Solvent effects on stereoselectivity: more than just an environment
Gianfranco Cainelli,* Paola Galletti and Daria Giacomini*
Received 2nd June 2008
First published as an Advance Article on the web 12th February 2009
DOI: 10.1039/b802815j
Stereoselectivity is a major topic in organic synthesis. Intensive investigations into the role of
solvents on diastereo- and enantioselective reactions, as well as temperature-dependent
measurements of diastereomeric and enantiomeric ratios, have shed light on the existence of
dynamic solvation effects. In this tutorial review, several examples of non-linear Eyring plots in
stereoselective nucleophilic additions, cycloadditions, photochemical and enzymatic reactions are
reported. Experimental data and spectroscopic analyses obtained in aliphatic and aromatic
hydrocarbons, halohydrocarbons, ethers and mixtures lead to the formulation of a hypothesis on
the inversion temperature phenomenon as being due to an equilibrium between distinct
solute–solvent clusters, which are the real reactive species in solution.
Introduction
From the early studies on asymmetric induction by Fischer up
to the excellent stereoselectivity obtained in the last two
decades by means of asymmetric organo- and metal-catalysis,
asymmetric organic synthesis is, nowadays, a main topic of
organic chemistry.1,2
Stereoselectivity is defined in a chemical reaction as the
preferential formation of one stereoisomer over another.3
The stereoselective control of a reaction is generally discussed
in terms of the geometry of reagents, reagent–catalyst
complexes and activated complexes. The effects of solvent on
stereoselectivity have been described by organic chemists, and
in particular several examples of solvent-dependent face selec-
tivity have been discussed. However, a proper rationalization
of these effects is far from being in sight. As an example, the
two most common qualitative models to rationalize the
diastereoselectivity in nucleophilic additions to a-chiralcarbonyl compounds, the Cram and Felkin-Ahn models,4
have two main limitations: (i) the reaction is considered to
occur in a vacuum, i.e. in the absence of any solvent, whereas
most reactions take place in a condensed liquid phase, and
(ii) each carbonyl compound has the same fixed conformation
and the attack comes from the less hindered direction. Thus,
these models deal only with the steric and electronic features of
activated complexes. In other words, only enthalpic differences
Fig. 1 Stereoselectivity as the ratio between the kinetic constants of
two competitive processes.
Department of Chemistry ‘‘G.Ciamician’’, University of Bologna,Via Selmi 2, 40126 Bologna, Italy. E-mail: [email protected];Fax: +39 051 2099456; Tel: +39 051 2099528
Paola Galletti, Gianfranco Cainelli and Daria Giacomini
Gianfranco Cainelli was born in Trento in 1932, became Dipl. Ing.in 1954 and Doctor der Technischen Wissenschaften at the ETHZurich working in the laboratory of Professor Oskar Jeger in1958. After two years as a postdoctoral fellow with Professor O.Jeger, he moved to the Politecnico of Milan as an Assistant ofProfessor A. Quilico. In 1968, he joined the Faculty of Pharmacyof the University of Bari as a full Professor of chemistry of naturalproducts. He has been Professor of organic chemistry at theUniversity of Bologna since 1971. His main research interestscover the synthesis of natural products.Paola Galletti was born in Bologna, Italy in 1969. She is aresearcher in organic chemistry at the University of Bologna, fromwhere she graduated in chemistry (1994) and received her PhD(1999). Her research interests are in the field of biocatalysis inorganic solvents and the synthesis of biologically-active compounds.
Daria Giacomini was born in SanMarino (R. S. M.) in 1958, graduated in chemistry in 1982 and obtained a PhD in chemistry fromthe University of Bologna in 1987. She is currently Associate Professor in organic chemistry at the University of Bologna. Herresearch interests cover stereoselectivity in the synthesis of biologically-active compounds.
990 | Chem. Soc. Rev., 2009, 38, 990–1001 This journal is �c The Royal Society of Chemistry 2009
TUTORIAL REVIEW www.rsc.org/csr | Chemical Society Reviews
Dow
nloa
ded
by U
nive
rsity
of
New
Ham
pshi
re o
n 28
Feb
ruar
y 20
13Pu
blis
hed
on 1
2 Fe
brua
ry 2
009
on h
ttp://
pubs
.rsc
.org
| do
i:10.
1039
/B80
2815
JView Article Online / Journal Homepage / Table of Contents for this issue
are mainly considered, whereas differential enthalpy and
differential entropy both contribute to the Gibbs energy.
For the formation of products in a stereoselective process, at
least two distinct reaction pathways are possible: starting from
the reagents A and B, two stereoisomers C or D are obtained
(Fig. 1). The aim of the preparative chemist is to achieve a
stronger differentiation between the two reaction paths to yield
only one isomer in a high selectivity.
The stereoselectivity of a process is essentially a kinetic
phenomenon, quantitatively expressed by the ratio of the
kinetic constants that result from all the steps necessary to
convert the reagents into the two stereoisomers.3
This very important relationship is not often exploited
by synthetic chemists, but it highlights the possibility of
controlling the stereoselectivity by changing the reaction
conditions that directly influence those rate constants, such
as the reaction solvent and temperature.5
Temperature dependence analyses reveal solvent
effects
In recent years, the observation of temperature effects in
connection with the Eyring theory has been used to gain a
better understanding of the factors influencing the stereo-
chemical outcome of various reactions. In particular, studies
of enantio- and diastereoselective reactions by varying the
reaction temperature have shed light on solvent effects and
stereoselectivity.6
An analysis of the logarithmic values of stereoselectivity
as a function of reciprocal temperature results in the linear
relationship in the equation reported in Fig. 2, known as
the modified Eyring equation,7 where ln(k/k0) is the natural
logarithm of stereoselectivity, and k and k0 are the overall rate
constants of the reactions leading to the two stereoisomers.
According to Eyring theory, when this relationship is
plotted, the slope corresponds to the difference in the overall
activation enthalpies (DDHa) and the intercept represents the
difference in the overall activation entropies (DDSa) (Fig. 2).
