Radical Cyclisation of α-Halo Aluminium Acetals: A Mechanistic Study

Article abstract of DOI:10.1002/chem.201504780

α-Bromo aluminium acetals are suitable substrates for Ueno-Stork-like radical cyclisations affording γ-lactols and acid-sensitive methylene-γ-lactols in high yields. The mechanistic study herein sets the scope and limitation of this reaction. The influence of the halide (or chalcogenide) atom X (X=Cl, Br, I, SPh, SePh) in the precursors α-haloesters, as well as influence of the solvent and temperature was studied. The structure of the aluminium acetal intermediates resulting from the reduction of the corresponding α-haloesters has been investigated by low-temperature ¹³C-INEPT diffusion-ordered NMR spectroscopy (DOSY) experiments and quantum calculations, providing new insights into the structures of these thermally labile intermediates. Oxygen-bridged dimeric structures with a planar Al₂O₂ ring are proposed for the least hindered aluminium acetals, while monomeric structures seem to prevail for the most hindered species. A comparison against the radical cyclisation of aluminium acetals derived from allyl and propargyl alcohols with the parent Ueno-Stork has been made at the BHandHLYP/6-311++G(d,p) level of theory, highlighting mechanistic similarities and differences.

Full text of DOI:10.1002/chem.201504780

DOI: 10.1002/chem.201504780
Full Paper
&
Radical Cyclisation
Radical Cyclisation of a-Halo Aluminium Acetals: A Mechanistic
Study
Romain BØnØteau,[a] Anne Boussonni›re,[a] Jean-Christophe Rouaud,[a] Jacques Lebreton,[a]
JØrôme Graton,[a] Denis Jacquemin,[a, b] Muriel Sebban,[c] Hassan Oulyadi,[c]
Ghanem Hamdoun,[c] Amber N. Hancock,[d] Carl H. Schiesser,[d] and Fabrice DØn›s*[a]
Dedicated to Professor Jean-FranÅois Normant on the occasion of his 80th birthday.
Abstract: a-Bromo aluminium acetals are suitable substrates
for Ueno–Stork-like radical cyclisations affording g-lactols
and acid-sensitive methylene-g-lactols in high yields. The
mechanistic study herein sets the scope and limitation of
this reaction. The influence of the halide (or chalcogenide)
atom X (X=Cl, Br, I, SPh, SePh) in the precursors a-haloest-
ers, as well as influence of the solvent and temperature was
studied. The structure of the aluminium acetal intermediates
resulting from the reduction of the corresponding a-halo-
esters has been investigated by low-temperature 13C-INEPT
diffusion-ordered NMR spectroscopy (DOSY) experiments
and quantum calculations, providing new insights into the
structures of these thermally labile intermediates. Oxygen-
bridged dimeric structures with a planar Al2O2 ring are pro-
posed for the least hindered aluminium acetals, while mono-
meric structures seem to prevail for the most hindered
species. A comparison against the radical cyclisation of
aluminium acetals derived from allyl and propargyl
alcohols with the parent Ueno–Stork has been made at the
BHandHLYP/6-311 + +G(d,p) level of theory, highlighting
mechanistic similarities and differences.
Introduction
resulting from the reduction of a-bromoesters with DIBAL-H
were stable enough at temperatures below À708C to engage
in a radical cyclisation in the presence of nBu3SnH, initiated
with Et3B/air. The process was high yielding with most of the
a-bromo precursors and gave access to g-lactols[2] and to
highly acid-sensitive methylene-g-lactols.[3] One of the main ad-
vantages of this reaction compared to the classical Ueno–Stork
reaction,[4,5] which uses a-haloacetals as the precursors for the
radical cyclisation, is that it allows for a direct access to g-lac-
tols or lactones without the need for strongly acidic condi-
tions. In that sense, it complements nicely the Ueno–Stork re-
action, which gives access to cyclic acetals, for which the reac-
tive aldehyde functionality is protected as an acetal, thus offer-
ing the possibility to pursue further functionalisation elsewhere
on the molecule. The aluminium acetals resulting from the cyc-
lisation process are reactive intermediates that could be oxi-
dised by simple aldehydes to produce g-lactones[6] and poly-
substituted butenolides.[7] So far no mechanistic study has
been reported. The structure of the aluminium acetal inter-
mediates of both the a-halo radical precursor and the radical
species that undergoes radical cyclisation are totally unknown.
In order to get some insights into the reaction mechanism of
this reaction, we conducted a systematic study that allowed us
to highlight the subtle differences in reactivity between differ-
ent types of precursors. Indeed, for similar structures of the a-
haloesters used as precursors for the preparation of the alumi-
nium acetals, the yields obtained in g-lactols strongly depend-
ed upon the nature of the halide or chalcogenide, the solvent,
Aluminium acetals are long-known intermediates in the
reduction of esters to aldehydes or alcohols with aluminium
hydrides. These species are thermally labile, but those resulting
from a reduction with diisobutylaluminium hydride (DIBAL-H)
proved to be sufficiently stable at low temperature to be
quenched to give either aldehydes (acidic quenching) or
mixed acetals (e.g., trapping with an acylating agent).[1] We
previously reported the use of these intermediates for the first
time in a radical reaction. We showed that aluminium acetals
[a] R. BØnØteau, A. Boussonni›re, J.-C. Rouaud, Prof. Dr. J. Lebreton,
Dr. J. Graton, Prof. Dr. D. Jacquemin, Dr. F. DØn›s
CEISAM UMR 6230 - UFR des Sciences et des Techniques
UniversitØ de Nantes, 2 rue de la Houssini›re BP
92208-44322 Nantes Cedex 3 (France)
E-mail: fabrice.denes@univ-nantes.fr
[b] Prof. Dr. D. Jacquemin
Institut Universitaire de France, 1, rue Descartes
75231 Paris Cedex 05 (France)
[c] Dr. M. Sebban, Prof. Dr. H. Oulyadi, G. Hamdoun
Normandie UniversitØ, COBRA UMR6014 and FR3038, UniversitØ de Rouen;
INSA de Rouen; CNRS,
76821 Mont-Saint-Aignan Cedex (France)
[d] Dr. A. N. Hancock, Prof. Dr. C. H. Schiesser
School of Chemistry and
Bio21 Molecular Science and Biotechnology Institute
The University of Melbourne, Victoria, 3010 (Australia)
Supporting information and ORCIDs for the authors for this article is avail-
able on the WWW under http://dx.doi.org/10.1002/chem.201504780
Chem. Eur. J. 2016, 22, 4809 – 4824
4809
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
and the temperature. Using both experimental (synthetic and
low-temperature NMR studies) and computational methods in
parallel, we were able to get useful information regarding the
plausible structures of these thermally labile intermediates.
Preliminary experiments highlighted a stark difference in the
radical cyclisation step between some precursors, and quan-
tum calculations indicate agreement. A comparison with the
classical Ueno–Stork reaction, which relies on the use of classi-
cal a-haloacetals, is also presented. Quantum calculations have
been conducted for both the classical Ueno–Stork cyclisation
and the radical cyclisation of aluminium acetals and transition
states were located at the BHandHLYP/6-311+ +G(d,p) level of
theory, allowing for a direct comparison between the two
systems. These results are presented herein.
Table 1. Screening of solvent for the radical cyclisation.
Entry
Substrate[a]
Solvent
T [8C][b]
Product
Yield [%][c]
1
2
3
4
1a
1a
1a
1a
toluene
CH2Cl2
n-hexane
THF
À73
À73
À73
À73
2
2
2
2
88Æ2
80Æ2
55
0[d]
[a] All reactions were carried out at À738C on a 2 mmol scale (and re-
peated 2–3 times), using 1.2 equiv of DIBAL-H (solution in the same sol-
vent as the solvent of reaction), 1.2–1.5 equiv of nBu3SnH and 0.3–
0.6 equiv of Et3B (1m solution in hexanes). [b] Temperature measured
inside the reaction flask. [c] Yields in isolated product, after purification
by chromatography on silica gel or Silica-KF. [d] Mainly over-reduction
was observed.
overall sequence was tested in toluene, methylene chloride,
and n-hexane. Preliminary experiments indicated that both tol-
uene and methylene chloride were suitable solvent for this
radical cyclisation (Table 1, entries 1 and 2). The yields in n-
hexane dropped to 55%, presumably due to the poor solubili-
ty of both the precursor and the aluminium acetal in this sol-
vent at low temperature (Table 1, entry 3). We therefore decid-
ed not to further investigate this solvent, although it might be
useful in combination with other non-protic solvents to reach
temperatures below À808C, which might be beneficial in
some cases for stability and diastereoselectivity. Polar solvents
such as THF led to the decomposition of the aluminium acetal
intermediate and, in this case, the alcohol resulting from over-
reduction was the major compound formed (Table 1, entry 4).
