while at 1 mmol g~1, only two acid sites per molecule, and at
2 mmol g~1 only ca. 1.5 acid sites per molecule are generated.
A possible explanation for this could be that antimony tri-
Ñuoride reacts with silanols to form some surface species,
which by polarisation of the surrounding hydroxyls give rise
to several BrÔnsted acid centres. Increasing the amount of
antimony triÑuoride decreases the amount of available
hydroxyls and hence the number of acid sites. An alternative
explanation could be the formation of SiÈOÈSbF species on
2
the surface, each accommodating three water molecules to
restore an octahedral coordination sphere in the case of the
0.5 mmol g~1 loading catalyst, while at higher loadings, Ñuo-
rine bridging surface species could lead to a decreased ability
to bind water molecules.
The nitrogen adsorption isotherms of the various HMS
supported catalysts are consistent with the structure of HMS
materials. All the prepared catalysts show a lower surface area
than that of the starting material (Table 2), which is under-
standable since the smallest pores may be blocked by anti-
mony triÑuoride. The average pore diameter should therefore
be larger than that of the parent material, which is the case
when the support was pre-treated at 120 ¡C, but not when it
was calcined. It is worth noting that, as studied on various
occasions in our laboratory, reÑuxing the supports in various
solvents does not a†ect their surface area to a signiÐcant
extent.
The catalysts show a high efficiency in catalysing the oxida-
tion of cyclohexanone to e-caprolactone. Excess cyclo-
hexanone was used to allow continuous azeotropic removal of
the water under reduced pressure. The catalyst was mixed
with cyclohexanone and heated to 70 ¡C before hydrogen per-
oxide was added drop-wise over 30 min, using a peristaltic
pump. 1H NMR was used to monitor the reaction because of
the formation of stable cyclohexanone peroxides, which
decompose in GC injectors to give caprolactone, cyclo-
hexanone and other products.13 As a consequence, it was not
possible to monitor cyclohexanone consumption (1H NMR
peaks of cyclohexanone and its peroxides overlap) and no
selectivity Ðgures could be obtained. Background reactions
with no catalyst and with underivatised HMS were carried
out. Under those conditions, cyclohexanone oxidation did not
occur to more than 2% (based on the amount of hydrogen
peroxide introduced) after 6 h. Homogeneous antimony tri-
Ñuoride did catalyse the reaction, but the kinetics of the oxi-
dation was much slower than with supported antimony
triÑuoride (Fig. 1).
Fig. 1 Rates of formation of caprolactone from cyclohexanone using
various SbF -based catalysts, at a loading of 1 mmol g~1 (yields
3
based on total amount of H O introduced).
2
2
Unfortunately, other reuse studies with di†erent batches of
HMS-E and HMS-C catalysts at various loadings failed to
conÐrm this behaviour. In every case, the catalyst gradually
lost activity from run to run, probably because of polymer
formation on the surface of the catalyst, as shown by DRIFT
and STA studies. A 19F NMR spectrum of the reaction liquor
after Ðltration did not show any Ñuorinated species, suggest-
ing that no antimony triÑuoride leached and that no Ñuori-
nated organics were produced during the reaction.
The e†ect of support treatment prior to reacting antimony
triÑuoride with its surface is very important for the catalytic
properties of the resulting material. With SbF supported on
3
calcined HMS, better yields of caprolactone can be obtained
than with SbF supported on HMS dried at 120 ¡C. Both the
3
water content of the support and the nature of its surface
before SbF grafting are important to the Ðnal catalytic
3
properties (if HMS materials behave like silica, the calcined
HMS surface should mainly present siloxane bridges,14 as
opposed to silanols for the HMS dried at 120 ¡C).
When HMS-C-SbF is used, caprolactone production stops
3
about 15 min after the end of hydrogen peroxide addition.
The decrease in caprolactone yield after that time is due to its
conversion to hydroxycaproic acid and low molecular weight
polycaprolactone (caprolactone oligomers). The analysis of
the reaction mixture at 60 min shows that its active oxygen
content is less than 1%, indicating that most of the hydrogen
peroxide has reacted.
At equivalent SbF loading and support pre-treatment,
3
HMS supported catalysts are more efficient than their Kiesel-
Variation of the quantity of catalyst (HMS-C-SbF ) at
3
gel analogues (Fig. 1), possibly because of their higher surface
area. The main by-products identiÐed were polycaprolactone,
hydroxycaproic acid and cyclohexanone peroxides.
In a set of reuse experiments, the HMS-E catalyst did not
lose its activity, but actually got slightly more active from one
run to the other, increasing the caprolactone yield from 27%
70 ¡C showed that the reaction rate is dependent on the
amount of catalyst used, but that the quantity of poly-
caprolactone and hydroxycaproic acid produced is even more
inÑuenced by this factor. Hence, larger quantities of catalyst
lead to poor caprolactone yield and high yields of hydroxy-
caproic acid and polycaprolactone (Fig. 2).
to 33% between the Ðrst and third use of HMS-E-SbF .
The e†ect of the SbF loading on the activity of the
3
3
HMS-C-SbF catalyst has also been investigated. When it is
3
increased to 2 mmol g~1, the initial rate of the reaction is
Table 2 Surface areas and average pore size diameters of the sup-
faster and the best caprolactone yield obtained at 70 ¡C is
40.3% (34.6% with a loading of 1 mmol g~1), but the amount
of by-products is also increased. On the other hand, with a
lower loading of 0.5 mmol g~1, even though the initial rate of
reaction is slightly slower, the yield of caprolactone after 45
min is comparable to that obtained with a 1 mmol g~1 cata-
lyst, and by-product formation is halved.
ported SbF catalysts
3
Loading/
mmol g~1
BET surface
area/m2 g~1
Average pore
diameter/nm
Material
Kieselgel 60
È
1
1
È
0.5
1
2
È
1
529
383.5
395
1151
835
843
712
1254
998
5.7
5.9
5.7
3.05
3
2.98
2.95
3.7
4.1
K-120-SbF
3
K-600-SbF
3
HMS-C
When the reaction is run at 90 ¡C (130 mbar), the efficiency
of the catalysts is even greater, enabling yields of ca. 80% of
caprolactone to be obtained in 45 min with a 1 mmol g~1
catalyst on HMS-C. When taking into account the amount of
polycaprolactone formed, it can be calculated that more than
88% of the hydrogen peroxide introduced (it is unfortunately
HMS-C-SbF
HMS-C-SbF
HMS-C-SbF
HMS-E
3
3
3
HMS-E-SbF
3
486
New J. Chem., 2000, 24, 485È488