Regioselective silylation of sugars through palladium nanoparticle-catalyzed silane alcoholysis

Article abstract of DOI:10.1021/ja026723v

Palladium (0)-catalyzed silane alcoholysis was applied to sugars for the first time using tertbutyldimethylsilane (TBDMS-H) and Ph₃SiH as the silanes. The catalyst is a colloidal solution of Pd(0) generated in situ from PdX₂ (X = Cl^-, OAc^-) and TBDMS-H in N,N-dimethylacetamide. The colloid has been characterized by dynamic light scattering and transmission electron microscopy and consists of catalytically highly active nanoparticles of ～2 nm diameter. The silane alcoholysis reaction is an effective method for the regioselective silylation of methyl and phenyl glycosides and generates hydrogen gas as the only side product. For many of the sugar substrates investigated, the distribution of regioisomers obtained is complementary to that of the traditional R₃SiCl/base (base = pyridine, imidazole) methodology and gives convenient access to the 3,6- rather than the 2,6-silylated pyranosides, obtained as the main product by the silyl chloride method. The method also allows a selective axial silylation of levoglucosan and 1,3,5-O-methylidene-myo-inositol. In an attempt to rationalize the observed regioselectivities, ab initio predictions (HF/3-21G*) have been made on the relative energies of some of the silylated products. They suggest that the observed regioselectivities do not reflect a kinetic vs thermodynamic product distribution but are induced by the silylation agent employed. Models for the possible origin of the observed regioselectivity in both silylation methods (silane- and silyl chloride-based) are discussed.

Full text of DOI:10.1021/ja026723v

Published on Web 08/10/2002
Regioselective Silylation of Sugars through Palladium
Nanoparticle-Catalyzed Silane Alcoholysis
Mee-Kyung Chung,§ Galina Orlova,§ John D. Goddard,§ Marcel Schlaf,*,§
Robert Harris,† Terrance J. Beveridge,† Gisele White,‡ and F. Ross Hallett‡
Contribution from the Guelph-Waterloo Centre for Graduate Work in Chemistry (GWC2),
Department of Chemistry and Biochemistry, NSERC Guelph Regional STEM Facility,
Department of Microbiology, and Guelph Biophysics Light Scattering Laboratory,
Department of Physics, UniVersity of Guelph, Guelph, Ontario, Canada, N1G 2W1
Received April 29, 2002
Abstract: Palladium(0)-catalyzed silane alcoholysis was applied to sugars for the first time using tert-
butyldimethylsilane (TBDMS-H) and Ph3SiH as the silanes. The catalyst is a colloidal solution of Pd(0)
generated in situ from PdX2 (X ) Cl-, OAc-) and TBDMS-H in N,N-dimethylacetamide. The colloid has
been characterized by dynamic light scattering and transmission electron microscopy and consists of
catalytically highly active nanoparticles of ∼2 nm diameter. The silane alcoholysis reaction is an effective
method for the regioselective silylation of methyl and phenyl glycosides and generates hydrogen gas as
the only side product. For many of the sugar substrates investigated, the distribution of regioisomers obtained
is complementary to that of the traditional R3SiCl/base (base ) pyridine, imidazole) methodology and gives
convenient access to the 3,6- rather than the 2,6-silylated pyranosides, obtained as the main product by
the silyl chloride method. The method also allows a selective axial silylation of levoglucosan and 1,3,5-
O-methylidene-myo-inositol. In an attempt to rationalize the observed regioselectivities, ab initio predictions
(HF/3-21G*) have been made on the relative energies of some of the silylated products. They suggest that
the observed regioselectivities do not reflect a kinetic vs thermodynamic product distribution but are induced
by the silylation agent employed. Models for the possible origin of the observed regioselectivity in both
silylation methods (silane- and silyl chloride-based) are discussed.
Introduction
butyldimethylsilyl chloride (TBDMS-Cl) and tert-butyldiphen-
ylsilyl chloride (TBDPS-Cl) have emerged as the silylation
Silyl ethers are versatile protecting groups for hydroxyl
functions and are routinely used in carbohydrate chemistry.1
The standard method for the silylation of a sugar hydroxyl
function is the reaction of the sugar substrate with a trialkylsilyl
chloride or mixed alkyl/arylsilyl chloride in the presence of
excess base, typically either imidazole in N,N-dimethylforma-
mide (DMF) solvent or pyridine, which acts as both the solvent
and base at the same time (eq 1).
reagents of choice, as the resulting silyl ethers strike the right
balance between stability under acidic or basic conditions and
ease of selective removal, e.g., by reaction with tetrabutylam-
monium fluoride4-7 or catalytic amounts of CBr4 in methanol.8
With TBDMS-Cl, a regioselective silylation of otherwise
unprotected methyl pyranosides of glucose, mannose, and
galactose has been achieved. Six-position monosilylated as well
as 2,6-disilylated and in some cases 2,4,6- and 2,3,6-trisilylated
pyranosides are accessible by this method in moderate to good
yields.9-11 When the reaction is mediated by equimolar amounts
of dibutyltin oxide, the silylation of methyl R-D-pyranosides of
glucose, mannose, and galactose with TBDMS-Cl gives the
6-monosilylated products in excellent yields.12
R-OH + R′R′′R′′′SiCl + B f
R-O-SiR′R′′R′′′ + HB+Cl- (1)
R ) sugar; R′,R′′,R′′′ ) alkyl or aryl;
B ) pyridine, imidazole
The use of trimethylsilyl chloride with this method allows
the rapid persilylation of sugars and polyols, rendering them
volatile enough for GC analysis.2,3 For synthetic purposes, tert-
(4) Corey, E. J.; Venkateswarlu, A. J. Am. Chem. Soc. 1972, 94, 6190-6191.
(5) Hanessian, S.; Lavallee, P. Can. J. Chem. 1975, 53, 2975-2977.
(6) Mulzer, J.; Scho¨llhorn, B. Angew. Chem., Int. Ed. Engl. 1990, 29, 431-
432.
* Address correspondence to this author. E-mail: mschlaf@uoguelph.ca.
§ Guelph-Waterloo Centre for Graduate Work in Chemistry, Department
of Chemistry and Biochemistry.
(7) Kraska, B.; Klemer, A.; Hagedorn, H. Carbohydr. Res. 1974, 36, 398-
403.
(8) Chen, M.-Y.; Lu, K.-C.; Lee, A. S.-Y.; Lin, C.-C. Tetrahedron Lett. 2002,
43, 2777-2780.
† NSERC Guelph Regional STEM Facility, Department of Microbiology.
‡ Guelph Biophysics Light Scattering Laboratory, Department of Physics.
(1) Collins, P.; Ferrier, R. Monosaccharides; John Wiley & Sons: Toronto,
1995.
(9) Franke, F.; Guthrie, R. D. Aust. J. Chem. 1977, 30, 639-647.
(10) Brandstetter, H. H.; Zbiral, E. HelV. Chim. Acta 1978, 61, 1832-1841.
(11) Halmos, T.; Montserrat, R.; Filippi, J.; Anonakis, K. Carbohydr. Res. 1987,
170, 57-69.
(2) Dutton, G. S. AdV. Carbohydr. Chem. Biochem. 1973, 28, 12-160.
(3) Andrews, M. A. Carbohydr. Res. 1989, 194, 1-19.
(12) Bredenkamp, M. W. S. Afr. J. Chem. 1995, 48, 154-156.
9
10508
J. AM. CHEM. SOC. 2002, 124, 10508-10518
10.1021/ja026723v CCC: $22.00 © 2002 American Chemical Society

Palladium Nanoparticle-Catalyzed Silylation of Pyranosides
A R T I C L E S
Chart 1
The transition metal-catalyzed silane alcoholysis reaction
shown in eq 2 presents an attractive alternative method for the
synthesis of silyl ethers that does not generate stoichiometric
amounts of pyridinium or imidazolium chloride but gives
hydrogen gas as the only side product, resulting in very simple
reaction protocols and workup procedures.
