Acetone condensation reaction on acid catalysts

Article abstract of DOI:10.1006/jcat.1998.2142

The condensation reaction of acetone on alumina and acid zeolites has been followed by FTIR. Under identical conditions, the reaction rate is faster on alumina, and the condensation goes beyond the formation of mesityl oxide. Zeolites without nonframework aluminum are poor catalysts. On HZSM-5 the reaction is about two orders of magnitude slower than on USY at 105°C. From these data, it appears that Lewis sites, even if they bound acetone less energetically than Bronsted sites, are responsible for the activation of the molecule. On alumina, the reaction would take place between gas phase acetone and acetone adsorbed on Lewis sites. On zeolites with nonframework aluminum and, thus, with Lewis sites, the reaction would involve acetone molecules adsorbed on Bronsted and Lewis sites, the activation occurring on the Lewis site.

Full text of DOI:10.1006/jcat.1998.2142

JOURNAL OF CATALYSIS 178, 188–197 (1998)
ARTICLE NO. CA982142
Acetone Condensation Reaction on Acid Catalysts
A. G. Panov and J. J. Fripiat1
Department of Chemistry and Laboratory for Surface Studies, University of Wisconsin-Milwaukee, Milwaukee, Wisconsin 53201
Received December 22, 1997; revised March 23, 1998; accepted April 28, 1998
seems to play a primary role in the bimolecular condensa-
tion (10). When a localized complex between acetone and
Brønsted acid sites is formed, the transformation in mesityl
oxide does not occur. By increasing the coverage or the
temperature of the system (100ꢀC), acetone dehydrates to
form mesityl oxide. The diacetone alcohol, the aldol dimer-
ization product of acetone is not observed (6–10).
The present study has been preceded by a paper dealing
mainly with the assignment of the spectral features of ad-
sorbed acetone and mesityl oxide. Mesityl oxide is easily
identified as the first observable condensation product. It
is, by the way, amusing that this identification is relatively
recent. For instance, Lercher (1), who has extensively stud-
ied the adsorption of acetone on several oxides and mixed
oxides using IR has wrongly assigned to carboxylate, bands
which, in fact, were due to mesityl oxide. Table 1 shows the
band assignments essential for the present work. It summa-
rizes the conclusions of the preliminary paper (14).
The condensation reaction of acetone on alumina and acid ze-
olites has been followed by FTIR. Under identical conditions, the
reaction rate is faster on alumina, and the condensation goes beyond
the formation of mesityl oxide. Zeolites without nonframework alu-
minum are poor catalysts. On HZSM-5 the reaction is about two or-
ders of magnitude slower than on USY at 105ꢀC. From these data,
it appears that Lewis sites, even if they bound acetone less energeti-
cally than Brønsted sites, are responsible for the activation of the
molecule. On alumina, the reaction would take place between gas
phase acetone and acetone adsorbed on Lewis sites. On zeolites with
nonframework aluminum and, thus, with Lewis sites, the reaction
would involve acetone molecules adsorbed on Brønsted and Lewis
c
sites, the activation occurring on the Lewis site.
ꢁ 1998 Academic Press
INTRODUCTION
Numerous studies have been devoted to the adsorption-
transformation ofacetone on oxides(1–3) and acid catalysts
(4–10). IR (1–7) and NMR (8–13) techniques were used to
describe adsorption complexes of acetone on surface sites.
The formation of an 1 : 1 adsorption complex between ace-
tone and Brønsted acid sites of HZSM-5 was reported by
several research groups (9, 12). Biaglow et al. (11) pointed
out that localized adsorption complexes exist on the surface
of acidic ZSM-5 zeolites at low coverages (less than one ace-
tone per one aluminum atom in framework position). At
higher coverages, acetone becomes more mobile which is
proven by the disappearance of spinning sidebands, present
at lower coverages, in the 13C NMR spectrum.
Both IR and NMR technique were also used to study the
reactivity of acetone complexes with zeolite acid sites to-
ward aldol condensation (6–10). The chemistry of aldol con-
densation of acetone over a variety of zeolites at coverages
higher than one acetone per Brønsted site and different
temperatures was detailed by Haw and co-workers (9). For
HZSM-5, formation of mesityl oxide was observed even at
room temperature. Biaglow et al. (10) studied the transfor-
mation of acetone over HZSM-5 zeolites at different cov-
erages and temperature. The mobility of adsorbed acetone
==
==
Besides the C O and C C stretching vibrations which
are the most important for this work, CH bendings are
also observed between 1470 and 1300 cmꢂ1. A band at
ꢃ1450 cmꢂ1 is characteristic of mesityl oxide. A shoulder at
1471 cmꢂ1 seems to be the result of the ring vibration of
mesitylene (14).
