Experimental design and the optimization of a polymer supported palladium complex for use in the Heck reaction

Article abstract of DOI:10.1016/j.tetlet.2004.09.016

Using a Design of Experiments approach the optimization of a Heck reaction catalyzed by a solid-supported sulfur-containing palladacycle was improved from 30% with large observable leaching to 88% with no observable leaching. Optimization of the Heck reaction of 4-bromoacetophenone with styrene by a polymer supported, sulfur-containing palladacycle, varying 6 factors at a total of 28 different levels, corresponding to 5760 different possibilities was undertaken. Conversion improved from 34%, with large observable leaching to 88% with no leaching. This was accomplished using a Design of Experiments approach facilitated by the Statistical Design Package, MODDE 7.0^TM.

Full text of DOI:10.1016/j.tetlet.2004.09.016

Tetrahedron
Letters
Tetrahedron Letters 45 (2004) 8239–8243
Experimental design and the optimization of a polymer
supported palladium complex for use in the Heck reaction
Catherine Anne McNamara,a Frank Kingb and Mark Bradleya,*
aDepartment of Chemistry, University of Southampton, Highfield, Southampton SO17 1BJ, UK
bJohnson Matthey Research and Technology, PO Box 1, Belasis Avenue, Billingham, Cleveland, TS23 1LB, UK
Received 22 July 2004; revised 25 August 2004; accepted 2 September 2004
Abstract—Optimization of the Heck reaction of 4-bromoacetophenone with styrene by a polymer supported, sulfur-containing pal-
ladacycle, varying 6 factors at a total of 28 different levels, corresponding to 5760 different possibilities was undertaken. Conversion
improved from 34%, with large observable leaching to 88% with no leaching. This was accomplished using a Design of Experiments
approach facilitated by the Statistical Design Package, MODDE 7.0TM
.
Ó 2004 Elsevier Ltd. All rights reserved.
Combinatorial chemistry has been one of the most rap-
idly expanding areas of chemistry in the past decade,
enabling the rapid, parallel synthesis and screening of
hundreds of thousands of compounds.1 However, the
optimization of a specific library synthesis, which can
involve numerous possibilities of reagents and reaction
conditions can often be far more time consuming than
carrying out the actual process of library synthesis
itself.2
rational, statistical fashion and various software pack-
ages have been designed to facilitate the use of this
methodology in a wide range of applications.5
The use of polymer supported catalysts and reagents has
been thoroughly reviewed6 and their use often facilitates
work-up and eradicates the need for time-consuming
purification steps. The possibility of recycling can also
appreciably reduce the cost and time of synthesizing,
or purchasing expensive or synthetically challenging
ligands.
Traditionally, chemists have addressed the issue of
optimizing a reaction by varying factors randomly or
more logically in a sequential, parallel fashion, some-
times entitled the OVAT (one variable at a time) or
the COST (changing one single factor at a time)
approach. These techniques are limited in that consecu-
tive, one-dimensional analyses do not completely utilize
the full, multi-dimensional experimental domain and
cannot efficiently account for interactions between fac-
tors.3 For this reason various groups have reported the
application of Design of Experiments (DoE) to optimize
reactions.4 This approach, which involves the systematic
variation of variables over a series of experimental
stages, aims to yield the maximum amount of informa-
tion from the minimum number of experiments, in a
In order to combine DoE with polymer supported cataly-
sis a system, which has been shown to work effectively
with scope for optimization was sought. Such a system
would be required to be affected by several factors,
which could be straightforwardly set at required levels,
and would require facile attachment of the catalytic
moiety onto a solid support. Soluble poly(ethylenegly-
col) (PEG) supported sulfur-containing palladacycles,
reported in 1999 by Bergbreiter et al. represent such a
catalytic system,7 which provide a range of ligands in
which the sulfur pincer, the resin type and the incorpo-
ration of a spacer group between the resin and the cata-
lytic moiety could be varied and readily prepared. In
addition to these ÔsupportÕ characteristics a variety of
reaction conditions such as: solvent, type of base, equiv-
alents of base and the incorporation of a tetraalkyl-
ammonium halide, phase transfer catalyst8 were all open
to optimization.