R-OH + R′R′′R′′′SiH [cat.]8 R-O-SiR′R′′R′′′ + H2(g)
R ) sugar; R′,R′′,R′′′ ) alkyl or aryl
(2)
The silane alcoholysis reaction is efficiently catalyzed by a
wide variety of transition metal complexes,13-26 the Lewis acid
B(C6F5)3,27 and most simply palladium, nickel, and ruthenium
metal.28 For a comprehensive review on the reaction, see
Lukevics and Dzintara.29
Building on the earlier work by Sommer and Lyons,28 who
first demonstrated the use of Pd(0) as a silane alcoholysis
catalyst, we have now found that a palladium nanoparticle
colloid generated in situ in N,N-dimethylacetamide (DMA)
solvent from PdX2 (X ) Cl-, OAc-) and tert-butyldimethyl-
silane (TBDMS-H) or triphenyl silane is an effective catalyst
for the regioselective silylation of sugars. For the first time, we
have applied this reaction to a series of methyl pyranosides of
glucose, mannose, galactose, xylose, and arabinose as well as
the derivatives 1,5-anhydro-xylitol, arabinitol, 1,6-anhydro-â-
glucopyranoside, and 1,3,5-O-myo-inositol and investigated the
regioselectivity and mechanism of the palladium-catalyzed silane
alcoholysis reaction in direct comparison to the classical silyl
chloride method.
Choice of Sugar Substrate. Because of their high natural
abundance and biological significance, we focused our study
on glucose, mannose, and galactose and some of their corre-
sponding pentoses and 1-deoxy-cyclitols. Due to mutarotation,
the reaction of free sugars with TBDMS-H/Pd(0) or TBDMS-
Cl/base gives intractable complex reaction mixtures and is
consequently not of synthetic value. Therefore, the methyl and
in one case phenyl glycosides as the simplest glycosides of the
sugars were employed as the first series of substrates. Also
investigated were 1,6-anhydro-â-glucopyranoside and 1,3,5-O-
methylidene-myo-inositol. For reference throughout this paper,
the structures of all sugar substrates investigated with their
numbering scheme are shown in Chart 1.
nonpolar solvents such as methylene chloride, toluene, and
pentane, which are incompatible with the highly polar and
lipophobic sugar substrates. In a recent study that employed
ethylene glycol and Et3SiH as a model system, we therefore
evaluated several silane alcoholysis catalysts in the highly polar
yet unreactive solvent DMA, which dissolves sugars and at the
same time is compatible with various transition metal-based
catalysts.30 From this study, Pd(0) generated in situ from either
PdCl2 or Pd(OAc)2 and the silane emerged as one of the most
active andssince no catalyst synthesis is involvedsmost easily
handled catalyst systems and was therefore selected for the
present study. The choice of TBDMS-H as the silane is based
on the properties of the well-established TBDMS-Cl-derived
silyl ethers discussed above and its low tendency for migration
when several hydroxyl functions are present.4,9-11
Choice of Solvent, Catalyst, and Silane. To date, catalytic
silane alcoholysis reactions have been carried out only in
(13) Doyle, M. P.; Higgins, K. G.; Bagheri, V.; Pieters, R. J.; Lewis, P. J.;
Pearson, M. M. J. Org. Chem. 1990, 55, 6082-6086.
(14) Barton, D. H.; Kelly, M. J. Tetrahedron Lett. 1992, 33, 5041-5044.
(15) Bedard, T. C.; Corey, J. Y. J. Organomet. Chem. 1992, 428, 315-333.
(16) Gregg, B. T.; Cutler, A. R. Organometallics 1994, 13, 1039-1043.
(17) Burn, M. J.; Bergman, R. G. J. Organomet. Chem. 1994, 472, 43-54.
(18) Lorenz, C.; Schubert, U. Chem. Ber. 1995, 128, 1267-1269.
(19) Chang, S.; Scharrer, E.; Brookhart, M. J. Mol. Catal. A 1998, 130, 107-
119.
Results
Silylation Reactions. All silane alcoholysis reactions were
carried out by first generating a black solution of the active
Pd(0) catalyst colloid by reacting PdX2 (3 mol %, X ) Cl-,
OAc-) with 0.3-3.3 equiv (with respect to sugar) of TBDMS-H
in DMA and then adding a second solution of the sugar in DMA.
The reaction progress was monitored by analytical TLC, and
completion of the reaction was indicated by the cessation of
hydrogen gas evolution and precipitation of bulk palladium
(20) Luo, X.-L.; Crabtree, R. H. J. Am. Chem. Soc. 1989, 111, 2527-2535.
(21) Oehmichen, U.; Singer, H. J. Organomet. Chem. 1983, 243, 199-204.
(22) Chalk, A. J. J. Chem. Soc., Chem. Commun. 1970, 847-848.
(23) Ojima, I.; Kogure, T.; Nihonyanagi, M.; Kono, H.; Inaba, S. Chem. Lett.
1973, 501-504.
(24) Corriu, R. J. P.; Moreau, J. J. E. J. Organomet. Chem. 1976, 114, 135-
144.
(25) Corriu, R. J.; Moreau, J. J. E. J. Chem. Soc., Chem. Commun. 1973, 38-
39.
(26) Blackburn, S. N.; Haszeldien, R. N.; Parish, R. V. J. Organomet. Chem.
1980, 192, 329-338.
(27) Blackwell, J. M.; Foster, K. L.; Beck, V. H.; Piers, W. E. J. Org. Chem.
1999, 64, 4887-4892.
(28) Sommer, L. H.; Lyons, J. E. J. Am. Chem. Soc. 1969, 91, 7061-7066.
(29) Lukevics, E.; Dzintara, M. J. Organomet. Chem. 1985, 295, 265-315.
(30) Chung, M.-K.; Ferguson, G.; Robertson, V.; Schlaf, M. Can. J. Chem. 2001,
79, 949-957.
9
J. AM. CHEM. SOC. VOL. 124, NO. 35, 2002 10509

A R T I C L E S
Chung et al.
Table 1. Isolated Yields for the Selective Monosilylation of 6-OH of Methyl Hexopyranosides with PdX2 (X ) Cl-, OAc-) 3 mol %)/
t
tBuMe2SiH and BuMe2SiCl/Base
amount of
TBDMS-H
(equiv)
yield with
TBDMS-H/PdX
yield with
TBDMS-Cl/base
(%)
catalyst
precursor
temp
(°C)
2
entry
sugar substrate
(%)
1
2
3
4
5
6
methyl R-D-Glu (1)
methyl â-D-Glu (2)
phenyl â-D-Glu (3)
methyl R-D-Man (6)
methyl R-D-Gal (7)
methyl â-D-Gal (8)
Pd(OAc)2
PdCl2
Pd(OAc)2
Pd(OAc)2
PdCl2
25
25
35
25
25
35
3.5
1.3
1.2
1.3
1.6
1.3
94
71
78
73
62
67
95;a 76;b 90c
74d
81e
84;c 77f
92c
Pd(OAc)2
72;g 77f
a Reference 32. b Reference 10. c Reference 12. d Reference 31. e This work, 1.2 equiv of TBDMS-Cl used, base ) imidazole. f Reference 33. g Reference
34.
Table 2. Isolated Yields for the Regioselective Disilylation of Methyl/Phenyl Hexopyranosides with the PdCl2 (3 mol %)/tBuMe2SiH System
t
(Results for BuMe2SiCl/Base Method in Parentheses Taken from Halmos et al.11 unless Indicated Otherwise)
yield of disilylated derivatives (%)
amount of TBDMS-H (mol equiv)
(TBDMS-Cl in parentheses)
yield of 6-position
monosilylated derivatives (%)
yield of trisilylated
derivatives (%) 2,3,6 and 2,4,6 silylated sugars (%)
total yield of
entry
sugar substrate
2,6
3,6
4,6
1
methyl R-D-Glu (1)
2.4 (2.2)
3.3 (3.2)
2.4 (2.2)
3.3 (3.2)
2.4 (2.2b)
3.3c
2.4 (2.2)
3.3 (3.2)
2.4 (2.2)
3.3 (3.2)
2.4 (2.2)
3.3 (3.2)
<5 (2)
<5
<5 (31)
<5 (-)
15 (-)
20
23 (-)
20 (-)
8 (-)
15 (-)
23 (-)
14 (-)
18 (70)
17 (42)
19 (20)
19 (28)
13 (29)
14
63 (11)
65 (15)
62 (20)
62 (29)
63 (48)
58
5 (-)
5 (-)
6 (-)
6 (1)
5 (-)
5
3 (9)
5 (18)
3 (-)
3 (-)
1 (17)
1
- (9)
- (19)
6 (10)
3 (14)
2 (8)
89 (92)
92 (75)
90 (71)
90 (74a)
96 (94)
97
95 (97)
90 (75)
88 (97)
92 (98)
92 (86)
96 (80)
2
3
4
5
6
methyl â-D-Glu (2)
phenyl â-D-Glu (3)
methyl R-D-Man (6)
methyl R-D-Gal (7)
methyl â-D-Gal (8)
9 (50)
62 (33)
1 (5)
11 (40)
36 (66)
35 (84)
30 (35)
59 (16) - (-)
39 (21) - (-)
40 (-)
- (-)
37 (43) - (-)
- (-)
36 (31d) 43 (-)
3 (49)
a Includes some tetrasilylated product. b This work. c Not determined for TBDMS-Cl. d 2,6-/3,6-Isomers not separated by Halmos et al.11
metal once all silane had been consumed. The precipitation of
the catalysts prevents a direct reuse of the palladium colloid,
but the metal can be fully recovered by filtration as the first
step of the very simple workup procedure. Concentration of the
solvent in vacuo directly gives the silylated sugars, which were
then separated into their respective isomers by flash column
chromatography or, in a few more difficult cases, by preparative
HPLC.
To allow a comparison of the classical silyl chloride with
the new silane alcoholysis method for the sugars listed in Chart
1, those substrates (mainly the pentose derivatives) for which
no literature data were available were also silylated using
TBDMS-Cl/imidazole, closely following the procedure de-
scribed by Halmos et al.11 Switching the solvent from N,N-
dimethylformamide (DMF) to DMA had no effect on yields or
product distributions.
Six-position monosilylated and disilylated sugars are acces-
sible with the silane alcoholysis method, with the product
distributions between the two being primarily dependent on the
amount of silane employed. Table 1 gives the conditions for
the monosilylation of the six hexopyranosides investigated and
compares them to yields achieved by previously published
procedures.10,12,31-34 The monosilylation reactions were empiri-
cally optimized by variation of three parameters: reaction
temperature, equivalents of silane, and catalyst precursors. In
four of the six cases, use of the somewhat less reactive Pd-
(OAc)2-derived catalyst (vide infra) proved to be the key to
limiting the reaction to the 6-monosilylated sugars as the main
products. In particular for 1, the monosilylated product 1a (see
Chart 1 for the numbering scheme) was isolated with this
catalyst, even if a large excess of silane was used. In all cases,
the remainder of the materials isolated consisted of unreacted
starting material and smaller amounts of disilylated sugars, both
of which were readily separated from the desired 6-monosilyl-
ated material by flash column chromatography. No monosilyl-
ated sugars with silylation in a position other than O-6 were
observed; i.e., the sterically most accessible primary alcohol
function is always silylated first by either method.
When the reactions were carried out with 2 equiv or more of
TBDMS-H, the disilylated sugars consistently become the
dominant products. Using 3.3 rather than 2.4 equiv of silane
had only a minor impact on the overall yields and product
distributions, and only very small amounts of trisilylated
products were observed. Table 2 summarizes the results of the
silane alcoholyis reaction for the six methyl and phenyl
pyranosides employed and compares them to those (given in
parentheses) obtained by Halmos et al. and ourselves through
the classical silyl chloride method. The identity of the isomers
was in each case established by 1H-COSY, 13C-JMOD, and 1H/
13C-HSQC NMR spectroscopy (see Experimental Section and
Supporting Information for exhaustive NMR data and spectra).
The most notable feature of the silane alcoholysis reaction is
that for the first four entries in Table 2, i.e., the glucose
derivatives 1-3 and the mannose derivative 6, the product
distributions of 2,6- and 3,6-disilylated sugars are complemen-
tary to those obtained through the silyl chloride/imidazole
methodsfor the former the 3,6- and for the latter the 2,6-
disilylated pyranoside is the main product. A regioselective 3,6-
silylation of 6 was also achieved by Halmos et al.11 by using a
combination of triethylamine and N,N′-dimethyl-4-aminopyri-
(31) Brandstetter, H. H. v.; Zribal, E. HelV. Chim. Acta 1980, 63, 327-343.
(32) Johnson, D. V.; Taubner, L. M. Tetrahedron Lett. 1996, 137, 605-608.
(33) Mark, E.; Zbiral, E.; Brandstetter, H. H. Monatsh. Chem. 1980, 111, 289-
307.
(34) Villalobos, A.; Danishefsky, S. J. J. Org. Chem. 1990, 55, 2776.
9
10510 J. AM. CHEM. SOC. VOL. 124, NO. 35, 2002

Palladium Nanoparticle-Catalyzed Silylation of Pyranosides
A R T I C L E S
regioisomer distribution and overall yield. As Figure 1 shows,
the only notable effect of raising the temperature from 25 to 60
or 75 °C was a slight increase in the amount of trisilylated sugar,
while the other isolated yields varied only within estimated the
error range of the isolation procedure (e5%).
We hypothesized that steric interactions between the
6-OTBDMS and/or the anomeric methyl group on the sugar
and either the active silylation species in the TBDMS-Cl/base
system (most likely [TBDMS-base]+) or the palladium metal
surface in the TBDMSH/Pd(0) system could be origin of the
observed differences in regioselectivity between the two meth-
ods, which should thus disappear if these interactions could not
occur. We therefore carried out a comparative study of the effect
of the presence of the 6-OTBDMS and anomeric substituents
on any regioselectivities in the silylation of the three secondary
ring hydroxyl functions using the two series methyl â-D-
glucopyranoside (2) f methyl â-D-xylopyranoside (4) f 1,5-
anhydro-xylitol (5) and methyl R-D-galactopyranoside (7) f
methyl â-D-arabinopyranoside (9) f 1,5-anhydro-arabinitol (10).
The second series is valid, because arabinose is known to exist
primarily in the 1C4 conformation.1 Methyl â-L-arabinopyrano-
side is then formally derived from methyl R-D-galactopyranoside
by replacing the 6-CH2OH function with hydrogen. Since all
other reagents are nonchiral, the enantiomeric relationship
between this enantiomer and the actually employed methyl â-D-
arabinopyranoside is irrelevant for the purpose of this study.
Figure 2 clarifies this stereochemical relationship.