The assignment of a band at ꢃ1700 cmꢂ1 to acetone (Ac)
interacting weakly with silanol comes from the observation
that the intensity of the zeolite OH stretching at 3740 cmꢂ1
decreases as absorption of acetone at ꢃ1700 cmꢂ1 increases
(14). This band is observed easily in USY and HZSM-5
zeolites. In H-mordenites in which the extent of dealumi-
nation is negligible, the main CO stretching vibrations of
absorbed Ac are at 1679 and 1645 cmꢂ1 with widths of 28
and 49 cmꢂ1, respectively. These bands are attributable to
–
–
two different kinds of bridging Si OH Al sites with differ-
ent acid strengths, the most shifted one being the strongest
(14). These are also observable in HZSM-5 and USY.
The width and frequency distribution of the CO stretch-
ing bands of absorbed acetone support the early conclu-
sion by Hair (15) that acetone is not a good probe for acid
sites. It would be physically more meaningful to consider a
continuum in the distribution of the strengths of electron
acceptor sites with which the carbonyl interacts, the spread
1 To whom correspondence should be addressed.
188
0021-9517/98 $25.00
c
Copyright ꢁ 1998 by Academic Press
All rights of reproduction in any form reserved.

ZEOLITE, ACETONE CONDENSATION, SYNERGY
189
frequencies at 2175 ꢄ 2 cmꢂ1 or at 2163 cmꢂ1 in HZSM-5
and USY (17). These observations suggest that carbon
monoxide interacts with two kinds of OH. Indeed, the
TABLE 1
==
Assignments of the C O Stretch of Acetone (Ac) and Mesityl
Oxide (MO) Adsorbed on Brønsted (B) and Lewis Acid Sites (L) on
Acid Catalysts at 105ꢀC. Zeolites Bridging OHs Acting as Brønsted
Sites Are Designated as B1 whereas the Weakly Acidic Bridging
OHs Are Designated as B2. WHH: Full Width at Half-Height
–
–
bridging Si OH Al are distributed among three popu-
lations, depending upon the number n of next-nearest-
neighbor Al atoms (0 < n < 3). Isolated Al sites (n = 0) are
bearing the most acidic OH (18, 19). The first and second
population (n = 1) are in cluster Si–1Al while the third one
is in cluster (Si–nAl with n > 1), either Si–2Al. . .Si–4Al.
Only the first and second populations are actual Brønsted
sites, in the sense that upon exposure to NH3 they trans-
form into NH4. The OHs in cluster (Si–nAl), n > 1 are not
Brønsted sites (no proton transfer to NH3) but they can
hydrogen-bond NH3.
Acetone
==
C
O stretch
(cmꢂ1
WHH
Sample and sites
)
(cmꢂ1
)
Physisorbed
SiOH
L
B2
B1
1714–1720
1700–1705
1690–1702
1679–1682
1652–1655
12–15
22
27
27
44
The B1 band at ꢃ1650 cmꢂ1 has been assigned to the
acetone carbonyl in a H-bonded complex with the most
acidic OH sites. The linear regression between the number
of acetone in the B1 band and the number of Brønsted sites
Mesityl oxide
==
==
C
O stretch
(cmꢂ1
WHH
(cmꢂ1
C
C stretch
(cmꢂ1
WHH
Sample and sites
)
)
)
(cmꢂ1
)
==
==
supports this agreement (14). The C O and C C band as-
signments of absorbed MO shown in Table 1 are in good
agreement with those in Ref. 6 (see two bottom rows) which
were obtained for Brønsted sites in HZSM-5. The bands
assigned to Lewis sites in Ref. 6 were obtained on a cal-
cined HZSM-5, and their frequencies are 24 cmꢂ1 lower
than those we have obtained on amorphous alumina or on
nonframework aluminas in USY.
B HMa
1617
1631
43
1557
1565
71
and HZSM-5b
B1; HYc
24
57
and B1; USYd
B2; HYe
1659
1632
1673
1676
18
30
21
21
1594
1565
32
55
42
42
B; USY
L; Aluminae
L; (USY)
1589–1603
1604
Now, turningour attention toward the catalyticactivation
of acetone, because the transformation of acetone by liquid
acids is well documented (20), the role of Brønsted acid
sites has been particularly emphasized (9, 10, 13). However,
the role of Lewis sites deserves some attention because it
is difficult to prepare an acid zeolite without having some
dealumination. Lewis sites are present on nonframework
alumina within the zeolite microcrystals (21). For example,
in hydrocarbon cracking and isomerization, the presence
of nonframework alumina (e.g., the presence of Lewis acid
sites) affects the catalytic activity (22).