The influence of the solvent on stereoselectivity can be
considered as a macroscopic effect related to solute–solvent
interactions at the molecular level. These interactions differ-
ently affect the reaction paths leading to the two stereoisomers
by changing the activation parameters and therefore the
stereoselectivity. The solvent effect can be so significant that
it can switch the stereoselectivity from enthalpy control (where
DDSa E 0) to entropy control (where DDHa E 0). As an
example, we observed such an effect on the facial diastereo-
selectivity of an aldol condensation under kinetic control.8
The reaction between lithium tert-butyl acetate and
2-phenylpropanal was studied in THF and n-hexane at
different temperatures, thus obtaining different syn and anti
aldol ratios. Data were analyzed according to the Eyring
equation, and values of DDHa and DDSa were obtained from
the slopes and intercepts of the linear plots of ln (syn/anti) vs. 1/T
(Fig. 3). In THF, DDSa was very small and only DDHa
determined the prevalence of the anti isomer. In contrast, in
n-hexane, where the enthalpy contribution was almost zero,
DDSa determined the diastereomeric ratio.
The differential entropy of activation (DDSa) reflects the
entropic difference between two diastereomeric activated
complexes, and it could depend on a different number of
microstates modulated by the reaction solvent. For enantio-
or diastereoselective reactions where small DDHas occur,
DDSa can be the determinant factor. Recent examples that
show constant selectivity with the variation of reaction tem-
perature have been reported in the literature by Inoue et al.9
and Gil et al.10 as ‘‘entropy-driven’’ processes.
It is difficult to rationalize solvation phenomena and to
discern the solvent effects on facial diastereoselectivity, but
some hypotheses could be considered. Accepting the effect of
solvents on conformational equilibria, two cases can be ratio-
nalised: (i) the solvent could enthalpically favour and
entropically disfavour an isomer whenever the solute–solvent
interactions are high in strength and/or in number, thus
stabilizing conformers with the lowest intramolecular steric
interactions; (ii) the solvent could entropically favour and
enthalpically disfavour an isomer whenever the solute–solvent
interactions are mild or low in number, which results in
the solvation being less ordered, thus making the system
conformationally less rigid.
It therefore appears clear that any theoretical model for the
analysis of stereochemical concerns completely looses its
predictive value by ignoring solvent effects.
Temperature-dependent studies of stereoselectivity often give
non-linear Eyring plots. Extreme values of ln(k/k0) vs. 1/TFig. 2 Stereoselectivity and thermodynamic parameters: a graphic
representation of the Eyring equation.
Fig. 3 Linear Eyring plots of the syn/anti diastereomeric ratios
obtained in an aldol reaction.
This journal is �c The Royal Society of Chemistry 2009 Chem. Soc. Rev., 2009, 38, 990–1001 | 991
Dow
nloa
ded
by U
nive
rsity
of
New
Ham
pshi
re o
n 28
Feb
ruar
y 20
13Pu
blis
hed
on 1
2 Fe
brua
ry 2
009
on h
ttp://
pubs
.rsc
.org
| do
i:10.
1039
/B80
2815
J
View Article Online
in the form of maxima and minima have been observed.
Fig. 4 reports an example by Collin et al. from the current
literature.11
The knowledge and understanding of such non-linear
behaviour is important for the optimization of selectivities,
especially the apex in the plot, which contains the most useful
information since, as in the example of Fig. 4, it represents the
point with the maximum available selectivity.
Non-linear Eyring plots consist of two distinct linear trends,
intersecting at a point which is identified as the so-called
inversion temperature (Tinv). This phenomenon was
thoroughly reviewed in 1991 by Scharf et al.,12 and it has been
observed quite frequently for different selective processes.
Pracejus and Tille first described a non-linear trend of
enantioselectivity in an asymmetric addition of alcohols to a
ketene catalyzed by chiral amines,13 and non-linear Eyring
plots have been observed in a large number of completely
different reactions: catalytic hydrogenations, photochemical
reactions like the Paterno–Buchi reaction, the Sharpless
dihydroxylation, Diels–Alder cycloadditions and several
addition reactions.14 Very recently, interesting examples of
non-linear Eyring plots in an enantioselective cycloaddition
have been reported by Buono et al.15
Fig. 5 and Fig. 6 show two recent examples of stereo-
selective Paterno–Buchi reactions with maxima16 or minima17
in their corresponding Eyring plots, respectively.
The phenomenon of inversion temperature and, more
importantly, its significance is a matter of debate.
In Scharf et al.’s studies, the observed inversion temperature
was explained by a reaction pathway with at least two stereo-
selective steps in the reaction mechanism that were differently
Fig. 4 A non-linear Eyring plot in an enantioselective Michael
addition (from: I. Reboule, R. Gil and J. Collin, Enantioselective
Conjugate Addition of Aromatic Amines to N-Alkenoyloxazolidinones
Catalyzed by Iodido(binaphtholato)samarium, Eur. J. Org. Chem.,
2008, 532–539. Copyright Wiley-VCH Verlag GmbH & Co. KGaA.
Reproduced with permission of Wiley).
Fig. 5 Non-linear Eyring plots in a regioselective cycloaddition
reaction (from: X.-M. Hei, Q.-H. Song, X.-B. Li, W.-J. Tang, H.-B.
Wang and Q.-X. Guo, Origin of a Large Temperature Dependence of
Regioselectivity Observed for [2 + 2] Photocycloaddition (Paterno–
Buchi Reaction) of 1,3-Dimethylthymine with Benzophenone and Its
Derivatives: Conformational Property of the Intermediary Triplet 1,4-
Diradicals, J. Org. Chem., 2005, 70, 2522. Reproduced with permis-
sion of the American Chemical Society).
Fig. 6 Non-linear Eyring plots in a diastereoselective cycloaddition
reaction (from: M. Abe, M. Terazawa, K. Nozaki, A. Masuyama and
T. Hayashia, Notable temperature effect on the stereoselectivity in
the photochemical [2 + 2] cycloaddition reaction (Paterno–Buchi
reaction) of 2,3-dihydrofuran-3-ol derivatives with benzophenone,
Tetrahedron Lett., 2006, 47, 2527. Reproduced with the permission
of Elsevier).
992 | Chem. Soc. Rev., 2009, 38, 990–1001 This journal is �c The Royal Society of Chemistry 2009
Dow
nloa
ded
by U
nive
rsity
of
New
Ham
pshi
re o
n 28
Feb
ruar
y 20
13Pu
blis
hed
on 1
2 Fe
brua
ry 2
009
on h
ttp://
pubs
.rsc
.org
| do
i:10.