Different reducing agents were tested in order to achieve
the reduction of bromoesters into the corresponding tetrahe-
dral intermediates. Among the various reducing agents tested,
the best results were obtained with DIBAL-H, as the more
reactive Red-Alꢁ led only to the displacement of the halogen
atom, whereas LiBHEt3 gave non-reproducible results (Table 2,
entries 1, 4 and 5).
Results and Discussion
We have chosen a model substrate for our study, with minimal
variation of the structure in order to exclude the effect of side
chain, and focus only on the influence of the substitution at
the a-position on the ester moiety. The variations in the
structures were thus limited to the nature of the halide (or
chalcogenide) and the number of alkyl substituents at the
carbon atom at the a-position. The general structure of these
substrates is shown in Scheme 1. The reaction conditions were
identical for all reactions, unless otherwise stated. Although
other factors may interfere in the cyclisation process, the yields
of the isolated products are likely to give indirect information
on the stability of the aluminium acetal intermediates.
We first decided to screen possible solvents for this radical
reaction. Protic solvents were not tested, as the first step re-
quires the use of an aluminium hydride. The feasibility of the
In the case of LiBHEt3, the hydride was well transferred to
the carbonyl group, as indicated on thin layer chromatography
(TLC) by the presence of the allylic alcohol corresponding to
the reduction of bromoester 1a, but the addition of nBu3SnH
and air did not lead to a clean conversion into lactol 2. In this
Table 2. Screening of initiator for the radical cyclisation of 1a at À738C
in toluene.
Entry
Hydride[a]
Equiv
Solvent
Initiator
Yield [%][c]
1
2
3
4
5
DIBAL-H
DIBAL-H
DIBAL-H
Red-Alꢁ[d]
1.2
2
3
1.1
1.1
toluene
toluene
toluene
toluene
Et3B/air
Et3B/air
Et3B/air
Et3B/air
–/air
90
91
88
0
[e]
LiBHEt3
toluene/THF
nd[f]
[a] All reactions were carried out at À738C on a 2 mmol scale (and re-
peated 2–3 times), using 1.2–1.5 equiv of nBu3SnH and 0.6 equivalents of
Et3B (1m solution in hexanes) and 2”1 mL of air (unless otherwise
stated). [b] Temperature measured inside the reaction flask. [c] Yields in
isolated product, after purification by chromatography on silica gel or
Silica-KF. [d] 60 wt.% solution in toluene. [e] 1.0m solution in THF. [f] Non-
reproducible results.
Scheme 1. General scheme for the cyclisation of aluminium acetals.
Chem. Eur. J. 2016, 22, 4809 – 4824
www.chemeurj.org
4810
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
case, no Et3B was added since it should be released in the
medium during the reduction step. We suspect in this case
that the presence of THF present in the commercially available
solution of LiBHEt3 is responsible for the degradation of the
aluminium acetal. This is in agreement with previous observa-
tions (Table 1, entry 4). The use of larger quantities of DIBAL-H
gave very similar results to those obtained in the presence of
only 1.2 equivalents, thus indicating that an excess of reducing
agent does not significantly decrease the thermal stability of
the acetal aluminium derived from 1a, at least for reactions
carried out in toluene at À738C (Table 2, entries 1–3).
tion. Indeed, in the reaction with a-chloroester 1d (Scheme 1)
and a-phenylthio derivative 1e, no traces of the corresponding
g-lactol 2 were observed. Because those derivatives are unlike-
ly to undergo halogen (or chalcogen) abstraction by the tin-
centred radical at À738C, the absence of reaction does not
give any information concerning the thermal stability of the
corresponding aluminium acetals. Nevertheless it allowed us to
exclude the possibility that Lewis acids present in the medium
(aluminium atoms) could facilitate the homolytic substitution
in a way that would allow for the use of these less reactive
precursors in this cyclisation process. We focused then our at-
tention on substrates 1a–c, and the related compounds 1 f–k,
which differ only by the presence of one or two substituents,
at the a-position. The results obtained at À738C in different
solvents are collected in Table 4.
We decided then to check whether initiators other than
Et3B/air could be used in this reaction. Et2Zn and iBu3Al gave
very similar results to Et3B in terms of yields (Table 3, entries 1–
3), but the reactions seemed to be slightly faster. Interestingly,
in the absence of any added initiator, the formation of g-lactol
2 was still observed and, under otherwise similar reaction con-
ditions, the product was isolated in about 50% (Table 3,
entry 4). It is likely that homolytic cleavage of the carbon–alu-
minium bond in either the aluminium acetal intermediate or
the remaining DIBAL-H (or both) occurred in the presence of
air, and led to carbon-centred radicals that could initiate the
radical chain.
Table 4. Influence of the solvent and substitution on the reduction-
cyclisation process.
Entry
Substrate[a]
Solvent
T [8C][b]
Product
Yield [%][c]
1
2
3
4
5
6
7
8
9
1a
1b
1c
1 f
1g
1h
1 f
1g
1h
1i
toluene
toluene
toluene
toluene
toluene
toluene
CH2Cl2
CH2Cl2
CH2Cl2
toluene
toluene
toluene
À73
À73
À73
À73
À73
À73
À73
À73
À73
À73
À73
À73
2
2
2
3
3
3
3
3
3
4
4
4
88Æ2
96Æ1
87Æ1
83Æ1
80Æ3
78Æ1
88
Table 3. Screening of initiator for the radical cyclisation of 1a at À738C
in toluene.
83
31Æ2
51Æ2
19Æ1
14Æ1
Entry
Initiator[a]
Equiv[b]
Product
Yield [%][c]
10
11
12
1
2
3
4
5
6
Et3B/air[d]
Et2Zn/air[e]
iBu3Al/air[f]
–/air
2”0.3
2”0.3
2”0.3
–
–
–
2a
2a
2a
2a
2a
2a
90
87
91
1j
1k
[a] All reactions were carried out at À738C on a 2 mmol scale (and re-
peated 2–3 times), using 1.2 equiv of DIBAL-H (solution in the same sol-
vent as the solvent of reaction), 1.2–1.5 equiv of nBu3SnH and 0.3–
0.6 equiv of Et3B (1m solution in hexanes). [b] Temperature measured
inside the reaction flask. [c] Yields in isolated product, after purification
by chromatography on silica gel or Silica-KF.
48Æ6
–/air[g]
36
–/air[h]
30
[a] All reaction were carried out at À738C on a 2 mmol scale (and repeat-
ed 2–3 times), using 1.2 equiv of DIBAL-H (solution in the same solvent
as the solvent of reaction), 1.2–1.5 equiv of nBu3SnH and 0.6 equivalent
of Initiator and 2”1 mL of air (unless otherwise stated). [b] Temperature
measured inside the reaction flask. [c] Yields in isolated product, after pu-
rification by chromatography on silica gel or Silica-KF. [d] 1m solution in
hexanes. [e] 15 wt.% in toluene. [f] 15 wt.% in toluene. [g] 2”20 mL of
air. [h] Slow addition of 3”2 mL of air (over 11 h).
When no alkyl substituent is present at the a-position (cf.
Scheme 1), very similar yields were obtained in toluene regard-
less of the nature of the halide or chalcogenide (Table 4, en-
tries 1–3). This indicates that the corresponding aluminium
acetal intermediates do not present a marked difference in
their thermal stability on the reaction time scale (5–8 h) at
À738C. The same general trend was observed with com-
pounds 1 f–h, which possess a single alkyl substituent at the
a-position (Table 4, entries 4–6), although lactol 3 was ob-
tained in slightly lower yields (ca. 78–83%) than for the corre-
sponding lactol 2 (87–96%). Unexpectedly, slightly higher
yields in lactol 3 were obtained from 1 f (X=Br) and 1g (X=I)
(88% and 83%, respectively) for the reactions carried out in
CH2Cl2 (Table 4, entries 7 and 8, vs. entries 4 and 5), while a sig-
nificant drop in the yield was observed in CH2Cl2 for selenyl
derivative 1h (X=SePh) (Table 4, entry 9).
