Combined experimental and theoretical mechanistic investigation of the barbier allylation in aqueous media

Article abstract of DOI:10.1021/jo800180d

(Chemical Equation Presented) The Barbier allylation of a series of para-substituted benzaldehydes with allylbromide in the presence of Zn, In, Sn, Sb, Bi, and Mg was investigated using competition experiments. In all cases, the slope of the Hammett plots

Full text of DOI:10.1021/jo800180d

Combined Experimental and Theoretical Mechanistic Investigation
of the Barbier Allylation in Aqueous Media
Johan Hygum Dam, Peter Fristrup,* and Robert Madsen*
Center for Sustainable and Green Chemistry, Building 201, Department of Chemistry,
Technical UniVersity of Denmark, DK-2800 Lyngby, Denmark
pf@kemi.dtu.dk; rm@kemi.dtu.dk
ReceiVed January 23, 2008
The Barbier allylation of a series of para-substituted benzaldehydes with allylbromide in the presence of
Zn, In, Sn, Sb, Bi, and Mg was investigated using competition experiments. In all cases, the slope of the
Hammett plots indicated a build-up of negative charge in the selectivity-determining step. For Zn, In,
Sn, Sb, and Bi, an inverse secondary kinetic isotope effect was found (kH/kD ) 0.75-0.95), which was
compatible with the formation of a discrete organometallic species prior to allylation via a closed six-
membered transition state. With Mg, a significantly larger build-up of negative charge along with a
small positive secondary kinetic isotope effect (kH/kD ) 1.06) indicated that the selectivity-determining
step was the generation of the radical anion of benzaldehyde. The reaction through a six-membered
transition state was modeled using density functional theory with the effect of solvent described by a
polarized continuum model. The calculated secondary deuterium isotope effects based on this mechanism
were found to be in good agreement with experimental values, thus adding further support to this
mechanistic scenario.
Introduction
however, the one-pot Barbier procedure has gained renewed
interest, especially for coupling of the more reactive allylha-
lides.6 Contrary to the Grignard reaction, the Barbier procedure
does not require strictly anhydrous solvents but can be performed
very effectively in aqueous media. In fact, the allylation of
aldehydes and ketones under the Barbier conditions usually
occurs faster and gives rise to a higher yield when water is used
as a (co)solvent.7 Furthermore, a number of metals are known
to participate in the coupling reaction in aqueous media
including zinc, indium, tin, manganese, antimony, bismuth, and
magnesium.8 Among these, the most widely used metals are
zinc and indium.7,9
The chemistry of organometallic nucleophiles began in the
middle of the 18th century with the synthesis of diethylzinc1
and the subsequent reaction with carbonyl groups.2 At the turn
of the century, Barbier described a one-pot reaction between a
ketone and an alkyl halide in the presence of magnesium metal.3
One year later, Grignard turned this reaction into a two-step
procedure by forming the organomagnesium reagent prior to
reaction with the carbonyl compound.4 This stepwise protocol
became the preferred method, and the Grignard reaction has
been one of the most important methods in organic chemistry
for forming a carbon-carbon bond.5 In the past two decades,
In spite of the many applications, the mechanism of the
Barbier reaction in aqueous media is not well-understood. In
the literature, it has been speculated as to whether the mechanism
follows a radical pathway or proceeds via a discrete allylmetal
* Corresponding authors. Fax: (+45) 45933968.
(1) Frankland, E. Ann. Chim. 1849, 71, 171-213.
(2) Frankland, E. Ann. Chim. 1863, 126, 109-113.
(3) Barbier, P. C. R. Acad. Sci. 1899, 128, 110-111.
(4) Grignard, V. C. R. Acad. Sci. 1900, 130, 1322-1324.
(5) Richey, H. G., Jr. Grignard ReagentssNew DeVelopments; Wiley:
New York, 2000.
(6) Blomberg, C. The Barbier Reaction and Related One-Step Processes;
Springer: Berlin, 1993.
(7) Li, C.-J. Chem. ReV. 2005, 105, 3095-3165.
10.1021/jo800180d CCC: $40.75 © 2008 American Chemical Society
Published on Web 03/20/2008
3228
J. Org. Chem. 2008, 73, 3228-3235

Barbier Allylation in Aqueous Media
species.8 It is known that the Barbier allylation in aprotic
solvents proceeds through a direct nucleophilic addition reaction
with zinc and tin,10 while allylmagnesium bromide reacts by a
single electron-transfer process.11 A recent computational study
has shown that allylzinc, -indium, -tin, -antimony, and -bismuth
species are more reactive toward the allylation of carbonyl
compounds than toward hydrolysis, while allylmagnesium
bromide shows a similar reactivity in the two reactions.12
Secondary deuterium kinetic isotope effect studies on the
allylation of benzaldehyde in aqueous media have indicated that
the reaction with zinc, indium, and tin proceeds through a rate-
determining polar addition with the allylmetal species, while
the allylation with magnesium and antimony follows a single
electron-transfer pathway on the metal surface in the magnesium
case and between the allylmetal and the aldehyde in the
antimony case.13 Further experimental evidence comes from
NMR studies that have shown that indium, tin, and antimony
react with allylbromide in water to form allylmetal intermedi-
ates.14 However, for the different metals, there are still a number
of unanswered questions with regard to the nature of the
allylmetal species and the involvement of single electron-transfer
processes. Therefore, we decided to perform a more systematic
study of the mechanism for the Barbier allylation in aqueous
media in the presence of various metals. Herein, we describe a
combined experimental and theoretical study with six metals
where we employed Hammett studies and isotopic labeling
experiments in combination with high-level computational
modeling.