Keywords: Design; Polymer; Supported; Palladium.
*
Corresponding author. Tel.: +44 2380593598; fax: +44
2380596766; e-mail: mb14@soton.ac.uk
0040-4039/$ - see front matter Ó 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tetlet.2004.09.016

8240
C. A. McNamara et al. / Tetrahedron Letters 45 (2004) 8239–8243
Bergbreiter has shown that tridentate, sulfur-containing
palladacycles with a phenyl, thiol pincer group attached
to PEG by an ether linkage were active catalysts in the
Heck reaction of aryl iodides with alkene acceptors,
although decomposition of the catalytic moiety was
observed. Ligands attached via an amido linkage were
also prepared, which also showed good activity in the
Heck reaction, this time with no observable decomposi-
tion. Soluble counterparts of the PEG supported cata-
lysts were however found to be ineffective in
promoting the reaction with aryl bromides.
Br
34% conversion
30% isolated yield
PS-Pd catalyst
DMF/NEt3
O
O
Scheme 2. Heck reaction of 4-bromoacetophenone and styrene.
mization in three stages, two screening experiments and
one optimization study with a smaller set of conditions.
With this in mind, four different thiol pincer groups
(phenyl, cyclohexyl, t-butyl and ethyl) attached to four
different resins (polystyrene (PS-0), polystyrene with a
six carbon chain spacer (PS-6), Tentagel (TG) and Argo-
pore (AP)) were synthesized as shown in Scheme 1, from
diol 1, which was synthesized in three steps from 5-hydr-
oxy-isophthalic acid as previously described by Van
Veggel and co-workers.9 Complexation with palladium
was achieved by reaction with Pd(MeCN)2 (BF4)2,
formed in situ from palladium chloride and silver
tetrafluoroborate.
In addition to the thiol and resin type other factors
shown in Table 1 were varied. Type of base was included
as a qualitative factor and the levels selected were con-
sidered to encompass the range of strong, medium and
weak inorganic, as well as soluble, bases.
Quantification of solvent is a major obstacle in DoE.
For example, the most common descriptor is the dielec-
tric constant, but often this is not sufficient, for example,
two solvents with very different properties such as hep-
tane and dioxane, have very similar values of 1.92 and
2.22, respectively. For this reason, in this study, an alter-
native approach where a large number of solvents were
incorporated as qualitative factors was undertaken. This
meant that model coefficients would not be associated
with a Ôweighting valueÕ and predictions on the optimum
solvent and base could be made with results in hand,
post experiment rather than before. Equivalents of base
and the incorporation of the phase transfer catalyst—
tetrabutylammonium bromide—were also included as
quantitative factors.
Resin 3a was used as the catalyst in the Heck reaction of
bromoacetophenone with styrene in NEt3/DMF, which
initially proceeded with only 34% HPLC conversion
with a large amount of observable palladium leaching
(Scheme 2).
To seek an improvement of this reaction an extensive
DoE investigation varying 6 factors at a total of 28 dif-
ferent levels, corresponding to 5760 different possibilities
was undertaken (if each quantitative variable repre-
sented three levels). As such a large number of factors
were to be studied, it was decided to carry out the opti-
Solvents were selected by plotting the lipophilicity of
27 common laboratory solvents versus the dielectric
constant using values reported by Musumarra and co-
workers10 and classing each according to the type of
functional group. Ten were then selected encompassing
all areas of the plot and functional group type (see sup-
porting information).