Modification of the amount and the rate of air introduced
did not allow us to improve the yields (Table 3, entries 5 and
6), and accordingly the use of an initiator proved to be neces-
sary in order to ensure high yields. The fact that the yields did
not exceed 50% in the absence of initiator was somewhat puz-
zling and we suspect that the presence of an initiator, such as
Et3B, Et2Zn or iBu3Al, protects the air-sensitive aluminium ace-
tals to some extent by consuming itself O2 present in the
medium, thus preventing a change in the structure of the alu-
minium acetals that seems to be detrimental to the thermal
stability of the species.[8]
Screening the nature of the halide (or chalcogenide) re-
vealed that only bromo, iodo and seleno derivatives (X=Br, I,
and SePh) could be used as precursors for the radical cyclisa-
Not surprisingly, compounds 1i–k with two alkyl substitu-
ents at the a-position were more difficult to reduce into the
corresponding aluminium acetals and the latter seem to be
Chem. Eur. J. 2016, 22, 4809 – 4824
www.chemeurj.org
4811
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
much less stable than the less hindered aluminium acetals ob-
tained from 1a–h. Here again, a stark contrast was observed
between the precursors, depending on the nature of the
halide or chalcogenide, as illustrated by the yields obtained at
À738C from 1i (X=Br), 1j (X=I) and 1k (X=SePh) (Table 4,
entries 10–12). Only bromoester 1i led to the corresponding g-
lactol 4 in an acceptable 50% yield at À738C, while the iodo
and phenylselenyl analogues 1j and 1k gave lactol 4 in very
low yields (below 20%).
to lower yields from 1 f–h, and only traces of lactol 3 (if any)
were obtained at À208C (Table 5, entries 10–12).
For the more hindered precursors 1i and 1j with two alkyl
substituents at the a-position, only the reaction carried out
from 1i (X=Br) led to lactol 4 (Table 5, entries 13 and 14).
From this study it appeared that both the substituents at
the a-position and the nature of the halide (or chalcogenide)
play a significant role on the stabilisation of the thermally
labile aluminium acetal intermediates. Moreover, although
probably innocent in the radical cyclisation itself, the solvent
also proved to play a role in this reaction. At this stage, all
these effects were difficult to rationalise and we decided to
have a look at the structure of the aluminium acetal inter-
mediates.
As expected, the general trend observed when varying the
temperature was that an increase of the temperature resulted
in lower yields (Table 5, entries 1–9). However, some differences
were observed depending on the nature of the precursor. For
instance, with compounds 1a–c, which have no alkyl substitu-
ents at the a-position, the increase of the temperature from
À738C to À408C, significantly decreases the reaction yield
with 1a and 1c (X=Br and X=SePh, respectively), but much
less with 1b (X=I). Indeed, the reaction carried out at À408C
led to g-lactol 2 in about 65% yield from 1a and 1c, while
a high yield of about 85% was still obtained from 1b (Table 5,
entries 1–3 vs. Table 4, entries 1–3). Surprisingly, the yields in g-
lactol 2 obtained at À208C from 1a were much higher than
those obtained from 1c, and more reproducible than with 1b
(Table 5, entries 4–6).
Conformational analysis and structure of the aluminium
acetals
DFT geometry optimisations at the PCM(toluene)-B3LYP/6-31+
G(d,p) level were carried out on slightly simplified models of
the experimentally studied aluminium acetals in order to es-
tablish conformational preferences of the reagents that could
help rationalising their experimental reactivity. Bromide, iodide
and phenylselenide derivatives were investigated, with one,
two, or no methyl substituents at the a-position.
As a result of its Lewis acid property, we made the assump-
tion that the aluminium atom in the aluminium acetals is likely
to be involved in Lewis acid/Lewis base interactions. Two
Lewis bases are present in the reaction medium, namely the
halide (or chalcogenide) and the second oxygen atom. Both in-
tramolecular and intermolecular interactions can be consid-
ered. Since we suspected steric hindrance to play a significant
role in the mode of complexation, we chose to keep, in the
theoretical model, two bulky isobutyl groups on the alumini-
um atom, as well as a secondary alcohol unit. The chiral allylic
alcohol moiety used in the experimental study was substituted
with an isopropyl group in order to minimise the number of
conformers and diastereoisomers to be modelled (Figure 1).
The analysis revealed two series of conformers in the mono-
Table 5. Influence of the temperature on the stability of the aluminium
acetals.
Entry
Substrate[a]
Solvent
T [8C][b]
Product
Yield [%][c]
1
2
3
4
5
6
7
8
9
10
11
12
13
14
1a
1b
1c
1a
1b
1c
1 f
1g
1h
1 f
1g
1h
1i
toluene
toluene
toluene
toluene
toluene
toluene
toluene
toluene
toluene
toluene
toluene
toluene
toluene
toluene
À40
À40
À40
À20
À20
À20
À40
À40
À40
À20
À20
À20
À40
À40
2
2
2
2
2
2
3
3
3
3
3
3
4
4
65Æ1
85Æ4
66Æ2
30Æ2
30Æ25[d]
0
43Æ1
41Æ5
13Æ1
12Æ1
0
0
42Æ2
1j
0
[a] All reaction were carried out at À738C on a 2 mmol scale (and repeat-
ed 2–3 times), using 1.2 equiv of DIBAL-H (solution in the same solvent
as the solvent of reaction), 1.2–1.5 equiv of nBu3SnH and 0.3–0.6 equiv of
Et3B (1m solution in hexanes). [b] Temperature measured inside the reac-
tion flask. [c] Yields in isolated product, after purification by chromatogra-
phy on silica gel or Silica-KF. [d] Non-reproducible results. The reaction
has been repeated 6 times and yields varied between 5–55%.
Contrary to what was observed at À738C in toluene with
precursors 1a–c and 1 f–h, which gave the cyclised com-
pounds in similar yields (87–96%, vs. 78–83%, respectively),
the reactions carried out at À408C indicated that the alu-
minium acetals prepared from 1 f–h showed a lower thermal
stability than those obtained from 1a–c (Table 5, entries 7–9
vs. entries 1–3). Increasing further the reaction temperature led
Figure 1. Possible conformations for the aluminium acetals at the
PCM(toluene)-B3LYP/6-31+G(d,p) level.
Chem. Eur. J. 2016, 22, 4809 – 4824
www.chemeurj.org
4812
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
meric structures, showing an intramolecular interaction be-
tween the aluminium atom and either the second oxygen
atom (Figure 1, structure A) or the halide/chalcogenide atom
(Figure 1, structure B).[9]
The distribution between the four- and five-membered che-
lates is dependent on the halide (chalcogenide) atom. If no
methyl substituent is located at the a-position, the bromide
and selenide derivatives strongly favour the Al···X interaction
(conformer B), the resultant stabilisation leading to almost
100% of such conformers (Table 6, entries 1 and 7). This prefer-
Figure 2. Example of dimeric structure of aluminium acetals (X=I;
R1 =R2 =H) at the PCM(toluene)-B3LYP/6-31+G(d,p) level.
Table 6. Distribution of the 4-membered vs. 5-membered chelates in the
aluminium acetal structures (PCM(toluene)-B3LYP/6-31+G(d,p) level at
195 K).
Table 7. Gibbs free energy of dimerisation DG8 of non-methylated,
monomethylated and dimethylated aluminium acetals, at the PCM(to-
luene)-B3LYP/6-31+G(d,p) level.
Entry
Halide X
R1
R2
DG8
B[b]
dAl-X[c]
[a]
[kcalmolÀ1
]
1
2
3
4
5
6
7
8
9
Br
Br
Br
I
I
I
PhSe
PhSe
PhSe
H
H
H
Me
H
H
Me
H
H
+1.94
+2.15
+3.23
À2.46
À1.03
+0.67
+5.47
+7.77
+7.22
99.7
2.74
2.77
2.70
3.24
3.09
3.04
2.67
2.67
2.66
Entry
Halide X
R1
R2
DG8
Me
Me
H
Me
Me
H
99.9
>99.9
0.2
13.3
92.7
>99.9
>99.9
>99.9
[kcalmolÀ1
]
1
2
3
4
5
6
7
8
9
Br
Br
Br
I
I
I
PhSe
PhSe
PhSe
H
H
H
Me
H
H
Me
H
H
À15.49
À8.99
+4.78
À6.33
À7.00
+8.00
À5.62
À2.65
+11.76
Me
Me
H
Me
Me
H
Me
Me
Me
[a] Gibbs energy difference between the most stable conformers A and B.
[b] Sum of the relative population calculated at 195 K for the various
conformers (see Experimental Part for details). [c] Length in Š.
Me
Me
Me
ence for the Al···X interaction also applies for the correspond-
ing monomethylated derivatives (Table 6, entries 2 and 8). In
contrast, the related iodine derivatives strongly favour the
Al···O interaction (conformer B) for both the non-methylated
and the monomethylated derivatives (Table 6, entries 4 and 5).