The aim of this paper is to reexamine the condensation
reaction. The reason for reexamining the problem stems
from the observation that on the surface of alumina, rich in
Lewis sites, condensation of acetone occurs very rapidly at
room temperature. It even proceeds to a very limited extent
at ꢂ60ꢀC. At room temperature, the reaction goes beyond
the formation of mesityl oxide because spectral features,
which could be caused by isophorone and aromatics have
been observed. Therefore, Lewis sites appear to be at least
as active as Brønsted sites, yet, it seems that this aspect has
been neglected in recent times.
a H-mordenite.
b HZSM-5.
c H near faugasite Y.
d Ultrastable Y.
e Nanosized alumina. SiOH represents the nonbridging OH group; its
OH stretch is usually between 3700 and 3750 cmꢂ1
.
of stretching frequency being from about 1700 to 1650 cmꢂ1
Lewis, as well as Brønsted sites, are included in this broad
distribution.
As stated earlier, mesityl oxide is a better probe than
acetone for acid sites because of the larger frequency shifts
.
==
==
of the conjugated C C, C O with respect to the electron
acceptor site strength. The relative shifts of the CO and CC
vibrations are affected to the same extent by this interac-
tion, but the bands are better resolved because they are
more spread out than in acetone.
The shifts of the C O and/or C C bands scale as the
strength of the Brønsted sites. The shifts in HZSM-5 and
HM are the largest, whereas in HY they are the smallest.
This is in agreement with activity for the steady-state trans-
formation of pentane. On HZSM-5, HM, USY, and HY pre-
treated at 500ꢀC, the activity scales as 100, 18.5, 13, and 0.85,
respectively (16). Earlier studies on the low-temperature
adsorption of carbon monoxide on acid or acid and dealu-
minated zeolites had shown two CO stretching bands with
==
==
Before continuing, we want to outline two apparently im-
portant omissions in the present contribution, which con-
cern either the formation of enol or of higher condensation
products. We have not been able to characterize enol, in
spite of a careful search. Moreover, we disagree with sev-
eral of the assignments of bands, which have been attribute

190
PANOV AND FRIPIAT
to enol and, which, according to us, belong to either mesityl
oxide or acetone. For instance, Hanson et al. (2) suggested
that a band at 1596 cmꢂ1 is caused by acetone enolate. We
TABLE 2
Number of Acid Sites on the Catalyst Used Here: L: Lewis and
Brt: Total Brønsted Obtained from NH3 Chemisorption (mmol/g)
(21, 24)
==
believe that this band is caused by the C C stretching of
MO on Lewis sites because it develops progressively with
Catalyst
Alumina
L
Brt
NMR Si/Al
a band at 1673 cmꢂ1 (C O on Lewis sites) upon adsorbing
==
MO on alumina.
0.4
—
—
4.9
5
Because the reaction temperature in this work never ex- USY
1.43
1.32
0.78
1.42
0.98
0.62
ceeds105ꢀC, the formation ofhigher condensation products
must have been limited. On alumina according to V. Bell
(8), the formation of isophorone occurs after 1.5 h at 72ꢀC,
and it is only above 200ꢀC that mesitylene appears. Hanson
HY(580)
USY5F
DHM700 and
VG700
USY10F
14
0.82
0.33
0.38
n/a
0.83
0.23
1.25
>1.6
0.83
14
59
et al. (2) have observed a strong IR band at 1638 cmꢂ1 on HM
9.8
HY
HZSM5
(2.55)a
17.5
adsorbing isophorone on ꢀ -alumina at room temperature.
ꢃ0
In several of our spectra, for instance, in adsorbing MO on
alumina, a shoulder is observed at that frequency, and it is
not that of an impurity. It could be caused by relatively
small amounts of physically adsorbed MO because it is
at ꢃ1635 cmꢂ1 that the CO stretching of MO shows up.
Among the objectives of this study, the determination
of the nature of the active sites is a first priority. The pri-
mary condensation reaction is a bimolecular reaction, and
as mentioned by others (6, 10), it has been suggested that it
occurred between acetone adsorbed on Brønsted sites(acti-
vated carbonyl group) and mobile (eventually physisorbed)
acetone. In the presence of Lewis sites, it could well happen
between acetone adsorbed on Brønsted sites and acetone
adsorbed on Lewis sites, the latter adsorption being much
less energetic than the first.
a Dealumination assumed to be negligible.
by 20 Torr oxygen at the same temperature. The 29Si NMR
spectra were recorded on samples reequilibrated with at-
mospheric moisture.
Reagents. Acetone (94.9% ) and mesityl oxide (99% )
were Aldrich products which were dried over 4A molec-
ular sieve (Ac) or MgSO4 (MO) and purified by repeated
freezing and thawing.
Instrument. The FTIR spectra were obtained with a
Perkin Elmer 1800 operating in the absorbance mode. The
spectral resolution was 2 cmꢂ1. The stainless steel cell is fit-
ted with KBr windows and can be outgassed up to 600ꢀC
or it can be cooled at ꢂ190ꢀC. The weight of the pellets is
about 14 mg/cm2. All spectra were recorded at the reaction
temperature.