SCHEME 1. General Scheme for Competitive Allylation of
Benzaldehydes
TABLE 1. Yields for Allylations in Scheme 1
yielda (%)
entry
X
Znb
Inc
Snc
Sbd
Bic
Mge
1
2
3
4
5
6
7
8
9
-CN
99 (94f) 90
94
93
98
-COOMe 98 (93f) 87 (99g) 82 (85g) 95
83 (88g) 25 (26g)
-CF3
-Cl
-H
-Me
-tBu
-OMe
-OBu
99 (91f) 98
96 (90f) 75 (80g) 95
96 (92f) 92
94
99
66 (90g) 94
60 (65g)
87 (92g) 84 (92g) 60 (62g)
69
88
49f
99 (95f) 75 (81g) 82 (93g) 50 (83g) 71 (84g) 14 (79g)
99 (92f) 56 (81g)
76 (82f)
63 (82g)
37 (72g)
18 (41g)
83 (90f)
a Yields determined by GC. b Performed in 1:1 THF/saturated aqueous
NH4Cl with acid-washed zinc. c Performed in 1:1 THF/H2O. d Performed
in 7:3 THF/0.5 M HCl with antimony metal prepared from reduction of
SbCl3 with NaBH4. e Performed with allyliodide in 4:1 DMF/0.1 M NH4Cl.
f Isolated yield. g Yield based on converted aldehyde.
with allylbromide (Scheme 1). Zinc, indium, tin, antimony,
bismuth, and magnesium were selected as the metals since they
display vastly different reactivities in aqueous media and may
react by different mechanisms.
First, reaction conditions had to be established for each of
the six metals. In most cases, THF was used as a cosolvent to
dissolve the para-substituted benzaldehydes. With zinc, the
highest yield of the homoallylic alcohols was obtained in a 1:1
mixture of THF and aqueous ammonium chloride (Table 1).8c
This mixture formed a two-phase system consisting of an
aqueous THF phase and a saturated ammonium chloride phase.
It was shown by GC that the aldehydes and the allylbromide
were present in the THF phase, while no compounds could be
detected in the ammonium chloride phase. The yields of the
homoallylic alcohols were lower when the reactions were
performed in a THF/water mixture, and this did not change by
using ultrasound. On the other hand, with indium, tin, and
bismuth, the THF/water mixture proved to give the best results.
All three metals reacted significantly more slowly than zinc,
and it was necessary to heat the reactions with tin and bismuth
to 60 °C. Antimony and magnesium resulted in a slower reaction
than all the other metals. With antimony, it was necessary to
use a solvent system of THF and 0.5 M hydrochloric acid.14c
Furthermore, antimony had to be prepared from antimony(III)
chloride and sodium borohydride16 to give a good conversion,
while commercially available antimony powder reacted more
sluggishly. It has been reported previously that antimony and
bismuth can be activated by potassium fluoride for the allylation
of aldehydes,8h,17 but in our hands, this procedure did not provide
a more reactive metal. Magnesium proved to be even more
troublesome, and we were unable to reproduce the original
conditions with allyliodide in either a THF/water mixture or in
a 0.1 M ammonium chloride solution.8g In our hands, these
reactions did not go to completion and mainly led to pinacol
Results and Discussion
Hammett Studies. We recently used a combination of
Hammett correlations and computational studies to investigate
the mechanism for the addition of functionalized organozinc
reagents to a series of para-substituted benzaldimines and found
that the rate-determining step most likely did not involve free
radicals.15 Prompted by this study, we chose to investigate the
Barbier reaction by allylating para-substituted benzaldehydes
(8) (a) Killinger, T. A.; Boughton, N. A.; Runge, T. A.; Wolinsky, J. J.
Organomet. Chem. 1977, 124, 131-134 (Zn). (b) Nokami, J.; Otera, J.;
Sudo, T.; Okawara, R. Organometallics 1983, 2, 191-193 (Sn). (c) Pe´trier,
C.; Luche, J.-L. J. Org. Chem. 1985, 50, 910-912 (Zn). (d) Li, C. J.; Chan,
T. H. Tetrahedron Lett. 1991, 32, 7017-7020 (In). (e) Wada, M.; Ohki,
H.; Akiba, K.-y. Bull. Chem. Soc. Jpn. 1990, 63, 1738-1747 (Bi). (f) Li,
C.-J.; Meng, Y.; Yi, X.-H.; Ma, J.; Chan, T.-H. J. Org. Chem. 1998, 63,
7498-7504 (Mn). (g) Zhang, W.-C.; Li, C.-J. J. Org. Chem. 1999, 64,
3230-3236 (Mg). (h) Li, L.-H.; Chan, T. H. Tetrahedron Lett. 2000, 41,
5009-5012 (Sb).
