Evaluation of reactor concepts for the continuous production of fine chemicals using the selective hydrogenation of cinnamaldehyde over palladium catalysts

Article abstract of DOI:10.1016/j.cattod.2013.12.051

Two reactor concepts based on the pellet string reactor (PSR) were evaluated using the continuous selective hydrogenation of cinnamaldehyde as a model reaction to characterize the suitability for the application in the production of fine chemicals or pharmaceuticals. Reaction studies were carried out in a single PSR and a scale-up concept consisting of 32 parallel channels (MPSR). The impact of liquid velocity, gas holdup and pressure on the conversion and selectivity were investigated for both reactors. It was found that the continuous reactors deliver a comparably high selectivity to a batch reactor when hydrodynamic conditions are chosen adequately to obtain plug flow. The results also indicate a comparable performance for both reactors in terms of reaction rate and selectivity. Thus, the concept seems to present a promising scale-up method based on single-channel experiments.

Full text of DOI:10.1016/j.cattod.2013.12.051

G Model
CATTOD-8842; No. of Pages7
ARTICLE IN PRESS
Catalysis Today xxx (2014) xxx–xxx
Contents lists available at ScienceDirect
Catalysis Today
journal homepage: www.elsevier.com/locate/cattod
Evaluation of reactor concepts for the continuous production of fine
chemicals using the selective hydrogenation of cinnamaldehyde over
palladium catalysts
∗
Anne Müller, Martina Ludwig, Martin Arlit, Rüdiger Lange
Chair of Chemical Reaction Engineering and Process Plant, Institute of Process Engineering and Environmental Technology, Faculty of Mechanical Science
and Engineering, Technische Universität Dresden, Germany
a r t i c l e i n f o
a b s t r a c t
Article history:
Two reactor concepts based on the pellet string reactor (PSR) were evaluated using the continuous
selective hydrogenation of cinnamaldehyde as a model reaction to characterize the suitability for the
application in the production of fine chemicals or pharmaceuticals. Reaction studies were carried out
in a single PSR and a scale-up concept consisting of 32 parallel channels (MPSR). The impact of liquid
velocity, gas holdup and pressure on the conversion and selectivity were investigated for both reactors.
It was found that the continuous reactors deliver a comparably high selectivity to a batch reactor when
hydrodynamic conditions are chosen adequately to obtain plug flow. The results also indicate a compa-
rable performance for both reactors in terms of reaction rate and selectivity. Thus, the concept seems to
present a promising scale-up method based on single-channel experiments.
Received 1 October 2013
Received in revised form 5 December 2013
Accepted 8 December 2013
Available online xxx
Keywords:
Pellet string reactor
Hydrodynamics
Batch-to-conti
Heterogeneous catalysis
Multiphase reaction
Cinnamaldehyde hydrogenation
© 2014 Elsevier B.V. All rights reserved.
1
. Introduction
Over the past two decades, chemical and pharmaceutical indus-
applications are rarer. A pellet string of large particles (8.2 mm) was
employed by Satterfield et al. [9] to investigate mass transfer lim-
itations for the hydrogenation of alpha-methylstyrene to cumene.
More recently, Kallinikos et al. [10,11] and Hipolito et al. [12] have
published results on the application of a pellet string reactor for
kinetic studies, e.g. for the hydrodesulphurization of heavy gasoil
fractions. They also showed that it is possible to obtain a very close
residence time distribution for two-phase flow in this type of reac-
tor [10,12]. A detailed characterization of the mass transfer and
hydrodynamic behavior for a three-phase reaction has been pre-
sented lately by Haase et al. [13] and Langsch et al. [14,15] based
on the hydrogenation of alpha-methylstyrene.
Regarding the good performance of the PSR, in addition to the
application in catalyst screening, the parallelization of multiple
packed channels in the millimeter range as a channel array presents
an interesting concept for the production of fine chemicals. For
instance, monolith substrates can form the inert channel walls and
are packed with spherical catalyst particles. A few examples of this
integrated packing have been presented by Kapteijn et al. [16],
Dautzenberg et al. [17] and Bauer et al. [2,13]. The latter has shown
a high performance for the hydrogenation of alpha-methylstyrene.
Different terms have been employed to describe this type of pack-
ing, including “structured trickle-bed reactor” [2,16], “composite
structured packing” [17] and “composite minichannel reactor”
[13]. In this work the term multi-channel pellet string reactor
(MPSR) will be used. It represents a scale-up concept based on
try aims to switch from batchwise mode of operation to a
continuous production in order to intensify the reaction process
in terms of increased product yield and quality, to improve pro-
cess safety, to reduce the amount of catalyst or to simplify scale-up
[
1]. While most of the reactions of the so-called “flow-chemistry”
contain only a liquid or gas phase, the integration of solid cata-
lyst still presents a major challenge in reaction engineering during
the realization of continuous three-phase reactions. Flow reactors
containing parallel flow channels which are packed with catalyst
particles, so called pellet string reactors, represent a promising
concept [2]. In contrast to mini or micro fixed bed reactors, the
catalyst particles are of the same dimension as the reactor diame-
ter. Therefore the risk of fluid maldistribution in the catalyst bed is
minimized.
The concept of the pellet string (PSR) reactor has been well-
known for a long time and has been employed for measuring mass
transfer coefficients, mostly in gas-phase reactions. Some examples
are the publications of Scott et al. [3], Lee et al. [4], Sharma et al. [5],
Solcova et al. [6,7] and Takacs et al. [8]. Works on gas–liquid–solid
∗ Corresponding author. Tel.: +49 351 463 35181; fax: +49 351 463 37757.
E-mail address: ruediger.lange@tu-dresden.de (R. Lange).
0
920-5861/$ – see front matter © 2014 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.cattod.2013.12.051
Please cite this article in press as: A. Müller, et al., Evaluation of reactor concepts for the continuous production
of fine chemicals using the selective hydrogenation of cinnamaldehyde over palladium catalysts, Catal. Today (2014),
http://dx.doi.org/10.1016/j.cattod.2013.12.051