A quadrupole mass spectrograph, linked to the infrared
cell by a leak valve, permits us to monitor the formation of
water or the disappearance of acetone from the gas phase.
EXPERIMENTAL
Catalysts. All catalysts used for the adsorption studies
have been described earlier. The nanosized amorphous, rich
in coordination defects, alumina has been prepared by a
sol–gel process (23) and calcined at 600ꢀC. Its Lewis acidity
has been carefully characterized (24–26). The total num-
The Reaction. The reaction temperature has been, in
ber of Lewis sites is 0.4 meq/g, and the surface area is be- all cases, 105ꢀC. Two different starting amounts of reagent
tween 300 and 350 m2/g. USY is a commercial (PQ) catalyst. were used. The so-called large dose corresponded to a
USY5F and USY10F are the results of further USY dealu- number of acetone molecules (ꢃ3 mmol/g) larger than
mination using NH4F (29). In USY5F both the numbers of the number of Brønsted sites (Table 2). The small dose
Brønsted and Lewis sites are reduced to less than 50% of (ꢃ0.4 mmol/g) is always smaller than the number of
the initial content in USY, whereas in USY10F the reduc- Brønsted sites. The large dose is larger than the sum of
tion reaches about 80% . The HZSM-5 sample is from PQ, Brønsted and Lewis sites.
and it was prepared as indicated in Ref. 24. The prepara-
The FTIR spectra were collected every 20 s in the case
tion and characterization of HM and DHM700 have been of alumina (4 scans) (fast reaction) and every 5 min in the
described in Ref. 26.
case of zeolites (20 scans) (slower reaction).
The formation of mesityl oxide was monitored in decon-
Table 2 shows the Si/Al ratios obtained from 29Si NMR
and the numbers of Lewis and Brønsted acid sites. In order voluting the spectra in order to record the integrated inten-
==
to minimize the dealumination process and, therefore, the sity of the C C stretch, which is observed between 1557
number of Lewis sites, HY, HM and HZSM5 were deammo- and 1565 cmꢂ1 on the B1 Brønsted sites or at ꢃ1600 cmꢂ1
niated at 400ꢀC in the IR cell before use. The dealuminated on Lewis sites. Examples of deconvolution and values of
samples were dried in the IR cell by outgassing at 500ꢀC the extinction coefficients were given in the preliminary
for 5 h under a 10ꢂ5 Torr residual pressure and pretreated paper (14).

ZEOLITE, ACETONE CONDENSATION, SYNERGY
191
==
It should be emphasized that the C C stretch in fur-
On alumina the reaction is fast even in the presence of
ther condensation products such as phorone could overlap smalldoses(Fig. 1A), whereason zeolite without nonframe-
==
with the C C bands assigned to mesityl oxide. Adsorbed work aluminum (Fig. 2B), there is practically no reaction. If
isophorone has been identified by a band at at 1638 cmꢂ1 the zeolite contains nonframework aluminum (NFAl), the
(2). The low vapor pressure of phorone makes it difficult to reaction rate is measurable even with small doses (Figs. 1B
study its spectrum in the adsorbed state. Condensation in and 2A). With large doses, the reaction proceeds quite
aromatics is also possible as suggested by Xu et al. (9).
rapidly on zeolites (Figs. 3 and 4). Still the absence of NFAl
Because the discussion will focus mainly on the initial slows down the reaction (Fig. 3B). On alumina, the reaction
rate of reaction and because the temperature did not exceed rate is too fast to be measured with large doses at 105ꢀC.
105ꢀC, it can be assumed that, for zeolites, the amount of
Note that, sometimes, the equilibrium between the gas
higher condensation products is negligible, with respect to phase and the solid is not reached at the time the initial
that of the main reaction product (e.g., mesityl oxide).
The consumption of acetone was followed by the de-
spectrum is taken. This is the case in Figs. 1B, 2A, and 4B.
The frequencies of the Ac CO and of the MO CO and
==
crease of the acetone C O stretching bands. The integrated CC stretching vibrations are indicated in the figures. They
intensities, converted into concentration using the extinc- differ slightly from those shown and assigned in Table 1
tion coefficients of CO on L, B1 and B2 and eventually SiOH corresponding to room temperature spectra. The shift of
ꢂ1
==
sites were obtained through the deconvolution of the com- the Mo C C band, for instance, is about 8 cm toward
plex band into three components (14). In fact, because the lower frequencies between room temperature and 105ꢀC
CO stretch of acetone on Lewis sites and on SiOH overlap, on USY.
