Detailed kinetic study of cumene isopropylation in a liquid-liquid biphasic system using acidic chloroaluminate ionic liquids

Article abstract of DOI:10.1016/j.jcat.2008.06.018

A kinetic study of cumene isopropylation has been carried out using the acidic ionic liquid 1-ethyl-3-methylimidazolium chloride ([EMIM]Cl)/AlCl₃ (molar ratio 1:2) as the catalyst phase in a semi-batch liquid-liquid biphasic reaction system. Kinetic models representing the alkylation reaction network have been established based on reaction temperature and propylene partial pressure variations. By comparing the results of the kinetic models with the measured concentrations in the liquid organic phase, the importance of the different product solubilities in the acidic ionic liquid became evident. Correction of the product concentrations in the organic phase based on a COSMO-RS calculation of the relative product solubilities in the acidic ionic liquid [EMIM][Al₂Cl₇] gave a remarkably good prediction of the reaction kinetics by the kinetic model. These findings demonstrate the suitability of COSMO-RS to predict the relative solubilities of different aromatic compounds in highly reactive catalytic systems.

Full text of DOI:10.1016/j.jcat.2008.06.018

Journal of Catalysis 258 (2008) 401–409
Contents lists available at ScienceDirect
Journal of Catalysis
www.elsevier.com/locate/jcat
Detailed kinetic study of cumene isopropylation in a liquid–liquid biphasic system
using acidic chloroaluminate ionic liquids
J. Joni a, D. Schmitt a, P.S. Schulz a, T.J. Lotz b, P. Wasserscheid a,∗
a
Lehrstuhl für Chemische Reaktionstechnik, Friedrich-Alexander Universität Erlangen-Nürnberg, Egerlandstr. 3, 91058 Germany
SI group-Switzerland GmbH, Kästeliweg 7, CH-4133 Pratteln, Switzerland
b
a r t i c l e i n f o
a b s t r a c t
Article history:
A kinetic study of cumene isopropylation has been carried out using the acidic ionic liquid 1-ethyl-
3-methylimidazolium chloride ([EMIM]Cl)/AlCl3 (molar ratio 1:2) as the catalyst phase in a semi-batch
liquid–liquid biphasic reaction system. Kinetic models representing the alkylation reaction network have
been established based on reaction temperature and propylene partial pressure variations. By comparing
the results of the kinetic models with the measured concentrations in the liquid organic phase, the
importance of the different product solubilities in the acidic ionic liquid became evident. Correction of
the product concentrations in the organic phase based on a COSMO-RS calculation of the relative product
solubilities in the acidic ionic liquid [EMIM][Al2Cl7] gave a remarkably good prediction of the reaction
kinetics by the kinetic model. These findings demonstrate the suitability of COSMO-RS to predict the
relative solubilities of different aromatic compounds in highly reactive catalytic systems.
Received 7 May 2008
Revised 12 June 2008
Accepted 15 June 2008
Available online 21 July 2008
Keywords:
Friedel–Crafts alkylation
Acidic ionic liquid
Kinetic investigation
COSMO-RS
© 2008 Elsevier Inc. All rights reserved.
1. Introduction
isopropylation and the catalyst was found to be active over more
than 15 h time on stream [10].
Surprisingly, despite all the above mentioned progress in the
Friedel–Crafts alkylation of aromatic substances is a major route
of producing alkylated arenes at industrial scale. The reaction is
catalysed by either Lewis or Brønsted acid catalysts and many ap-
plications still use homogeneous aluminium(III) chloride [1]. A ma-
jor disadvantage of the current available homogeneous processes
is the need to separate the reaction mixture from the catalyst
through a hydrolysis step in which the catalyst is decomposed and
extracted to the aqueous phase. Alkylations using heterogeneous
catalysts mostly give quite low conversion and require relatively
extreme reaction conditions. The latter often trigger oligomerisa-
tion [2] or even coke formation [3].
use of acidic ionic liquid systems, there are—to the best of our
knowledge—no detailed kinetic studies in the field of ionic liquid
catalysed Friedel–Crafts alkylations. In this paper we present a first
kinetic study of the isopropylation of cumene catalysed by Lewis-
acidic imidazolium chloroaluminate melts. The aim of the study is
to provide important information for the further optimisation of
the reaction system towards a potential commercial application.
2. Experimental and methods
2.1. Chemicals
The application of acidic ionic liquids has opened the possibility
to immobilise the acidic active species in an ionic phase, immisci-
ble with the liquid product layer. The investigated reactions in such
systems include the alkylation of benzene and toluene with alkyl
halide [4,5], linear alkyl benzene production [6], alkylation of aro-
matics using alkenes [7] and alkylation of isobutene with 2-butene
(refinery alkylation) [8]. Recent developments showed that ionic
liquids of the type [cation][(CF3SO2)2N]/AlCl3 are able to dissolve
significantly higher amounts of AlCl3 than the conventional sys-
tems of type [cation][Cl]/AlCl3 [9]. Furthermore, application of such
system in a continuous loop reactor was demonstrated for toluene
Cumene of 99.0% purity (Acros Organic) was applied as aro-
matic compound together with propylene 2.8 (99.8%) from Linde
as alkylating agent. Water content of cumene was confirmed to be
below 20 ppm.