(9) We recently used both metals for allylation of carbohydrate aldehydes
in aqueous media, see: (a) Håkansson, A. E.; Palmelund, A.; Holm, H.;
Madsen, R. Chem.sEur. J. 2006, 12, 3243-3253. (b) Hansen, F. G.;
Bundgaard, E.; Madsen, R. J. Org. Chem. 2005, 70, 10139-10142. (c)
Palmelund, A.; Madsen, R. J. Org. Chem. 2005, 70, 8248-8251. (d)
Keinicke, L.; Madsen, R. Org. Biomol. Chem. 2005, 3, 4124-4128.
(10) (a) Yamataka, H.; Nishikawa, K.; Hanafusa, T. Bull. Chem. Soc.
Jpn. 1992, 65, 2145-2150. (b) Yamataka, H.; Nishikawa, K.; Hanafusa,
T. Bull. Chem. Soc. Jpn. 1994, 67, 242-245.
(11) Gajewski, J. J.; Bocian, W.; Harris, N. J.; Olson, L. P.; Gajewski,
J. P. J. Am. Chem. Soc. 1999, 121, 326-334.
(12) Chung, L. W.; Chan, T. H.; Wu, Y.-D. Organometallics 2005, 24,
1598-1607.
(13) (a) Gajewski, J. J.; Bocian, W.; Brichford, N. L.; Henderson, J. L.
J. Org. Chem. 2002, 67, 4236-4240. (b) Lucas, P.; Gajewski, J. J.; Chan,
T. H. Can. J. Anal. Sci. Spectrosc. 2003, 48, 1-6.
(14) (a) Chan, T. H.; Yang, Y. J. Am. Chem. Soc. 1999, 121, 3228-
3229. (b) Chan, T. H.; Yang, Y.; Li, C. J. J. Org. Chem. 1999, 64, 4452-
4455. (c) Li, L.-H.; Chan, T. H. Can. J. Chem. 2001, 79, 1536-1540.
(15) Keinicke, L.; Fristrup, P.; Norrby, P.-O.; Madsen, R. J. Am. Chem.
Soc. 2005, 127, 15756-15761.
(16) Ren, P.-D.; Jin, Q.-H.; Yao, Z.-P. Synth. Commun. 1997, 27, 2761-
2767.
(17) Smith, K.; Lock, S.; El-Hiti, G. A.; Wada, M.; Miyoshi, N. Org.
Biomol. Chem. 2004, 2, 935-938.
J. Org. Chem, Vol. 73, No. 8, 2008 3229

Dam et al.
FIGURE 2. Hammett plot for the zinc-mediated allylation of para-
substituted benzaldehydes. σ values were obtained from Hansch et al.21
FIGURE 1. Kinetic data for the competitive zinc allylation.
developed for reactions in which a cation was developed.19
Although this fit to the σ+ values usually implies the involve-
ment of a cation, the slope of the line is positive for both sets
of σ values, indicating that a small negative charge is being
developed at the benzylic position. The methoxy and butoxy
substituents cause the reaction to proceed more slowly than
would be expected from the standard σ values, which suggests
that for these substituents, there could be a change in the rate-
determining step.20 In any case, the poor correlation with the
σ• values indicates that a radical intermediate is not involved
in the allylation reaction.
The method applied for the zinc allylation also was used in
the competitive allylations with indium, tin, antimony, bismuth,
and magnesium (i.e., the disappearance of the starting material
was monitored during the progress of the reaction). As compared
to the zinc allylation, these reactions were all slower and
required several hours to reach full conversion. As a result, it
was not necessary to add the metals in portions, and the entire
portion of metal was therefore added at the beginning of each
experiment. For tin and bismuth, the reaction took 1-4 h to
initiate and then went to completion in 3-4 h.
The full Hammett plots are included in the Supporting
Information. For the four metals (indium, tin, antimony, and
bismuth), a good correlation was obtained using ordinary σ
values, although the methoxy substituent again deviated from
the straight line (Figure 3). The slopes of the lines are positive
with a slightly larger F value than previously observed with
zinc, indicating a higher build-up of negative charge in the
selectivity-determining step. The reactions with magnesium only
led to the desired allylation product for the para substituents
Me, Cl, CF3, and COOMe, and only in the former three cases
were there reasonable GC yields (see Table 1). The Hammett
plot for magnesium indicates the build-up of a significant
negative charge in the benzylic position and clearly distinguishes
this metal from the other five that are all very similar.
The similarity between the first five metals is somewhat
surprising when the differences in reaction conditions are taken
into account, but this similarity gives a strong indication that a
common mechanism is operating. The relatively small slope of
the lines indicates that only a partial negative charge is amassed
in the transition state, which is what would be expected from a
closed, six-membered transition state (TS). For magnesium, the
coupling and reduction of the aldehyde even with a variety of
different magnesium sources. Other solvent mixtures were
investigated, and it was found that 0.1 M ammonium chloride
in either ethanol or DMF gave full conversion and fewer
byproducts. The best results were obtained in a 4:1 mixture of
DMF and 0.1 M ammonium chloride.