1039
/B80
2815
J
View Article Online
affected by temperature. The Tinv was explained by the
existence of two levels of selectivity, and a change in the
dominance of activation enthalpy differences and activation
entropy differences occurring at different levels of selection in
different temperature domains. In this approach, the abrupt
change in stereoselectivity at the Tinv is due to a change of the
rate-determining stereoselective step.12
For Ridd et al., a sudden change in the dominance of
selection levels is unlikely. Their interpretation of the non-
linear behaviour is based on changes of the rate-determining
steps from one level of selection to the other, but only in one of
the two temperature domains is a single selection step really
rate-determining, while the remaining region may be viewed as
representing a transitional regime, in which none of the single
steps in the mechanism is rate-determining.18
Heller et al. have shown that Scharf and Ridd’s two
arguments are in fact not contradictory. Both interpretations
are possible within defined limits and by considering a
temperature-dependent change in the relative concentration
of two intermediates, which then leads to selective product
formation.19
Following Ridd et al.’s hypothesis, Gypser and Norrby
considered the non-linear relative rate plots in terms of their
absolute rate behaviour.20 They highlighted the ‘‘transition
region width’’ and ‘‘inversion temperature’’ as important
features in their interpretation of the relative rate plots of
the corresponding absolute rate plots. Their theoretical efforts
showed that inversion points for relative rates are physically
significant if the inversion temperatures for the absolute rates
are identical. If the inversion points for the absolute rates
differ, they have no physical meaning.
We have extensively studied the diastereofacial selectivity of
a-hydroxy aldehydes and silylimines, and have observed
non-linear Eyring plots in many cases. Based on experimental
data, we proposed a different interpretation of the inversion
temperature that mainly depends on solvation:
Tinv could constitute a transition temperature between two
different solute–solvent clusters, which act as distinct supra-
molecules with different thermodynamic properties, reactivity
and therefore stereoselectivity.21
Our proposal on the solvent-dependent nature of Tinv is
based upon two main lines of evidence obtained in a detailed
study on the diastereoselective addition of n-BuLi to
2-phenylpropanal and O-(TBS)-mandelic aldehyde in a series
of linear hydrocarbons:
(i) each aldehyde–solvent couple generates a characteristic
inversion temperature.
(ii) for each aldehyde, Tinv depends on the chain length of
the solvent.22
Measurements of the diastereomeric ratio of n-BuLi
addition to 2-phenylpropanal in a homologous series of linear
hydrocarbons showed some striking regularities in their
corresponding Eyring plots (Fig. 7).
The pattern is consistent with the chain length of the
solvent: the longer the hydrocarbon chain, the lower the de
and the higher the Tinv. In all cases, Tinv constitutes a break
point, leading to two sets of activation parameters, one for
T 4 Tinv and one for T o Tinv. At low temperatures,
the slopes of the regression lines (enthalpy contribution)
flatten; the differential entropy therefore modulates the stereo-
selectivity, and the effect of solvent chain length on face
selectivity is due to an entropic control. On the other hand,
Fig. 7 (a) Non-linear Eyring plots for the diastereoselective addition of n-BuLi to 2-phenylpropanal in hydrocarbons, and a correlation between
Tinv values and the melting points of the solvents. (b) Correlations between Tinv and melting point, and Tinv and the chain length of hydrocarbons
for the diastereoselective addition of n-BuLi to O-(TBS)-mandelic aldehyde in hydrocarbons.
This journal is �c The Royal Society of Chemistry 2009 Chem. Soc. Rev., 2009, 38, 990–1001 | 993
Dow
nloa
ded
by U
nive
rsity
of
New
Ham
pshi
re o
n 28
Feb
ruar
y 20
13Pu
blis
hed
on 1
2 Fe
brua
ry 2
009
on h
ttp://
pubs
.rsc
.org
| do
i:10.
1039
/B80
2815
J
View Article Online
at T 4 Tinv, lengthening the solvent chain increases the slope,
thus increasing the enthalpic contribution to the
diastereomeric ratio.
All efforts to correlate inversion temperatures with classical
solvent parameters, such as relative permittivity and
viscosity,23 have failed. Interestingly, we found a good corre-
lation between Tinv values and hydrocarbon melting points.
This correlation was confirmed, even in the case of O-(TBS)-
mandelic aldehyde, upon addition of n-BuLi, where Tinv
correlates better when considering the odd and even series of
linear hydrocarbons (Fig. 7(b)). A plot of the inversion
temperature vs. the number of solvent carbon atoms exhibits
the same alternating feature as that of the hydrocarbon
melting points. It is known that even- and odd-numbered
linear hydrocarbons differ in their crystalline packing, and
this fact causes the melting point alternation.24 A similar
alternation has been observed for the Tinv values obtained
with 2-phenylpropanal and O-(TBS)-mandelic aldehyde,
suggesting a relationship with the phase modification that
occurs upon melting.
An innovative explanation can be then formulated: Tinv
could constitute a sort of transition between two ‘phases’
which, in the case of solutions, could be represented by two
distinct solute–solvent clusters (Fig. 8). At a temperature
higher than Tinv, one solute–solvent cluster is present in
solution, generating a linear Eyring plot; at a temperature
lower than Tinv, another solvation cluster is present, generat-
ing a second linear plot. Thus in a non-linear Eyring plot, the
Tinv results from the intersection of the two linear plots is
generated by two independent supramolecular species that are
in equilibrium at Tinv. These two dynamic solvation clusters
behave like two different molecules, with different thermo-
dynamic parameters and therefore different stereoselectivities.
In this hypothesis, Tinv represents the interconversion
temperature between two supramolecules, and does not imply
any change in the rate-determining step or in the reaction
mechanism.