==
only three components were used. The sum of the concen-
On alumina (Fig. 1) besides the line assigned at the C C
trations in acetone adsorbed on these sites is called Acs stretch of MO on Lewis sites (at ꢃ1591 cmꢂ1), two bands
(s of adsorbed acetone, whatever the site).
are growing at 1620 and 1471 cmꢂ1, respectively. A weak
band at ꢃ1471 cmꢂ1 is observed in zeolites with large con-
tents in NFAl (Figs. 3A and 4A), but the corresponding
condensation product is formed in much smaller amounts.
Water is a product of the reaction which can poison the
RESULTS
Figures 1–4 show examples of the evolution of the FTIR
spectra over time during the reaction of large and small catalyst by reacting with Lewis acid sites. Not only does
doses of acetone at 105ꢀC on different catalysts. These H2O chemisorb on L sites where it competes with acetone,
eight sets of spectra are representative for all the experi- but also upon dissociative chemisorption it decreases the
ments which will be described and the results which will number of coordinately unsaturated aluminum in alumina
be discussed in the present contribution. From the decovo- and in the nonframework alumina species in zeolites (28).
lution of these spectra and using the specific absorbances The mass 18 signal is easily detected by the quadrupole;
reported earlier (14), follow the quantitative data reported it increases with time at the same rate as the acetone
in Figs. 5 and 6 and discussed in the last section.
43 and 58 signals disappear, but it is always smaller than
FIG. 1. Evolution of infrared spectra of small doses (see text) of Ac adsorbed on alumina (A) and on DHM700 (B) at 105ꢀC with time. In
(A) the spectra are recorded every 20 s (4 scans per spectrum) and in (B) they are repeated every 5 min (24 scans per spectrum). The initial spectrum
is always indicated by a dotted line; the final one, by a solid line.

192
PANOV AND FRIPIAT
FIG. 2. Evolution of infrared spectra of small doses of Ac adsorbed on USY (A) and on HY (B) at 105ꢀC with time. The spectra are recorded every
5 min (24 scans per spectrum).
expected. In other words, a fraction of the water is adsorbed fact, it is impossible to distinguish between the initial OH
or chemisorbed. The obvious domain of frequency to look shifted upon interaction of the various proton acceptors in
for water is in the bending deformation region (e.g., near the system and “new” OH in the stretching region.
1620 cmꢂ1). A band is observed in this region, especially
The progressofthe reaction willbe appreciated bythe de-
==
on catalysts in which the yield in MO is high (Figs. 1A, 2B, velopment of the band assigned either to C C on Brønsted
==
3A and 4A). Unfortunately, the region of the H2O bend- sites in zeolites or to C C on Lewis sites on alumina. On
==
ing mode is also that of the C O stretching of MO on bridging OH sites, as shown in Table 1, depending upon
Brønsted sites (Table 2). As a consequence, the formation the strength of the site, the band is between ꢃ1550 and
of water cannot be monitored. We have observed on USY, 1580 cmꢂ1
.
for instance, that the intensity of the 1621 cmꢂ1 band, if
assigned to MO, overestimates the yield in MO. This goes the band encompassing the C O stretch of Ac interact-
In order to obtain information on the Ac consumption,
==
==
in the right direction, if the MO C O stretching and H2O ing with the B1 and B2 sites was integrated. The integrated
bending overlap.
intensity between ꢃ1698 and 1715 cmꢂ1 was considered
Without too optimistic anticipation, we have tried to ob- as representing Ac interacting either with Lewis sites or
serve eventual new OH stretching signals above 3000 cmꢂ1 terminal silanols, namely, Ac in weaker interaction with
as reaction proceeds on alumina or zeolite with NFAl. In the surface but still adsorbed at 105ꢀC. On alumina, Ac
FIG. 3. Evolution of infrared spectra of large doses of Ac adsorbed on USY (A) and HY (B) at 105ꢀC with time. (See Fig. 1 caption.)

ZEOLITE, ACETONE CONDENSATION, SYNERGY
193
FIG. 4. Evolution of infrared spectra of large doses of Ac adsorbed on USY10F (A) and on DHM700 (B) at 105ꢀC with time. (See Fig. 1 caption.)
is adsorbed on Lewis sites only, and the CO stretch is at condensation in higher polymer is responsible for the de-
==
1695 cmꢂ1 (Fig. 1A).
crease of the MO C O band intensity. In phorone, for in-
==
The evolutions of the surface concentrations over time stance, there is one C O in the condensation product for
are shown in Figs. 5 and 6. For zeolites, the sum of the three acetone molecules, against one for two in mesityl ox-
amounts of reagent and of the reaction product (multiplied ide. It would be harzardous to speculate on variations like
by 2) on the surface reaches a constant level after about the ones just mentioned because they accumulate all the
100 min (Figs. 5A and 6B).
On alumina, the same trend is observed if the C C vibra-
experimental errors.