O
O
O
OTBDMS
a), b), c)
O
O
O
O
d), e)
THPO
O
OTHP
OH
OH
Br
Br
O
1
The factors and response (conversion) were entered into
the design wizard in MODDE 7.0,5c,11 which generated
a D-Optimal, screening design with a total of 40 exper-
iments, including three replicated centre points. Reac-
OH
O
RS
N
H
R1
f), g)
h), i)
Pd
Cl
S
R
RS
SR
R1
R1
Table 1. Factors to be varied in the first screening experiment
R
R
3a-d PS-0 4a-d PS-6
5a-d TG 6a-d AP
(16 different resins)
Factor
Type
Levels
2a Ph 2b c-hex
2c tBu 2d Et
Resin
Thiol
Base
Qualitative
Qualitative
Qualitative
Qualitative
4
4
PS-0, PS-6, TG, AP
Ethyl, c-hexyl, t-butyl, phenyl
NaOH, Na2CO3, NaOAc, NEt3
Heptane, dibutyl ether, CHCl3,
ethyl acetate, 1,4-dioxane,
1-pentanol, 3-methyl-2-
butanone, NMP, DMF, MeCN
1 and 4
4
10
Scheme 1. Synthesis of polymer supported palladacycles: (a) DHP,
p-PTS, DCM, 2.5h, quant; (b) TBAF, THF, 18h, 76%; (c)
BrCH2CO2Et, K2CO3, KI, MeCN, reflux, 6h, 68%; (d) i. p-ToSH,
EtOH, rt, 18h, ii. Amberlyst A-15 EtOH, reflux, 18h, 67%; (e) i. MsCl,
Et3N, DCM, 0°C, 50min, ii. LiBr, acetone, reflux, 18h, 69%; (f) RSH,
K2CO3, DMF, reflux, 18h, 82–100%; (g) LiOH, THF/H2O (4/1), 2h,
82–95%; (h) resin, DIC, HOBt, DCM, o/n; (i) i. Pd(MeCN)2(BF4)2,
MeCN, reflux o/n, ii. brine, DCM, 30min.
Solvent
Base eq Quantitative
PTCa eq Quantitative
0 and 2
a PTC = phase transfer catalyst.

C. A. McNamara et al. / Tetrahedron Letters 45 (2004) 8239–8243
8241
tions were conducted in a Radleys Carousel with modi-
fied thinner tubes, in random order at 70°C. A leaching
assessment screen, which involved assigning a value
according to the observable leaching, was also formu-
lated. Results from this initial screening experiment
(see supporting information for full table of results
and coefficient overview plot) showed conversions after
24h ranged from 1% to 94% and a wide range of leach-
ing was observed. MODDE fitted a model to the data,
which gave an R2 of 0.83 and 0.82 for conversion and
leaching, respectively. The normal effect plots for both
responses are shown in Figure 1, which indicate that
the most positive significant factor on conversion was
the use of DMF, however this corresponded to the sec-
ond largest negative effect on leaching. NaOH as base
gave the worst response, with high levels of leaching.
indicated that the presence of the phase transfer catalyst
had a positive effect on the conversion although for this
experiment the levels had been set at 0 and 2 therefore,
except for in the case of the centre points, it had been
used in excess (2equiv). It was deemed desirable to
investigate the use of this material at lower levels and
for the second screening stage this factor was set at levels
of 0.1 and 2.
By considering the solvent as a qualitative factor it had
been hoped that correlations between the model coeffi-
cients and their physio-chemical properties could be
established in order that predictions for the desirable
parameters could be made. Unfortunately no such rela-
tionships could be detected although some conclusions
could be drawn including; the reaction proceeded in a
wide range of solvents with different functionalities,
except for CHCl3, the only chlorinated solvent. The
worst leaching was observed for nitrogen-containing sol-
vents with high dielectric constants.
Contour plots (see supporting information) indicated
that the presence of a phase transfer catalyst (PTC)
had a positive effect on the conversion but no effect on
leaching. In contrast the equivalents of base had no
effect on conversion but led to increased leaching.
For the second screening stage it was decided to select 5
out of the initial 10 solvents; chloroform, pentanol, ethyl
acetate and MeCN were disregarded because of their
negative effect on conversion. DMF was retained
because of its positive effect on conversion although
NMP was omitted because it was considered similar to
DMF.
Although minor, the effect of the thiol pincer group
showed that the electron rich, bulky c-hexyl and t-butyl
thiols gave better conversions than their smaller, ethyl
counterparts, with the lowest levels of leaching observed
for the c-hexyl moiety. Finally Argopore and Tentagel
supported catalysts gave higher conversions than those
immobilized on polystyrene however, these also gave
the highest levels of leaching. PS-6 gave the least leach-
ing but also the lowest conversion response.