1-ethyl-3-methylimidazolium chloride ([EMIM]Cl) was pur-
chased from Solvent Innovation GmbH, Cologne and used without
further purification. Toluene (BASF) was used after reaction to
dilute the samples for GC analysis. Aluminium(III) chloride was
purchased from Merck and used without further purification.
For measuring the solubility of cumene (99% from Acros Or-
ganic), 1,3-diisopropylbenzene (96% from Alfa Aesar) and 1,3,5-tri-
isopropylbenzene (>97% from Alfa Aesar) in the chloroaluminate
ionic liquids by 1H-NMR spectroscopy, deuterated dichloromethane
from Fluka was applied as NMR solvent.
Corresponding author. Fax: +49 9131 8527421.
*
E-mail address: wasserscheid@crt.cbi.uni-erlangen.de (P. Wasserscheid).
0021-9517/$ – see front matter © 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2008.06.018

402
J. Joni et al. / Journal of Catalysis 258 (2008) 401–409
Fig. 1. Reactor set up for liquid–liquid biphasic isopropylation of cumene/meta-xylene using acidic ionic liquid.
2.2. Kinetic experiments
Table 1
Required propylene partial pressure for various reaction temperatures to adjust
20 mol% propylene in the organic phase
Experiments were carried out in a 600 mL Parr autoclave (see
Fig. 1) equipped with a PARR four blade-gas entrainment stirrer of
3.2 cm diameter. Prior to the reaction, the autoclave was stored
Temperature
Propylene partial pressurea
◦
50
70
90
C
C
C
3.6 bar
4.7 bar
6.3 bar
8.5 bar
◦
◦
◦
in an oven at 90 C for 24 h to avoid any moisture contamination
during reaction. The organic phase consisted at the beginning of
the reaction of the extraction solvent and the aromatic substrate
(in a molar ratio 9:1). The reaction was carried out in a glass liner
placed into the autoclave. The glass liner was pre-heated before
◦
110
C
a
Estimated using ASPEN PLUS simulator using NRTL thermodynamic model.
◦
use in an oven (90 C) and was filled previously with argon to
the organic phase, the total reaction volume was considered to be
constant. Each sample taken from the reactor was further diluted
with toluene and neutralised (using a mixture of solid magnesium
sulphate and sodium carbonate) before being analysed by gas chro-
matography.
avoid moisture contamination during weighing.
The catalyst was prepared by mixing a 2:1 molar ratio of AlCl3
◦
and [EMIM]Cl respectively at 70–80 C in a Schlenk flask until
complete dilution of AlCl3 and a clear liquid was obtained. The de-
sired amount of the catalyst phase was then placed into the glass
liner using a syringe. Once the glass liner was placed inside the au-
toclave, the reactor was evacuated using a vacuum pump at room
temperature for about two to three minutes followed by argon
flushing. This procedure was repeated three times. The reactor was
then heated up to the desired temperature using an external elec-
trical heating jacket. Upon reaching the desired temperature, the
reactor was connected to the propylene bottle with a defined out-
let pressure and a sample was taken for reference. The propylene
partial pressure was set in such way that approximately 20 mol%
of propylene existed in the saturated organic phase at each tem-
perature. The required propylene partial pressure can be seen from
Table 1.
The reaction started (t = 0) when the gas entrainment stirrer
was turned on. To achieve a good gas entrainment and saturation
of the liquid phase the stirring rate was set to 1200 rpm. In or-
der to get a first qualitative information on the reaction progress,
an analytical balance was used to monitor the weight loss of the
propylene bottle. Samples of 1 mL were taken through a capillary
tube from the organic phase and since the total amount required
for these GC analyses was very small compared to the volume of
2.3. GC analysis
Samples were analysed by means of gas chromatography (Var-
ian CP-3900) using a WCOT fused silica column (0.21 mm diameter
and 50 m). The gas chromatograph was equipped with a Flame
◦
Ionisation Detector (FID) working at 270 C.
The molar amount of each substance in the analysed reaction
volume was calculated based on the mass fraction of each sub-
stance, the mass fraction of cyclohexane as standard multiplied by
a correction factor as seen in Eq. (1) with CF being the correction
factor of the GC and Mw being the molecular weight of the anal-
ysed substance.
%mi · mstandard
ni = CF ·
.
(1)
%mstandard · Mw,i
Since there were only negligible volumes of samples taken out of
the reactor during the kinetic studies, the amount of standard sub-
stance in the reactor can be considered to be constant throughout
the reaction.

J. Joni et al. / Journal of Catalysis 258 (2008) 401–409
403
2.4. Data processing
reaction coordinates, respectively. The first subscript refers to the
isopropylation reaction of cumene (subscript = 1) while the second
subscript refers to the single, double and triple alkylation reaction,
respectively,
The kinetic model was fitted based on the molar amount/time
profiles for each substance involved in the model. The modelled or-
dinary differential equations for the reaction network were solved
using the explicit Runge–Kutta pair of Bogacki and Shampine [11].