The difference between the sum of the concentration in
==
tion of MO is considered, but the sum of twice the surface reaction products and the dose introduced initially is the
MO concentration plus the surface reagent concentration result of what remains in the gas phase. The decrease of
starts decreasing after about 5 min reaction. A decrease the 58 molecular mass signal with respect to time, shown by
==
of the MO C O stretching band intensity occurs beyond the dotted line in Fig. 5A, indicatesthat, indeed, the reaction
5 min; it is not observed for zeolites. Most probably, the is over when the sum of the reagent plus the product reach
ꢅ
==
==
C
FIG. 5. Balance: (AcS + 2 C C) surface concentration (᭺), to be compared with the variation with time of the surface concentrations in C
(᭜) and in CO (♦) of MO and in Ac(s) (᭹). (- - - -) Ac in gas phase. (A) Alumina, (B) USY10F. The total amount of acetone introduced in the IR cell
was 0.5 mmol/g catalyst for alumina and 0.4 mmol/g catalyst for USY10F.

194
PANOV AND FRIPIAT
ꢅ
==
==
FIG. 6. Balance: (Ac + 2 C C) surface concentration (᭺), to be compared with the variation of the surface concentrations in C C (᭜) of MO
and in Ac(s) (᭹) with time. (A) USY, (B) HM. The total amount of acetone introduced in the IR cell was 3 mmol/g catalyst.
the initial amount of reagent (e.g., ꢃ0.4 mmol/g catalyst for with coordinately unsaturated aluminum (the Lewis sites)
the small dose).
kills the catalyst (28). It has been shown, for instance, that
For the large doses, there is always an excess of reagent in an alumina similar to that used in this work and calcined
in the gas phase amounting to about 20% of the total trans- at 600ꢀC, the concentration in pentacoordinated alumina
formation. There always remains a fair amount of acetone decreases from 27 to 9% upon exposure to the atmospheric
in the gas phase at the time the yield in MO levels out. The humidity at room temperature. Similar observations have
classical explanation for the leveling is that the reaction is been made for nonframework aluminum in zeolites (18, 9).
poisoned by the reaction product. Here, it would probably On the contrary, water has a negligible effect on the cata-
be more correct to say that the catalyst is product poisoned. lytic activity of the Brønsted sites (9).
Because mesityl oxide is more basic than acetone, it is more
strongly adsorbed by the acid sites, and water in reacting room temperature on Brønsted (C C stretch at 1560 cm
Figure 7A showsthat mesityloxide on USY isadsorbed at
ꢂ1
==
)
FIG. 7. (A) Adsorption of mesityl oxide at room temperature (—) or at 105ꢀC after a few minutes (– –) or after a longer time (- - - -) on USY.
(B) Increase of the C C band intensity at ꢃ1590 cmꢂ1 or decrease of the quadrupole signal at mass 58 in the gaseous phase in the presence of alumina
==
(small dose) at 105ꢀC.

ZEOLITE, ACETONE CONDENSATION, SYNERGY
195
The situation for zeolite is very different. The rate is con-
veniently represented as a linear function of the ratio
TABLE 3
Reaction Rate at Time t = 1 min for Ro; Mof (mesityl oxide yield
at the end of the reaction); and Acs [amount of adsorbed acetone
(all sites) after 1-min reaction]. VAc/VN2 Is a Measurement of the
Availability of the Porous Structure to Acetone: It is the Ratio
of Acetone Adsorbed at Saturation, at Room Temperature, to the
Volume of N2 at Saturation at ꢂ196ꢀC
(AcS)(Acw)/MOs,
[1]
as shown in Fig. 8. AcS stands for Ac strongly adsorbed by
contrast with Acw, which stands for Ac weakly adsorbed.
At first sight, the correlations are not impressive, but it must
be remembered that the fraction in Eq. [1] is obtained from
three independent measurements of AcS, Acw, and MOs.
The carbonyl stretching vibration of AcS is between
ꢃ1650 and 1690 cmꢂ1, which means (Table 1) that not only
acetone on Brønsted site B1, but also on site B2 (e.g., the
weakly acidic bridging OH groups) is involved. The car-
bonyl stretching vibration of Acw is between 1690 and
1705 cmꢂ1, thus belonging to acetone adsorbed on Lewis
Ro
(ꢁmol/g/s)
MOf
(mmol/g)
Acs
(mmol/g)
Catalyst
VAc/VN2
Aluminaa
USY
3.59
1.98
1.12
1.05
0.4
0.09
0.91
0.70
0.66
—
0.52
n/a
0.97
1.17
0.82
HY(580)
USY5F
DHM700 and
VG700
USY10F
HM
0.41
0.37
0.4
0.33
0.27
0.01
0.34
0.39
0.69
0.44
—
0.33
0.31
1.08
1.72
—
0.31
0.3
0.67
n/a
–
sites and/or the terminal Si OH. MOs is the surface mesityl
HY
HZSM5
oxide adsorbed on Brønsted sites corresponding to the
==
0.59
C
C stretch between 1550 and 1565 cmꢂ1. We may account
a Small dose; all other experiments were carried out with large doses.