Finally, the Al···X interaction (X=Br, I, PhSe) was found to be
favoured for the more sterically hindered dimethylated deriva-
tives, regardless the nature of the X atom (Table 6, entries 3, 6
and 9). In these cases, the intramolecular AlÀX distances are
slightly shorter than those in the non-methylated and mono-
methylated derivatives.
oured on the grounds of thermodynamics, regardless of the
nature of the halide or chalcogenide. However, significant dif-
ferences in energy have been calculated depending on the
nature of the X group (X=Br, I, PhSe). For instance, the dimeri-
sation process appeared to be much more favourable for the
non-substituted a-bromo aluminium acetal, than for the corre-
sponding iodo- and phenylselenyl derivatives (Table 7, en-
tries 1, 4 and 7). Although less pronounced, this trend is still
followed for the monomethylated aluminium acetals (Table 7,
entries 2, 5 and 8). At this stage, despite favourable orientation
in the optimised structures, the participation of the halide to
form dimeric oxygen-bridged structures with a five-coordinate
aluminium centre[9a,10c] could be ruled out. Indeed, the distance
between the aluminium and the halogen atoms in the dimers
appeared to be significantly longer than those observed in the
monomers.
In the case of simpler and thermally stable aluminium
alkoxides, numerous dimeric structures have been reported
previously (157 dimeric structures in the 2015 version of the
Cambridge Structural Database).[10] As
a consequence, it
appeared necessary to consider such dimeric structures for the
thermally unstable a-haloaluminium acetals.[11] Energy minima
were located for the dimeric structures with a four-membered
Al2O2 ring, regardless the initial conformation (structure A or B)
of the monomeric species (Figure 2).[9c] This allowed us to cal-
culate the Gibbs free energies (DG8) of the dimerisation for dif-
ferent aluminium acetals.
By contrast, a,a-dimethylated aluminium acetals appeared
to be less stable in the dimeric than in the monomeric form
(Table 7, entries 3, 6 and 9)[12] and this is consistent with the
increase in steric hindrance on the surroundings of the alu-
minium atom. Other hypotheses (six- and eight-membered
chelates) seemed to be unstable, as none of the optimised ge-
ometries converged to these structures, but rather to the
previous four-membered chelates. Trimeric structures were not
considered for these species as they seem very unlikely due to
steric hindrance,[13] and the presence of dialkylaluminium deriv-
atives (and not trialkoxides species).[14]
The results presented in Table 7 clearly indicate that the di-
merisation processes from monomeric aluminium acetals are
exergonic reactions for the least sterically hindered aluminium
acetals, those presenting none, or only one methyl substituent
at the a-position. In these cases, dimeric structures are fav-
Chem. Eur. J. 2016, 22, 4809 – 4824
www.chemeurj.org
4813
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Determination of the structure by NMR spectroscopy
were prepared in one step from allyl alcohol and commercially
available 1-13C-bromoacetic acid (>99% 13C) and 1-13C-bromo-
propionic acid (>99% 13C), respectively (Scheme 2a). The sele-
noester 5c was prepared in two steps from 1-13C-bromopro-
pionic acid (Scheme 2b). Finally, gem-dimethyl bromoester 5d
was obtained in two steps from 1-13C-2-methylpropionic acid
(>99% 13C) (Scheme 2c). Three 13C-labelled internal standards
were synthesised in 2–3 steps from commercially available 2,6-
dibromoanisole (Scheme 2d).
The DFT methods gave us good indication that different types
of structures could be expected depending on the nature of
the substituents of the aluminium acetals. Of course, the ther-
modynamic criterion is insufficient to guarantee the presence
of dimers in solution under our reaction conditions. Indeed,
considering the highly crowded environment at the aluminium
atom, the dimerisation process, although favoured thermody-
namically for some substrates, might be kinetically beyond
reach at low temperature. Consequently, we considered the
possibility of accessing the structure of the aluminium acetals
using diffusion-ordered NMR spectroscopy (DOSY) experiments
at low temperature. Diffusion-ordered spectroscopy of various
1
6
nuclei (e.g., H, 13C, Li, 31P) proved to be extremely useful for
the determination of the aggregation state of lithiated spe-
cies.[15] We hoped that 13C-INEPT DOSY with internal referen-
ces[15c,16] would allow us to determine the aggregation state of
the different species in solution. The first step of this study
was to measure 1H and 13C spectra of the aluminium acetals re-
sulting from the reduction of a-haloesters with DIBAL-H. For
the sake of simplicity in the analysis of the 13C spectra, 13C-
labelled a-haloesters (enriched at the carbonyl group) were
prepared.[17] The 13C NMR spectra were measured with delays
allowing for quasi-quantitative experiments.
The synthesis of 13C-labelled compounds that could be used
in this study as an internal reference system were undertaken,
keeping in mind that these references should be easy to
access and should not interact with the different aluminium
species in solution (aluminium acetals, DIBAL-H). At the same
time, these 13C-labelled compounds should give characteristic
signals in a region of the NMR spectrum free from other sig-
nals. We selected anisole derivatives, the Lewis base properties
of which were expected to be low enough not to interact with
the aluminium acetals (the latter proved to be unstable in THF
solution, vide supra). The reduction of various a-haloesters was
carried out at À738C in [D8]toluene, using a commercially
available solution of DIBAL-H in toluene, then the solution was
rapidly transferred into the NMR sample, pre-cooled at À808C
(see Supporting Information for details). The 1D (1H, 13C, Jmod)
and 2D (13C-INEPT DOSY) spectra were measured at À708C,
the viscosity of the solution being too high at lower tempera-
tures. The samples were measured in the absence and in the
presence of internal references, to ensure that the anisole de-
rivatives (see below) did not affect the stability of the alumini-
um acetal species. Moreover we verified that these anisole de-
rivatives had no interaction with DIBAL-H, which would result
in dramatic errors in the molecular weight of the species.
We decided then to prepare 13C-labelled bromoesters, carry
out the reduction with DIBAL-H and try to get an estimate of
their molecular weight thanks to the diffusion ordered spec-
troscopy (13C-INEPT DOSY). For the sake of simplification, we
decided to introduce the enriched 13C at the carbonyl group in
the bromoester. This would make the synthesis of the precur-
sor easier and we expected the signal for the 13C correspond-
ing to the acetal centre to appear in a region of the NMR spec-
trum almost free from other signals. The precursors 5a and 5b
Scheme 2. Preparation of 13C-labelled precursors 5a–d and 13C-labelled
internal references 6–8.
The reduction of 5a was initially carried out with 1.2 equiva-
lents of DIBAL-H (commercially available solution, 1.2m in tolu-
ene) but for the rest of the study 2 equivalents of reducing
agent were used to ensure complete reduction and to avoid
any changes in the relative integration of the signals during
the DOSY experiments that could be caused by the conversion
of unreactive bromoester, thus introducing significant error in
the determination of the molecular weight. Although changes
in the structure have been observed for organoaluminium de-
rivatives obtained from aminoacids depending on the reactant
molar ratio,[9d,18] no changes in the 13C NMR spectra were ob-
served for the reduction of 5a with 1.2 or 2 equivalents of
DIBAL-H, suggesting the existence of the same species at both
molar ratio. For practical convenience, the reduction of bro-
moester 5a was then carried out in [D8]-toluene at À738C with
2 equivalents of DIBAL-H (1.2m in toluene) at a concentration
slightly higher than those usually used in our experiments for
the sequence reduction/radical cyclisation.[19] The solution was
Chem. Eur. J. 2016, 22, 4809 – 4824
www.chemeurj.org
4814
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
then quickly transferred into a pre-cooled NMR tube and the
13C spectrum of the reaction mixture was then measured (see
Supporting Information for details). We were pleased to obtain
a very clean 13C NMR spectrum, which clearly indicated com-
plete consumption of the starting material and formation of
two new species, as indicated by the presence of two intense
signals at 97.21 and 97.05 ppm (Figure 3), a region characteris-
tic of the acetal carbon atom. Theoretical calculations of the
13C NMR spectra estimate the chemical shift for the C(OR)OAl
carbon atom to be in the region 100 ppm, while the corre-
sponding aluminium alkoxides CH2ÀOAl should appear near
70 ppm (see Supporting Information for details). The relative
integration of the two signals gave an approximate 2:1 ratio
for the two species. The presence of two signals is consistent
with the presence of two diastereomeric dimeric species
formed by the dimerisation of a chiral, racemic monomeric alu-
minium acetal in a moderately selective dimerisation process
(Figure 4).[20]
The results of the INEPT 13C-DOSY experiment allowed us to
rule out the existence of monomeric species in solution for the
non-substituted aluminium acetal derived from 5a. A good cal-
ibration line (y=7.2545xÀ67.36, r2 =0.99) was established by
plotting the logarithm of the molecular weight (MW) of the
13C-labelled references (6, 7, 8, and [D8]-toluene), versus the
logarithm of their diffusion coefficient values measured from
the 13C INEPT-DOSY experiment (Figure 3). Diffusion coefficients
values measured at 97.21 and 97.05 ppm led then, by
extrapolation, to molecular weight consistent with dimeric
species (see Supporting Information for details).