Heptane, dibutyl ether, methyl-butanone, dioxane and
DMF were therefore selected as the five solvents for
further investigation. The factors and their levels for
the second screening experiment are summarized in
Table 2.
For the second screening experiment it was decided to
retain all levels of thiol pincer group and resin but PS-
6 was excluded as it had the worst effect on conversion.
As it had no effect on the conversion but a negative effect
on leaching the equivalents of base was fixed at 1. Anal-
ysis of the results from the first screening experiment
Again the factors and responses were entered into
MODDE, which generated 28 necessary experiments to
fulfil the design criteria. The experiments were carried
out in blocks of 12, again in random order, (see supporting
Conversion
Leaching
0.98
0.98
sol (DMF)
sol(dibutylether)
0.95
0.95
pha
bas(NaOAc)
0.9
0.9
)
sol(NMP
sol(chloroform)
sol(dibutylether)
sol(pentanol)
0.8
0.8
)
sol(methyl-butanone
thi(c-hexyl)
res(argopore)
sol(ethylacetate)
0.7
0.6
0.5
0.4
0.3
thi( butyl)
0.7
t-
bas(Na2CO3)
thi(c-hexyl)
res(polystyrene 6)
res(tentagel)
0.6
pha
res(tentagel)
bas(NaOAc)
bas(NaOH)
0.5
0.4
0.3
sol(MeCN)
ba2
)
sol(methyl-butanone
)
sol(dioxane
thi(t-butyl)
res(argopore)
thi(ethyl)
)
thi(ethyl
bas(Na2CO3)
sol(ethylacetate)
ba2
0.2
0.2
sol(dioxane)
sol(NMP)
sol(pentanol)
res(polystyrene 6)
0.1
0.1
sol(DMF)
sol(chloroform)
0.05
0.05
bas(NaOH)
sol(MeCN)
0.02
0.02
-5.00 0.00
5.00 10.00
-1.00
0.00
1.00
2.00
Effects
Effects
Figure 1. Normal effect plot for the first screening experiment (res = resin, sol = solvent, bas = base, pha = equivalents of phase transfer catalyst,
ba2 = equivalents of base).

8242
C. A. McNamara et al. / Tetrahedron Letters 45 (2004) 8239–8243
Table 2. Factors varied in the second screening experiment
expensive and has a superior loading to the other two).
It was also decided to fix the use of NaOAc as base as
this had the most positive effect on both responses. Thus
the other two factors varied were solvent and thiol and
two levels for each were then decided upon, these were
cyclohexyl and t-butyl due their positive effect on con-
version. With regard to solvent, dioxane and heptane
were retained as these had positive effects on conversion.
Factor
Type
Levels
Resin
Thiol
Base
Qualitative
Qualitative
Qualitative
Qualitative
4
4
PS-0, TG, AP
Ethyl, c-hexyl, t-butyl, phenyl
NaOAc and NEt3
Heptane, dibutyl ether,
1,4-dioxane, 3-methyl-2-
butanone, DMF
4
10
Solvent
PTC (eq) Quantitative
0 and 2
The factors investigated in the response surface mode-
ling experiment are shown in Table 3. This gave 16
experiments to fulfil the design criterion, in these experi-
ments no leaching assessment was undertaken as almost
none was observed. This time the R2 for the model was
0.81 (Table 4).
information for results and coefficient overview plot). A
model was fitted, with an R2 of 0.79 and 0.82 for conver-
sion and leaching, respectively (Fig. 2).
For this experiment, the most significant positive effect
on conversion was the use of a phase transfer catalyst,
closely followed by the use of the cyclohexyl thiol pincer
group. The most significant, positive effect on leaching
was the use of heptane as solvent followed by the use
of dioxane. Of the three resins the only positive response
for conversion was for polystyrene, Tentagel gave a posi-
tive effect on leaching. With regard to the four thiols
only cyclohexyl and ethyl had positive effects, on con-
version and leaching, respectively, the others showed
negative effects. Interestingly the phenyl pincer, which
is the most commonly used had the most negative effect
on both responses.