for the linearity of the reaction rate in zeolite with the ratio
shown in Eq. [1] in the following way. Let Ace be an enol
which is the classical reaction intermediate (20). The Lewis
(ꢅ) and Brønsted (ᮀ) sites are covered with acetone or
Ac + ꢅ → Acꢅ and Ac + ᮀ → Acᮀ. Then
as well as on Lewis (C C stretch at 1603 cmꢂ1) sites. At
==
ꢀ
==
105 C, the intensity of the C C stretch on the Brønsted
sites increases noticeably. The mobility of MO increases at
105ꢀC, and the adsorption on the Brønsted sites is proba-
bly more energetic than on the Lewis sites. Consequently,
even if MO is formed on Lewis sites, the spectra will show
it adsorbed on Brønsted sites. Exactly what type of site
is poisoned by the mesityl oxide has to be studied when
analyzing the experimental results. As shown in Table 3,
the maximum amount of mesityl oxide obtained from the
acetone condensation reaction is smaller than the surface
concentration in acid sites.
k1
Acꢅ
→ Ace + ꢅ
[2]
[3]
Ace + Acᮀ → Pᮀ
k2
Ace + Pᮀ → Ac + Pᮀ
[4]
k3
2Ac + ᮀ → Pᮀ
[5]
Step [2] corresponds to the activation step of acetone into
enol by ꢅ. Acᮀ would be AcS. Thus, Acw could be identified
as Acꢅ. From [3] comes
DISCUSSION
==
On alumina, Fig. 7B shows that the C C surface con-
d(Pᮀ)
= k3(Ace)(Acᮀ).
centration and the intensity of the quadrupole mass signal
for mass 58 follow opposite linear variation with ln t and
reaches a constant value at t ' 2.5 min, then the system
runs out of reagent. A kinetic law in ln t suggests that the
rate obeys the Elovitch (28) relationship. Here, the rate of
adsorption and the rate of transformation are the same and
governed by an equation such as
dt
Step [3] is that suggested by Biaglow et al. (10). Step [4]
corresponds, for zeolites, to the deactivation of enol by the
reaction product. From the steady state for Ace comes
dAce/dt = 0 = k1(Acꢅ) ꢂ k2(Ace)(Acᮀ) ꢂ k3(Ace)(Pᮀ)
and finally,
dn
= a exp(ꢂꢂn),
dt
dP
ᮀ
where ꢂ and a are a function of temperature. A gas phase
molecule hitting the surface is transformed into MO after
colliding with a molecule activated by a Lewis site.
There is too little information to draw conclusions for the
reaction kinetics on alumina, except that the reaction rate
is much faster than on zeolite and the Lewis sites on the
surface of alumina activate acetone energetically.
= k1k2(Acᮀ)(Acꢅ)/[k2(Acᮀ) + k3(Pᮀ)].
dt
The latter equation accounts for the observation in
Fig. 8A, B if k2 is smaller than k3. Step [3] is the rate-
limiting step. Steps [2] and [3] represent the synergy be-
tween Lewis and Brønsted sites. Thus, the activation would

196
PANOV AND FRIPIAT
FIG. 8. Reaction rate expressed in function of the ratio Ac(w) Ac(S)/MO on USY (A), USY5F (B), HM (C), and USY10 (D). Large dose at 105ꢀC.
Two different sets of results are shown for USY in (A).
occur on Lewis sites. In zeolites, these sites would be on face density in Brønsted sites is shown in Fig. 9B. Also,
nonframework alumina.
the reaction is “stoichiometric” in the sense that water kills
This hypothesis has multiple consequences. First, there the catalytic sites. By comparing data in Tables 2 and 3, it
should be a relationship between the rate of the reaction can be seen that the maximum yield in MO, that is MOf, is
and the surface density in Lewis sites. This is suggested by always lower than the initial number of Lewis sites (ex-
Fig. 9A, whereas the absence of correlation with the sur- cept in USY5F where it is slightly higher). The synergy
FIG. 9. Reaction rate Ro (after 1 min reaction) on zeolites expressed vs. the number of Lewis sites (A) or Brønsted sites (B). The black square in
(A) is the reaction rate observed for two experiments on H-mordenite, dealuminated at 700ꢀC.

ZEOLITE, ACETONE CONDENSATION, SYNERGY
197
implies that both types of acid sites are spatially close to one
another.
3. Anderson, J. A., and Rochester, C. H., J. Chem. Soc., Faraday Trans. 1
85, 1129 (1989).