Figure 4. 13C NMR spectrum of aluminium acetal derived from 13C-labelled
bromoester 5b, measured at 203 K.
The reduction of monomethyl a-bromoester 5b with DIBAL-
H at À708C gave a more complex system. The aluminium
acetal species could also be observed in the 13C NMR spec-
trum, with characteristic signals in the region of 90–100 ppm.
Here again, 13C NMR spectrum with J-modulation of spin echo
Figure 3. 13C INEPT-DOSY of aluminium acetal derived from 5a.
Chem. Eur. J. 2016, 22, 4809 – 4824
www.chemeurj.org
4815
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
(Jmod) indicated that the signals present in this region corre-
sponded to the 13C-labelled CH (and not CH2). The presence of
(at least) six signals in the region 98–102 ppm tends to support
the hypothesis of the diastereomeric dimeric structures for this
monomethyl aluminium acetal, in agreement with the first-
principle calculations. By comparison with the 13C NMR spec-
trum obtained for the reduction of 5a, for which a certain
level of diastereoselectivity was observed for the dimerisation
process (vide infra, relative integration ca. 2:1), the relative in-
tegration of the different signals (see peaks 2 and 7 in
Figure 4) tends to indicate that the reduction of a-substituted
bromoester 5b proceeded with a high level of stereocontrol.
Unfortunately, the complexity and the overlapping of 13C sig-
nals in the range 98–102 ppm prevented an accurate estima-
tion of the formula weights for each peak. However, the analy-
sis of the 13C INEPT-DOSY considering the global integration of
the six (or seven) signals between 98 and 102 ppm led to a
molecular weight of 460, which is intermediate between the
theoretical molecular weights of the monomeric and dimeric
species (see Supporting Information for details).
ent if aluminium acetal C decomposed, but as it is not 13C-la-
belled, the corresponding signal for the ÀCH2O(Al) would be
weak. More likely, heterodimer I (MW=478), might contribute
to the six (or seven) signals observed on the 13C NMR spectrum
of aluminium acetal derived from 5b. The contribution of hete-
rodimer I to the mixture of species, if possible, should depend
on the initial stoichiometry between the reactants.
In contrast to the aluminium acetals derived from bromoest-
ers 5a and 5b, we were unsuccessful in the preparation and
observation of the aluminium acetal derived from selenoester
5c (X=SePh). This failure is in agreement with the results ob-
tained in the radical cyclisation of 1h, for which a dramatic
drop in the yield was observed between À738C and À408C
(Table 4, entry 6 vs. Table 5, entry 9). It is likely that the rapid
warm up during the transfer from the reaction flask led to the
decomposition of the aluminium acetal formed in the reaction
flask at À708C. The absence of experimental data, together
with the moderate exothermicity of the dimerisation process
calculated at the PCM(toluene)-B3LYP/6-31+G(d,p) level
(Table 7, entry 8) do not allow us to make conclusions concern-
ing the aggregation state of aluminium acetals derived from
a-substituted selenoesters. In other words, for these species,
the existence of monomeric forms cannot be excluded. This, as
we will see later, can have dramatic consequences on the
diastereoselectivity of the radical cyclisation process.
Heterodimeric structures G and H (Scheme 3) could be ob-
tained by reaction between aluminium acetal C and aluminium
alkoxide F and D, respectively. Both D and F result from the
decomposition of the aluminium intermediate C, with D and E
being the primary products and F being obtained by over-re-
duction of E with the excess of DIBAL-H. Although possible,
the heterodimeric structures G and H have been ruled out in
the case of monomethyl a-bromoester 5b, because no intense
signals characteristic for a ÀCH2O(Al) species were observed in
the region 70 ppm (H cannot be formed without other species
(such as G) containing at least one 13C-labelled CH2O(Al)
moiety being present). The homodimer (not shown) formed
from the diisobutylaluminium alkoxide D would also be pres-
Finally, we investigated the case of the challenging substrate
5d (X=Br; R1 =R2 =Me). Because of its sterically crowded envi-
ronment and the moderate yields obtained from 1i in the radi-
cal cyclisation even at À738C (see Table 4, entry 10), it was ex-
pected that the observation of aluminium acetal species would
be difficult. This was indeed the case and we obtained non-re-
producible results. However, in one of our attempts at prepar-
ing the expected aluminium acetal species, we were able to
observe a mixture between unreacted a-bromoester 5d, an
aluminium acetal species, and signals assigned to the alumini-
um alkoxide resulting from the over-reduction of 5d. The ob-
servation of one main CH signal in the Jmod spectrum tends
to support the presence of a monomeric structure for the alu-
minium acetal derived from a,a-disusbtituted bromoesters, in
agreement with our calculations. However, at this stage a heter-
odimeric structure similar to dimer I (see Scheme 3) cannot be
definitely ruled out.
With these experimental and theoretical data in hand, we
started to investigate the energy profiles for the radical
cyclisation and located transition states for the different cases.
Transition states for the radical cyclisation
Aluminium acetals are potentially reactive species by one of
their oxygen atoms, the aluminium alkoxide moiety being nu-
cleophilic enough to react with an acylating agent and to form
mixed acetals.[21] Therefore, the reaction between non-cyclised
or cyclised aluminium acetals and the tributyltin halide gener-
ated during the course of the reaction must be envisaged. In
order to ensure that the species involved in the radical cyclisa-
tion step were truly aluminium acetals (and not tributyltin ace-
tals), we carried out a 119Sn NMR analysis of the crude reaction
Scheme 3. Possible heterodimeric structures.
Chem. Eur. J. 2016, 22, 4809 – 4824
www.chemeurj.org
4816
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
mixture. This study clearly indicates the presence of Bu3SnBr
and Bu6Sn2 as the two major components of the reaction mix-
ture, with very little amounts (if any) of tin alkoxide derivatives
(see Supporting Information for details). We can conclude that,
because no cyclic tin acetals were observed under these reac-
tion conditions, it is very unlikely that linear tin acetal were
formed to some extent as intermediates during the radical
cyclisation process. Accordingly, only aluminium acetals will be
considered as possible intermediates in the following
calculations.
between the aluminium atom and the halogen (or chalcogen)
atom. Although this mode of complexation results in a slightly
longer carbon–halogen (or carbon–chalcogen) bond, thus sug-
gesting a weakening of these bonds, the halogen (or chalco-
gen) abstraction might be disfavoured on structure B due to
the presence of the bulky isobutyl groups on the aluminium
atom. In this case, the halogen abstraction would probably de-
viate from the ideal attack angle, which was previously deter-
mined to be nearly collinear in the case of bromine and iodine
(Figure 6).[23] The deviation from this ideal situation should
result in an increase in the energy required to break this bond
(not calculated). Accordingly, the halogen atom abstraction
might be easier on either monomeric structures A, or on the
least stable dimeric structures. At this point, we reasoned that
the minor structure A is the reactive species in solution and, in
this case a monomeric radical involving a four-membered
chelate was considered.
With this information in hand, and having established the
structure for some of the aluminium acetals obtained during
the reduction with DIBAL-H, we then turned our attention to
the radical cyclisation itself. Two different cases have to be
considered; the cyclisation of monomeric species and the cycli-
sation of dimeric species. In the case of dimeric aluminium ace-
tals, the optimised geometries obtained at the PCM(toluene)-
B3LYP/6-31+G(d,p) level showed that the halogen (or seleni-
um) atom of the precursors for the radical cyclisation points
outside of the structure, which makes them relatively accessi-
ble for abstraction by the tin-centred radical in a SH2 homolytic
substitution process.[22] Since these aluminium acetals have
been proved to be more stable in a dimeric form and because
these structures allow for halogen abstraction (X=Br, I) by the
tin-centred radical to proceed with a nearly collinear approach,
it appears unlikely that the radical cyclisation would require
the aluminium acetal to rearrange into a monomeric structure
prior to halogen abstraction in these cases. It seems even
more unlikely that once the halogen atom abstraction has
been achieved, the dimeric carbon-centred radical has enough
time to collapse and give a monomeric structure before the 5-
exo-trig cyclisation takes place. Accordingly, dimeric structures
will be considered for the search of transition states for the
radical cyclisation of non-methylated and monomethylated de-
rivatives. The presence of the four isobutyl groups in the di-
meric aluminium acetals makes the environment at the four-
membered ring formed by the dimerisation process so crowd-
ed (see Figure 2) that the possibility of the two radical cyclisa-
tion that take place on the dimers to be correlated seems very
unlikely. Instead, two totally independent radical cyclisation
events are believed to take place in these structures.