In a given experiment, in the explorable T range, if two
solvent–solute clusters were present, only one Tinv could be
observed. However, there are Eyring plots with two Tinv values
and thus three solvation supramolecules are present in the
range. We observed this phenomenon when n-BuLi was
added to O-(TBS)-mandelic aldehyde in n-hexane.22 Our
interpretation of the solvation-dependent nature of Tinv fits
well, even in those cases of non-linear Eyring plots with
multiple Tinv values, as well as in some other unusual temp-
erature behaviours, like that observed in the diastereoselective
allylation of the O-(TBS)-N-triisopropylsilylimine of lactal
(Fig. 9). This imine showed a reversal of diastereoselectivity
with temperature, going from 97% of the anti isomer at
�90 1C to 65% of the syn isomer at 60 1C. Moreover, in the
temperature range �40 to �46 1C, we have an abrupt jump in
diastereoselectivity.25
This interruption marks two linear trends, one at a higher T
and one at a lower T, with quite similar slopes but different
intercepts, meaning that the jump is derived exclusively from a
change in DDSa. This break in the Eyring plot can be readily
explained in terms of the interconversion of solute–solvent
clusters, such as at the inversion temperature, and, in our
opinion, it can be effectively considered a proper Tinv.
To strengthen the solvation hypothesis on Tinv, we reported
a study on the diastereofacial selectivity of EtMgBr addition
to (2S)-O-(tert-butyldimethylsilyl)lactal. Analysis of the
temperature dependence of the diastereomeric ratio gave
non-linear Eyring plots with peculiar Tinv values for each
solvent (Fig. 10).26
We observed that solute–solvent interactions govern a
diastereomeric switch from the anti to the syn isomer upon
changing from THF to an anisole solvent system. In THF, the
temperature scarcely affects the diastereofacial selectivity over
a wide T range until Tinv, after which the diastereomeric ratio
(dr) rapidly decreases; the flattest part of the plot, and con-
sequently the steady diastereomeric excess, is derived from a
small value of DDHa, and thus the good diastereoselectivity
can be entirely attributed to an entropic effect. Diethylether
and di-n-pentylether have parallel Eyring plots at T 4 Tinv,
whereas branched ethers run almost parallel in all T ranges:
this means that only DDSa contributes to differentiate the
diastereomeric ratio between the reaction solvents. In spite of
difficulties in modelling entropy, our results clearly demon-
strate how deep its influence on diastereofacial selectivity
can be.Fig. 8 Tinv in an Eyring plot as the interconversion temperature
between two solute–solvent clusters.
Fig. 9 The Eyring plot for the diastereoselective addition toO-(TBS)-
imine of (2S)-lactal in THF.
994 | Chem. Soc. Rev., 2009, 38, 990–1001 This journal is �c The Royal Society of Chemistry 2009
Dow
nloa
ded
by U
nive
rsity
of
New
Ham
pshi
re o
n 28
Feb
ruar
y 20
13Pu
blis
hed
on 1
2 Fe
brua
ry 2
009
on h
ttp://
pubs
.rsc
.org
| do
i:10.
1039
/B80
2815
J
View Article Online
There has been remarkable progress in the development of
electronic theories of molecules in solution in the last two
decades,27–29 but efficient and accurate treatment of solvation
remains a problem in molecular modelling, despite its prime
importance. Molecular clusters are nowadays the preferred
models to study both the energetics and dynamics of molecules
in a tailored molecular environment,30,31 and in this context
we developed the idea of the dynamic solvation cluster nature
of the inversion temperature.
It is difficult to formulate a detailed microscopic model for
solvation cluster interconversion at Tinv and, where the experi-
mental structural information in solution is missing, one can
rely solely on hypotheses. The interconversion of solvation
clusters at Tinv could involve an internal rearrangement of
the solvation cluster, or a solvent shell reorganization with
transfer of solvent molecules from the solute–solvent cluster to
the medium.
The experimental results now available strongly suggest that
for all molecules in solution, even in the presence of weak
solvation forces such as in hydrocarbons, the active chemical
species is not the bare molecule but molecule–solvent clusters,
whose structure and reactivity depend on the temperature; the
inversion temperature reveals effects on the stereoselectivity by
distinct solvation clusters.
Spectroscopic evidence for solvation cluster theory
on Tinv
To support our interpretative model of inversion temperature,
we considered spectroscopic techniques. We started with 13C
NMR spectroscopy because chemical shifts are affected by
intermolecular interactions with the solvent. We recorded 13C
NMR spectra of some aldehydes in d8-THF or d14-n-hexane at
several temperature values and noted an evolution of the
CQO chemical shift(s) with temperature.32 After full assign-
ment, all spectra showed a unique set of signals in the
temperature range explored, which ensured the presence of a
population-weighted average of rapidly interconverting
conformers. Fig. 12 reports dCQO vs. T for (S)-O-(TBS)-lactal
in deuterated THF and n-hexane. In both solvents, the
carbonyl chemical shift decreased upon warming. This carbonyl
shielding could be interpreted as a loss of electron density:
upon raising the temperature, a looser solvation may be
envisaged. Upon switching from THF to n-hexane, a shift of
Dd = 4 was observed for the CQO signal: the ether solvent
enhances the shielding thanks to a higher electron density on
the carbonyl group than the hydrocarbon. Moreover, on
closer inspection, it could be recognized that both plots were
composed of two linear segments, intersecting at a tempera-
ture that we named TNMR (Fig. 11). The data for dCQO vs. T
were analyzed by a linear least-squares fitting, and a residual
analysis was applied to evaluate the number of linear trends
and the value of the intersecting temperature (TNMR). More
specifically, the complete {dCQOi, Ti} data set was first esti-
mated by fitting a single regression line. The data pair corres-
ponding to the largest absolute value residual was used to
partition the data set into left and right subsets. The two
groups of points were then separately fitted with a least-
squares straight line. For the n-hexane solution, TNMR
occurred at �68 1C, whereas this figure was 0 1C in THF.
These TNMR numerical values are close to the inversion
temperatures of the diastereoselective nucleophilic addition to
(S)-O-(TBS)-lactal: with n-BuLi in THF, Tinv = +1 1C, with
tBuLi in n-hexane, Tinv = �81 1C, respectively.32
We found equivalence between TNMR and Tinv values
several times; for instance, in the Diels–Alder reaction in
toluene (Fig. 12; Tinv = �61 1C, TNMR = �62 1C).
Fig. 10 Non-linear Eyring plots of the diastereomeric ratios in the diastereoselective addition of EtMgBr to O-(TBS)-lactal in a series of ethers.
Fig. 11 Non-linear plots of 13CQO chemical shift vs. temperature for
O-(TBS)-lactal in deuterated solvents.