The Hammett studies were performed as a series of competi-
tion experiments, where the reaction progress was determined
by measuring the disappearance of aldehyde using naphthalene
as the internal standard.18 If the reactions were of first order in
aldehyde, the concentrations of the aldehydes will follow, for
example, eq 1, where the subscript 0 stands for initial concentra-
tion, X is the para-substituted aldehyde, and H is benzaldehyde.
X0
H0
ln
) k ln
(1)
( )
( )
rel
X
H
The competition experiments were carried out with 1 equiv
of benzaldehyde, 1 equiv of one of the six different para-
substituted benzaldehydes, 1-2 equiv of allylbromide, and 1-2
equiv of the metal. Zinc was the first metal to be investigated,
and the experiments were performed by adding zinc in portions
of approximately 0.1 equiv because the reaction was otherwise
too fast to be adequately monitored. The reaction was also highly
exothermic and will lead to extensive formation of byproducts
if the temperature is not maintained at room temperature. It was
not possible to extend the Hammett plot to include para-
nitrobenzaldehyde or -dimethylaminobenzaldehyde since these
aldehydes with strongly electron-withdrawing or -donating
substituents were resistant to the allylation conditions. In all
cases, the correlation coefficient was better than 0.99, when ln-
(X0/X) was plotted against ln(H0/H) (Figure 1).
The obtained krel values were used to construct the Hammett
plot in Figure 2. When applying the regular σ values, a plot
with a small positive slope was obtained for the electron-
withdrawing substituent (right-hand side in Figure 2). However,
this line could not be extended to include the electron-donating
substituents, where a significantly steeper line was obtained.
Surprisingly, we found a linear correlation when using σ+
values, in spite of the fact that these values were originally
(18) For other reactions that have recently been investigated by Hammett
studies, see: (a) Fristrup, P.; Johansen, L. B.; Christensen, C. H. Catal.
Lett. 2008, 120, 184-190. (b) Sa´nchez, R. S.; Zhuravlev, F. A. J. Am. Chem.
Soc. 2007, 129, 5824-5825. (c) Yi, C. S.; Yun, S. Y. J. Am. Chem. Soc.
2005, 127, 17000-17006. (d) Lane, B. S.; Brown, M. A.; Sames, D. J.
Am. Chem. Soc. 2005, 127, 8050-8057. (e) Fristrup, P.; Le Quement, S.;
Tanner, D.; Norrby, P.-O. Organometallics 2004, 23, 6160-6165.
(19) Brown, H. C.; Okamoto, Y. J. Am. Chem. Soc. 1958, 80, 4979-
4987.
(20) For a discussion of non-linear Hammett plots, see: Schreck, J. O.
J. Chem. Educ. 1971, 48, 103-107.
3230 J. Org. Chem., Vol. 73, No. 8, 2008

Barbier Allylation in Aqueous Media
Using selected ion monitoring (SIM) for the molecular ions of
both benzaldehyde and d-benzaldehyde (m/z ) 106 and 107,
respectively), it was possible to quantify the disappearance of
these aldehydes relative to the molecular ion of the internal
standard (naphthalene, m/z )128). This experiment gave a
SDKIE value of 0.92, which corroborates the slightly higher
value (0.95) obtained in the indirect competition experiment.
Also, for antimony, there are significant differences in the
SDKIE value determined here (0.75) and the one reported earlier
by Lucas et al. (1.07).13b However, in their case, a 2 M potassium
fluoride solution was used, whereas we used a 0.5 M solution
of hydrochloric acid, which may significantly alter the pathway
of the reaction.
In contrast to the other five metals, the use of magnesium
resulted in a normal rather than an inverse SDKIE value. The
exact value was 1.04, which is very close to the value obtained
by Lucas et al. (1.06).13b The finding of a normal SDKIE can
be taken as an indication that a radical anion forms in the
selectivity-determining step, in sharp contrast to all the previ-
ously investigated metals where the SDKIE value below 1
clearly indicates that C-C bond formation is rate-limiting.
Computational Study. In conjunction with the experimental
study, we performed a computational study in an attempt to
increase our understanding of this fundamental reaction. Al-
though the metal-mediated allylation reaction is one of the basic
transformations in organic synthesis, it is only recently that it
has attracted interest from the theoretical community.24 In
general, allylmetal species can be considered to be highly
fluxional, and there are several different equilibria that may be
operating, including the well-known equilibrium between η1-
and η3-coordination modes. The most well-studied of these
equilibria is undoubtedly the Schlenk equilibrium for Grignard
reagents,25 but in addition to this, several different monomeric,
oligomeric, or even polymeric complexes have been suggested,
in particular, for indium.14a
In the present study, we limited ourselves to the mono-allyl
species, which can be considered to be the simplest possible
model for the previously mentioned complexes. As to whether
this mono-allyl species is indeed the molecular entity responsible
for the observed reaction is beyond the scope of the current
study, but it is possible that such coordinatively unsaturated
species can allow for a facile pre-coordination of the substrate.