Accordingly, and for the sake of calculation cost, the search for
transition states was carried out on the simplified systems
depicted in Figure 5.
Figure 6. Halogen (chalcogen) abstraction by the tin-centred radical on
monomeric aluminium acetals.
In contrast to the parent Ueno–Stork reaction, for which the
stereoselectivity can be easily determined from the ratio of
products, it was difficult to have access the diastereoselectivity
of the radical cyclisation involving aluminium acetals. Indeed,
in this case the aqueous work-up delivers a lactol, for which
the hemiacetal centre is highly epimerisable. One alternative
could be to trap the cyclic aluminium acetal to form a mixed
acetal,[21] but both the reaction conditions for the trapping and
the low stability of the resulting products (see below) would
result in a lack of accuracy in the determination of the levels of
stereoselectivity. For these reasons, we decided to compare
the energy profiles of the radical cyclisation involving alumini-
um acetals with those involving a-bromoacetals (Ueno–Stork
reaction). The comparison was made for both the 5-exo-trig
and 5-exo-dig processes. Although calculations have already
been reported by some of us[24] for the Ueno–Stork cyclisation
onto carbon–carbon double bonds, for the sake of comparison
these systems have been (re)computed at the BHandHLYP/6-
311+ +G(d,p) level of theory to be consistent with the
aluminium acetal models.
The case of monomeric aluminium acetals is more complex.
Indeed, for the more hindered gem-dimethyl-substituted alu-
minium acetals, the favoured structure B implies complexation
5-exo-dig Cyclisations for Ueno–Stork acetals and dimeric
aluminium acetals
The search for transition states was systematically performed
directly at the BHandHLYP level of theory with the basis set
6-311+ +G(d,p). For all optimised structures, frequency
Figure 5. Structure used in the calculations to locate transition states for the
radical cyclisation.
Chem. Eur. J. 2016, 22, 4809 – 4824
www.chemeurj.org
4817
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
calculations were performed to ascertain their nature (minima
or transition states). A rapid conformational analysis was
performed for both the transition states and minima in order
to locate the most stable structures for each series.
(+7.5 kcalmolÀ1). The much lower difference in energy be-
tween TSAAA and TSAAB compared to the corresponding
Ueno–Stork cyclisation (0.3 kcalmolÀ1 vs. 2.9 kcalmolÀ1 for the
Ueno–Stork cyclisation) indicates that even at À788C, both
transition states TSAAA and TSAAB do contribute to the forma-
tion of the cyclised compounds, a significant qualitative differ-
ence between the two reactions.
Two pseudo-chair transition states were located for the 5-
exo-dig Ueno–Stork cyclisation, the most stable presenting
a pseudo-axial disposition for the OCH3 group. This is in good
agreement with the related 5-exo-trig Ueno–Stork cyclisations
previously investigated by Schiesser and Renaud.[24] The most
stable transition state (TSUEA) lies at +7.5 kcalmolÀ1 above the
radical precursor (Gibbs free energy calculated at 195 K in the
gas phase). The pseudo-equatorial transition state (TSUEB) ap-
pears about 2.9 kcalmolÀ1 above TSUEA (see Supporting Infor-
mation for details). Similarly, two pseudo-chair transition states
were located for the radical cyclisation of aluminium acetal
derivatives onto a triple bond (Figure 7). The transition state
TSAAA (+9.5 kcalmolÀ1) with a pseudo-axial CÀO(Al) bond is
5-exo-trig Cyclisations for Ueno–Stork acetals and dimeric
aluminium acetals
Cyclisation onto double bond was then studied. As previously,
we conducted the analysis of the simple OCH3 acetal used in
the Ueno–Stork reaction. Four transition states were located at
the BHandHLYP/6-311+ +G(d,p) level (see Supporting Infor-
mation for details), namely pseudo-chair axial-cis TSUEC, twist
axial-trans TSUED, pseudo-chair equatorial-cis TSUEE, and equa-
torial-trans TSUEF. Here again, the results are consistent with
published data,[24] and in agreement with the levels of diaste-
reoselectivity observed for these types of radical cyclisation.[25]
still more stable than transition state TSAAB (+9.8 kcalmolÀ1
)
with a pseudo-equatorial CÀO(Al) bond, but significantly more
destabilised than for the corresponding Ueno–Stork acetal
Figure 7. Energy profile for the radical 5-exo-dig cyclisation of aluminium acetals (BHandHLYP/6-311+ +G(d,p) level at 195 K). The values indicated on the
energy profile refer to the difference in electronic energy DE (top), electronic+zero-point energy correction D(E+ZPE) (middle), and Gibbs free energy DG8
(bottom, in red) with the uncyclised radical GSAATriple.
Chem. Eur. J. 2016, 22, 4809 – 4824
www.chemeurj.org
4818
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
The two more stable transition state TSUEC and TSUED present
a pseudo-axial disposition of the OCH3 group allowing for
a better overlapping between the oxygen lone-pair and the
s*CÀO bond (anomeric effect).[26] Transition states TSUEC and
TSUED were found 5.8 kcalmolÀ1 and 6.9 kcalmolÀ1, respective-
ly, above the reactants. The two pseudo-chair transition states
TSUEE (+8.5 kcalmolÀ1) and TSUEF (+10.7 kcalmolÀ1) lead to
cyclic acetals with the trans and cis relative configuration, re-
spectively, but they are much more energetic than the most
stable transition state TSUEC to contribute significantly to the
diastereoselectivity of this cyclisation process. The thermo-
chemical data calculated at 195 K indicated a 1.2 kcalmolÀ1
gap between TSUEC related to the cis isomer and TSUED related
to the trans one. The DFT energy profile was found consistent
with the high levels of stereoselectivity observed at low
temperature for 5-exo-trig Ueno–Stork cyclisations.
instability of the cyclic mixed acetals.[25a] Although it is difficult
to conclude at this stage, and despite the yields for these trap-
ping reactions, it seems that the cis/trans (or trans/cis) selectivi-
ty was significantly lower than that observed for the corre-
sponding Ueno–Stork reaction. It is worth noting that only the
radical cyclisation involving a dimeric structure for the alumini-
um acetal is consistent with the moderate level of stereo-
selectivity observed. Indeed, calculations performed with
monomeric species predicted that the trans configuration
should be strongly favoured in this case (vide infra).
The energy profile for the 5-exo-trig cyclisation of aluminium
acetals resembles that of the Ueno–Stork cyclisation, with two
main differences: firstly, only pseudo-chair transition states
were located and no twist conformation was found; secondly,
the second more stable transition state TSAAD (leading to the
trans isomer) is no longer having the alkoxide substituent in
the pseudo-axial position (Figure 8). In contrast, it is a pseudo-
chair transition state with the CÀO(Al) bond in the pseudo-
equatorial position, located only 0.3 kcalmolÀ1 above the most
stable transition state TSAAC, that leads to the cis isomer. The
next higher transition states TSAAE (giving the trans isomer)
and TSAAF (giving the cis isomer) were found 1.8 kcalmolÀ1
and 2.6 kcalmolÀ1, respectively, above TSAAC. These values are
consistent with the low levels of stereocontrol that have been
observed during the radical cyclisation of aluminium acetals.
As previously noted for the 5-exo-dig cyclisation of aluminium
acetal, the 5-exo-trig process for the aluminium acetal series
showed a significantly higher activation barrier (1.6 times
greater) than that of the corresponding Ueno–Stork cyclisation.
To further support the structure of the aluminium acetals,
the hypothesis of a monomeric structure was investigated for
the least hindered aluminium acetal. Three transition states
were located for the cyclisation of monomeric aluminium ace-
tals onto a C=C bond. TSAAG corresponding to the lowest acti-
vation barrier was found to lead to a cyclic aluminium acetal
with a trans configuration (Figure 9). The two other transition
states TSAAH and TSAAI both lead the cis isomer. The significant
energy difference of 1.3 kcalmolÀ1 between monomeric transi-
tion states TSAAG and TSAAH (compared to 0.3 kcalmolÀ1 only
in the hypothesis of dimeric structures, and 1.2 kcalmolÀ1 for
the highly stereoselective Ueno–Stork cyclisation) suggests
that a highly diastereoselective radical cyclisation could take
place. These figures appeared to be inconsistent with the low
stereoselectivity observed after trapping of the cyclic alumini-
um acetal with benzoyl fluoride. Altogether, the calculations
made to determine the favoured structure of the starting bro-
moacetals, the low-temperature 13C-INEPT DOSY experiments,
the apparent diastereoselectivity of the cyclisation process on
simple substrates and the DFT energy profiles converge to sup-
port the hypothesis of a dimeric structure for the aluminium
acetals obtained from non-substituted a-bromoesters.