This journal is �c The Royal Society of Chemistry 2009 Chem. Soc. Rev., 2009, 38, 990–1001 | 995
Dow
nloa
ded
by U
nive
rsity
of
New
Ham
pshi
re o
n 28
Feb
ruar
y 20
13Pu
blis
hed
on 1
2 Fe
brua
ry 2
009
on h
ttp://
pubs
.rsc
.org
| do
i:10.
1039
/B80
2815
J
View Article Online
From all of the collected data, TNMR and Tinv emerge as two
independent experimental observations related to a specific
solute–solvent couple, thus reinforcing our hypothesis on the
solvation-dependent nature of Tinv.
More recently, variable temperature NMR measurements
were coupled with the circular dichroism (CD) and ultraviolet
(UV) spectra of solutions of chiral aldehydes in the absence of
any reaction.33
UV-vis absorption and CD are two techniques also affected
by solute–solvent interactions, CD being generally the more
sensitive. In collaboration with Spada et al.,33 we evaluated the
UV and CD spectra of a-chiral aldehydes in n-decane and
n-dodecane solution upon changing the temperature, and
compared the results with 13C NMR variable temperature
analyses and the observed Tinv values of diastereoselective
reactions. As an example, in Fig. 13, ellipticity and absorbance
at selected wavelengths for (S)-O-(TBS)-lactal in n-dodecane,
and CQO chemical shift, were plotted against T. All the plots
presented break points, denoting a change of slope. It is worth
noting that all break temperatures (named TUV, TCD and
TNMR) matched well with Tinv, within experimental error,
being obtained from the diastereoselective addition of n-BuLi
to (S)-O-(TBS)-lactal in n-dodecane.
Small differences in numerical values could be accounted for
by the fact that TNMR, TCD and TUV refer to temperature-
dependent changes in solvation of the reactant in the ground
state, whereas the Tinv refers to the effect of that solvation
change on the differential activation energy (DDGa) of the two
diastereomeric reaction paths. Moreover, TNMR was detected
in a deuterated solvent (d26-n-dodecane). Thus, TNMR, TCD,
Fig. 12 Non-linear Eyring plots in a diastereoselective Diels–Alder cycloaddition.
Fig. 13 Temperature-dependent data for O-(TBS)-lactal in n-dodecane: (a) 13C NMR chemical shift of CQO, (b) CD ellipticity, (c) anti/syn ratio
upon n-BuLi addition and (d) UV absorbance.
996 | Chem. Soc. Rev., 2009, 38, 990–1001 This journal is �c The Royal Society of Chemistry 2009
Dow
nloa
ded
by U
nive
rsity
of
New
Ham
pshi
re o
n 28
Feb
ruar
y 20
13Pu
blis
hed
on 1
2 Fe
brua
ry 2
009
on h
ttp://
pubs
.rsc
.org
| do
i:10.
1039
/B80
2815
J
View Article Online
TUV and Tinv are effectively independent experimental results
of the same phenomenon that is directly linked to the aldehyde–
solvent pair. TNMR, TCD and TUV reveal the presence of an
equilibrium between two aldehyde–solvent clusters, which
exist in specific temperature domains, and Tinv reveals the
equilibrium between the clusters, which has important effects
on the stereoselectivity.
Solvation modelling and the inversion temperature
The phenomenon of inversion temperature and its experi-
mentally-observed dependence on the substrate–solvent pair
is not easy to rationalize intuitively, nor to be predicted using
quantum modelling, not only in the traditional ‘‘one molecule
in a vacuum approach’’, but also using fixed conformation
continuum or discrete models with a few explicit solvent
molecules to account for the solvent effect. Indeed, on the
one hand, important differences in inversion temperature are
found in different linear alkanes with essentially the same
relative permittivity, but on the other, the good conforma-
tional sampling in an equilibrium-condensed phase that would
be necessary to undertake this task is, from a computational
point of view, beyond the current capability of quantum
chemical calculations.
In collaboration with Zannoni et al., we have tried to
predict the p-facial selectivity of solvation with an atomistic
simulation.34 Using molecular dynamics simulations, we
compared the facial solvation of (R)-2-phenylpropanal in
n-pentane and n-octane at various temperatures with the
diastereoselectivity data for the n-BuLi addition to (R)-2-
phenylpropanal, which gave non-linear Eyring plots with
peculiar inversion temperatures (Fig. 14).
Extensive molecular dynamics computer simulations for
(R)-2-phenylpropanal in n-pentane and n-octane were
performed with the aim of examining the distribution of
solvent molecules around the solute and the properties of
the chiral solute in the explicit medium. It was found that
the close presence of a chiral center determined different
average solvent densities on the two faces of the p-bond and
that the density depends on the solvent. From this point of
view, n-pentane and n-octane showed remarkable differences,
which were likely to affect the diastereoselectivity of a nucleo-
philic reaction if we think of the solvent density as a measure
of the accessibility of a face. The higher diastereoselectivity
reported in n-pentane can be rationalized as a higher accessi-
bility of the face that leads to the anti product.
As this concept is not sufficient in itself to explain the
presence of an inversion temperature, we then focused on a
physical property of the solute: the conformationally-averaged
molecular chiral index G0S, which can be determined from the
simulation data.34 We found that ln G0S exhibits a non-linear
temperature dependence in both solvents, with break tempera-
tures, TCI (Fig. 15), that are in good agreement with the
experimental Tinv values (n-pentane: TCI = �50 1C and
Tinv = �69 1C, n-octane: TCI = 1 1C and Tinv = 15 1C;
Fig. 14).
We believe that this MD simulation could lead to a viable
approach to predict Tinv values and other subtle solvent effects
in a number of stereoselective reactions.
The effects of solvent mixtures on stereoselectivity
and Tinv
The development and design of industrial processes requires
the selection of the most adequate solvent to increase the
productivity, reduce waste and improve the whole product
cycle, according to the demand for sustainable and environmentallyFig. 14 Eyring plots of diastereoselective addition to (R)-2-phenyl-
propanal.