The low F value determined in the Hammett studies indicates
that only a slight build-up of negative charge takes place in the
TS, which is in line with a closed Zimmermann-Traxler
transition state.26,27 Furthermore, as the Hammett correlations
were not linear with the σ• values, the involvement of radicals
in the selectivity-determining step can be ruled out. As a
consequence, we limited the computational investigation to
molecules that exist as closed-shell singlets. The oxidation state
for each of the metals, which is a necessary prerequisite to obtain
FIGURE 3. Hammett plots for the allylations with zinc, indium, tin,
antimony, bismuth, and magnesium.
significantly larger slope indicates that a significant negative
charge is present in the selectivity-determining step. This would
be expected, for instance, if the generation of a radical anion in
the benzylic position is the selectivity-determining step as was
suggested previously,8g which is then followed by a fast
allylation reaction. This mechanism is supported by two earlier
Hammett studies where substituted benzophenones were reacted
with t-butylmagnesium chloride21 and allylmagnesium bro-
mide,22 respectively. When t-butylmagnesium chloride was
employed, a Hammett F value of 3.0 was found, which is similar
to the result in the current study and in line with the formation
of a radical anion by electron transfer from the Grignard reagent.
On the contrary, when preformed allylmagnesium chloride was
used, the reaction rate was completely unaffected by the nature
of the para substituents, which gave a slope of zero in the
Hammett plot.23 The results obtained in the current study
indicate that under Barbier conditions, the radical anion of
benzaldehyde is generated first, as indicated by the high F value,
and that the following allylation is neither rate- nor selectivity-
determining.
Secondary Deuterium Kinetic Isotope Effect (SDKIE). To
add further mechanistic support to the closed, six-membered
TS, a study of the SDKIE was undertaken. An inverse SDKIE
(kH/kD < 1) value is a strong indication of a polar addition as
the rate-determining step,11,13 whereas a normal SDKIE (kH/kD
> 1) value would be compatible with the formation of a radical
species. By means of two separate competition experiments,
the relative rate of allylation for benzaldehyde and d-benzal-
dehyde could be determined relative to p-tolualdehyde. The
comparison of the rates was carried out indirectly since standard
GC is unable to differentiate between d-benzaldehyde and
benzaldehyde. With zinc, the SDKIE was determined to be as
large as 0.83, which suggests a relatively late TS. With tin,
antimony, and bismuth, almost equally large values were
determined (kH/kD ) 0.85 (Sn), 0.75 (Sb), and 0.85 (Bi),
respectively), whereas the use of indium resulted in only a minor
effect (kH/kD ) 0.95), suggesting that a much earlier TS is
(24) Ye, J.-L.; Huang, P.-Q.; Lu, X. J. Org. Chem. 2007, 72, 35-42.
(25) (a) Tammiku-Taul, J.; Burk, P.; Tuulmets, A. J. Phys. Chem. A
2004, 108, 133-139. (b) Ehlers, A. W.; van Klink, G. P. M.; van Eis, M.
J.; Bickelhaupt, F.; Nederkoorn, P. H. J.; Lammertsma, K. J. Mol. Model.
2000, 6, 186-194. (c) Axten, J.; Troy, J.; Jiang, P.; Trachtman, M.; Bock,
C. W. Struct. Chem. 1994, 5, 99-108.
operating for this metal. In a related study by Lucas et al.,13b
a
significantly lower SDKIE was found for indium (kH/kD ) 0.82),
and due to the obvious disagreement, we decided also to perform
a direct determination of the SDKIE for indium with GC-MS.
(26) Zimmerman, H. E.; Traxler, M. D. J. Am. Chem. Soc. 1957, 79,
1920-1923.
(27) Recently, a six-membered TS also was computed for the palladium-
catalyzed electrophilic allylation of aldehydes with allylstannanes, see: (a)
Piechaczyk, O.; Cantat, T.; Me´zailles, N.; Le Floch, P. J. Org. Chem. 2007,
72, 4228-4237. (b) Solin, N.; Kjellgren, J.; Szabo´, K. J. J. Am. Chem.
Soc. 2004, 126, 7026-7033.
(21) Hansch, C.; Leo, A.; Taft, R. W. Chem. ReV. 1991, 91, 165-195.
(22) Holm, T.; Crossland, I. Acta Chem. Scand. 1971, 25, 59-69.
(23) Yamataka, H.; Matsuyama, T.; Hanafusa, T. J. Am. Chem. Soc. 1989,
111, 4912-4918.
J. Org. Chem, Vol. 73, No. 8, 2008 3231

Dam et al.
TABLE 2. Characteristic Distances and Imaginary Eigenfrequencies for Six-Membered Transition States (Gas Phase)
entry
metal (oxidn state)
no. of Br-
C-C distance (Å)
M-C distance (Å)
imaginary eigenfrequency (cm-1)
1
2
3
4
5
6
Zn(II)
In(III)
Sn(II)
Sn(IV)
Sb(III)
Bi(III)
1
2
1
3
2
2
2.18
2.10
2.28
1.95
2.03
2.04
2.16
2.28
2.39
2.33
2.33
2.41
270i
309i
240i
366i
327i
333i
SCHEME 2. General Outline of Structures Included in the
Computational Study
meaningful results, was not known a priori; thus, we decided
to investigate several different possibilities, including zinc(II),
indium(I), indium(III), tin(II), tin(IV), antimony(III), and bis-
muth(III). As the experimental studies clearly indicate that the
reaction with magnesium proceeds through an entirely different
mechanism, we have not attempted to include this metal in the
modeling study.28
Two conformations (chair or boat) are possible for the cyclic
TS, but previous work on related systems revealed the chair
conformation as being energetically favored,15 which prompted
us to exclude the boat conformations in the current study. With
benzaldehyde as the substrate, the phenyl moiety can occupy
either pseudo-axial or pseudo-equatorial positions in the six-
membered cyclic transition state, and not surprisingly, we found
the pseudo-equatorial position to be favored. To obtain an
overall neutral model system, auxiliary bromide atoms were used
as ligands for the metal atom (Scheme 2).29
FIGURE 4. Energy profiles for allylation in the presence of zinc(II).