We then turned our attention to the determination of the
energy profile for the related radical cyclisation of aluminium
acetals. As mentioned previously, one of the issues here was
the loss of the stereochemical information during the work-up
of the reaction, due to the rapid epimerisation of the anomeric
centre in the final lactol. Insights into stereoselectivity of the
cyclisation process were sought by attempting to trap the
cyclic aluminium acetal intermediate. The radical cyclisation
was carried out with bromoester 5a (not 13C-labelled) under
our standard reaction conditions, then the resulting cyclic alu-
minium acetal was trapped by benzoyl fluoride in the presence
of pyridine and 4-dimethylaminopyridine (DMPA).[21c,d] These re-
action conditions proved to give slightly better results in our
case than the use of acetic anhydride, and cyclic acetal 9 was
obtained in a modest 65% yield for the reaction carried out in
toluene as a 66:34 mixture of diastereomers (Scheme 4). Slight-
Scheme 4. Attempt at determining of the diastereoselectivity of the
cyclisation process for aluminium acetals.
ly lower yield and selectivity were obtained in CH2Cl2 (41%,
d.r.=60:40). These diastereomeric ratios are to be considered
with great care because on the one hand, modest yields were
obtained both in toluene and methylene chloride and, on the
other hand, the ratio were found to change during purification
by flash chromatography over silica gel, probably due to the
5-exo-dig and 5-exo-trig Cyclisations for monomeric,
gem-dimethyl aluminium acetals
The energy profiles for the radical cyclisation of monomeric,
gem-disubstituted aluminium acetals onto carbon–carbon
Chem. Eur. J. 2016, 22, 4809 – 4824
www.chemeurj.org
4819
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Figure 8. Energy profile for the radical 5-exo-trig cyclisation of aluminium acetals (BHandHLYP/6-311+ +G(d,p) level at 195 K). The values indicated on the
energy profile refer to the difference in electronic energy DE (top), electronic+zero-point energy correction D(E+ZPE) (middle), and Gibbs free energy DG8
(bottom, in red) with the uncyclised radical GSAADouble. Erel refers to the energy gap between the different transition states and most stable transition state
TSAAC.
triple and double bonds were calculated at the same level of
theory as for the dimeric structures. Both in the cases of 5-exo-
dig (Figure 10) and the 5-exo-trig cyclisation processes
(Figure 11) the lowest activation barriers for the monomeric,
gem-dimethyl aluminium acetals were found to be about 2.5–
3.0 kcalmolÀ1 higher than those for the cyclisation of the corre-
sponding non-substituted dimeric acetals. This tends to indi-
cate that the radical cyclisation of monomeric, gem-dimethyl
aluminium acetals is significantly more difficult than the corre-
sponding cyclisation of dimeric non-substituted aluminium
acetals. For the 5-exo-dig cyclisation process a single transition
state TSAAJ (+12.6 kcalmolÀ1) was located (Figure 10), while
two transition states TSAA
K ) and TSAAL
(+12.1 kcalmolÀ1
(+12.4 kcalmolÀ1) were found for the 5-exo-trig cyclisation pro-
cess (Figure 11). In the radical cyclisation onto a C=C bond,
a slight preference for the formation of the cyclic aluminium
Chem. Eur. J. 2016, 22, 4809 – 4824
www.chemeurj.org
4820
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Figure 9. Energy profile for the 5-exo-trig cyclisation of non-substituted aluminium acetals with the hypothesis of a monomeric structure (BHandHLYP/6-
311+ +G(d,p) level of theory at 195 K). The values indicated on the energy profile refer to the difference in electronic energy DE (top), electronic+zero point
energy correction D(E+ZPE) (middle), and Gibbs free energy DG8 (bottom, in red) with the uncyclised radical GSAAMonomer. Erel refer to the energy gap
between the different transition states and most stable transition state TSAAG.
acetal with the trans configuration was observed (0.2 kcal
molÀ1 between TSAAK and TSAAL).
ene or methylene chloride as solvent. Notably, some solvent
effects have been observed with mono-substituted aluminium
acetals depending on the nature of the halide. Among the ini-
tiators used in this study, diethylzinc, triisobutylaluminium and
triethylborane were found to give good results. In the absence
of any additional initiator, the radical reaction was also found
to take place but the moderate yields obtained in this case
suggest that the presence of oxygen might lead to a change
in the structure of the tetrahedral intermediates into less
stable aluminium acetal species. Low-temperature 13C-INEPT
DOSY NMR experiments allowed us to gain insights into the
structures of these aluminium species and, together with DFT
methods, to propose two different structures depending on
the nature of the aluminium acetal considered. Both the
Conclusion
The structure and the reactivity of aluminium acetals derived
from a-haloesters have been studied in detail. Experimental
data indicate that both the nature of the halide (or chalcoge-
nide) and the substitution at the a-position play a key role in
the stability of the tetrahedral intermediates and/or in the effi-
ciency of the radical cyclisation. Various initiators were tested,
together with various reducing agents and solvents. The best
conditions for the ionic reduction/radical cyclisation sequence
consist in the use of DIBAL-H as the hydride donor, with tolu-
Chem. Eur. J. 2016, 22, 4809 – 4824
www.chemeurj.org
4821
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Figure 10. Energy profile for the radical 5-exo-dig cyclisation of monomeric, gem-dimethylaluminium acetals (BHandHLYP/6-311+ +G(d,p) level at 195 K). The
values indicated on the energy profile refer to the difference in electronic energy DE (top), electronic+zero-point energy correction D(E+ZPE) (middle), and
Gibbs free energy DG8 (bottom, in red) with the uncyclised radical GSAA-monomer-Me2-Triple.
nature of the halide (or chalcogenide) and the substitution at
Acknowledgements
the a-position proved to play a role, with dimeric structures
with a four-membered chelate being preferred for the non-
The authors thank the “Minist›re de la Recherche et de l’En-
substituted aluminium acetals, while monomeric structures
proved to be more stable for gem-dimethyl-substituted alumi-
nium acetals. Energy profiles for the radical cyclisation (5-exo-
dig and 5-exo-trig processes) were calculated at the BHandH-
LYP/6-311 + +G(d,p) level of theory and compared to the
energy profiles for the Ueno–Stork cyclisation modelled at the
same level of theory. Strong differences were observed be-
tween the two types of cyclisations, with the Ueno–Stork cycli-
sation being predicted to be much more stereoselective and
presenting significantly lower activation barriers. Beside the
two extreme cases (non-substituted and gem-disubstituted alu-
minium acetals), the DFT calculations also indicated the pres-
ence of dimeric structures for the mono-substituted aluminium
acetals. This was confirmed by low-temperature NMR experi-
ments. The latter also suggested a high level of diastereoselec-
tivity for the reduction of a-bromoesters with DIBAL-H.
seignement SupØrieur” (MESR grant for A.B.) and the “Agence
Nationale de la Recherche” (grant ANR-10-JCJC-0712 “Radic[Al]”
for R.B., and ANR Blanc 13 BS07-0010-01 “UZFul-Chem” for J.-
C.R.) for funding. F.D and R.B. thank Drs. J. Maddaluno, A. Harri-
son-Marchand and M. Durandetti (COBRA, Rouen) for helpful
discussions and for offering access to the facilities for the prep-
aration of the aluminium acetals used in for the NMR study.
F.D. warmly acknowledges the Australian research council and
the University of Melbourne for financial support and access to
calculation facilities. D.J. acknowledges the European Research
Council (ERC) and the RØgion des Pays de la Loire for financial
support in the framework of a Starting Grant (Marches 278845)
and the LumoMat Project, respectively. This research used re-
sources of 1) the GENCI-CINES/IDRIS, 2) CCIPL (Centre de Calcul
Intensif des Pays de Loire), and 3) a local Troy cluster.
Keywords: aluminium acetals · computational methods · NMR
spectroscopy · radical cyclisation · reaction mechanisms
Chem. Eur. J. 2016, 22, 4809 – 4824
www.chemeurj.org
4822
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Figure 11. Energy profile for the radical 5-exo-trig cyclisation of monomeric, gem-dimethylaluminium acetals (BHandHLYP/6-311+ +G(d,p) level at 195 K). The
values indicated on the energy profile refer to the difference in electronic energy DE (top), electronic+zero-point energy correction D(E+ZPE) (middle), and
Gibbs free energy DG8 (bottom, in red) with the uncyclised radical GSAA-monomer-Me2-Double.