The fitting procedure was carried out by iterating the reaction or-
der and reaction rate constant in order to minimise the sum of
square error between the calculated model and the experimental
values. This was carried out using a non-linear optimisation algo-
rithm as described in the literature [12].
k1–1;ε1–1
C6H5(C3H7) + C3H6 −−−−−−→ C6H4(C3H7)2,
k1–2;ε1–2
C6H4(C3H7)2 + C3H6 −−−−−−→ C6H3(C3H7)3,
C6H3(C3H7)3 + C3H6 −−−−−−→ C6H2(C3H7)4.
k1–3;ε1–3
(2)
For simplicity sake let A = cumene, B = diisopropylbenzene (DIPB),
C = triisopropylbenzene (TIPB) and D = tetraisopropylbenzene
(TeIPB). The reaction rate expression for each reaction can now
be written as follows:
2.5. NMR solubility experiment
Solubility measurements of aromatic compounds in chloroalu-
minate ionic liquids were carried out using 1H NMR spectroscopy.
The samples were prepared by mixing the aromatic compound
under investigation (cumene, diisopropylbenzene or triisopropyl-
benzene, respectively) and cyclohexane in an equimolar amount.
5 g of each of these organic solutions were intensively mixed (24 h,
500 rpm to ensure equilibrium conditions) with 5 g of the chloroa-
luminate ionic liquid. After stirring, the ionic liquid phase was
separated under inert atmosphere and analysed by 1H NMR us-
ing a JEOL ECX400 spectrometer (400 MHz).
r1 = k1–1 · CAnA with k1–1 = k1 · Cnp1
,
,
.
ꢀ
ꢀ
propylene
r2 = k1–2 · CBnB with k1–2 = k2 · Cnp2
ꢀ
ꢀ
propylene
r3 = k1–3 · CCnC with k1–3 = k3 · Cnp3
(3)
ꢀ
ꢀ
propylene
ꢀ
The modified reaction rate constant k contains the propylene con-
centration term as it was fed continuously throughout the reaction
and assumed to be constant over time. A mass balance of the liq-
uid organic phase for a discontinuous system gives equation:
dnA
= −r1 · Vr,
2.6. Diffusion coefficient measurements using NMR-diffusion ordered
spectroscopy (DOSY)
dt
dnB
= (r1 − r2) · Vr,
dt
The diffusion coefficients of cumene, 1,3-diisopropylbenzene
and 1,3,5-triisopropylbenzene in the chloroaluminate ionic liq-
uid were determined using NMR-Diffusion Ordered Spectroscopy
(DOSY) technique. This method has been widely applied in de-
termining diffusion coefficient of organic matters as well as self
diffusion coefficient of ion-pairs in ionic liquids as described in
various literatures [13]. For the DOSY measurements samples of
2.5 g of each of the aromatic substance with 2.5 g of the chloro-
aluminate ionic liquid were prepared. After stirring the mixture for
24 h the molecular diffusion of the investigated aromatic molecule
was observed by applying magnetic field gradients. Based on the
aromatic molecule movement from one magnetic field zone to an-
other at a given temperature one can determine the diffusion co-
efficient of the observed aromatic substance in ionic liquid phase.
For our DOSY measurements, 0.1 s and 0.003 s were used as dif-
dnC
= (r2 − r3) · Vr.
(4)
dt
Moreover the molar amount of each substance at a given time can
be described as functions of reaction coordinate:
nA(t) = nA0 − ε1–1(t),
nB(t) = nB0 + ε1–1(t) − ε1–2(t),
nC(t) = nC0 + ε1–2(t) − ε1–3(t).
Substituting Eqs. (5) and (3) into Eqs. (4) gives a set of coupled
differential equations as shown in Eq. (6):
(5)
dε1–1
(1−nA)
ꢀ
nA
= k1–1 · Vr
· (nA,0 − ε1–1
) ,
dt
dε1–2
fusion time and gradient time respectively. Magnetic fields of 30–
(1−nB)
ꢀ
nB
−1
= k1–2 · Vr
· (nB,0 + ε1–1 − ε1–2
· (nC,0 + ε1–2 − ε1–3
)
,
.
300 mT min
were applied.
dt
dε1–3
(1−nC)
ꢀ
nC
= k1–3 · Vr
)
(6)
3. Results and discussions
dt
To get the activation energy and the reaction order of cumene
isopropylation, the above mentioned model was fitted to the ex-
Kinetic investigation of cumene isopropylation was carried out
◦
◦
in the temperature range between 50 C and 110 C. Based on the
experimental protocol (see above), the following assumptions for
the modelling were applied:
◦
◦
◦
perimental results at reaction temperature of 50 C, 70 C, 90 C
◦
and 110 C. Interestingly, our experiments revealed a significant
discrepancy between the molar amount of consumed cumene and
the sum of all alkylated products analysed in the organic phase (for
details see supporting information). When these experimental val-
ues were directly fitted to the model, noticeable “delay” behaviour
was observed. This can be seen in Fig. 2 as the shaded area I, II
and III for DIPB, TIPB and TeIPB, respectively.
We assumed that the delayed, S-shaped profile of the alky-
lated products can be explained by the fact that reaction products
first accumulate in the acidic ionic liquid phase before they diffuse
out into the liquid organic phase where samples were taken and
analysed. The different behaviour for cumene would be explained—
according to the same assumption—by the fact that cumene and
the ionic liquid phase had been in contact during reaction start up
and therefore considered to be in equilibrium with the ionic liquid
phase right from the start of the reaction.
1. Propylene is assumed to reach its saturated concentration in
the liquid organic phase due to the effectiveness of the gas
entrainment stirrer and the applied high mixing rate.
2. Since propylene is present in excess in the liquid organic phase
(in comparison to cumene) and it is fed continuously, all reac-
tions are assumed to be not limited by propylene availability.
3. The amount of cumene is very low in comparison to the sol-
vent so that the propylene concentration in the liquid organic
phase can be assumed to be constant.
3.1. Kinetic model for the isopropylation of cumene
The reaction network of cumene isopropylation is described in
Eq. (2) with k and ε being the reaction rates constant and the

404
J. Joni et al. / Journal of Catalysis 258 (2008) 401–409
3.2. Experimental and COSMO-RS-based estimation of aromatic
solubility in a non-acidic [EMIM]Cl/AlCl3 mixture
However, in the case of acidic chloroaluminate ionic liquid, this
procedure was impossible due to the fact that the acidic ionic liq-
uid catalyst phase started to react with the aromatic substances
under investigation during the time of the NMR solubility experi-
ment. A typical 1H NMR spectrum of the ionic liquid phase for a
cumene/cyclohexane-acidic ionic liquid ([EMIM]Cl/AlCl3 = 1:2) sys-
To confirm our assumption we aimed to determine the solubil-
ity of all relevant aromatic compounds in the acidic chloroalumi-
nate ionic liquid. As a suitable method to determine the solubility
of organic substances in ionic liquids we chose 1H NMR spec-
troscopy. The method allows to compare the signal intensity of a
characteristic ionic liquid cation signal with a characteristic signal
of the dissolved aromatic compound and to determine in this way
the molar solubility.
◦
tem is shown in Fig. 3c (spectrum was taken at 19 C 30 min after
addition of the ionic liquid to the organic mixture). Remarkably,
almost no cumene could be detected in this spectrum as can be
seen by the missing “fingerprint” of the tertiary carbon of the iso-
propyl group. Moreover, appearance of the single peak at 7.34 ppm
◦
◦
Fig. 2. Fitting result for cumene isopropylation at 50 C taking only the products analysed from the liquid organic phase. Treact = 50 C; pTot = 3.6 bar; [EMIM]Cl/AlCl3 = 1:2;
mol mL−1; solvent = cyclohexane.
−4
[cumene] = 8.3 × 10
(a)
Fig. 3. Typical ionic liquid phase 1H NMR of cyclohexane–cumene–ionic liquid system in equilibrium at room temperature: (a) cumene reference, (b) ionic liquid =
[EMIM]Cl/AlCl3 = 1:1, (c) ionic liquid = [EMIM]Cl/AlCl3 = 1:2.

J. Joni et al. / Journal of Catalysis 258 (2008) 401–409
405
(b)
(c)
Fig. 3. (continued)
indicates benzene formation by cumene dealkylation. This result
emphasizes the difficulty of solubility measurement in highly reac-
tive systems.
As one attractive possibility to overcome these practical diffi-
culties of solubility measurements in highly reactive systems we
identified the option to model the partition coefficient of each aro-
matic substance using the Conductor-like Screening Model for Real
Solvent (COSMO-RS) [14].
However, there is—to the best of our knowledge—no report so
far in the literature describing the usefulness of COSMO-RS to pre-
dict the solubility of organic substances in chloroaluminate ionic
liquids. Therefore, we had to confirm in first place the reliability of
this method for chloroaluminate ionic liquids.
It is important to note that the dealkylation reaction observed
in these NMR solubility experiments does not play a significant
role in the kinetic experiments. Almost no formation of benzene
was observed in the kinetic experiments so that all reactions in the
kinetic experiments can be assumed to be irreversible. The reason
for this difference is that the NMR solubility measurements have
been recorded in absence of propylene while all kinetic experi-
ments have been carried out in the presence of excess propylene.

406
J. Joni et al. / Journal of Catalysis 258 (2008) 401–409
Table 2
Comparison of COSMO-RS estimated partition coefficient with experimental resultsa
No.
Aromatic (Ar)
Experimental values
COSMO-RS estimation
b
c
XAr,org
XAr,org
XAr,IL
Xsolvent,IL
end
Pi
Si
γAr,org
γAr,IL
Pi
Si
init
end
end
1
2
3
Cumene
1,3-diisopropylbenzene
1,3,5-triisopropylbenzene
0.502
0.502
0.500
0.465
0.504
0.511
0.210
0.067
0.020
0.089
0.073
0.059
0.453
0.134
0.039
1.000
0.296
0.087
0.090
0.035
0.014
2.338
4.087
5.731
0.039
0.009
0.003
1.000
0.223
0.065
a
b
c
Solvent = cyclohexane; IL = [EMIM]Cl/AlCl3 = 1:1; organic mixture: IL = 1:1 mass ratio.
Si = relative solubility based on individual partition coefficient = Pi /P j
.
Estimated partition coefficient from COSMO-RS based on the activity coefficient according to thermodynamical equilibrium criteria where
x
i,phase 1/xi,phase 2
=
γ
i,phase 2/γi,phase 1.
[EMIM]Cl + AlCl3 ꢀ [EMIM][AlCl4]
[EMIM][AlCl4] + AlCl3 ꢀ [EMIM][Al2Cl7]
[EMIM][Al2Cl7] + AlCl3 ꢀ [EMIM][Al3Cl10
]
Scheme 1. Various ionic species present in an equilibrium condition of [EMIM]Cl/
AlCl3 mixture [16].
To achieve this, we decided to compare experimental NMR solu-
bility data for neutral, non-reactive [EMIM][AlCl4] with COSMO-RS
derived partition coefficients for the same ionic liquid under the
◦
same conditions of the experiment (19 C).
The COSMO-RS simulation incorporates two crucial assump-
tions. The first assumption is that the modelled ionic liquid phase
consist of only [EMIM][AlCl4] species, which is very well fulfilled
for a [EMIM]Cl/AlCl3 mixture in a 1:1 molar ratio [15,16]. The
second assumption considers cyclohexane to be completely im-
miscible with the ionic liquid phase. Also this pre-condition can
be regarded to be in good agreement with the experimental re-
ality. Table 2 shows the results of our COSMO-RS modelling. It
can be seen that COSMO-RS gives only a very crude estimation
of the absolute experimental partition coefficient values. How-
ever, the relative partition coefficient values (Pi/Pcumene) given by
COSMO-RS simulation represent the experimental values remark-
ably well.
Fig. 4. Relative solubility of cumene, DIPB, TIPB and TeIPB in [EMIM][Al2Cl7] vs. cy-
clohexane as calculated at various temperatures using COSMO-RS. Solvent = cyclo-
hexane; ionic liquid = [EMIM][Al2Cl7].
the acidic ionic liquid drastically. Moreover, our calculations indi-
cate that the solubility of each aromatic substance in the acidic
ionic liquid phase increases slightly with temperature.
This fit between the COSMO-RS predicted and the experi-
mentally determined, relative partition coefficients of cumene,
1,3-diisopropylbenzene and 1,3,5-triisopropylbenzene in the neu-
tral [EMIM][AlCl4] ionic liquid, justified the next step in our study,
namely the COSMO-RS simulation of the relative partition coef-
ficients for an acidic, reactive ionic liquid. Using the COSMO-RS
estimated, relative partition coefficients for the acidic ionic liq-
uid system (which cannot be experimentally verified, see above)
we aimed to support our hypothesis that the dissolved alkylation
products in the acidic ionic liquid phase account for the unusual
S-shaped product curves observed in the experiments.
The relative solubility (Si) of DIPB, TIPB and TeIPB in the acidic
ionic liquid phase has been calculated using equation:
Pi
Si =
,
i = DIPB, TIPB or TeIPB.
(7)
PDIPB + PTIPB + PTeIPB
The next step in our efforts to explain the experimentally ob-
served S-shape behaviour in the concentration/time plot was to
use the calculated relative partition coefficients and relative sol-
ubilities to correct the product analyses from the organic phase
during the kinetic experiments. The corrected molar amount of
each alkylated product at each reaction temperature was obtained
by proportionally distributing the observed missing molar amount
of feedstock (cumene) using the relative solubility factors (Si) cal-
culated via the COSMO-RS method. The corrected molar amount of
each substituted product for cumene isopropylation is provided in
the supporting information.
It is important to point out that in such reacting system an
equilibrium condition between organic phase and the ionic liquid
phase is very unlikely to be reached. However, by implement-
ing the relative solubilities as correction factors we estimate the
relative driving forces for mass transfer of one specific aromatic
compound from the reactive phase into the organic phase to be
proportional to the distribution of this specific substance between
the ionic liquid and the organic phase.
3.3. COSMO-RS based partition coefficients estimation for highly acidic
ionic liquid systems
Unfortunately, the situation in the acidic ionic liquid of inter-
est, [EMIM]Cl/AlCl3 (molar ratio: 1:2) is significantly more complex
than in the case of the neutral ionic liquid. This is due to the fact
that the acidic chloroaluminate system consists of a mixture of an-
ions [15] (see Scheme 1).
For our COSMO-RS simulation we decided to simplify the com-
plex ionic liquid anion mixture and considered the acidic ionic
liquid to be pure [EMIM][Al2Cl7]. This simplification appeared ac-
ceptable as a 1:2 molar ratio of [EMIM]Cl and AlCl3 (as used in
our experiments) is known to contain almost 80 mol% of the anion
−
In fact, by applying the solubility corrected values for the for-
mation of aromatic products in the cumene isopropylation system,
the kinetic model (see Eq. (6)) shows a much better agreement
than before (compare Figs. 2 and 5).
[Al2Cl7] and only 20 mol% of other anionic species [16].
Fig.
4 shows the key result of the COSMO-RS calculation,
namely the fact that an increasing degree of alkylation at the aro-
matic ring decreases the solubility of that aromatic compound in

J. Joni et al. / Journal of Catalysis 258 (2008) 401–409
407
(a)
(b)
(c)
(d)
Fig. 5. Model-fitted results for cumene isopropylation reaction using corrected datasets based on proportionality factor estimated by COSMO-RS. [EMIM]Cl/AlCl3 = 1:2;
mol mL−1; Cat:cumene = 1:25: (a) Treact = 50 C, pTot = 3.6 bar; (b) Treact = 70 C, pTot = 4.7 bar; (c) Treact = 90 C,
−4
◦
◦
◦
solvent = cyclohexane; [cumene] = app. 8.3 × 10
◦
pTot = 6.3 bar; (d) Treact = 110 C, pTot = 8.5 bar.
Table 3
the individual experiments while concentrations and concentration
ratios were kept constant in all experiments.
Fitted kinetic parameters for cumene isopropylationa
As expected, the observed reaction rate constants increase for
Cumene → DIPB
DIPB → TIPB
TIPB → TeIPB
higher consecutive alkylation reaction which is in agreement with
the fact that the already alkylated products have a higher reactivity
due to the higher electron density in their aromatic ring.
Reaction order
Reaction rate
0.99
1.04
1.05
−4 −1
−3 −1
−3 −1
5.82 × 10
s
1.91 × 10
s
2.68 × 10
5.93 × 10
3.20 × 10
7.72 × 10
s
◦
constant (50 C)
The formal activation energies for the reaction network were
determined by plotting the logarithmic value of the reaction rate
constant against 1/T in an Arrhenius diagram. It is important to
note that the temperature dependency of propylene solubility is
included in the here determined EA,eff value. Remarkably, all re-
action steps show relatively low EA,eff values in the liquid–liquid
biphasic system (see Fig. 6). Due to the fact that we consider a
pseudo-one-component system (where propylene solubility and its
temperature dependency are contained in the reaction rate con-
stant of the model) the formal effective activation energy values of
the biphasic alkylation reaction presented in this work cannot be
directly compared to other published examples of homogeneous
alkylation reactions [17a] and solid acid catalysed alkylation sys-
−4 −1
−3 −1
−3 −1
Reaction rate
8.23 × 10
1.71 × 10
1.80 × 10
s
4.65 × 10
1.92 × 10
3.49 × 10
s
s
◦
constant (70 C)
−3 −1
−3 −1
−3 −1
Reaction rate
s
s
s
◦
constant (90 C)
−3 −1
−3 −1
−3 −1
Reaction rate
constant (110 C)
s
s
s
◦
a
For reaction condition refer to Fig. 5.
The same correction procedure was therefore carried out for
all other reaction temperatures. The results are also displayed in
Fig. 5. By fitting the model to the COSMO-RS corrected values,
the reaction order of cumene, diisopropylbenzene and triisopropyl-
benzene for the main, first consecutive and second consecutive
alkylation reaction respectively was found to be first order (see
Table 3).
For proper interpretation of the data in Fig. 5 it should be noted
that the varying molar amounts of aromatic substances at initial
condition (t = 0) are due to slightly varying reaction volumes in
tems [17b] for which activation energies have been reported in the
−1
range of 40 to 90 kJ mol
.
In order to find out whether the observed, relatively low EA,eff
values have to be attributed to mass transfer effects an estimation
of Hatta number (Ha) [18] was carried out for our liquid–liquid
biphasic alkylation catalysis. The mathematical form of the Hatta
number for a first order reaction is given in Eq. (8). In our case,

408
J. Joni et al. / Journal of Catalysis 258 (2008) 401–409
Fig. 8. Diffusion coefficient of cumene, DIPB and TIPB in chloroaluminate ionic liquid
(EMIMCl:AlCl3 = 1:1) at various temperature modelled as D(T) = Do exp(−ED/RT ).
Fig. 6. Effective activation energy (EA,eff) for main, first and second consecutive alky-
lation for cumene isopropylation reaction.
Diffusion coefficients of the aromatic substances in the ionic
liquid phase were measured using NMR-diffusion ordered spec-
troscopy (DOSY). Again, the DOSY measurements to determine the
diffusion coefficients of the aromatic compounds in the ionic liquid
were performed for the neutral ionic liquid ([EMIM]Cl/AlCl3 = 1:1),
since cumene reaction were inevitable during the time of the DOSY
measurements in acidic systems.
Fig. 8 shows the diffusion coefficient of cumene, 1,3-diiso-
propylbenzene and 1,3,5-triisopropylbenzene in the neutral chloro-
aluminate ionic liquid as a function of temperature (with ED being
the activation energy of diffusion and Do being the maximum
diffusion coefficient). Using this Arrhenius function, diffusion co-
efficients of the aromatic components were calculated for the rel-
◦
◦
evant temperature of 90 C. For cumene a Dcumene,IL,90 of 6.1 ×
C
−10
−1
10
m2 s
was obtained in this way.
The mass transfer constant (ki,IL) is strongly dependent on the
ionic liquid’s droplet size. This results from a complex interplay of
reactor configuration (i.e. stirrer type and dimensions, reactor ge-
ometry, etc.) and physico-chemical properties of the ionic liquid
itself (i.e. surface tension under reaction conditions). Newman cor-
relation [19] as shown in Eq. (9) provides a rough estimation for
ki,IL assuming that there is no convective flow inside the ionic liq-
uid droplets.
Fig. 7. Cumene conversion in strongly diluted reaction system at various stirring
◦
rate. pTot = 6.3 bar; Treact = 90 C; [EMIM]Cl/AlCl3 = 1:2; Cat:cumene = 1:50; [cu-
−4
−1
.
mene] = 8.3 × 10
mol mL
the Hatta number represents the ratio of chemical reaction rate
in the ionic liquid phase versus the rate of mass transport of the
aromatic feedstock into the ionic liquid phase.
Di,IL
ki,IL = 6.58
.
(9)
d32
ꢀ
To estimate the mean sauter diameter (d32) of a highly dispersed
medium the empiric correlation of Shinnar [20] can be applied:
1
Ha =
·
k · Di,IL
(8)
ki,IL
−0.6
d32,max = 0.15 · Dstirrer · We
(10)
with ki,IL-mass transfer coefficient of aromatic component “i” in
the ionic liquid phase (m s−1); k-intrinsic reaction rate constant
of pseudo-first order reaction (s−1); Di,IL-diffusion coefficient of
aromatic component “i” into the ionic liquid phase (m2 s−1).
Since intrinsic reaction rate constant is required for such es-
timation, additional experiments with highly diluted organic and
with d32,max—maximum sauter diameter (mm); Dstirrer—stirrer
diameter (mm); We—Stirrer’s Weber number = (ρsolvent · N2
·
D3stirrer)/γinterface; ρsolvent—dispersant’s density (kg m−3); N—stir-
ring rate (rps); γinterface—interfacial tension between dispersant
(cyclohexane)
and
dispersed
substances
(ionic liquid) =
◦
|γsolvent − γIL| (Nm−1).
catalyst phases were carried out at 90 C. By varying the stirring
rate of the gas entrainment stirrer (see Fig. 7) we could demon-
strate that the reaction rate (i.e. cumene depletion rate) in these
highly diluted systems are independent on stirring rate for rates
greater than 800 rpm. Consequently, reaction rate constants deter-
mined under these conditions can be interpreted as intrinsic reac-
tion rates that are not influenced by any mass transfer influence.
Properties values for cyclohexane were obtained from Perry’s
chemical engineering handbook [21]. The surface tension of the
chloroaluminate ionic liquid was taken from a publication by Tong
et al. [22]. With these data the calculated Sauter diameter ac-
cording to Eq. (10) is in the range of 140–170 μm. Entering this
value into Eq. (9) one obtains a mass transfer coefficient (ki,IL
)
−5
−1
The following intrinsic reaction rate constants were determined in
of 2.37 × 10
m s
and together with the diffusion coefficient
−3
this way: Isopropylation of cumene: 1.32 × 10
s
−1, isopropyla-
and isopropylation of
(Di,IL) and the reaction rate constant (k) the Hatta number can be
calculated. The Hatta number calculated in this way for the con-
version of cumene to diisopropylbenzene was found to be 0.03 to
−3 −1
tion of diisopropylbenzene: 2.01 × 10
s
−3 −1
triisopropylbenzene: 2.17 × 10
s
.

J. Joni et al. / Journal of Catalysis 258 (2008) 401–409
409
0.05 showing that the reaction rate was not limited by mass trans-
fer into the ionic liquid phase. This was also the case for the first
and the second consecutive alkylation reaction.
that propylene solubility in acidic chloroaluminate ionic liquids in-
creases with decreasing temperature [23].
Consequently, the observed, relatively low activation energy of
the isopropylation reactions cannot be attributed to mass transfer
influences on the observable reaction rate. A convincing explana-
tion for the low formal activation energy arises from the fact that
the temperature dependency of propylene solubility in the ionic
liquid phase is included in the EA,eff value in our case. Propylene
pressure was adjusted to provide excess propylene in the organic
and in the ionic liquid layer (for more information see supporting
information), but temperature dependent propylene concentration
in the ionic liquid will still have a strong influence on the re-
Acknowledgments
The authors would like to thank Dipl- Ing. Viktor Ladnak for
support and many scientific discussions. Moreover, we would like
to thank the SI group, Pratteln, Switzerland for financial support.
Supporting information
Supporting information for this article may be found on Sci-
enceDirect, in the online version.
ꢀ
action rate (in our model on the reaction rate coefficients k1−x).
Please visit DOI: 10.1016/j.jcat.2008.06.018.
In former work by Eichmann [23], it was demonstrated that the
propylene solubility in chloroaluminate ionic liquids can increase
with decreasing temperatures and this effect was used in a similar
argument to explain even the observed negative activation energies
found for the Ni-catalysed dimerisation of propylene in slightly
acidic chloroaluminate ionic liquids.
References
[1] O. Ulmann, Ulmann’s Encyclopedia of Industrial Chemistry, vol. B8, in: fifth ed.,
Wiley–VCH, 1995, p. 267.
[2] W.W. Kaeding, J. Catal. 120 (1989) 409.
[3] (a) P.A. Arroyo, C.A. Henriques, E.F. Sousa-Aguiar, A. Martinez, J.L.F. Monteiro,
Stud. Surf. Sci. Catal. 130C (2000) 2555;
4. Summary
(b) C.A. Querini, E. Roa, Appl. Catal. A 163 (1997) 199;
(c) J.L. Sotelo, M.A. Uguina, J.L. Valverde, D.P. Serrano, Appl. Catal. A 114 (1994)
273.
This work provides detailed kinetic information on liquid–liquid
biphasic alkylation reactions catalysed by acidic ionic liquids. It
was found that the model of pseudo first order consecutive re-
actions could be fitted satisfyingly to the experimental data only
after taking into account the product share dissolving into the
acidic catalyst medium. Since measuring the product solubility in
the highly reactive catalyst phase was not possible in a direct man-
ner, we calculated the product distribution in the acidic ionic liq-
uid phase by estimating the partition coefficient using COSMO-RS
model. Two major conclusions can be drawn from the good agree-
ment that was found between the kinetic model and the corrected
experimental data:
[4] (a) J.A. Boon, J.A. Levisky, J.L. Pflug, J.S. Wilkes, J. Org. Chem. 51 (1986) 480;
(b) C.P. DeCastro, E. Sauvage, M.H. Valkenberg, J. Catal. 196 (2000) 86.