All energies are relative to the isolated reactants (Zn-allyl complex
and benzaldehyde).
With zinc as the metal, the formation of a pre-reactive
complex from benzaldehyde and an σ-allylzinc fragment was
exothermic by 68 kJ/mol (gas phase), but the inclusion of
entropic contributions (Gibbs free energy, 298 K) reduced this
to 16 kJ/mol due to the reduction in molecularity. The use of a
polarized continuum solvation model (THF) reduced the exo-
thermicity for the coordination to 18 kJ/mol (SCF energy),
whereas by including the gas-phase entropic contributions, the
coordination was endergonic by 34 kJ/mol. The pre-complex
can undergo C-C bond formation through a six-membered
cyclic TS with a C-C bond length of 2.19 Å and a Zn-C bond
length of 2.16 Å (Table 2, entry 1). The structure of the six-
membered ring remains essentially unchanged upon introduction
of implicit solvation (parameters for THF), with the only
noticeable change being the movement of the coordinating
bromine atom into a pseudo-axial position. The energy profiles
for the allylation reaction with zinc are plotted in Figure 4, with
the isolated σ-allylzinc species and benzaldehyde as references.
The solvation model selectively stabilizes the post-complex
(product) due to the formation of more localized charges on
both zinc and oxygen, and this effect is also present to some
FIGURE 5. Illustration of the cyclic, six-membered TS for the
allylation of benzaldehyde with zinc (gas phase).
degree in the TS. The locked six-membered TS has fewer
degrees of freedom than both reactant and product (Figure 5),
resulting in a higher Gibbs free energy, but the overall
transformation is energetically favored in all cases.
On the basis of the similarity of the Hammett plots, we also
desired to find similarities between the different metals in the
computational study. With metallic zinc, the oxidative addition
was expected to give exclusively the zinc(II) oxidation state,
whereas for indium, both indium(I) and indium(III) oxidation
states must be considered. Initially, we decided to characterize
the reaction with indium(III). However, this resulted in a pre-
complex and a post-complex with virtually identical energies
(In(III), Figure 7). The TS could be located in the gas phase
and by using the solvation model and was found to have a
structure similar to that found for zinc, albeit with slightly
different bond distances (In-C: 2.28 Å and C-C: 2.10 Å,
Figure 6 and Table 2, entry 2). Although indium(III) is
frequently employed in allylation reactions, NMR studies of
the Barbier allylation in water have suggested that the initially
formed allylindium(I) complex reacts directly with the aldehyde
instead of participating in further oxidative addition reactions.14a
As a result, we decided also to characterize the In(I)-mediated
pathway.
(28) For a DFT investigation on the addition of Grignard reagents to
carbonyl compounds, see: Yamazaki, S.; Yamabe, S. J. Org. Chem. 2002,
67, 9346-9353.
(29) To estimate the influence of the ligands on zinc, we also performed
an investigation of the Zn-mediated allylation with two water molecules as
ligands. In addition, the entire system was embedded in the PCM solvation
model for water. The structures of the transition states were found to be
very similar, documenting that the models investigated here indeed are
relevant for the reactions carried out in aqueous media (XYZ data and
solvation energies are included in the Supporting Information).
Unfortunately, all attempts to locate a TS for the indium(I)-
mediated pathway were unsuccessful. Instead, we performed a
3232 J. Org. Chem., Vol. 73, No. 8, 2008

Barbier Allylation in Aqueous Media
FIGURE 6. Structure of the TS with In(III) and two auxiliary bromide
atoms.
FIGURE 9. Structure of the TS with tin(II) and one bromide.
FIGURE 10. Structure of the TS obtained with bismuth in the gas
phase. The corresponding TS with antimony is virtually identical, except
for a slightly shorter M-C bond length.
FIGURE 7. Energy profiles for the C-C bond formation via a six-
membered TS. All energies are composite energies.
the TS with the smaller and more highly charged tin(IV) has
shorter Sn-C distances, whereas the overall more endergonic
tin(II) pathway has an earlier TS with a longer C-C distance
(Figure 9 and Table 2, entries 3 and 4).
Antimony and bismuth are both located in group 15 of the
periodic table and would be expected to behave similarly.
Indeed, the structures obtained in the computational study were
quite similar, and energies were consistently within 5-10 kJ/
mol, which prompted us to include only the structure of the TS
obtained with bismuth (Figure 10). As expected, for both metals,
the overall transformation is exergonic in the gas phase by 36
kJ/mol (Sb) and 30 kJ/mol (Bi), respectively. Inclusion of the
solvation model lowered the exergonicity to 10 kJ/mol for
antimony, while for bismuth, the reactant and product complexes
now had equal energies.