[1] For a review on the reactivity of aluminium acetals, see: A. Bousson-
ni›re, R. BØnØteau, J. Lebreton, F. DØn›s, Eur. J. Org. Chem. 2013, 7853–
7866.
[10] F. H. Allen, Acta Crystallogr. Sect. B 2002, 58, 380–388. For some exam-
ples of aluminium alkoxides with a planar Al2O2 ring, see: a) T. Mole,
Aust. J. Chem. 1966, 19, 373–379; b) B. Cetinkaya, P. B. Hitchcock, H. A.
Jasim, M. F. Lappert, H. D. Williams, Polyhedron 1990, 9, 239–243; c) A.
Pietrzykowski, T. Skrok, S. Pasynkiewicz, M. Brzoska-Mizgalski, J. Zachara,
R. Anulewicz-Ostrowska, K. Suwinska, L. B. Jerzykiewicz, Inorg. Chim.
Acta 2002, 334, 385–394; d) S. Szumacher, A. R. Kunicki, I. Madura, J. Za-
chara, J. Organomet. Chem. 2003, 682, 196–203; e) S. Yoon, S. H. Kim, J.
Heo, Y. Kim, Inorg. Chem. Commun. 2013, 29, 157–159.
[11] The formation of dimeric aluminium species obtained by addition of di-
ethylaluminium benzenethiolate onto aldehydes has been suggested
on the basis of 13C NMR analyses, see: G. Bastug, S. Dierick, F. Lebreux,
I. E. Marko, Org. Lett. 2012, 14, 1306–1309.
[2] A. Boussonni›re, F. DØn›s, J. Lebreton, Angew. Chem. Int. Ed. 2009, 48,
9549–9552; Angew. Chem. 2009, 121, 9713–9716.
[3] A. Boussonni›re, R. BØnØteau, N. Zimmermann, J. Lebreton, F. DØn›s,
Chem. Eur. J. 2011, 17, 5613–5627.
[4] a) Y. Ueno, K. Chino, M. Watanabe, O. Moriya, M. Okawara, J. Am. Chem.
Soc. 1982, 104, 5564–5566; b) Y. Ueno, K. Chino, M. Watanabe, O.
Moriya, M. Okawara, J. Chem. Soc. Perkin Trans. 1 1986, 1351–1356;
c) G. Stork, R. Mook, Jr., S. A. Biller, S. D. Rychnovsky, J. Am. Chem. Soc.
1983, 105, 3741–3742; d) G. Stork, P. M. Sher, J. Am. Chem. Soc. 1983,
105, 6765–6766; e) G. Stork, P. M. Sher, J. Am. Chem. Soc. 1986, 108,
303–304; f) G. Stork, P. M. Sher, H.-L. Chen, J. Am. Chem. Soc. 1986, 108,
6384–6385; g) G. Stork, Bull. Chem. Soc. Jpn. 1988, 61, 149–154.
[5] For a review, see: X. J. Salom-Roig, F. DØn›s, P. Renaud, Synthesis 2004,
1903–1928.
[12] For examples of monomeric, sterically hindered aluminium alkoxides,
see: M. D. Healy, D. A. Wierda, A. R. Barron, Organometallics 1988, 7,
2543–2548.
[13] Only very small aluminium alkoxides such as Me2AlOMe have been re-
ported to exist as trimeric structures, see: a) D. A. Drew, A. Haaland, J.
Weidlein, Z. Anorg. Allg. Chem. 1973, 398, 241–248; b) A. Haaland, O.
Stokkeland, J. Organomet. Chem. 1975, 94, 345–352.
[6] R. BØnØteau, J. Lebreton, F. DØn›s, Chem. Asian J. 2012, 7, 1516–1520.
[7] R. BØnØteau, C. F. Despiau, J.-C. Rouaud, A. Boussonni›re, V. Silvestre, J.
Lebreton, F. DØn›s, Chem. Eur. J. 2015, 21, 11378–11386.
´
[8] J. Lewinski, J. Zachara, P. Gos, E. Grabska, T. Kopec, I. Madura, W. Marci-
[14] J. Pauls, B. Neumüller, Z. Anorg. Allg. Chem. 2000, 626, 270–279.
[15] a) I. Keresztes, P. G. Williard, J. Am. Chem. Soc. 2000, 122, 10228–10229;
b) D. Li, C. Sun, P. G. Williard, J. Am. Chem. Soc. 2008, 130, 11726–11736;
c) D. Li, R. Hopson, W. Li, J. Liu, P. G. Williard, Org. Lett. 2008, 10, 909–
911; d) G. Kagan, W. Li, R. Hopson, P. G. Williard, Org. Lett. 2009, 11,
4818–4821; e) A. Botana, P. W. A. Howe, V. Car, G. A. Morris, M. Nilsson,
J. Magn. Reson. 2011, 211, 25–29.
niak, I. Prowotorow, Chem. Eur. J. 2000, 6, 3215–3227.
[9] For examples of dialkylaluminium five-membered chelates derived from
aminoacids, whose structure have been determined by X-ray diffraction
studies, see: a) R. Kumar, M. L. Sierra, J. P. Oliver, Organometallics 1994,
13, 4285–4293; b) T. Gelbrich, E. Hecht, K. H. Thiele, J. Sieler, J. Organo-
met. Chem. 2000, 595, 21–30; c) J. F. Allan, W. Clegg, M. R. J. Elsegood,
K. W. Henderson, A. E. McKeown, P. H. Moran, I. M. Rakov, J. Organomet.
[16] D. Li, I. Keresztes, R. Hopson, P. G. Williard, Acc. Chem. Res. 2009, 42,
270–280.
´
Chem. 2000, 602, 15–23; d) J. Lewinski, I. Justyniak, J. Zachara, E. Trat-
kiewicz, Organometallics 2003, 22, 4151–4157.
[17] C. Su, R. Hopson, P. G. Williard, J. Org. Chem. 2013, 78, 11733–11746.
Chem. Eur. J. 2016, 22, 4809 – 4824
www.chemeurj.org
4823
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
´
[18] J. Lewinski, I. Justyniak, P. Horeglad, E. Tratkiewicz, J. Zachara, Z. Ochal,
[23] C. H. Schiesser, B. A. Smart, T.-A. Tran, Tetrahedron 1995, 51, 3327–3338.
[24] O. Corminboeuf, P. Renaud, C. H. Schiesser, Chem. Eur. J. 2003, 9, 1578–
1584.
Organometallics 2004, 23, 4430–4437.
[19] Note: We have checked that no change in the NMR spectrum was ob-
served for the reduction of 5a at the concentration used in the radical
cyclisation.
[20] The (R,S) dimeric structure for Table 6, entry 1 was found to be about
1.2 kcalmolÀ1 less stable than the corresponding (R,R) dimer in DFT cal-
culations. Despite the fact that the potential energy surfaces have clear-
ly not been explored, it rationalise the presence of two different species
in solution in a ratio that deviates from 1:1.
[21] a) D. J. Kopecky, S. D. Rychnovsky, J. Org. Chem. 2000, 65, 191–198;
b) D. J. Kopecky, S. D. Rychnovsky, Org. Synth. 2003, 80, 177–183; c) N.
Sizemore, S. D. Rychnovsky, Org. Synth. 2012, 89, 143–158; d) Y. Zhang,
T. Rovis, Org. Lett. 2004, 6, 1877–1879.
[25] a) A. L. J. Beckwith, D. M. Page, J. Org. Chem. 1998, 63, 5144–5153; b) F.
Villar, P. Renaud, Tetrahedron Lett. 1998, 39, 8655–8658; c) F. Villar, O.
Equey, P. Renaud, Org. Lett. 2000, 2, 1061–1064; d) F. Villar, T. Kolly-
Kovac, O. Equey, P. Renaud, Chem. Eur. J. 2003, 9, 1566–1577.
[26] The preference for the pseudo-axial position is in contradiction with
the classical model for 5-exo-trig radical cyclisations, see: a) A. L. J. Beck-
with, C. H. Schiesser, Tetrahedron 1985, 41, 3925–3941; b) D. C. Spell-
meyer, K. N. Houk, J. Org. Chem. 1987, 52, 959–974.
Received: November 26, 2015
[22] a) J. E. Lyons, C. H. Schiesser, J. Chem. Soc. Perkin Trans. 2 1992, 1655–
1656; b) C. H. Schiesser, B. A. Smart, Tetrahedron 1995, 51, 6051–6060.
Published online on February 18, 2016
Chem. Eur. J. 2016, 22, 4809 – 4824
www.chemeurj.org
4824
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim