A new, efficient route to titanium-silsesquioxane epoxidation catalysts developed by using high-speed experimentation techniques

Article abstract of DOI:10.1002/1521-3773(20010216)40:4<740::AID-ANIE7400>3.0.CO;2-C

High-speed experimentation techniques have been applied in the synthesis and testing of silsesquioxane-based titanium catalysts (see picture) for the epoxidation of alkenes. Different solvents and organotrichlorosilanes were employed in the optimization of the hydrolytic condensation of silanes to open silsesquioxane structures. This new, fast, and simple synthesis route yields catalysts whose performance is comparable to the best known silsesquioxane-based titanium catalysts.

Full text of DOI:10.1002/1521-3773(20010216)40:4<740::AID-ANIE7400>3.0.CO;2-C

COMMUNICATIONS
A New, Efficient Route to Titanium ±
Silsesquioxane Epoxidation Catalysts
Developed by Using High-Speed
Experimentation Techniques**
Paolo P. Pescarmona, Jan C. van der Waal,
Ian E. Maxwell, and Thomas Maschmeyer*
Titanium centers dispersed on, and supported by, various
forms of silica are catalysts with remarkable properties for
partial oxidations.[1] The character of the active site varies with
the silica support used. Although it has been established that
in the best (on the basis of activity per gram of titanium)
heterogeneous catalysts the Ti centers are all four-coordi-
nate,[2±4] it is still unclear whether, for optimum performance,
the active catalytic species requires the specific configuration
of four siloxy groups, or if being partially hydrolyzed to, say,
one hydroxy and three siloxy ligands is equally good.
Chemical modeling studies by various groups[5±7] have shown
that Ti catalysts with fewer than four siloxy groups should also
be active as catalysts. These considerations indicate that there
is still some residual uncertainty regarding the active site,
despite the many experimental and computational studies
carried out to resolve it.[8, 9] Indeed, several mechanisms, or
variations of one underlying mechanism, might operate
depending on the precise configuration of the Ti site. Hence,
different types of Ti centers might exhibit the desired
characteristics.
For these catalysts, which are difficult to characterize,
organic ± inorganic hybrid compounds have proved to be
useful models. Especially silsesquioxanes[10±12] (RSiO1.5)a-
(H2O)0.5b (R is an organic group; a ꢀ 1, b ꢀ 0) have been a
focus of attention as model compounds for silica. Of interest
for catalysis are those silsesquioxanes that contain Si OH
groups, that is partially condensed structures to which a
catalytically active compound can be complexed.[13, 14] The
incompletely condensed silsesquioxane trisilanol 1 (a ˆ 7,
b ˆ 3) has recently been reported as a precursor for the
soluble, titanium model compound 2 (Scheme 1), which
results in an active catalyst for the epoxidation of alkenes.[5±7]
Silsesquioxane 1 can be obtained by the slow hydrolytic
condensation of the corresponding monosilane RSiX3 [Eq. (1)
and (2); R ˆ organic group; X ˆ Cl, OMe, OEt; a ‡ b ˆ 2n;
n ˆ 1,2,3...;
Scheme 1. Complexation of titanium to the incompletely condensed
silsesquioxane R7Si7O12H3 (R ˆ cyclohexyl, cyclopentyl), modeling the
grafting of titanium on silica.
b ꢁ 3a].[15] Besides compound 1, the reaction may produce
other silsesquioxane structures, some of which may also
contain silanol groups that could react with titanium species to
produce catalytic materials.
The preparation of compound 1 is time-consuming and
requires a number of purification steps.[15] Here we describe
the application of high-speed experimentation techniques[16]
to optimize the preparation of silsesquioxanes as precursors
for Ti catalysts active in the epoxidation of alkenes with
peroxides. The aim is to identify a faster and cheaper way to
synthesize silsesquioxane precursors for Ti catalysts. Since
silsesquioxanes other than 1 may also show relevant catalytic
activity after reaction with titanium, the synthesis of the
silsesquioxane precursors was optimized not as a function of
the yield in compound 1, but rather as a function of the
product yield due to the catalytic activity of the Ti silses-
quioxanes in the epoxidation of 1-octene. The observed
activity is thus the combined activity of all these compounds
(vide infra).
Different factors may influence the reaction equilibria that
determine which silsesquioxane structures and in what
amounts are produced by the hydrolytic condensation of the
organosilane RSiX3.[10, 12] Among those factors, the solvent
and the nature of the organic group R are known to play a
predominant role. Both the solvent and the R group influence
the kinetic rate constant and the solubility of intermediates
and products. The R group also affects the thermodynamic
stability of the different silsesquioxane structures.
A number of solvents and trichlorosilanes were varied
systematically to investigate a wide range of parameters. The
chosen solvents are water-miscible, polar solvents: acetone,
acetonitrile, methanol, and tetrahydrofuran (THF). The
selected trichlorosilanes RSiCl3 were either those usually
employed (R ˆ cyclohexyl and cyclopentyl) or those for which
the corresponding open silsesquioxane 1 has not yet been
reported in the literature (R ˆ phenyl, methyl, n-octyl, allyl)
but which were readily available.
In the preparation of the silsesquioxane-based catalysts,
two general types of behavior regarding the solvents and the
organotrichlorosilanes can be identified:
a) Catalysts derived from silsesquioxane structures that
were synthesized in acetonitrile show the highest catalytic
activity, followed, in decreasing order of activity, by those
from acetone, methanol, and tetrahydrofuran (Figure 1A).
Interestingly, acetonitrile is more effective than acetone, the
RSiX3
‡
3H2O !RSi(OH)3
‡
3HX
(1)
aRSi(OH)3 > (RSiO1.5)a(H2O)0.5b
‡
(1.5a 0.5b)H2O
(2)
[*] Prof. Dr. T. Maschmeyer, P. P. Pescarmona, Dr. J. C. van der Waal
Laboratory of Applied Organic Chemistry and Catalysis
DelftChemTech, TU Delft
Julianalaan 136, 2628 BL Delft (The Netherlands)
Fax : (‡31)15-278-4289
E-mail: th.maschmeyer@tnw.tudelft.nl
Dr. J. C. van der Waal, Dr. I. E. Maxwell
Avantium Technologies B.V.
Zekeringstraat 29, 1014 BV, P.O. Box 2915, 1000 CX Amsterdam
(The Netherlands)
[**] P.P.P. is grateful for a studentship from TUD.
740
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 2001
1433-7851/01/4004-0740 $ 17.50+.50/0
Angew. Chem. Int. Ed. 2001, 40, No. 4

COMMUNICATIONS
regardless of the solvent used during the
synthesis (Figure 1A). This trend is in good
agreement with the literature, where cyclo-
pentyl- and cyclohexyltrichlorosilanes are
reported to form incompletely condensed
silsesquioxanes in high yields.[15] In princi-
ple, the nature of the organic group might
influence which silsesquioxane structures
are produced and in which ratios,[10] as well
as the electronic and spatial properties of
the actual catalyst. Considering that the
organic groups are linked to the titanium
center through a Si O unit, electronic
effects should have a negligible influence
on the catalytic properties and, since the
organic groups point away from the titani-
um center, steric effects are probably also
not relevant. The principal effect of the
organic group is, therefore, considered to
be its role in determining which silsesquiox-
ane structures are formed during the hydro-
lytic condensation and in which amounts.
The trend in the catalytic activity, as a
function of the solvent used in the hydro-
lytic condensation, is not influenced by the
nature of the organic group R. The activity
trend as a function of the organic group is
likewise not affected by the solvent. There-
fore, regarding these trends, the solvent and
organic group can be considered as inde-
pendent variables.
The activities of the catalysts synthesized
using high-speed experimentation techni-
ques were compared with those of
Ti(OiPr)4 and of the Ti ± cyclopentylsilses-
quioxane catalyst 2 produced by the reac-
tion of pure cyclopentylsilsesquioxane 1
with Ti(OiPr)4. The high-speed experimen-
tation catalysts with R ˆ cyclohexyl, cyclo-
pentyl, phenyl, and methyl show an activity
higher than Ti(OiPr)4, confirming the for-
mation of Ti ± silsesquioxanes as the active
species. Among these catalysts, the one
derived from the combination of cyclo-
pentyltrichlorosilane and acetonitrile dis-
plays the highest catalytic activity for the
epoxidation of 1-octene (Figure 1A). This
activity is 87% of that of the Ti ± cyclo-
pentylsilsesquioxane catalyst 2. The rele-
Figure 1. Activities of Ti ± silsesquioxane catalysts (obtained by combining different trichlorosi-
lanes and solvents, using water (A) or a 0.3m HCl solution (B) as hydrolyzing agent) relative to Ti ±
cyclopentylsilsesquioxane catalyst 2 (activity set at 1) in the epoxidation of 1-octene.
vance of this result lies in the fact that the synthesis of our
silsesquioxane precursors does not require any purification
process and is much less time-consuming than the synthesis of
compound 1. Scaling up of the experiment to 100-mL scale
yielded a catalyst with an identical activity, thus demonstrat-
ing that the small-scale high-speed experiment is comparable
to conventional experimentation.
solvent commonly reported for the synthesis of incompletely
condensed silsesquioxane 1.[15] This result can be explained in
terms of electric properties: amongst the solvents used,
acetonitrile has the highest dipole moment and dielectric
constant. Therefore, it can stabilize silsesquioxane species
containing Si OH groups, favoring their synthesis over that of
the less polar, completely condensed silsesquioxanes.
Two 1:1 molar mixtures of n-octyl- and methyltrichlorosi-
lane and of cyclohexyl- and phenyltrichlorosilane were used
to investigate possible synergetic effects between two differ-
b) With respect to the silanes employed, the highest
epoxidation activities were observed in the order cyclo-
pentyl > cyclohexyl > phenyl > methyl > allyl ꢂ n-octyl ꢂ 0,
Angew. Chem. Int. Ed. 2001, 40, No. 4
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 2001
1433-7851/01/4004-0741 $ 17.50+.50/0
741

COMMUNICATIONS
ent organotrichlorosilanes in the formation of incompletely
condensed structures (Figure 1A). The n-octyl/methyl mix-
ture gave a very low epoxidation activity. As in the case of the
pure n-octyl compound, the hydrolytic condensation pro-
duced a gel, generated by micelle formation due, probably, to
the slower hydrolysis rate of long-chain alkylsilanes.[12] Water
retained in this gel would be inimical to the Ti insertion. The
cyclohexyl-/phenyltrichlorosilane mixture showed considera-
ble activity, suggesting the possibility of synergy between
trichlorosilanes with different organic R groups.
Trying to further optimize the reaction, we investigated a
variation of the ternary composition of the three most active
silanes (cyclopentyl, cylcohexyl, and phenyl) in the best
solvent (acetonitrile) (Figure 2). Although the highest epox-
idation activity was found for the pure cyclopentyltrichloro-
silane (i.e. the experiment did not lead to the identification of
tional insight in the system under study. Some general trends
with respect to the roles of both the solvent and the
organotrichlorosilane have been identified, and it has been
shown that the synthesis of silsesquioxane catalysts is favored
using a neutral rather than an acidified (HCl) solution. In
future experiments, we plan to identify the most active Ti-
silsesquioxane by a deconvolution of the complex silsesquiox-
ane mixture. Moreover, we aim to use high-speed exper-
imentation for further optimization of the synthesis of the
silsesquioxane precursors, for the optimization of the Ti-
insertion step, and for the introduction of aqueous hydrogen
peroxide as oxidant.
Experimental Section
Experiments were performed on a parallel synthesis workstation[17] by
using four different solvents (acetone, acetonitrile, methanol, and tetrahy-
drofuran) and different silanes RSiCl3 (R ˆ cyclohexyl, cyclopentyl,
phenyl, methyl, n-octyl, allyl, and mixtures of these). In
a typical
experiment, 6 Â 2 mL aliquots of each of the four solvents were dispensed
in a rack containing 4 mL glass tubes, followed by the addition of 340 mmol
of each of the six silanes to the solvent-containing reaction vessels in such a
way that 24 individually different silane solutions were prepared. Hydrol-
ysis of the silane was initiated by the addition of 0.5 mL of either water or a
0.3m HCl solution to the reaction vessel and placing the vessel array on an
orbital shaker at 508C for 18 h. After removal of the solvent and of the
excess water and hydrochloric acid in a vacuum centrifuge, the samples
were stored under argon. Titanium insertion was performed by dissolving
the crude silsesquioxane mixture in THF (2 mL) under argon, followed by
the addition of titanium isopropoxide. After 5 h at 508C, the tetrahydro-
furan was removed by means of the vacuum centrifuge and the samples
were stored under argon.
The catalytic activity of the materials obtained was determined by the
epoxidation of 1-octene with tert-butyl hydroperoxide (TBHP). The
reaction was performed by adding to each dried sample a solution of
TBHP (ꢂ0.36 mmol) in cyclohexane (TBHP:catalyst ratio ˆ 7.5) and
1-octene (0.0144 mol), which acted both as reactant and solvent (1-
octene:catalyst ratio ˆ 300). Prior to dispensing, 2 mL of decane was added
for each 98 mL of 1-octene to serve as an internal standard in the following
GC analysis. Samples were taken after the reaction had been allow to
proceed for 5 h at 608C and were analyzed on a UNICAM Pro GC using a
CP-Sil-5B column. The epoxidation reaction proceeded very slowly at
room temperature for several days; thus, the yield increased until a plateau
was reached. The reported activities[18] were obtained by normalizing the
GC peak area for 1,2-epoxyoctane by means of the internal standard. Since
the experiments were performed in 1-octene also as a solvent, it has not
been possible to realize a mass balance. The reaction was completely
selective in the synthesis of 1,2-epoxyoctane: no other products were
detected. The conversion selectivity of TBHP towards 1,2-epoxyoctane lies
between 80 and 94%.