Figure 1. Product distribution and yields of the silylation of methyl R-D-
glucopyranoside (1) with the TDBMS-H/PdCl2 system as a function of
temperature.
dine (DMAP) instead of imidazole as the base. Through control
reactions they established that for this sugar the initial product
is, in fact, the 3,6-silylated isomer, which under the imidazole
reaction conditions isomerizes to the 2,6-silylated isomer, a
process that the authors relate to the cis relationship of the OH-2
and OH-3 groups in 6. In the presence of NEt3/DMAP, the
isomerization is suppressed, and 6b is obtained as the main
product. Under the same conditions (i.e., NEt3/DMAP/TBDMS-
Cl), Halmos and co-workers reported that for the glucose and
galactose substrates 1, 2, 7, and 8, both the 2,6- and 3,6-isomers
are formed in ∼40% yield in each case; i.e., essentially no
regioselectivity and also no isomerization was observed. Control
experiments under our base-free silane alcoholysis conditions
also showed no isomerization for any of the sugars, including
6b and 6c. Applying the silane alcoholysis method to the two
methyl D-galactopyranosides 7 and 8 (entries 5 and 6 in Table
2), the 3,6-isomers were still the more abundant products
formed; however, the regioselectivity bias of 3,6 over 2,6 was
only marginal (e7%) and only for sugar 7 exceeds the yield of
3,6-isomer obtainable through the silyl chloride method. With
all other parameters unchanged, the much diminished relative
regioselectivity for either silylation method must therefore be
a consequence of the axial orientation of OH-4 in the galactose-
derived substrates.
Table 3 summarizes the results of silylations by both methods
for the two series. The first three entries show a clear preference
for 3-silylation by the silane alcoholysis method, while the silyl
chloride method gave mixtures of the 2- and 4-silylated sugar
as the main product. In fact, for 9 the preference for 3-silylation
is much higher than for the parent compound 7 in the galactose-
derived series (see entry 5 in Table 2). The 2- and 4-silylated
sugars 5b and 5d are enantiomers to each other and thus cannot
be separated by conventional chromatography. Preparative
HPLC allowed us to separate the other 2- and 4-silylated sugars
4b and 4d and 9b and 9d, respectively, and unambiguously
assign their structures by NMR spectroscopy. Unfortunately,
we were not able to separate compounds 10a-c by HPLC or
other chromatography techniques, leaving the comparison
incomplete for the second series of sugars. The NMR spectrum
of the isomeric mixture of 10 is too complex for meaningful
With 1 as the model compound and PdCl2 as the catalyst
precursor, we also tested the temperature dependence of
Figure 2. Stereochemical relationship between methyl R-D-galactopyranoside (7) and methyl â-D-arabinopyranoside (9).
9
J. AM. CHEM. SOC. VOL. 124, NO. 35, 2002 10511

A R T I C L E S
Chung et al.
Table 3. Effect of the Removal of 6-CH2OH and the Anomeric Substituents on the Product Distribution for Glucose and Galactose
t
Derivative: Silylation with 1.2 Equiv of TBDMS-H/-Cl (Isolated Yields from the BuMe2SiCl/Imidazole Method (This Work) in Parentheses)
monosilylated derivatives (%)
entry
sugar substrate
2 + 4a
3
disilylated derivatives (%)
total silylated compounds (%)
1
methyl â-D-Xyl (4)
20 [10 + 10]
(42 [14 + 28])
23 (racem.)
(41)
27 [22 + 5]
(54 [36 + 18])
57
(29)
44
(23)
43
4
(16)
18
(20)
12
(18)
8
(38)
81
(87)
85
(84)
82
(81)
42
(63)
2
3
4
1,5-anhydro-Xyl (5)
methyl â-D-Ara (9)
1,5-anhydro-Ara (10)
(9)
34; 1.1:1.0:0.8 mixture of three inseparable isomers
(25; 0.3:0.9:1.0 mixture of three inseparable isomers)
a Separated by prep. HPLC.
isomer assignments, and only the relative ratio of isomers could
be extracted from the data. Nevertheless, the sum of these
experiments leads us to conclude that steric interactions due to
the presence of 6-OTBDMS and/or the anomeric methyl group
cannot explain the observed differences in regioselectivity
between the two methods.
the silylation isomers are very small (less than 5 RT), and the
resulting equilibrium constants cluster very closely around unity.
The energies and resulting equilibrium constants in Table 5 were
calculated in vacuo, i.e., ignoring solvent effects. For the
glucose- and galactose-derived isomers, lying at the opposite
extremes of calculated ∆∆G values, we therefore recalculated
the energies, taking the presence of DMA solvent into account
through a continuum model (ꢀDMA ) 37.78, a0 ) 6.10-
6.29).35-37 Within the continuum model, the influence of the
solvent is marginal, with energy differences of 3.376 and -1.556
kcal/mol for the 1b/1c and 7b/7c pairs, respectively, resulting
in the equilibrium constants given in the last column of Table
5, which are also essentially unity. We cannot exclude the
possibility that, in a specific solvation model, larger energy
differences between the respective isomers might emerge;
however, we would anticipate them to be small, given the small
differences in topology between the disilylated isomers. Also,
such a specific model would require the consideration of defined
orientations and hydrogen bonds between the sugar substrate
and an unknown number of DMA molecules in the first and
possibly second solvation sphere of each isomer, which
constitutes an enormous challenge beyond our present compu-
tational abilities. In summary, the calculations suggest that the
observed regioselectivities do not reflect intrinsic thermody-
namic and kinetic product distributions, but under the reaction
conditions are a function of the silylation agent alone rather
than the substrate.
A second hypothesis is that the observed differences in
regioselectivities may reflect a kinetic vs thermodynamic product
distribution, with the uncatalyzed silyl chloride reaction giving
a thermodynamic and the palladium-catalyzed silane alcoholysis
reaction giving a kinetic product distribution. If this were true,
and taking into account that our control experiments established
that no isomerization takes place under the silane alcoholysis
conditions, reacting a 6-silylated sugar substrate with less than
1 equiv of TBDMS-H should yield exclusively the 3,6-silylated
product. Table 4 shows the results of this experiment for methyl
R-D-glucopyranoside (1) as the model compound. The ratio of
1c/1b isolated is very similar, whether 0.3 or 1.2 equiv of silane
was used, suggesting that the two isomers were formed
simultaneously at approximately equal rates; i.e., on the basis
of this experiment, one cannot assign a kinetic or thermodynamic
preference to either one of them.
Table 4. Product Distribution in the Pd(0) Nanoparticle-Catalyzed
Reaction of 6-tert-Butyldimethylsilyl-methyl-R-D-glucopyranoside
with 0.3 and 1.2 Equiv of TBDMS-H
yields of disilylated
sugars (%)
amount of
ratio
yield of recovered
TBDMS-H used (equiv)
2,6
3,6
4,6
3,6/2,6
starting material (%)
Using both Ph3SiH and TBDMS-H as the silane, we also
applied our method to the conformationally locked sugar
derivatives 1,6-anhydro-â-glucopyranoside and 1,3,5-O-meth-
ylidene-myo-inositol. Table 6 summarizes the results of these
experiments. With the exception of the 2-silylated and 2,4-
disilylated inositols of entry 1,38 all compounds in Table 6 are
new. Most significantly, the silane alcoholysis method gives
access to the racemic 4-TBDMS-1,3,5-O-methylidene-myo-
inositol in ∼40% yield. This substitution pattern has, to our
knowledge, previously not been achieved on this sugar and may
be of high synthetic value in the preparation of selectively
phosphorylated inositols, which play an important biological
role in cell membranes and as secondary messengers.
0.3
1.2
2.5
15
7.5
53
<1
3.0
3.5
85
10
4
To further test this notion, we determined the equilibrium
constants of the hypothetical isodesmic isomerization reaction
defined in Figure 3 for the methyl D-pyranosides 1, 4, 6, 7, and
9 by calculating their total energies and applying the van’t Hoff’s
equation. The energies in Table 5 were calculated ab initio at
the HF/3-21G* level and include zero-point and thermal free
energy corrections. No imaginary frequencies were observed
in the calculations. The differences in free enthalpy between
With Ph3SiH as the silane, the addition of 2 equiv of Proton
Sponge (1,8-bis(dimethylamino)naphthalene) per palladium
results in higher yields, as the base scavenges HCl generated
(35) Kirkwood, J. G. J. Chem. Phys. 1934, 2, 351.
(36) Onsager, L. J. Am. Chem. Soc. 1936, 58, 1486-1493.
(37) Wong, M. W.; Frisch, M. J.; Wiberg, K. B. J. Am. Chem. Soc. 1991, 113,
4776-4782.
Figure 3. Isodesmic isomerization reaction between 2-/2,6- and 3-/3,6-
silylated methyl R-D-pyranosides.
(38) Angyal, S. J. Carbohydr. Res. 2000, 325, 313-320.
9
10512 J. AM. CHEM. SOC. VOL. 124, NO. 35, 2002

Palladium Nanoparticle-Catalyzed Silylation of Pyranosides
A R T I C L E S
Table 5. Total Energies of and Energy Differences and Resulting Equilibrium Constants for the Isodesmic Isomerization Reaction between
2-/2,6- and 3-/3,6-Silylated Methyl D-Pyranosides 1, 4, 6, 7, and 9 Calculated at the HF/3-21G* Levela
b
c
d
silylated sugar
total free energy ∆G (kcal/mol)
∆∆G (kcal/mol)
K
iso at 298 K
solvent-corrected Kiso at 298 K
2,6-di-TBDMS-methyl-R-D-Glu (1b)
3,6-di-TBDMS-methyl-R-D-Glu (1c)
2-TBDMS-methyl-â-D-Xyl (4b)
3-TBDMS-methyl-â-D-Xyl (4c)
2,6-di-TBDMS-methyl-R-D-Man (6b)
3,6-di-TBDMS-methyl-R-D-Man (6c)
2,6-di-TBDMS-methyl-R-D-Gal (7b)
3,6-di-TBDMS-methyl-R-D-Gal (7c)
2-TBDMS-methyl-â-D-Ara (9b)
3-TBDMS-methyl-â-D-Ara (9c)
-1105145.36
-1105148.20
-633828.78
-633829.07
-1105142.76
-1105142.32
-1105150.92
-1105148.07
-633827.71
-633829.07
2.8458
0.2868
0.9952
0.9995
1.0007
1.0048
1.0027
0.9943
n.d.
-0.4424
-2.8494
-1.5964
n.d.
1.0026
n.d.
a In vacuo. b Energies in hartrees were calculated to five decimal places with nine significant digits and are converted to kcal/mol by multiplying with
627.51 kcal/a.u. and rounded in this table. c ) ∆G(2-/2,6-) - ∆G(3-/3,6-). d By K ) e-∆G/RT
.
Table 6. Silylation of 1,6-Anhydro-â-glucopyranoside and 1,3,5-O-Methylidene-myo-inositol with TBDMS-H or Ph3SiH and 3 mol % PdCl2 at
t
60 °C (Isolated Yields from the BuMe2SiCl/Base Method in Parentheses)a
monosilylated derivatives (%)
entry
sugar substrate
silane
amount of silane (equiv)
2-eq.
4-ax.
2,4-disilylated derivatives (%)
total yield (%)
1
1,3,5-O-methylidene-Inositol (12)
TBDMS-H
1.2
(1.0)
2.4
2.3
2.3
6
(74)
-
38 (rac.)
(-)
37 (rac.)
(-)
52 (rac.)
57 (rac.)
90
81
(74)a
91
57
90
2
3
4
1,3,5-O-methylidene-Inositol (12)
1,3,5-O-methylidene-Inositol (12)
â-1,6-anhydro-glucopyranose (11)
TBDMS-H
Ph3SiHb
39 (rac.)
-
Ph3SiHb
-
a Reference 38; base employed is 2,6-lutidine. b Two equivalents of Proton Sponge with respect to PdCl2 added.
Table 7. Results of the Characterization of the Palladium Colloids
by TEM and DLS
during the in situ generation of the Pd(0) catalyst (vide infra),
which otherwise leads to some loss of silane from the product.
catalyst precursor
Pd(OAc)2
Nature of the Catalyst. In their seminal work, Sommer and
Lyons found that, in methylene chloride, xylene, or pentane
solvent, palladium supported on charcoal, palladium metal itself,
and bulk palladium metal, generated in situ from PdCl2 and
various trialkyl silanes as the reducing agent, all form active
catalysts.28 Employing optically active silanes, Sommer and
Lyons also established that, for the first two of these catalysts,
the alcoholysis proceeds with inversion of stereochemistry at
the silicon, while for the third the reaction leads to racemic
products, which they attributed to the presence of an acidic
racemizing agent, most likely HCl. When the catalyst is
moderated by the presence of small amounts of NEt3, complete
inversion of the silane is again observed. In both cases, the
authors note that bulk palladium metal immediately precipitates
from the solution to form the active catalyst.
PdCl2
Pd(OAc)2
large
small
TEM data
no. of particles sampled
temperature (°C)
mean size (nm)
ESD
100
25
2.04
0.57
1.03
3.4
213
25
62.21
33.74
21.83
207.71
88
80
60
25.66
6.02
12.75
42.69
3.16
0.64
1.97
5.2
min. size (nm)
max. size (nm)
DLS data
mean size (nm)
1.83
44.8
2.46
reaction of Pd(OAc)2 in the same solvent (DMA) using an
excess of NaOAc as the reducing agent, respectively, we
suspected that the catalytically active black solutions generated
with silane in DMA may, in fact, also represent a palladium
colloid, and we decided to characterize it in detail by dynamic
light scattering (DLS) and transmission electron microscopy
(TEM).
Table 7 summarizes the results of the TEM and DLS analysis
of three representative colloids formed from PdCl2 at 25 °C
and from Pd(OAc)2 at 25 and 60 °C in the presence of a 40-
fold excess of TBDMS-H. The particle size distributions were
reproducible within the error ranges indicated in the table. TEM
images of three representative samples formed under these
conditions are shown in Figure 4. From the data, the size of
particles formed is a function of the counterion and temperature.
With the chloride counterion, a nanoparticle colloid of very
small particles and narrow size distribution was obtained. Figure
5 directly compares the size distribution measured for this
We observed the same behavior when using either methylene
chloride or acetonitrile as the solvent; however, the resulting
palladium metal was not active or was only marginally active
for the further silylation of secondary hydroxyl functions in
6-TBDMS-methyl-R-D-glucopyranoside, a sugar substrate soluble
in these solvents. In contrast, we found that the addition of
TBDMS-H to PdX2 (X ) Cl-, OAc-) suspended in DMA
solvent did not lead to the precipitation of bulk metal. Instead,
black solutions of homogeneous appearance were formed, which
in the presence of excess silane are catalytically highly active
for the silylation of DMA soluble sugars and stable for months
with no precipitation of agglomerates of palladium metal, even
when stored at - 20 °C.
Prompted by earlier reports by Fowley et al.39 and Reetz et
al.40,41 that described the formation of palladium colloids either
by the reaction of homogeneous solutions of Pd(hfacac)2 (hfacac
) CF3COCHCOCF3) in heptane, diethyl ether, methylene
chloride, or acetone with Et3SiH and other silanes or by the
(39) Fowley, L. A.; Michos, D.; Luo, X.-L.; Crabtree, R. H. Tetrahedron Lett.
1993, 34, 3075-3078.
(40) Reetz, M. T.; Maase, M. AdV. Mater. 1999, 11, 773-777.
(41) Reetz, M. T.; Westermann, E. Angew. Chem., Int. Ed. 2000, 39, 165-168.
9
J. AM. CHEM. SOC. VOL. 124, NO. 35, 2002 10513

A R T I C L E S
Chung et al.
Figure 4. (a) TEM image of a colloid formed from PdCl2 at 25 °C. (b) TEM image of a colloid formed from Pd(OAc)2 at 25 °C. (c) TEM images of a
colloid formed from Pd(OAc)2 at 60 °C. Representative smaller particles on top, larger ones at bottom. The top image is a direct enlargement of a section
of the lower one.
data sets indicates that the appearance of the TEM images is
not an artifact of the sample preparation, but a genuine
representation of the catalyst colloid in solution. When Pd(OAc)2
was used as the colloid precursor at room temperature, much
larger particles, as shown in Figure 4b, with a much wider size
distribution were obtained, in which most of the larger particles
visually appear to be agglomerates of smaller ones of 20-40
nm diameter. At 60 °C, Pd(OAc)2 resulted in a colloid that
showed a bimodal particle size distribution, with the smaller
particles, shown in the enlarged section of Figure 4c, by TEM
being close in size to those formed from PdCl2. The DLS results
for the acetate-derived colloid formed at room temperature do
not show good agreement with the TEM data, which, as already
pointed out by Watzky and Finke,42 illustrates the fact that DLS
is essentially a statistical method that gives reliable results only
for monodisperse sets of particles. A much better agreement of
Figure 5. Comparison of particle size distributions obtained by TEM and
DLS in a Pd(0) colloid generated from PdCl2/TBDMS-H at 25 °C in DMA.
the DLS with the TEM data is observed for the acetate-derived
colloid formed at 60 °C, as its DLS behavior is dominated by
the much more abundant smaller particles.
colloid by TEM with those resulting from DLS. Since DLS
allows catalyst characterization under the same conditions as
the actual reaction, the excellent agreement between the two
(42) Watzky, M. A.; Finke, R. G. J. Am. Chem. Soc. 1997, 119, 10382-10400.
9
10514 J. AM. CHEM. SOC. VOL. 124, NO. 35, 2002

Palladium Nanoparticle-Catalyzed Silylation of Pyranosides
A R T I C L E S
Scheme 1
sugar products obtained by silane alcoholysis was not affected
by the type of catalyst precursor; e.g., the silylation of methyl
R-D-glucopyranoside with Pd(OAc)2-derived colloids gives,
within error range ((3%), the same yields and regioisomer
distribution as the reaction with a PdCl2-derived catalyst if the
reaction temperature is raised to 60 °C in order to compensate
for the slightly lower activity of the acetate colloid. This leads
us to conclude that TBDMS-Cl cannot be the active silylation
species and that the activation of palladium black by small
amounts of TBDMS-Cl must be a surface activation effect,
possibly the removal of an oxide coating of the palladium
powder by the thermodynamically favorable formation of
silicon-oxygen bonds. The lower reactivity of the catalyst in
the presence of base is then a consequence of the adsorption of
the base or its protonated form to the palladium surface. DABCO
should show stronger adsorption than the Proton Sponge due
to its more rigid structure and the more pronounced directionality
of the free electron pair on nitrogen. This interpretation is
consistent with Reetz’s observation of a monolayered coating
of palladium nanoparticles by tetraalkylammonium salts.43
The different particle sizes with different counterions under
otherwise identical conditions point to different nucleation and/
or growth rates as a function of the counterion present. In this
context, the fact that the colloid is stable only as long as free
silane is present suggests that the silane rather than the
counterions stabilize the colloid against agglomeration. A more
detailed study of the palladium colloid that addresses its
mechanism of formation and stabilization as well as the
elemental composition and surface structure of the nanoparticles
will be presented elsewhere.
Finally, there is a qualitative correlation between the particle
size and the observed catalytic activity; i.e., the colloid formed
from PdCl2 at 25 °C is much more active than that formed from
Pd(OAc)2 at the same temperature. In turn, the colloid formed
at 60 °C from Pd(OAc)2, containing particles close in size to
those of the PdCl2-derived colloid, showed a comparable, only
slightly lower catalytic activity. Combined with the observation
that bulk palladium metal is essentially inactive for the silylation
of sugars, this suggests that it is primarily the available surface
area that determines the overall catalytic activity.
In light of the results of the calculations (see Table 5), the
different regioisomer distributions realized by the two different
silylation methods are not intrinsically related to the sugar
structure but are a consequence of different modes of interaction
between the sugar and either the heterogeneous palladium/
solution interface in the silane alcoholysis reaction or the active
silylation reagent derived from silyl chlorides and base, most
likely of the type [R3Si-NR3′]+ in molecular disperse homo-
geneous solution.
Discussion
For the silyl chloride method, we considered two possible
origins for the observed regioselectivities: chelation of a
hypervalent silicon in the transition state of the reaction by cis-
related hydroxyl functions in the sugars or activation of
individual hydroxyl functions toward electrophilic attack by [R3-
Si-NR3]+ through intramolecular hydrogen bonds. As shown
in Figure 6, a chelation of the postulated silylating agent [R3-
Si-NR′3]+, here formulated as a six-coordinate transition state,
should be possible between any cis-related hydroxyl functions.
Eliminating the protonated base [HNR′3]+, analogous five-
coordinate intermediates as local minima along the reaction
coordinate could also be formulated. In either case, the hydroxyl
groups involved are the OH-2 and OCH3-1 oxygens for methyl
R-D-gluco- and galactopyranosides, OH-2 and OH-3 for methyl
R-D-mannopyranoside, the OCH3-1 and OH-2 in methyl â-D-
arabinopyranoside, and also the OH-4 and OH-3 in the galac-
topyranosides. A chelation with one or both ring oxygens could
also explain the high preference for 2-silylation in the inositol
derivative 12 with TBDMS-Cl, while for any conceivable trans-
1,2-diequatorial diol units such a chelation appears less likely.
The notion of a chelation-controlled regioselectivity correlates
well with the observed isomer yields in Tables 2 and 3. The
substrates 2, 3, 4, 5, and 8, in which no chelation involving
OH-2 is expected, show much less preference for the 2-/2,6-
over the 3-/3,6-isomer, typically less than 10%, while for
substrates 1, 6, 7, and 9 that can chelate through OH-2, the
2,6-isomer dominates. An overall higher reactivity of OH-2
against electrophilic attack by benzoyl and methylsulfonyl
Possible Reaction Mechanisms and Origin of the Regio-
selectivity in the Sugar Silylations. In a control experiment,
we found that powdered palladium black could be activated in
situ by a small amount of TBDMS-Cl (equimolar to palladium)
to give a moderately active silane alcoholysis catalyst, which
however did not show colloid formation. Together with the
initial observation that powdered palladium metal did not show
catalytic activity for the silylation of the sugar substrates, this
hinted that, with the PdCl2-generated colloids, chloride might
play an active role in the silane alcoholysis. In particular, we
considered the possibility that TBDMS-Cl might be formed
catalytically in a low steady-state concentration at the palladium
surface and constitute the actual reactive species, and that the
reaction could follow the path presented in Scheme 1.
Since the mechanism in Scheme 1 requires the transient
presence of free HCl, it should be inhibited by the addition of
a base. We therefore also conducted silane alcoholysis reactions
with 1 as the model substrate in the presence of 2 equiv (with
respect to PdCl2) of Proton Sponge or DABCO (1,4-diazabicyclo-
[2.2.2]octane). In the presence of the first base, the reactions
had to be heated to 60 °C in order to achieve a catalytic
silylation, and the isolated yields of 1b and 1c were somewhat
lower, at 16% and 47%, respectively (compare with the yields
in Table 2), while in the presence of DABCO only 6-monosi-
lylated products were found; i.e., base inhibited the reaction.
However, the reaction also proceeded with Pd(OAc)2, for
which an HOAc-mediated catalytic cycle analogous to Scheme
1 is unlikely, as the silicon-oxygen bond in TBDMS-OAc is
unreactive against alcohols once formed due to its high bond
energy. Also, the product distribution of regioisomers in the
(43) Reetz, M. T.; Helbig, W.; Quaiser, S. A.; Stimming, U.; Breuer, N.; Vogel,
R. Science 1995, 267, 367-369.
9
J. AM. CHEM. SOC. VOL. 124, NO. 35, 2002 10515

A R T I C L E S
Chung et al.
t
Figure 6. Possible induction of regioselectivity by hypervalent silicon chelation in the transition state of the BuMe2SiCl/base method.