Fig. 15 A snapshot of the united atom n-pentane molecules partially
or totally contained in the first solvation sphere around the carbonyl
carbon (n-pentane, 180 K). Molecular chiral index G0S vs. T for (R)-2-
phenylpropanal obtained from MD simulations in hydrocarbons
(from: R. Berardi, G. Cainelli, P. Galletti, D. Giacomini, A. Gualandi,
L. Muccioli and C. Zannoni, Can the p-Facial Selectivity of Solvation
Be Predicted by Atomistic Simulation?, J. Am. Chem. Soc., 2005, 127,
10699. Reproduced with permission of the American Chemical
Society).
This journal is �c The Royal Society of Chemistry 2009 Chem. Soc. Rev., 2009, 38, 990–1001 | 997
Dow
nloa
ded
by U
nive
rsity
of
New
Ham
pshi
re o
n 28
Feb
ruar
y 20
13Pu
blis
hed
on 1
2 Fe
brua
ry 2
009
on h
ttp://
pubs
.rsc
.org
| do
i:10.
1039
/B80
2815
J
View Article Online
friendly technologies. The use of mixed solvents is almost
ubiquitous, ranging from petrochemical to pharmaceutical
industries. It appears interesting then to evaluate the impact
of solvent mixtures on stereoselectivity and Tinv.
We have investigated the effect on diastereoselectivity of
solvent mixtures with respect to pure solvents in the addition
of n-BuLi to 2-phenylpropanal (Fig. 16).35
In binary mixtures of hydrocarbons, Tinv was still observed
but it was generally higher than that obtained in pure solvents.
This result suggests that solvation clusters with a mix of
solvents could be thermodynamically more stable than those
occurring in the corresponding pure solvents, so that their
interconversion requires more energy. It is interesting to note
that in a ternary mixture composed of n-pentane, n-hexane
and n-decane in a 1 : 1 : 1 molar ratio, Tinv disappeared, and
the diastereoselectivity was slightly higher than that obtained
in each individual pure solvent. Since at high T the DDHa is
lower than in pure solvents, the higher selectivity reached in
ternary mixtures is due to a larger DDSa contribution.
Studying the correlation of Tinv with the n-hexane/n-decane
molar fraction, it was observed that the greatest change in Tinv
occurred at the extremities of the graph close to the pure
solvents, whereas over a large mole fraction range, Tinv was
relatively unaffected by the composition of the mixture
(Fig. 17). The consistency of Tinv could be due to the presence
of two solvation clusters with microstructures independent of
the composition of the bulk solvent. This is quite interesting
because hydrocarbons, which should exert only weak and
non-specific interactions with a neutral dipolar solute, could
reveal more structured solvation shells than is currently
believed. Interpolation of the experimental data allowed us
to evaluate the molecular composition of the solvation shell,
which was a ratio of 5 : 2 n-hexane/n-decane per mole of
2-phenylpropanal.
The study on solvent mixtures was extended to mixtures of
n-hexane and 5 mol% of a tertiary amine. We examined the
role of several amines as co-solvents in the addition of n-BuLi
to 2-phenylpropanal and (S)-O-(TBS)-lactal, which have been
revealed to have strong effects on diastereofacial selectivity
(Fig. 18).36 The presence of such trialkylamines in n-hexane
substantially enhanced the preferential formation of the anti
isomer for 2-phenylpropanal across the T range, allowing an
increased diastereoselectivity with respect to that in pure
n-hexane or pure THF.
The possibility to improve diastereofacial selectivity by
adding a tertiary amine as a co-solvent to a hydrocarbon is
quite an interesting result in view of the potential synthetic and
industrial applications, and for the possible replacement
of THF as a solvent in organometallic synthesis with its
associated cost and safety issues.
The presence of Tinv values in Eyring plots, which differ
when changing the amine, confirm the direct involvement of
solute–solvent clusters, which are subject to temperature-
dependent reorganization phenomena. The mixture with
Et3N presented Tinv as a jump, as in the case of the allylation
of lactalimine (cf. Fig. 9). Moreover, Tinv values in n-hexane/
amine mixtures occur at higher values than in pure n-hexane,
Fig. 16 Eyring plots for diastereoselective addition to 2-phenylpropanal in pure solvents and solvent mixtures.
Fig. 17 The Tinv dependence on solvent composition for the addition
of n-BuLi to 2-phenylpropanal.
998 | Chem. Soc. Rev., 2009, 38, 990–1001 This journal is �c The Royal Society of Chemistry 2009
Dow
nloa
ded
by U
nive
rsity
of
New
Ham
pshi
re o
n 28
Feb
ruar
y 20
13Pu
blis
hed
on 1
2 Fe
brua
ry 2
009
on h
ttp://
pubs
.rsc
.org
| do
i:10.
1039
/B80
2815
J
View Article Online
as in the case of binary mixture n-hexane/n-decane (Fig. 16),
thus indicating the greater stability of solvation clusters in
mixtures.
The addition of n-Bu3N as a co-solvent changed the
temperature dependence, even for the diastereoselective
addition of (cyanomethyl)lithium to 2-phenylpropanal
(Fig. 19).37 In this case, we observed an opposite temperature
behaviour in the diastereomeric ratio for reactions in n-hexane
vs. n-hexane/n-Bu3N and methylcyclohexane. We have
already observed opposite concavities for the Eyring plots of
cyclic vs. linear hydrocarbons in the addition of n-BuLi to
2-phenylpropanal.21
The change in concavity is derived from a change in the
slope of one of the two linear parts of the plot, and hence it
depends on a change in the enthalpic contribution to the
stereoselectivity. However, this quite interesting aspect de-
serves deeper investigation, and dedicated MD simulations
could disclose differences in microscopic solvation that may
account for this linear vs. cyclic solvent effect.38
The inversion temperature in enzymatic reactions
Over the past few years, the use of enzymes in organic
synthesis has become increasingly important due to the basic
discovery that enzymatic reactions can occur in organic
solvents as well as in aqueous media. Indeed, enzymatic
selectivity is remarkably dependent on the organic solvent,
and many attempts have been made to explain this influence.39
However, less attention has been paid to the effect of
temperature on enzymatic selectivity.
We have reported an example of inversion temperature and
the corresponding TNMR found in the enzymatic resolution of
a b-lactam by Penicillin-G Acilase (Fig. 20). It is interesting to
note that the highest enantioselectivity can be achieved at
higher temperatures, whereas at lower temperatures, the
stereoselectivity is less dependent on T, thus revealing once
more a control by DDSa.