Figure 2. Activity in the epoxidation of 1-octene as a function of the ratio
of cyclopentyl-, cylcohexyl- and phenyltrichlorosilanes in the initial
silsesquioxane synthesis mixture (darkness of color / activity).
a more active catalyst), the surface of the activity diagram
shows some interesting features, underpinning the utility of
the experiment: various mixtures yield catalysts with higher
activities than those from pure cyclohexyl- and phenyltri-
chlorosilanes. Such behavior would have been difficult to
identify without a combinatorial approach.
During the hydrolysis of the trichlorosilanes, hydrochloric
acid is produced [Eq. (1)], which could influence the amounts
of silsesquioxanes generated.[10]. To check this potential pH
effect, the experiment was performed by using either H2O or a
0.3m HCl/water solution to hydrolyze the silanes. The trends
observed with both are very similar (Figure 1) and a higher
activity is found when using a neutral water solution (70%
greater on average); thus extra hydrochloric acid is not
beneficial.
The other catalysts (Ti(OiPr)4 and Ti ± silsesquioxane 2) were tested by
using the same amount of titanium and the same experimental conditions
as described above.
Received: August 28, 2000 [Z15709]
[1] B. Notari, Adv. Catal. 1996, 41, 253.
[2] G. Bellussi, M. S. Rigutto, Stud. Surf. Sci. Catal. 1994, 85, 177.
[3] T. Maschmeyer, F. Rey, G. Sankar, J. M. Thomas, Nature 1995, 378,
159.
[4] M. C. Klunduk, T. Maschmeyer, J. M. Thomas, B. F. G. Johnson,
Chem. Eur. J. 1999, 5, 1481.
[5] T. Maschmeyer, M. C. Klunduk, C. M. Martin, D. S. Shephard, J. M.
Thomas, B. F. G. Johnson, Chem. Commun. 1997, 1847.
[6] M. Crocker, R. H. M. Herold, A. G. Orpen, Chem. Commun. 1997,
2411.
[7] a) H. C. L. Abbenhuis, S. Krijnen, R. A. van Santen, Chem. Commun.
1997, 331; b) S. Krijnen, H. C. L. Abbenhuis, R. W. J. M. Hanssen,
We have shown that high-speed experimentation can be a
valuable technique in the optimization of catalysts. A very
active silsesquioxane-based catalyst for the epoxidation of
1-octene (cyclopentyltrichlorosilane hydrolyzed in acetoni-
trile) has been prepared by a much faster and cheaper method
than conventionally possible. In addition to identifying the
optimal synthesis conditions within the screened parameters,
the large number of experiments performed by using high-
speed experimentation techniques enabled us to gain addi-
742
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 2001
1433-7851/01/4004-0742 $ 17.50+.50/0
Angew. Chem. Int. Ed. 2001, 40, No. 4

COMMUNICATIONS
important goal.[1] b-peptides are resistant to cleavage by most
proteases and are therefore of considerable interest as
building blocks in drug design.[2] The groups of Seebach,
Gellman, and DeGrado have demonstrated that oligomers
derived from b-amino acids such as 1 and 2 can adopt
predictable conformations that are a direct consequence of
the nature of the monomeric building blocks 1 and 2
(Scheme 1).[3] We reasoned that incorporation of two aro-
matic rings into 2 would constrain the amino acid backbone of
oligomeric derivatives, creating foldamers with novel and
well-defined hydrogen-bonding patterns.
J. H. C. van Hooff, R. A. van Santen, Angew. Chem. 1998, 110, 374;
Angew. Chem. Int. Ed. 1998, 37, 356.
[8] L. Marchese, T. Maschmeyer, E. Gianotti, S. Coluccia, J. M. Thomas, J.
Phys. Chem. B 1997, 101, 8836.
[9] a) P. E. Sinclair, G. Sankar, C. R. A. Catlow, J. M. Thomas, T.
Maschmeyer, J. Phys. Chem. B 1997, 101, 4232; b) D. Tantanak,
M. A. Vincent, I. H. Hiller, Chem. Commun. 1998, 1031.
[10] M. G. Voronkov, V. I. Lavrentꢁyev, Top. Curr. Chem. 1982, 102, 199.
[11] F. J. Feher, T. A. Budzichowski, Polyhedron 1995, 14, 3239.
