhibits a nonlinear temperature dependence in both solvents

with chirality break temperatures TCI which are in a good agreement with the experimental selectivity

inversion temperatures (pentane: TCI=223 K, Tinv=204 K; octane TCI=274 K, Tinv=288 K). The MD

simulations thus show the existence of an inversion temperature which can be ascribed to the interaction

of the aldehyde with the solvent and it is, in this respect, independent on the nature of the reactant. The

chiral index <G0S> hence results in a sensitive probe for temperature and solvent effects.

Although the detailed nature of the specific solvent-solute interactions responsible for the inversion of

selectivity still has to be investigated in general terms, we believe that the approach developed here,

with full inclusion of the solvent, provides an important opening and warrants a systematic study on a

larger class of solvents and reactions.

ACKNOWLEDGMENTS

We gratefully acknowledge financial support from MIUR through FIRB project “Solvent and

temperature effect on stereoselective processes in organic synthesis”, from University of Bologna and

from INSTM.
16
SUPPORTING INFORMATION

1. Experimental part.
2. Table A. Percentage concentration of anti and syn diastereoisomers in the nBuLi addition to
aldehyde 1 in n-pentane and n-octane.
3. Table B. Fit parameters of the DFT torsional potential.

REFERENCES

[1] Reichardt, C. Solvents and Solvent Effects in Organic Chemistry; 3rd ed., Wiley-VCH:
Weinheim, 2003.

[2] For recent examples see for instance: (a) Roland, S.; Mangeney, P. Eur. J. Org. Chem. 2000,
611; (b) Cainelli, G.; Giacomini, D.; Galletti, P. Eur. J. Org. Chem. 1999, 61. (c) Crimmins, M.
T.; Choy A. L.; J. Am. Chem. Soc. 1997, 119, 10237. (d) Murray, R. W.; Singh, M.; Williams, B.
L.; Moncrieff, H. M.; Tetrahedron Lett. 1995, 36, 2437. (e) Saito, T.; Kawamura, M.;
Nishimura, J. Tetrahedron Lett. 1997, 38, 8231. (f) Wipf, P.; Jung, J-K. Angew. Chem., Int. Ed.
Engl. 1997, 36, 764. (g) Denmark, S. E.; Nakajima, N.; Nicaise, O. J-C. J. Am. Chem. Soc. 1994,
116, 8797. (h) Reetz, M. T.; Stanchev, S.; Haning, H. Tetrahedron 1992, 48, 6813.

[3] Cram, D.; Kopecky, K. J. Am. Chem. Soc. 1952, 74, 5828.

[4] Cherest, M.; Felkin, H.; Prudent, N. Tetrahedron Lett. 1968, 2199.

[5] Ahn, N.T.; Eisenstein, O. Nouv. J. Chem. 1977, 1, 61.

[6] (a) Eyring, H. J. Phys. Chem. 1935, 3 , 107. (b) Glasstone, S.; Laidler, K. J.; Eyring, H. The
Theory of Rate Processes. McGraw-Hill: New York, 1941, chapter 4.

[7] For a review see: Buschmann, H.; Scharf, H.-D.; Hoffmann, N.; Esser, P. Angew. Chem. 1991,
103, 480; Angew. Chem., Int. Ed. Engl. 1991, 30, 477.

[8] (a) Cainelli, G.; Giacomini, D.; Galletti, P. Chem. Commun. 1999, 567; (b) Cainelli, G.;
Giacomini, D.; Galletti, P.; Marini, A. Angew. Chem. 1996, 108, 3016, Angew. Chem. Int. Ed.
Engl. 1996 , 35, 2849.

[9] (a) Fleischer, J. M.; Gushurst, A. J.; Jorgensen, W. L. J. Org. Chem. 1995, 60, 490; (b) Tomoda,
S.; Senju, T. Chem. Commun. 1999, 621.

17
[10] Sicinska, D.; Paneth, P.; Truhlar, D. G. J. Phys. Chem. B 2002, 106, 2708.

[11] see, e.g., Cramer, C. Essentials of Computational Chemistry ; Wiley: Chichester, 2002.

[12] Cainelli, G.; Giacomini, D.; Galletti, P.; Orioli, P. Angew. Chem. Int. Ed. Engl. 2000, 39, 523.

[13] Cainelli, G.; Giacomini, D.; Galletti, P.; Orioli, P. Eur. J. Org. Chem., 2001, 4509.

[14] Cainelli, G.; Galletti, P.; Pieraccini, S.; Quintavalla, A.; Giacomini, D.; Spada, G. P. Chirality,
2004, 16, 50.

[15] Frenkel D.; Smit, B. Understanding Molecular Simulations: From Algorithms to Applications;
Academic Press: San Diego, 1996.

[16] Pasini, P.; Zannoni, C., Eds. Advances in the computer simulations of liquid crystals; Kluwer:
Dordrecht, 2000.

[17] Botuha, C.; Haddad, M.; Larchevêque, M. Tetrahedron: Asymmetry 1998, 9, 1929.

[18] Nath, S.K.; Escobedo, F.A.; de Pablo, J. J. J. Chem. Phys. 1998, 108 , 9905.

[19] Cornell, W. D.; Cieplak, P.; Bayly, C. I.; Gould, I. R.; Merz, K. M., Jr.; Ferguson, D. M.;
Spellmeyer, D. C.; Fox, T.; Caldwell, J. W.; Kollman, P. A. J. Am. Chem. Soc. 1995, 117, 5179.

[20] Koch, W.; Holthausen, M. C. A Chemist’s Guide to Density Functional Theory, Wiley-VCH
Verlag GmbH: Weinheim, 2001.

[21] Earl, D. J.; Wilson, M. R. J. Chem. Phys. 2003, 119, 10280.

[22] Maris, A.; Caminati, W. Phys. Chem. Chem. Phys. 2003, 5, 2795.

[23] Procacci, P.; Paci, E.; Darden, T.; Marchi, M. J. Comput. Chem. 1997, 18, 1848.

[24] Berardi, R.; Muccioli, L.; Zannoni, C. ChemPhysChem 2004, 5, 104

[25] see, e.g., Dorfmüller,Th.; Pecora, R. Rotational dynamics of small and macromolecules;
Springer-Verlag: Berlin, 1987.

[26] Ruch, E.; Runge, W.; Kresze, G. Angew. Chem., Int. Ed. Engl 1973, 20, 12

[27] Ruch, E., Angew. Chem., Int. Ed. Engl 1977, 65, 16

[28] Buda, A.; Mislow, K., J. Am. Chem. Soc. 1992, 114, 6006.

18
[29] Petitjean, M., Entropy 2003, 5, 271.

[30] Solymosi, M.; Low, R.J.; Grayson, M.; Neal, M. P. J. Chem. Phys. 2002, 116, 9875.

[31] Neal, M. P.; Solymosi, M.; Wilson, M. R.; Earl, D. J. J. Chem. Phys. 2003, 119, 3567.

19
20