On the basis of our interpretation, the presence of a Tinv,
indicates the presence of two dynamic substrate–solvent
Fig. 18 Eyring plots of the diastereoselective addition of n-BuLi to 2-phenylpropanal in n-hexane and tertiary amines 5% mixtures.
Fig. 19 Eyring plots for the diastereoselective addition of (cyanomethyl)lithium to 2-phenylpropanal in n-hexane, methylcyclohexane and
n-Bu3N 5% mixtures.
This journal is �c The Royal Society of Chemistry 2009 Chem. Soc. Rev., 2009, 38, 990–1001 | 999
Dow
nloa
ded
by U
nive
rsity
of
New
Ham
pshi
re o
n 28
Feb
ruar
y 20
13Pu
blis
hed
on 1
2 Fe
brua
ry 2
009
on h
ttp://
pubs
.rsc
.org
| do
i:10.
1039
/B80
2815
J
View Article Online
clusters, which affect the thermodynamic parameters and
therefore the enantioselectivity, even of enzymatic reactions.40
As in the case of diastereoselective processes, the nature of
these solvation clusters and their behavior with temperature
should not depend on the reaction but mainly on the
solute–solvent pair. To verify this hypothesis, we compared
the enantioselectivity behaviour with temperature of the
enzymatic (bio-catalyst) and a non-enzymatic (chemo-catalyst)
acylation reactions of some racemic and meso alcohols.41 We
selected lipases as bio-catalysts and chiral proline-based
diamines as chemo-catalysts. We then examined the Eyring
plots of enantioselectivity obtained with the chemo- and
bio-catalysts in the same reaction solvent. Fig. 21 reports an
example with 2-phenylpropanol in THF.
All three plots present Tinv values, and the numerical values
for the bio- and the chemo-catalyzed reactions are fairly
similar. This result confirms the dependence of Tinv on
solute–solvent clustering phenomena and not on the reaction
considered, even in the case of enzymatic reactions.
We then confirmed the solute–solvent clustering phenomena
with variable temperature 13C NMR spectroscopy. Fig. 21
shows plots of selected chemical shifts of 2-phenylpropanal in
d8-THF vs. absolute temperature. More than one carbon atom
showed non-linear behaviour with very similar TNMR values.
A direct comparison showed a close correspondence between
Tinv obtained from the bio- and chemo-catalyzed reactions in
THF, and the TNMR values.
As already discussed, Tinv and TNMR are independent
experimental observations of the same phenomenon of solva-
tion clustering. The results obtained therefore confirm that the
Tinv is quite a general phenomenon that is independent of the
particular reaction system, including the nature of the catalyst,
i.e., enzyme or chiral diamine, but tightly related to the nature
of the substrate–solvent pair.
Nakamura et al. observed that lipases recognize not only the
structure of the substrate, but also that of the solvent, assum-
ing that a solvent influences enzymatic reactions as a molecule
rather than as a bulk medium.42 Klibanov and co-workers
Fig. 20 Non-linear Eyring plots of enzymatic enantioselectivity (E) in the resolution of a racemic azetidinone and its corresponding TNMR.
Fig. 21 The Eyring plot for the enantioselective acetylation of
2-phenylpropanol with lipase and chiral diamines, and the chemical
shifts for selected carbons in 2-phenylpropanol vs. temperature in
deuterated THF.
1000 | Chem. Soc. Rev., 2009, 38, 990–1001 This journal is �c The Royal Society of Chemistry 2009
Dow
nloa
ded
by U
nive
rsity
of
New
Ham
pshi
re o
n 28
Feb
ruar
y 20
13Pu
blis
hed
on 1
2 Fe
brua
ry 2
009
on h
ttp://
pubs
.rsc
.org
| do
i:10.
1039
/B80
2815
J
View Article Online
have stated that the solvent dependence of enzymatic prochiral
selectivity can be attributed primarily to changes in the relative
solvation modes of the substrate in the transition state.43 Our
interpretation is not in contrast with these observations, but
our comparative studies have confirmed that solute–solvent
clusters and their temperature-dependent dynamics have a
fundamental influence on stereoselectivity in chemical, as well
as in enzymatic, reactions.
Concluding remarks
Our interest in the stereoselectivity of chemical and enzymatic
reactions challenged us to investigate solvation effects. Given
the enormous interest in improving diastereo- and enantio-
selectivity, particularly in the synthesis of biologically-active
compounds, our results focus attention on a set of important
topics and open questions:
� Temperature-dependent studies on stereoselective reac-
tions can reveal dynamic solvent effects, with the presence of
the inversion temperature phenomenon.
� For all reactions in solution, the reactive species is not the
bare solute molecule but a solute–solvent cluster whose struc-
ture and reactivity depends on temperature.
� Any theoretical model for the analysis of stereoselective
processes completely looses its predictive value by ignoring
solvent effects.
� Entropy could be an important parameter in determining
stereoselectivity, even though it is generally ignored in models
of asymmetric induction; in some cases, it can be the unique
cause of the observed selectivity.
� Stereoselectivity can be improved by using mixtures of
solvents or co-solvents, essentially because of an entropic
control.
Acknowledgements
We are grateful to Prof. Spada and Prof. Zannoni for their
fruitful collaboration. We thank Dr Francesca Paradisi for
helpful discussions and comments. We thank our graduate and
postdoctoral co-workers, whose names are cited in the references,
for their valuable contribution to the work of our group. We are
also grateful to MIUR for supporting this research.
References
1 As a recent and general review, see: E. Negishi, Bull. Chem. Soc.Jpn., 2007, 80, 233–257 and references cited therein.
2 As a recent and general review, see: H. Pellissier, Tetrahedron,2007, 18, 2249–2293 and references cited therein.
3 A. D. McNaught and A. Wilkinson, IUPAC Compendium ofChemical Terminology: The Gold Book, Blackwell Science, Oxford,UK, 2nd edn, 1997.
4 E. L. Eliel, S. H. Wilen and L. N. Mander, Stereochemistry ofOrganic Compounds, Wiley, New York, 1994.
5 I. W. M. Smith, Chem. Soc. Rev., 2008, 37, 812–826.6 G. Cainelli, D. Giacomini and P. Galletti, Chem. Commun., 1999,567–572.