The normalized coefficient plot for these results is shown
in Figure 3, which shows that the c-hexyl thiol pincer
group and the use of dioxane as solvent gave superior
conversions while the reaction showed curvature with
regard to phase transfer catalyst that is it had a quad-
ratic dependence and conversion peaked at intermediate
levels. The normalized coefficient plot also shows that
the effect of the phase transfer catalyst depended on
the solvent, in that increased levels caused conversion
levels to increase more with heptane than with dioxane.
From the optimization experiments it was clear that the
best conditions were with the cyclohexyl thiol pincer lig-
Due to the higher number of experiments required, for
response surface designs it is desirable to minimize the
number of factors and their levels. It was thus decided,
for the final experimental phase to vary three factors
at two levels.
Table 3. Factors varied in the response surface modeling experiment
Factor
Thiol
Type
Levels
Qualitative
+1
À1
+1
À1
+1
À1
c-Hexyl
t-Butyl
1
From the results of the second screening experiment it
was clear that the quantity of the phase transfer catalyst
had to be further investigated. It was decided to fix the
resin as polystyrene as this gave the most positive effect
on conversion (pleasing as this resin is considerably less
PTC
Quantitative
Qualitative
0.1
Dioxane
Heptane
Solvent
Conversion
Leaching
pha
sol(heptane)
0.95
0.95
0.9
0.9
)
thi(c-hexyl)
sol(dioxane
0.8
0.8
0.7
0.6
0.5
0.4
0.3
res(polystyrene)
thi(ethyl)
0.7
0.6
0.5
0.4
sol(heptane)
bas(NaOAc)
thi(c-hexyl)
sol(dioxane)
bas(NaOAc)
res(polystyrene)
thi(t-butyl)
thi(t-butyl)
res(argopore)
res(argopore)
0.3
sol(methyl-butanone)
sol(methyl-butanone)
0.2
0.2
thi(ethyl)
pha
0.1
0.1
0.05
0.05
sol(DMF)
sol(DMF)
0
2000
4000
-2.00 0.00
2.00
4.00
Effects
Effects
Figure 2. Normal effects plot for the second screening experiment.

C. A. McNamara et al. / Tetrahedron Letters 45 (2004) 8239–8243
8243
Table 4.
part, the reaction showed a quadratic dependence
concerning the level of phase transfer catalyst.
Consideration of the factors using this approach not
only afforded substantially improved conditions but also
led to an overall, greater understanding of the reaction.
Run
Thiol
PTC eqa
Solvent
Conversion (%)
1
c-Hexyl
c-Hexyl
c-Hexyl
t-Butyl
c-Hexyl
c-Hexyl
t-Butyl
c-Hexyl
c-Hexyl
t-Butyl
c-Hexyl
t-Butyl
c-Hexyl
c-Hexyl
t-Butyl
t-Butyl
0.55
0.55
0.1
1
Heptane
Heptane
Dioxane
Heptane
Heptane
Dioxane
Dioxane
Dioxane
Heptane
Heptane
Heptane
Dioxane
Dioxane
Heptane
Dioxane
Heptane
76
70
61
69
50
91
85
73
85
79
69
48
42
88
25
7
2
3
4
5
0.1
0.55
0.55
1
6
Acknowledgements
7
8
We thank the EPSRC and Johnson Matthey for a
CASE award.
9
1
10
11
12
13
14
15
16
0.55
0.55
1
Supplementary data
0.1
0.55
0.1
0.1
Supplementary data associated with this article can be
found, in the online version, at doi:10.1016/j.tetlet.
2004.09.016.
a PTC = phase transfer catalyst.
References and notes
0.50
0.00
1. Sanchez-Martin, R. M.; Mittoo, S.; Bradley, M. Curr.
Top. Med. Chem. 2004, 4, 653.