FIGURE 8. Energy profiles calculated for the indium(I)-mediated
allylation. Each data point represents a fully minimized structure having
that particular C-C bond length fixed.
Determination of Kinetic Isotope Effect. All theoretical
determinations of kinetic isotope effects are based on the
calculated frequencies using analytical gradients in the gas
phase. The simplest approach uses only the differences in zero-
point energies (ZPEs) and derives a relative speed of reaction
for the two isotopomers using a simple Boltzmann-type expres-
sion.30 However, with the current advances in theoretical
methods, it has even become possible to evaluate the differences
in Gibbs free energies that in addition to the previously
mentioned ZPE also account for the differences in all vibrational
modes of the molecule. A closed six-membered transition state
is expected to be accompanied by an inverse KIE due to the
increased sp3-character of the carbonyl carbon atom, which
results in a weakening of the bending vibrations for the attached
hydrogen or deuterium. In all cases, the analysis was based on
series of energy minimizations with the C-C bond length fixed
between 3.6 and 1.6 Å in intervals of 0.2 Å (Figure 8).
On the basis of the composite energies, it seems that a pre-
reactive complex was formed with a C-C distance of ap-
proximately 3.2 Å, which undergoes allylation through a TS
situated at a C-C distance of about 2.4 Å with a very low
activation energy of only 5 kJ/mol. This clearly illustrates the
high reactivity of this very electron-rich In(I)-allyl species, which
is in accordance with the scenario outlined previously, where
the formation of the In(I)-allyl species is slow but the following
reaction with aldehyde takes place almost instantaneously.
With tin, we investigated both a tin(II)- and a tin(IV)-mediated
allylation pathway since NMR studies have shown that both
may operate in aqueous solution.14b Also, here, the lower
oxidation state seems to be more reasonable (Sn(II), Figure 7),
although the two TS are structurally rather similar. As expected,
(30) Westheimer, F. H. Chem. ReV. 1961, 61, 265-273.
J. Org. Chem, Vol. 73, No. 8, 2008 3233

Dam et al.
the calculated differences in Gibbs free energies upon going
from the pre-reactive complex to the TS, which were derived
from the analytical frequency calculations in the gas phase. Data
for the deuterated analogues were obtained using the Hessian
from the non-deuterated analogue by simply changing the
cyclic TS under development of a small negative charge in the
benzylic position. A computational study further supported the
involvement of a closed, six-membered transition state, and in
addition, the calculated SDKIE value based on this mechanism
was found to be in good agreement with the experimental values.
2
hydrogen isotope (1H f H).
For indium and tin, both low oxidation states (In+ and Sn2+
)
and high oxidation states (In3+ and Sn4+) were considered
theoretically, and for indium, the lower oxidation state gave
the best agreement with the experimental results, whereas for
tin, the results indicated that the higher of the two oxidation
states was involved. When magnesium was employed, under
the Barbier conditions, both experimental and theoretical data
supported the formation of a radical anion as the selectivity-
determining event. We believe that detailed information on the
allylation of aldehydes by the Barbier procedure conveyed in
this work may prove to be useful for further improvements of
the reaction conditions and thereby stimulate additional ap-
plications of this method.
For zinc, the SDKIE value was calculated to be 0.91 (at 298
K), which is slightly higher than the experimental value derived
from the competition experiments (0.86). For indium(III), the
calculated SDKIE value was determined to be 0.89, which is
relatively close to the theoretical value determined for zinc but
significantly lower than the experimental value determined for
indium (0.95). However, since we were only able to locate the
TS for the indium(III)-mediated reaction, this discrepancy
between experiment and theory indicates the involvement of
an indium(I)-mediated pathway as suggested by NMR studies.14a
With a more electron-rich In(I)-allyl species, the C-C bond
formation process could be extremely facile, and the corre-
sponding TS would then have significant early character, which
is expected to result in a very small SDKIE value. With tin as
the metal, reaction pathways for both tin(II)- and tin(IV)-
mediated reactions were characterized. The agreement with
experiment (kH/kD ) 0.85) was better for the tin(IV)-mediated
pathway, where the calculated SDKIE value was indeed 0.85
(T ) 60 °C), whereas for the tin(II)-mediated pathway, the
calculated SDKIE value was slightly higher, 0.93 (T ) 60 °C).
This is an indication that the tin(IV)-mediated pathway is
operating in these reactions, although we cannot rule out the
involvement of a parallel tin(II)-mediated pathway as suggested
earlier.14b For antimony and bismuth, only the M(III) oxidation
states were characterized, which were the only oxidation states
that were deemed plausible, and in addition, this is also in line
with previous investigations using NMR spectroscopy.14c With
antimony, the SDKIE value was calculated to be 0.77 at 298
K, which is in good agreement with the experimental value
(0.75). For bismuth, the experiments were conducted at 60 °C
(kH/kD ) 0.85), and when this temperature was used in the
Boltzmann expression, the theoretical value was 0.89, which is
only slightly higher than the experimental value. It is particularly
noteworthy that the SDKIE value for antimony is in good
agreement with the experimental value, and we see no indication
of a radical pathway in this reaction as suggested earlier.13b Since
the experiments with magnesium point to the formation of a
radical anion of benzaldehyde as the initial, selectivity-determin-
ing event, we also attempted to calculate the resulting theoretical
SDKIE value for this transformation. Consequently, the radical
anion of benzaldehyde was optimized, and frequencies were
calculated for both isotopomers. The SDKIE value was calcu-
lated here to be 1.11 (298 K), in semiquantitative agreement
with the experimental value, thus further supporting the forma-
tion of a radical anion when this metal is employed.