G Model
CATTOD-8842; No. of Pages7
ARTICLE IN PRESS
2
A. Müller et al. / Catalysis Today xxx (2014) xxx–xxx
Table 1
Main catalysts properties.
Catalyst
Pd/Al2O3 (A)
Pd/Al2O3 (B)
Metal content (by manufacturer) [wt.%]
0.5
3.0
250
291.27
139.40
0.0037
42.97%
20.9%
5.4
1.0
1.9
170
270.03
148.38
0.0037
41.03%
25.9%
4.3
−
3
Mean particle diameter [10 m]
Estimated thickness of active layer [10 m]
−
6
a
3
Total pore volume [mm /g]
a
2
Pore surface area [m /g]
Modal pore diameter [10 m]
a
−6
a
Porosity
b
Metal dispersion
b
−9
Metal particle size [10 m]
Fig. 1. Reaction network for the hydrogenation of cinnamaldehyde over palladium
2
3
Geometric specific surface area [m /m ]
2000.0
3157.89
catalyst.
a
Determined by mercury porosimetry.
Determined by CO chemisorption.
b
the numbering-up from one channel to several parallel channels
of the same dimension. This approach is generally applied in mini
and micro reactors to enlarge the production capacity while still
preserving the benefits of the small dimensions [18].
was assured by monitoring the inlet and outlet temperature and
regarding the moderate reaction enthalpy and low conversion.
The following procedure was applied for all experiments. First the
pressure was adjusted by feeding hydrogen into the system and
adjusting the pressure controller. Then the reactor was flushed
with liquid to ensure full wetting of the catalyst bed. Afterwards
the desired liquid flow rate was set at the pump. Once the sys-
tem temperature reached a stationary point, the hydrogen flow
was set by the mass flow controllers. The system was allowed to
reach a steady state, usually after 60 min. This was checked by ana-
lyzing liquid samples taken at the reactor outlet at several times
during a run (every 10–20 min). Sample analysis was performed
offline by an Agilent 6890N GC equipped with an HP-1 column and
an FID detector. First reaction studies were carried out to evalu-
ate the performance of the PSR. During the experiments the flow
regime was observed so that a regular slug flow was maintained at
all times. Studies on the influence of different process parameters
were carried out, including the effect of liquid velocity, gas holdup
and pressure. Superficial liquid velocity uL,s and gas velocity uG,s
were calculated based on the reactor cross-sectional area SR and
Whereas reaction studies to characterize mass transfer have
been carried out with different hydrogenation reactions, so far no
investigations on the application of this reactor concept for a selec-
tive reaction have been published. This work will therefore focus
on the investigation of a PSR for the selective hydrogenation of cin-
namaldehyde (CAL) to hydrocinnamaldehyde (HCAL) as a model
reaction to prove the beneficial reactor performance. The synthe-
