C.R. Ho et al. / Journal of Catalysis 365 (2018) 174–183
179
supports the conclusion that acetone and DAA are in pseudo-
equilibrium. In contrast to dimerization of H6-acetone, dimerization
proceed via an E1cB mechanism at low temperatures in which an
enolate species is formed via -hydrogen abstraction by a basic
a
of D6-acetone (
D
Grxn = +9.7 kcal/mol) is more facile thermodynam-
surface oxygen atom. The hydroxyl group in DAA is subsequently
eliminated to form MO and water. E1cB mechanisms are observed
commonly in homogeneous aldol dehydration reactions with
NaOH and have been proposed for alcohol dehydration over solid
base catalysts such as MgO [67–69]. A central question in the
E1cB mechanism is whether proton abstraction or subsequent car-
banion decomposition is rate-limiting. This issue can be addressed
by noting that that proton abstraction of DAA and acetone both
ically. This is due to a change in hybridization geometry from sp2
(acetone ꢀ carbonyl) to sp3 (DAA ꢀ hydroxyl) [59]. The out-of-
plane bending mode is stronger for sp3 geometry, which leads to a
larger reduction in zero point energy upon H ? D substitution. For
a given H6-/D6-acetone partial pressure, the calculated gas-phase
equilibrium concentration of D12-DAA is 1.3x times greater than
the concentration of H12-DAA. This translates to a predicted equilib-
rium isotope effect of kH/kD ꢄ 0.67, which agrees qualitatively with
the experimental observation. Therefore, the observed inverse iso-
tope effect is attributed to a change in the equilibrium constant
between acetone and DAA. Similar isotope effects have been
observed for other carbonyl addition reactions [60,61].
occur at the a-carbon next to the carbonyl group. Since enolate for-
mation of acetone is rapid and reversible over HAP, it suggests that
proton abstraction of DAA is rapid as well. This conclusion is in
agreement with the observed inverse isotope effect, indicating that
C-H bond cleavage is not rate-limiting. Thus, DAA and its enolate
are in pseudo-equilibrium and the subsequent elimination of
OHꢀ to form MIBK is slow and irreversible.
Several related studies of aldol condensation reactions over HAP
and other metal oxides report that dehydration of the aldol inter-
mediate is facile [21,61,62]. However, these studies were con-
ducted at higher temperatures (>533 K) compared to the current
work (373 K). Monomolecular dehydration is favored entropically
due to the formation of two products from one reactant. Thus,
dehydration of the addition product becomes more difficult at
lower temperatures relative to other steps such as enolate forma-
tion and CAC coupling. This is likely why many liquid phase
studies of acetone condensation (which are typically conducted
at T < 373 K) report that aldol dehydration is slow compared to
the formation of the addition product [63,64].
In-situ titration experiments were performed to determine the
site requirements for acetone condensation (Fig. 4). Co-feeding
pyridine (1 kPa) does not significantly affect MIBK formation rates,
which indicates that acid sites are weak and cannot bind pyridine
strongly, consistent with DRIFTS studies of adsorbed pyridine
reported by Hill et al. [65] By contrast, introducing CO2 decreases
rates by more than an order of magnitude, suggesting that the
rate-limiting dehydration of DAA mainly occurs over basic sites.
At first glance, this may seem contradictory to findings by Kibby
et al., who found that alcohol dehydration over HAP requires acid
sites [66]. However, those reactions were performed with unacti-
vated alcohols at much higher temperatures (500–600 K). The
dehydration of DAA is more facile than typical alkanols because
The results reported to this point show that the surface oxygen
atoms of HAP are responsible for proton abstraction from DAA.
Therefore, we hypothesized that increasing the basic strength of
these oxygen atoms should enhance the overall reaction rate. This
was tested by measuring acetone condensation rates over the var-
ious ion-exchanged HAP catalysts. Since these catalysts contain
both Ca2+ sites and M2+ sites, the extent of exchange needs to be
considered to determine the true effect of the exchanged cation
on reaction rates. Calculations show that the bonding between
CaAO in HAP is mostly ionic with localized electron density [55].
Substitution of Ca2+ with another divalent cation affects the partial
charge of the adjacent oxygen atoms, but does not significantly
influence the electronics of oxygen atoms in the second shell
because of the ionic nature of the bonds. If the electron density
away from the substituted site is invariant with the substitution,
then the apparent reaction rates measured over the ion-
exchanged samples can be decoupled linearly into the sum of
two contributions: rates over the native Ca2+ sites and rates over
the exchanged M2+ sites. The rates over the native Ca2+ sites are
measured over the original Ca-HAP sample. For the ion-
exchanged samples, the rates over the exchanged M2+ sites can
be extracted by knowing the fractional surface coverage of Ca2+
and M2+. The assumption of local electron density does not hold
when substituting Ca2+ with Pb2+ because of the strong covalent
PbAO bonds [70]. The apparent MIBK formation rate over Pb-
HAP is over 140 times slower than Ca-HAP even though only 90%
of the Ca sites are replaced with Pb, showing that activity of Ca
the C AH bond is weakened by the nearby carbonyl group, which
a
makes the
a-hydrogen more prone to attack by a base. The ease of
a-hydrogen abstraction suggests that DAA dehydration can
sites decreases with substitution of Ca2+ by Pb2+
.
MIBK formation rates of each sample were normalized by the
surface area or basic site density (as determined by CO2-TPD) after
accounting for the extent of exchange and are plotted in Fig. 5.
Since the electron density of the oxygen atoms are influenced by
the neighboring cations, the electronegativity of the cation, which
is a measure of the propensity to attract electrons, can be used as a
descriptor of basicity. As the basicity of the catalyst increases from
Pb2+ to Sr2+, the reaction rates also increase. However, even though
Ba-HAP is more basic than Sr-HAP, it is three times less active, sug-
gesting that either an appropriate balance of acid and base strength
is necessary, or a more detailed descriptor is needed. The measured
activation energies are in relative agreement with the rates. With
increasing basicity, the apparent activation energy decreases from
16 kcal/mol for Cd-HAP to 11 kcal/mol for Sr-HAP and then
increases to 14 kcal/mol for Ba-HAP.
DFT calculations were performed to gain more insight into the
mechanism of DAA dehydration. The OH-terminated (0 1 0) plane
on HAP was chosen as the model surface because it is preferentially
exposed under the used synthesis conditions (co-precipitation
method, high pH) [49]. Furthermore, studies by Tsuchida et al.
[31] and Rodrigues et al. [71] have shown that HAP catalysts with
Fig. 4. Effect of titrant on MIBK formation rate. Reaction conditions: PAcetone = 5 kPa,
H2 = 5 kPa, PTitrant = 1 kPa, T = 373 K, massHAP = 0.01 g, massPd/SiO2 = 0.04 g.
P