Figure 7. Possible intramolecular hydrogen bonding between cis-related hydroxyl functions in pyranosides 1-12.
groups has previously also been observed for gluco- and
galactopyranosides, while for mannopyranosides OH-3 was most
reactive.44 As mentioned earlier, Halmos et al.11 found that,
depending on the base employed, either the 2,6-silylated (6b)
or 3,6-silylated mannose derivative (6c) is formed, with the latter
isomerizing to the former in an imidazole/DMF mixture, but
not in triethylamine/DMF. This observation suggests that, at
least for 6, specific hydrogen bonding between the substrate
and the various solvent/base combinations employed create large
enough energy differences between the isomers to distinguish
kinetic and thermodynamic products, overlaying the very small
energy differences presented in Table 5.
A specific solvation of the substrate through hydrogen bonds
between the sugar and the solvent and its impact on intramo-
lecular hydrogen bonds in the substrate could thus also explain
the observed regioselectivities. Intramolecular hydrogen bonds
in sugars are weak, typically on the order of 1-3 RT (i.e., less
than 1.8 kcal/mol), but are well established by 1H NMR studies,
even in DMSO-d6, which is even more polar (ꢀ ) 48.9) than
DMA (ꢀ ) 37.78)45-47 and have also been proposed on the basis
of solution-state IR48,49 and theoretical studies.50 Intramolecular
hydrogen bonds are also inferred to be the origin of regiose-
lectivity in the DMAP-catalyzed acetylation of octyl hexopy-
ranosides in CHCl3.51 A SIMPLE NMR study by Angyal and
Christofides46 concluded that, at room temperature and any given
time, only 5-10% of a dissolved sugar will have intramolecular
hydrogen bonds, reflecting competition from binding to the
solvent and the fact that their relative energies are on the order
of thermal motion in the system, i.e., RT. Following the Curtin-
Hammett principle, the low abundance of intramolecularly
bonded hydroxyls is, however, irrelevant, provided they react
faster than nonbonded ones. This would indeed be expected for
oxygen atoms in hydroxyl functions that act as proton donors
in hydrogen bonds and that should therefore display an enhanced
nucleophilicity.
Possible modes of activation of individual hydroxyl functions
by intramolecular hydrogen bonding in the substrates 1, 6, 7,
9, and 12 are shown in Figure 7. Analogous to the alternative
chelation model, the structures in Figure 7 assume that effective
hydrogen bonding is possible only between cis-related hydroxyl
functions; e.g., a hydrogen bond is likely to occur between the
OH-2 and the oxygen in OCH3-1 in the R-forms of gluco- and
galactopyranoside, but not in the â-forms. For each of the sugars
6, 7, and 9, two hydrogen bond orientations are conceivable
and shown in Figure 7, leading to a nucleophilic activation of
either O-2 or O-3 in 6, or O-3 or O-4 in 7 and 9, respectively.
The R/â-pairs of sugars 1/2 and 7/8 provide a direct test of the
hydrogen-bonding model: For both pairs, the yields in Table 2
show that there is a strong preference of 2-silylation by silyl
chloride for the R-form, but essentially no selectivity for the
â-form. The latter is also true for the phenyl â-glucopyranoside
3. The preference for 2-silylation in 6 and 9 suggests that the
hydrogen-bond-activating OH-2 (top row in Figure 7) may be
more prevalent. For the xylose substrates 4 and 5, no hydrogen
bond influence is expected due to the all-trans orientation of
the hydroxyl functions. In 4, positions 3 and 4 are preferentially
silylated, possibly reflecting steric hindrance by the anomeric
methyl group. Effectively no regioselectivity is observed with
5, in which the symmetry-equivalent positions 2 and 4 and the
(44) Haines, A. H. AdV. Carbohydr. Chem. 1976, 3, 11-110.
(45) Christofides, J. C.; Davies, D. B. J. Chem. Soc., Perkin Trans. 2 1987,
97-102.
(46) Angyal, S. J.; Christofides, J. C. J. Chem. Soc., Perkin Trans. 2 1996, 1485-
1491.
(47) Bernet, B.; Vasella, A. HelV. Chim. Acta 2000, 83, 995-1021.
(48) Bartsch, J.; Prey, V. Liebigs Ann. Chem. 1968, 717, 198-204.
(49) Zhbankov, R. G. J. Mol. Struct. 1992, 270, 523-539.
(50) Woods, R. J.; Szarek, W. A.; Smith, V. H. J. Am. Chem. Soc. 1990, 112,
4732-4741.
(51) Kurahashi, T.; Mizutani, T.; Yoshida, J.-I. J. Chem. Soc., Perkin Trans. 1
1999, 465-473.
9
10516 J. AM. CHEM. SOC. VOL. 124, NO. 35, 2002

Palladium Nanoparticle-Catalyzed Silylation of Pyranosides
A R T I C L E S
position 3 are, within error range, all silylated in just over 20%
yield. The data are incomplete for 10 due to the separation
problems noted above. In the inositol derivative 12, the
intramolecular hydrogen bond between OH-2 and the adjacent
ring oxygens indicated in Figure 7 may be the cause of the high
selectivity for 2-silylation in this substrate.
to the established silyl chloride method. The nanoparticle nature
of the catalysts is a prerequisite for the high catalytic activity
observed. In several cases, the new method gives convenient
access to isomers that are formed either only as minor
components or not at all by the silyl chloride method. Studies
to extend the method to disilanes and to explore its tolerance
to the presence of various functional groups in the sugar are
presently under way.
A rational explanation for the origin of the observed 3-/3,6-
preference with the silane alcoholysis method is much more
elusive. Because the reaction occurs on the surface of palladium
nanoparticles at active sites of presently unknown structure and
elemental composition, it is exceedingly difficult to gain further
meaningful mechanistic insights, in particular with respect to
the observed regioselectivities of the sugar silylations. It is,
however, reasonable to postulate that the same silane activation
mechanism operates on the nanoparticles as on bulk palladium
Experimental Section
General. All synthetic experiments were performed under a dry
argon atmosphere by standard Schlenk-tube techniques. Sample solu-
tions for dynamic light scattering (DLS) and transmission electron
microscopy (TEM) were prepared inside an inert-gas drybox under
argon atmosphere.
Electron microscopy was carried out on a LEO 912AB operating at
100 kV with a liquid nitrogen anticontaminator in place. Digital images
were collected using a 1K × 1K PROSCAN CCD camera and processed
using the measurement software in the SIS EsiVision program. Both
negatively stained (using 2% w/v uranyl acetate) and unstained samples
were imaged for measurement comparison. All figures in this paper
are of unstained samples, so that the region of electron density in each
particle is attributed to the palladium.
catalyst. As originally proposed by Sommer and Lyons,52
a
“backside” attack of the alcohol nucleophile on a silane activated
by either dissociative or nondissociative chemisorption onto the
surface leads to silicon-oxygen bond formation with inversion
at the silicon, as proven for chiral silanes by these authors.
There is a strong correlation between the position of ring
oxygens and the preferentially silylated hydroxyl function: With
the exception of galactose, it is the secondary hydroxyl function
most spatially remote from the ring oxygens in the sugar
substrates, i.e., OH-3 in the pyranoses and one of the axial
hydroxyls in 12 (i.e., OH-4), that is most reactive in the silane
alcoholysis reaction. One conceivable explanation is that the
sugar preferably reacts out of an orientation that positions the
ring oxygen as far away as possible from the palladium surface,
which automatically orients OH-3 toward the surface. The effect
could therefore originate from an electrostatic repulsion between
the nonbonding electron density on the ring oxygens and the
palladium surface. The preferential silylation of OH-3 by the
Pd(0)/silane system occurs regardless of the presence of an
anomeric methyl group and C-6 but appears to be modulated
by their presence, possibly due to steric influences. Removing
C-6 (see results in Tables 2 and 3) increases the preference for
silylation of OH-3 vs OH-2 from 3.3:1 in 2 to 5.7:1 in 4 and
from essentially no preference with a ratio of 1:1.1 in 7 to 1:2
in 9. All these considerations, however, still fail to give a rational
explanation for the lack of selectivity observed with the
galactose substrates 7 and 8, and the nature of the effect of the
axial hydroxyl OH-4 is unclear.