[12] P. G. Harrison, J. Organomet. Chem. 1997, 542, 141.
[13] T. Maschmeyer, J. M. Thomas, A. F. Masters, NATO ASI Ser. Ser. C
1997, 498, 461.
[14] H. C. L. Abbenhuis, Chem. Eur. J. 2000, 6, 25.
[15] a) F. J. Feher, D. A. Newman, J. F. Walzer, J. Am. Chem. Soc. 1989,
111, 1741; b) F. J. Feher, T. A. Budzichowski, R. L. Blanski, K. J.
Weller, J. W. Ziller, Organometallics 1991, 10, 2526.
[16] Formerly, high-speed experimentation was commonly referred to as
parallel combinatorial chemistry and high-speed screening. For a
review over these techniques, see P. P. Pescarmona, J. C. van der Waal,
I. E. Maxwell, T. Maschmeyer, Catal. Lett. 1999, 63, 1.
[17] Experiments were performed by using a liquid handling system,
developed in-house by Shell SRCTA, and a 24-tube based rack with
the dimensions of the 96-well plate. The equipment and the
experimental time were kindly donated by Shell and later by
Avantium Technologies.
[18] The values for the activities are the averages of the results from
different experiments. The deviation of each set of results from the
average lies in a range between 0 and 10% of the average value. To
test the robustness of our equipment, our results were also compared
to the results from similar experiments performed at Shell Research
Laboratories Amsterdam; the results were in good agreement.
Design and Synthesis of Foldamers Based on an
Anthracene Diels ± Alder Adduct**
Scheme 1.
Jeffrey D. Winkler,* Evgueni L. Piatnitski,
John Mehlmann, Jiri Kasparec, and Paul H. Axelsen*
We describe herein the preparation of the protected b-
amino acid 9 from 6 (Scheme 2), the Diels ± Alder adduct of
anthracene and dimenthyl fumarate, and the synthesis and
study of both the dipeptide 4 and the tetrapeptide 5. The
presence of the dihydroanthracene moieties leads to a high
level of structural definition in both 4 and 5. The structure of 4
exhibits a unique level of organization. The NMR-derived
solution structure of 5 reveals a twelve-membered ring helical
conformation with a molecular volume of about 1200 Š.
The synthesis of the tetracyclic b-amino acid monomer is
outlined in Scheme 2. Reaction of anthracene with dimenthyl
fumarate under Lewis acid catalysis leads to the formation of
6 in 88% yield and 99% ee.[4] Conversion of 6 to the
corresponding dimethyl ester 7[5] in acidic methanol[6] fol-
lowed by saponification of 7 under mild basic conditions
produces a statistical mixture of the desired monoacid 8, along
with the starting diester 7 and the corresponding diacid (not
shown) which can be efficiently recycled. Curtius reaction of
the acyl azide derived from 8 gave the N-Boc amino ester 9 in
34% yield over three steps. Reaction of 9 with NaOH gave the
corresponding acid 10, whereas exposure of 9 to trifluoro-
acetic acid led to the formation of the deprotected amino ester
11. Condensation of 12, the acid chloride derived from 10,
with 11 gave dipeptide 4 in 79% overall yield from 9.
Dedicated to Professor Ronald Breslow
on the occasion of his 70th birthday
The design and synthesis of peptides and peptide analogues
with predictable and reproducible folding patterns is an
[*] Prof. Dr. J. D. Winkler, E. L. Piatnitski, J. Mehlmann, J. Kasparec
Department of Chemistry, University of Pennsylvania
231 South 34th Street, Philadelphia, PA 19104-6323 (USA)
Fax : (‡1)215-573-6329
E-mail: winkler@sas.upenn.edu
Prof. Dr. P. H. Axelsen
Department of Pharmacology, University of Pennsylvania (USA)
[**] We thank Professor Samuel H. Gellman (University of Wisconsin)
and Professors William F. DeGrado and Stanley J. Opella (University
of Pennsylvania) for valuable discussions. We also thank Ms. Lori
Krim and Mr. Daniel Macks for their experimental and computational
contributions to the early stages of this work. Generous financial
support from Boehringer± Ingelheim and Alfred P. Bader (J.K.:
Alfred E. Bader Fellow, 1996 ± 1998) is gratefully acknowledged.
Supporting information for this article (synthetic procedures and
spectroscopic data for dipeptide 4, tetrapeptide 5, and all synthetic
intermediates) is available on the WWW under http://www.ange-
wandte.com or from the author.
Angew. Chem. Int. Ed. 2001, 40, No. 4
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 2001
1433-7851/01/4004-0743 $ 17.50+.50/0
743