7 S. Glasstone, K. J. Laidler and H. Eyring, The Theory of RateProcesses, McGraw-Hill, New York, 1941, ch. 4, p. 153.
8 G. Cainelli, P. Galletti, D. Giacomini and P. Orioli, TetrahedronLett., 2001, 42, 7383–7385.
9 Y. Inoue, H. Ikeda, M. Kaneda, T. Sumimura, S. R. L. Everitt andT. Wada, J. Am. Chem. Soc., 2000, 122, 406–407 and referencescited therein.
10 O. Benali, M. A. Miranda, R. Tormos and S. Gil, J. Org. Chem.,2002, 67, 7915–7918 and references cited therein.
11 I. Reboule, R. Gil and J. Collin, Eur. J. Org. Chem., 2008, 532–539.12 H. Buschmann, H.-D. Scharf, N. Hoffmann and P. Esser, Angew.
Chem., Int. Ed. Engl., 1991, 30, 477–515.13 H. Pracejus and A. Tille, Chem. Ber., 1963, 854–865.14 D. Heller and H. Buschmann, Top. Catal., 1998, 5, 159–176 and
references cited therein.15 O. Pardigon, A. Tenaglia and G. Buono, J. Mol. Catal. A: Chem.,
2003, 196, 157–164.16 X.-M. Hei, Q.-H. Song, X.-B. Li, W.-J. Tang, H.-B. Wang and
Q.-X. Guo, J. Org. Chem., 2005, 70, 2522–2527.17 M. Abe, M. Terazawa, K. Nozaki, A. Masuyama and T. Hayashi,
Tetrahedron Lett., 2006, 47, 2527–2530.18 K. J. Hale and J. H. Ridd, J. Chem. Soc., Perkin Trans. 2, 1995,
1601–1605.19 D. Heller, H. Buschmann and H.-D. Scharf, Angew. Chem., Int.
Ed. Engl., 1996, 35, 1852–1854.20 A. Gypser and P.-O. Norrby, J. Chem. Soc., Perkin Trans. 2, 1997,
939–943.21 G. Cainelli, D. Giacomini, P. Galletti and A. Marini, Angew.
Chem., Int. Ed. Engl., 1996, 35, 2849–2852.22 G. Cainelli, D. Giacomini, P. Galletti, P. Orioli and F. Paradisi,
Eur. J. Org. Chem., 2000, 3619–3626.23 A solvent viscosity dependence of the diastereoselectivity in a
Paterno–Buchi reaction has been reported: A. G. Griesbeck,S. Bondock and M. S. Gudipati, Angew. Chem., Int. Ed., 2001,40, 4684–4687.
24 A. T. Balaban, D. J. Klein, N. H. March, M. P. Tosi andM. Ausloos, ChemPhysChem, 2005, 6, 1741–1745.
25 G. Cainelli, D. Giacomini, P. Galletti and A. Quintavalla, Eur. J.Org. Chem., 2002, 3153–3161.
26 G. Cainelli, D. Giacomini, P. Galletti and P. Orioli, Eur. J. Org.Chem., 2001, 4509–4515.
27 C. Reichardt, Solvent and Solvent Effects in Organic Chemistry,Wiley-VCH Verlag GmbH & Co. KgaA, Weinheim, 3rd edn, 2003.
28 A. A. Neufeld, J. Chem. Phys., 2005, 122(16), 1641101–16411012.29 M. Orozco and F. J. Luque, Chem. Rev., 2000, 100, 4187–4225.30 M. Sedlak, J. Phys. Chem. B, 2006, 110, 4329–4338.31 A. W. Castleman, Jr and K. H. Bowen, Jr, J. Phys. Chem., 1996,
100, 12911–12944.32 G. Cainelli, D. Giacomini, P. Galletti and P. Orioli, Angew. Chem.,
Int. Ed., 2000, 39, 523–527.33 G. Cainelli, P. Galletti, S. Pieraccini, A. Quintavalla, D. Giacomini
and G. Spada, Chirality, 2004, 16, 50–56.34 R. Berardi, G. Cainelli, P. Galletti, D. Giacomini, A. Gualandi,
L. Muccioli and C. Zannoni, J. Am. Chem. Soc., 2005, 127,10699–10706.
35 G. Cainelli, P. Galletti, D. Giacomini, P. Orioli, A. Polidoro andM. C. Righetti, Helv. Chim. Acta, 2000, 83, 1951–1961.
36 G. Cainelli, D. Giacomini, P. Galletti and A. Quintavalla, Eur. J.Org. Chem., 2003, 1993–2000.
37 G. Cainelli, P. Galletti, D. Giacomini, A. Gualandi andA. Quintavalla, Tetrahedron, 2005, 61, 69–75.
38 For a mathematical discussion about maxima and minima in non-linear Eyring plots, see: D. Heller, H. Buschmann andH. Neumann, J. Chem. Soc., Perkin Trans. 2, 1999, 175–181.
39 G. Carrea and S. Riva, Angew. Chem., Int. Ed., 2000, 39,2226–2254.
40 For solute–solvent dynamics and energetics in enzyme catalysis,see: M. H. M. Olsson and A. Warshel, J. Am. Chem. Soc., 2004,126, 15167–15179.
41 G. Cainelli, P. Galletti, D. Giacomini, A. Gualandi andA. Quintavalla, Helv. Chim. Acta, 2003, 86, 3548–3559.
42 K. Nakamura, M. Kinoshita and A. Ohno, Tetrahedron, 1995, 51,8799–8808.
43 For the de-solvation model of organic solvent effects on enzymaticreactions, see: T. Ke, C. R. Wescott and A. M. Klibanov, J. Am.Chem. Soc., 1996, 118, 3366–3374.
This journal is �c The Royal Society of Chemistry 2009 Chem. Soc. Rev., 2009, 38, 990–1001 | 1001
Dow
nloa
ded
by U
nive
rsity
of
New
Ham
pshi
re o
n 28
Feb
ruar
y 20
13Pu
blis
hed
on 1
2 Fe
brua
ry 2
009
on h
ttp://
pubs
.rsc
.org
| do
i:10.
1039
/B80
2815
J
View Article Online