Computational Details
All calculations were performed using density functional
theory (B3LYP functional31), in combination with the LACVP*
basis set as incorporated in the Jaguar program package.32
The LACVP* basis set used the Los Alamos effective core po-
tential (ECP) and basis set for the description of heavy atoms (Zn,
In, Sn, Sb, Bi, and Br),33 whereas all lighter atoms (H, C, O, and
Mg) were described using the 6-31G* basis set.34 Solvation
was described using the Poisson-Boltzmann self-consistent reaction
field model (PB-SCRF)35 with parameters suitable for tetrahydro-
furan (dielectric constant: 7.43 and probe radius: 2.524). The
difference between the results obtained in the gas phase and
those obtained using a solvation model can be large, whereas the
difference between two different solvation models is small. The
solvation model for water thus was expected to give virtually
identical results, which was verified for Zn-allyl addition to
benzaldehyde. The PB-SCRF is a continuum solvation model,
where the molecule is put into a reaction field consisting of surface
charges on a solvent-accessible surface constructed using a
hypothetical spherical solvent probe molecule with the indicated
radius.36 The solution-phase structures were fully optimized to allow
comparison to gas-phase structures. Vibrational frequencies were
calculated for the gas-phase structures using analytical expressions,
whereas the corresponding frequencies for the solution phase only
were available using numerical derivatives. The numerical frequency
calculations are very time-consuming, and in addition, a lower
accuracy must be expected; therefore, this approach is intractable.
Instead, to obtain an approximate Gibbs free energy in the solution
phase (composite energy), we decided to use the solution-phase
energy and add the vibrational contribution from the gas phase
(∆Gsolv ) ∆Esolv + ∆Ggas - ∆Egas). We found that it is advanta-
geous to use the energy for the fully optimized solution-phase
(31) (a) Lee, C.; Yang, W.: Parr, R. G. Phys. ReV. B: Condens. Matter
Mater. Phys. 1988, 37, 785-789. (b) Becke, A. D. J. Chem. Phys. 1993,
98, 5648-5652. (c) Becke, A. D. J. Chem. Phys. 1993, 98, 1372-1377.
(32) Jaguar, version 6.5, release 106; Schrodinger, LLC: New York,
2005.
Conclusion
The metal-mediated allylation of aldehydes under Barbier
conditions was investigated with both experimental and theoreti-
cal methods. For zinc, indium, tin, antimony, and bismuth,
linear Hammett plots were obtained with regular σ values,
whereas radicals did not seem to be involved in the rate-
determining step. A small positive slope was found for
the Hammett correlation, which is a strong indication that a
discrete organometallic allylmetal species is present in the
rate-determining step, which subsequently reacts in a closed,
(33) Hay, P. J.; Wadt, R. W. J. Chem. Phys. 1985, 82, 299-310.
(34) Hehre, W. J.; Ditchfield, R.; Pople, J. A. J. Chem. Phys. 1972, 56,
2257-2261.
(35) (a) Tannor, D. J.; Marten, B.; Murphy, R.; Friesner, R. A.; Sitkoff,
D.; Nicholls, A.; Ringnalda, M.; Goddard, W. A., III; Honig, B. J. Am.
Chem. Soc. 1994, 116, 11875-11882. (b) Marten, B.; Kim, K.; Cortis, C.;
Friesner, R. A.; Murphy, R. B.; Ringnalda, M. N.; Sitkoff, D.; Honig, B. J.
Phys. Chem. 1996, 100, 11775-11788.
(36) For a discussion of implicit solvation models, see: Cramer, C. J.
Essentials of Computational Chemistry: Theories and Models; Wiley: New
York, 2002.
3234 J. Org. Chem., Vol. 73, No. 8, 2008

Barbier Allylation in Aqueous Media
structures (Esolv
37 rather than using single-point solvation energies.
For transition states, one single imaginary eigenfrequency was
obtained in all cases, which upon visualization was found to
correspond to the expected vibration of the forming C-C bonds.
)
and Green Chemistry is funded by the Danish National Research
Foundation.
Supporting Information Available: Experimental procedures,
characterization of compounds, and experimental data from kinetic
runs. Cartesian coordinates, SCF energies, and Gibbs free energies
for structures calculated in the gas phase and PB-SCRF energies
for structures calculated using the solvation model. This material
is available free of charge via the Internet at http://pubs.acs.org.
Acknowledgment. The Lundbeck Foundation is gratefully
acknowledged for financial support. The Center for Sustainable
(37) Ahlquist, M.; Fristrup, P.; Tanner, D.; Norrby, P.-O. Organometallics
2006, 25, 2066-2073.
JO800180D
J. Org. Chem, Vol. 73, No. 8, 2008 3235