4. McManus, J. C., Harano, Y., and Low, M. J. D., Can. J. Chem. 47, 2545
(1969).
Finally, a major difference between alumina and zeolite
comes from the fact that in alumina, the gas phase acetone is
probably involved, whereas on zeolite, condensation occurs
between acetone adsorbed on two different kinds of sites.
The influence of the gas phase is indirect because of the
equilibrium between Acw and the gas phase.
5. Dao, T. V., Kitaev, L. E., Topchieva, K. V., Kubasov, A. A., and Ratov,
A. N., Vestnik Mosk. Universiteta, Khimiya 26, 65 (1971).
ˇ
6. Kubelkova´, L., Cejka, J., and Nova´kova´, J., Zeolites 11, 48 (1991).
7. Kubelkova´, L., and Nova´kova´, J., Zeolites 11, 822 (1991).
8. Bell, V. A., and Gold, H. S., J. Catal. 79, 286 (1983).
9. Xu, T., Munson, E. J., and Haw, J., J. Am. Chem. Soc. 116, 1962 (1994).
ˇ
10. Biaglow, A. I., Sepa, J., Gorte, R. J., and White, D., J. Catal. 151, 373
(1995).
CONCLUSIONS
11. Biaglow, A. I., Gorte, R. J., Kokotailo, G. T., and White, D., J. Catal.
148, 779 (1994).
12. Sepa, J., Lee, C., Gorte, R. J., Kassab, E., Evleth, E. M., Jessri, M.,
Allavena, M., and White, D., J. Phys. Chem. 100, 18515 (1996).
13. Kubelkova, L., Kotria, J., Florian, J., Bolom, T., Fraissard, J., Heribout,
L., and Doremieux-Morin, C., 11th Intern. Congress on Catalysis 101,
761 (1996).
14. Panov, A. G., and Fripiat, J. J., Langmuir. [submitted]
15. Hair, M. L., “Infrared Spectroscopy in Surface Chemistry,” p. 149.
Dekker, New York. 1967.
16. Hong, Y., Gruver, V., and Fripiat, J. J., J. Catal. 161, 766 (1996).
17. Gruver, V., Panov, A., and Fripiat, J. J., Langmuir 12, 2505 (1996).
18. Blumenfeld, A. L., Coster, D., and Fripiat, J. J., J. Phys. Chem. 99, 1581
(1995).
19. Blumenfeld, A. L., and Fripiat J. J., Topics in Catalysis 4, 119 (1997).
20. Farcasiu, D., Ghenciu, A., and Miller, G., J. Catal. 134, 118 (1992).
21. Hong, Y., and Fripiat, J. J., Microporous Materials 4, 323 (1995).
22. Hong, Y., Gruver, V., and Fripiat, J. J., J. Catal. 150, 421 (1994).
23. Coster, D., and Fripiat, J. J., Chem. Materials 5, 1204 (1993).
24. Yin, F., Blumenfeld, A. L., Gruver, V., and Fripiat, J. J., J. Phys. Chem.
B. 101, 1824 (1997).
The condensation of acetone on mesityl oxide is a bi-
molecular reaction between acetone activated on a Lewis
site and acetone adsorbed on bridging OH zeolites. The
turnover per Lewis sites is on the order of 10ꢂ3 reactions
per site and per second at 105ꢀC, and it is constant for all
zeolites. On alumina, it seems that the reaction occurs be-
tween a molecule in the gas phase and a molecule activated
on a Lewis site. In addition, the condensation goes further
than mesityl oxide.
ˇ
On zeolites and at 105ꢀC, the extent of the further
condensation reaction is much smaller. On both zeolites
with nonframework aluminum and alumina, the catalyst
gets “product poisoned” by MO and, also, perhaps, by
water.
ACKNOWLEDGMENTS
25. Coster, D., Blumenfeld, A. L., Gruver, V., and Fripiat, J. J., J. Phys.
Chem. 98, 6201 (1994).
26. Gruver, V., and Fripiat, J. J., J. Phys. Chem. 98, 8549 (1994).
27. Panov, A. G., Gruver, V., and Fripiat, J. J., J. Catal. 168, 321
(1997).
This work has been made possible by DOE grant DOE-FG02-90
ER1430. We want to thank the referees for interesting comments.
REFERENCES
28. Coster, D. J., Fripiat, J. J., Muscas, M., and Auroux, A., Langmuir 11,
2615 (1995).
1. Lercher, J. A., Z. Phys. Chem. N. F. 129, 209 (1982).
2. Hanson, B. E., Wieserman, L. F., Wagner, G. W., and Kaufman, R. A., 29. Roginski, S. S., in “Adsorption und Katalyse in homogen Oberfla¨-
Langmuir 3, 549 (1987). chen.” Akademie Verlag, Berlin. 1958.