[5] B.J. Piersma, M. Merchant, in: C.L. Hussey, S.N. Flengas, J.S. Wilkes, Y. Ito (Eds.),
Proc. 7th Int. Symp. Molten Salts, Houghton Coll., New York, 1990, p. 805.
[6] L. Xiao, K.E. Johnson, Can. J. Chem. 82 (2004) 491.
[7] (a) X. Sun, S. Zhao, Petrol. Sci. 3 (2006) 60;
(b) K. Qiao, C. Yokoyama, Chem. Lett. 33 (7) (2004) 944.
[8] K. Voo, V.V. Namboodiri, R.S. Varma, P.G. Smirniotis, J. Catal. 222 (2004) 511.
[9] N. Brausch, A. Metlen, P. Wasserscheid, Chem. Commun. (2004) 1552.
[10] V. Ladnak, N. Hofmann, N. Brausch, P. Wasserscheid, Adv. Synth. Catal. 349
(2007) 719.
[11] (a) P. Bogacki, L.F. Shampine, Appl. Math. Lett. 2 (1989) 1;
(b) J.R. Dormand, P.J. Prince, J. Comp. Appl. Math. 6 (1980) 19.
[12] (a) J.C. Lagarias, J.A. Reeds, M.H. Wright, P.E. Wright, SIAM J. Optimization 9
(1998) 112;
(a) COSMO-RS does indeed represent a useful tool to predict the
relative solubilities of the different alkylated aromatic com-
pounds in acidic chloroaluminate ionic liquids. The COSMO-RS
calculations allow correcting the data from the experimental
analysis of the organic product phase in a reasonable manner.
(b) The large difference in the solubilities of the alkylated aro-
matic compounds in the acidic chloroaluminate ionic liquid is
indeed the major reason for the observed “delay behaviour”
when taking into account only the product analyses from the
organic phase. The COSMO-RS calculations showed a signifi-
cantly lower solubility of aromatic compounds with increas-
ing degree of alkylation thus providing a reasonable fore-
cast for the experimentally observed enhanced selectivity to
monoalkylated products in the liquid–liquid biphasic ionic liq-
uid system.
(b) R. Fletcher, M.J.D. Powell, Comput. J. 6 (1963) 163;
(c) D. Goldfarb, Math. Comput. 24 (1970) 23.
[13] (a) A. Macchioni, G. Ciancaleoni, C. Zuccaccia, D. Zuccaccia, Chem. Soc. Rev. 37
(2008);
(b) H. Tokuda, K. Hayamizu, K. Ishii, M.A.B.H. Susan, M. Watanabe, J. Phys.
Chem. B 109 (2005) 6103;
(c) N. Karger, T. Vardag, H.-D. Lüdemann, J. Chem. Phys. 100 (1994) 11.
[14] A. Klamt, F. Eckert, Fluid Phase Equilib. 172 (2000) 43.
[15] (a) Z.J. Karpinski, R.A. Osteryoung, Inorg. Chem. 23 (1984) 1491;
(b) A.A.K. Abdul-Sada, A.M. Greenway, K.R. Seddon, T. Welton, Org. Mass. Spec-
trom. 28 (1993) 759;
(c) P. Wasserscheid, W. Keim, Angew. Chem. Int. Ed. 39 (21) (2000) 3772.
[16] H.A. Øye, M. Jagtoyen, T. Oksefjell, J.S. Wilkes, Mater. Sci. Forum 73 (1991) 183.
[17] (a) I.M. Kolesnikov, Chem. Tech. Fuel Oil 40 (2004) 403;
(b) A. Corma, V. Martinez-Soria, E. Schnoeveld, J. Catal. 192 (2000) 163;
U. Sridevi, B.K.B. Rao, N.C. Pradhan, S.S. Tambe, C.V. Satyanarayana, B.S. Rao,
Ind. Eng. Chem. Res. 40 (2001) 3133.
[18] M. Baerns, A. Behr, A. Brehm, J. Gmehling, H. Hofmann, U. Onken, A. Renken,
Technische Chemie, in: Wiley–VCH, Weinheim, 2006, p. 134.
[19] A.B. Newman, Trans. AIChE 27 (1931) 310.
Despite the fact that a low formal activation energy was ob-
served for the investigated isopropylation reactions, mass transfer
effects on the observed reaction rate can be ruled out for our ex-
periments as the estimated Hatta number for the reaction system
is significantly smaller than 0.3. A reasonable explanation for the
observed low formal activation energies is given by the known fact
[20] R. Shinnar, Fluid Mech. 10 (1961) 259.
[21] P.E. Liley, G.H. Thomson, D.G. Friend, T.E. Daubert, E. Buck, in: R.H. Perry, D.W.
Green, J.O. Maloney (Eds.), Perry’s Chemical Engineers’ Handbook, seventh ed.,
McGraw–Hill, Australia, 1997, p. 2-1.
[22] J. Tong, Q.S. Liu, W.G. Xu, D.W. Fang, J.Z. Yang, J. Phys. Chem. B 112 (2008).
[23] M. Eichmann, Dissertation, RWTH, Aachen, 1999, p. 179.