2. Pilipauskas, D. R. Med. Res. Rev. 1999, 19, 463.
3. Carlson, R. Design and Optimization in Organic Synthesis;
Elsevier: Amsterdam, 1992.
thi(t-butyl)
thi(c-hexyl)
pha
sol(dioxane)
sol(heptane)
pha*pha
4. (a) Chen, J. J.; Nugent, T. C.; Lu, C. V.; Kondapally, S.;
Giannousis, P.; Wang, Y.; Wilmot, J. T. Org. Process Res.
Dev. 2003, 7, 313; (b) Jamieson, C.; Congreve, M. S.;
Emiabata-Smith, D. F.; Ley, S. V.; Scicinski, J. J. Org.
Process Res. Dev. 2002, 6, 823; (c) Jamieson, C.; Congreve,
M. S.; Emiabata-Smith, D. F.; Ley, S. V. Synlett 2000,
1603; (d) Gooding, O. W.; Vo, L.; Bhattacharyya, S.;
Labadie, J. W. J. Comb. Chem. 2002, 4, 576; (e) Owen, M.
R.; Luscombe, C.; Lai, L. W.; Godbert, S.; Crookes, D.
L.; Emiabata-Smith, D. Org. Process Res. Dev. 2001, 5,
308; (f) Evans, M. D.; Ring, J.; Schoen, A.; Bell, A.;
Edwards, P.; Berthelot, D.; Nicewonger, R.; Baldino, C.
M. Tetrahedron Lett. 2003, 44, 9337.
pha*sol(dioxane)
pha*sol(heptane)
-0.50
-1.00
Conversion
Figure 3. Normalized coefficient plot for the optimization experiment.
and in dioxane. MODDE predicted the optimum equiva-
lents of phase transfer catalyst to be 0.65 and under
these conditions the predicted yield was 89%.
5. (a) Multisimplex, Grabitech Solutions AB, Trafikgatan 52,
856 44 Sundsvall, Sweden; (b) Design Expert, Stat-Ease
Inc, 2021 E Hennepin Avenue, Ste 480, Minneapolis, MN
55413-2726; (c) Modde, Woodside House, Woodside
Road, Winkfield, Windsor, Berkshire SL4 2DX.
6. (a) McNamara, C. A.; Dixon, M. J.; Bradley, M. Chem.
Rev. 2002, 102, 3275; (b) Dickerson, T. J.; Reed, N. N.;
Janda, K. D. Chem. Rev. 2002, 102, 3325; (c) Leadbeater,
N. E.; Marco, M. Chem. Rev. 2002, 102, 3217; (d) Song, C.
E.; Lee, S. G. Chem. Rev. 2002, 102, 3495.
7. Bergbreiter, D. E.; Osburn, P. L.; Liu, Y. S. J. Am. Chem.
Soc. 1999, 121, 9531.
8. (a) Jeffery, T. Tetrahedron 1996, 52, 10113; (b) Amatore,
C.; Azzabi, M.; Jutand, A. J. Am. Chem. Soc. 1991, 113,
8375.
The reaction was repeated three times under the follow-
ing conditions: resin—polystyrene 0; thiol—c-hexyl;
base—NaOAc; equivalents of base—1, equivalents of
phase transfer catalyst 0.65 and solvent—dioxane.
Yields obtained were 87%, 88% and 91% with no
observable leaching.
In conclusion, an extensive DoE investigation varying 6
factors at a total of 28 different levels, corresponding to
5760 different possibilities, resulted in the Heck reaction
of 4-bromoacetophenone with styrene being improved
from 34% conversion in a DMF/NEt3 system with large
observable leaching of catalyst to 88% with no observa-
ble leaching in a dioxane/NaOAc system. Although
overall fairly insignificant the choice of resin was poly-
styrene and the incorporation of the cyclohexyl thiol
pincer group proved far superior to the phenyl counter-
9. Huck, W. T. S.; Van Veggel, F. C. J. M.; Reinhoudt, D. N.
J. Mater. Chem. 1997, 7, 1213.
10. Ballistreri, F. P.; Fortuna, C. G.; Musumarra, G.; Pavone,
D.; Scire, S. Arkivoc 2002, 54.
11. Musumarra, G.; Scire, S. J. Chemometr. 2001, 15, 199.