An electrostatic repulsion between ring oxygens and the
catalyst surface can also explain the preference for axial
4-silylation of the inositol deriative 12, but this result could
also be due to steric interactions between the methylidene bridge
and the silane-coated palladium surface.
Finally, as with the silyl chloride method, a specific solvation
of the sugar, in which the nucleophilicity of individual hydroxyl
functions is enhanced by hydrogen bonding, is conceivable, with
the presence of the solid/liquid interface imposing a different
near-order on the solvent spheres around the sugar molecules
than in the silyl chloride method that could then result in the
observed different regioslectivities.
Dynamic light scattering (DLS) was used to determine the size
distributions of the suspended colloids prepared as described below.
For each run, 100 µL of concentrated palladium colloidal suspension
was transferred to a square (optics quality) cuvette (QS, Hellma,
Concord, ON) containing 4 mL of the appropriate solvent. Incident
light (of wavelength 532 nm) was provided by a 50 mW Nd:YAG
laser (DPSS 532, Coherent, Santa Clara, CA), and light scattered at
90° was correlated using a BI9000AT autocorrelator (Brookhaven
Instruments Corp., Holtsville, NY) with software 9KDLSW Ver.1.35.
Correlation function analysis was performed to obtain number distribu-
tions that could be compared to data obtained from electron microscopy.
A detailed description of the program TRIMIE (compiled using Lahey
F95, Incline Village, NV) used for this task has been described by
Hallett et al.53 Input parameters for solvent viscosity and refractive index
were 2.141 cP and 1.438, respectively, obtained from the CRC
Handbook of Chemistry and Physics.
NMR spectra were recorded on Bruker Avance 400 and 600 MHz
instruments in deuterated solvents obtained from CIL Inc. Isomer
assignments of the silylated sugar species were made on the basis of
2D COSY, NOESY, and HSQC spectra (see Supporting Information
for details). Elemental analyses were performed by M-H-W Labora-
tories, Phoenix, AZ. Sugar substrates, tBuMe2SiH, and metal salts were
purchased from Aldrich Chemical Co. and Strem Chemicals, respec-
tively, and used as received. 1,5-Anhydro-xylitol and 1,5-anhydro-
arabinitol were prepared according to literature procedures.54 Methyl
â-D-arabinopyranoside was prepared by Fischer glycosidation with
methanol and purified by recrystallization from ethanol/ethyl acetate.
N,N-Dimethylacetamide (DMA) and N,N-dimethylformamide (DMF)
were obtained from Caledon Laboratories Ltd. or Fisher Scientific, dried
by vacuum distillation from BaO, and subsequently stored over activated
4 Å molecular sieves. Flash column chromatography was performed
on wet-packed Merck silica gel 60F at 1 psi static pressure, set by a
break-through valve at the column head. HPLC separations were
performed on a Varian ProStar system fitted with UV and RI detectors
using 41 × 250 mm columns. Details for the chromatographic
separations of the individual sugars are given in the Supporting
Information.
General Procedure for the Silylation of Sugar Substrates Using
PdX2 (X ) Cl-, OAc-) as Catalyst Precursors. A DMA (3 mL)
Conclusion
The palladium(0) nanoparticle-catalyzed silylation of sugars
by silane alcoholysis of TBDMS-H is an attractive alternative
(53) Hallett, F. R.; Craig, T.; Marsh, J.; Nickel, B. G. Can. J. Spectrosc. 1989,
34, 63-70.
(54) Regeling, H.; Zwanenburg, B.; Chittenden, G. J. F.; Rehnberg, N.
Carbohydr. Res. 1993, 244, 187-190.
(52) Sommer, L. H.; Lyons, J. E. J. Am. Chem. Soc. 1967, 89, 1521-1522.
9
J. AM. CHEM. SOC. VOL. 124, NO. 35, 2002 10517

A R T I C L E S
Chung et al.
solution of the sugar substrate (1.0 mmol) was prepared in a one-necked
round-bottom flask fitted with a Schlenk stopcock. BuMe2SiH (0.3-
suspended in dry DMA (6 mL). After 10 min, 100 µL of this solution
was diluted with 4 mL of DMA to give a sample solution with an
effective palladium concentration of 1.25 × 10-4 mol/L.
t
3.3 equiv depending on the experiment performed) was added directly
to a second Schlenk flask containing the catalyst PdX2 (3 mol % with
respect to sugar) in 3 mL of DMA and stirred for 1 min to give a
black colloidal solution. Subsequently the sugar substrate solution was
added, the apparatus connected to a gas bubbler, and the reaction stirred
until hydrogen gas evolution stopped (12-24 h) and TLC indicated
completion of the reaction. DMA was removed by rotary evaporation
at 45 °C under full oil-pump vacuum. Flash column chromatography
of the residue with ethyl acetate and hexane in various ratios removes
the catalyst and yields the silylated sugars as well separated isomers.
See Supporting Information for detailed spectral and other data on
individual compounds.
t
Preparation of Samples for TEM. BuMe2SiH (1.2 mmol) was
added to PdX2 (0.03 mmol, X ) Cl-, OAc-) suspended in dry DMA
(6 mL). After the solution was stirred for 10 min, 3 mL of sample
solution was collected. The black solution was then removed from the
glovebox and the specimen mounted on Formvar/carbon 200-mesh
copper grids by floating the grid face down on the sample suspension
for 30 s and blotting off the excess on a filter paper. To stain the
samples, they were next floated on a drop of 2% w/v uranyl acetate
for 10 s and then blotted.
Acknowledgment. Funding for this study was provided by
the Natural Science and Engineering Research Council of
Canada (NSERC), the Canadian Foundation for Innovation
(CFI), the Ontario Innovation Trust Fund (OIT), Varian Canada
Inc., and the University of Guelph. Electron microscopy was
performed in the NSERC Guelph Regional STEM Facility,
which is partially funded by an NSERC Major Facilities Access
Grant.
Silylation of Sugar Substrates Using Pd(0) Black and tBuMe2SiCl
as the Catalyst Precursor. The procedure is identical to the one above,
with the exception that Pd(0) powder is used. A suspension of Pd(0)
in DMA rather than a nanocolloid is formed. Five minutes after all
reagents had been combined, a catalytic amount (6 mol % with respect
to palladium, i.e., 0.06 mmol) of tBuMe2SiCl was added to the reaction
flask. Further procedure and workup were as described above.
Silylation of Pentopyranosides Using tBuMe2SiCl. (This procedure
is essentially identical to the one reported by Halmos et al.11) A DMA
or DMF (3 mL) solution of the sugar substrate (1.0 mmol) and imidazole
(2.5 mmol) was prepared in a round-bottomed flask with a Schlenk
Supporting Information Available: Comprehensive collec-
1
tion of 400 MHz H and 13C NMR data (COSY, HSQC) with
images of spectra for all compounds and peak assignments;
conditions for chromatographic isolation of the compounds and
elemental analysis data for new compounds and molecular
modeling coordinates for the compounds listed in Table 5 (PDF).
This material is available free of charge at http://pubs.acs.org.
t
stopcock. Another DMA (3 mL) solution of BuMe2SiCl (1.2 mmol)
was prepared in a second Schlenk flask. The silyl chloride solution
was added dropwise to the pentopyranosides and imidazole solution.
After 24 h, the reactions were worked up as above.
t
Preparation of Samples for Dynamic Light Scattering. BuMe2-
SiH (1.2 mmol) was added to PdX2 (0.03 mmol, X ) Cl-, OAc-)
JA026723V
9
10518 J. AM. CHEM. SOC. VOL. 124, NO. 35, 2002