sis of a large number of fine chemicals, particularly in the field of
flavor and fragrance chemistry and pharmaceuticals, involves the
selective hydrogenation of unsaturated carbonyl intermediates as a
critical step. The heterogeneously catalyzed hydrogenation of CAL
over a palladium catalyst in the solvent toluene yields HCAL as
main product and hydrocinnamyl alcohol (HCOL) as side product.
In accordance with other publications, the intermediate cinnamyl
alcohol (COL) cannot be detected [19–22]. As the unsaturated alco-
hol is the main desired product demanded from the industry,
another catalyst, like platinum, should also be tested in the future.
Kinetic studies with palladium catalysts have shown that HCOL is
formed only via the COL route and not from HCAL. Thus, the reaction
network is formed by two parallel reactions (Fig. 1).
˙
the volumetric flow rate V for the prevailing reaction temperature
TR and pressure pR (Eqs. (1)–(3)) (STP – standard temperature TSTP
and pressure pSTP).
In this contribution, results from first experimental studies in a
single PSR will be presented. For comparison reaction kinetics will
be determined in a batch reactor system. In all cases commercial
egg-shell catalysts are employed (0.5% or 1.0% Pd/␥-alumina). The
scale-up to multiple channels (MPSR) will be carried out and first
conclusions regarding the impact of the different process parame-
ters are deducted.
Superficial liquid velocity:
V˙ L
uL,s =
(1)
SR
Superficial gas velocity:
˙
V˙
V
G,STP pSTP · TR
G
uG,s
=
=
(2)
(3)
SR
SR pR · T
STP
2
. Experimental
Gas holdup:
uG,s
2.1. Chemicals and gases
HG = u
+ uL,s
G,s
Reaction studies were carried out with a solution of 50 kg/m3
3
(
380 mol/m ) trans-cinnamaldehyde (C H O, 98%, abcr) in toluene
9
8
Table 2
(
C7H , >99.8%, Fisher Scientific). Two different commercial Pd/␥-
Experimental conditions for both reactor concepts.
8
Al O egg-shell catalysts were used (0.5% Pd/␥-Al O , 3.0 mm (A);
2
3
2
3
PSR
MPSR
1
.0% Pd/␥-Al O , 1.8–2.0 mm (B)). The main catalyst properties are
2 3
◦
TR [K]
pR [bar]
uL,s [10 m/s]
uG,s [10 m/s]
353
353
10
0.8, 1.6, 2.4
0.7 . . . 7.3
0.45, 0.6, 0.75
380
given in Table 1. Prior to reaction the catalyst was reduced at 210 C
and 10 bar (a) for 4 h under flowing hydrogen. All chemicals and
commercial catalyst particles were used as received. Gases used
are of 5.0 purity (Air Liquide).
10, 20, 30
10, 20, 40, 80
9 . . . 68
0.5, 0.65, 0.8
380
−3
−
3
HG
3
cCAL,0 [mol/m ]
Re = uTP,s·ꢀL·d/ꢁL
131 . . . 716
301 . . . 1652
0.906
13 . . . 82
29 . . . 186
2
.2. Reaction studies
Re* = uTP,s·ꢀL·d/(ꢁL·(1 − εbed))
2
h
Bo = ꢀL · g · d /ꢂ
0.653
2.25·10 . . .1.46·10
−
4
−3
−1
−5
−4
−4
Ca = uTP·ꢁL/ꢂ
4.05·10 . . .2.22·10
Reaction studies were carried out isothermally at 353 K and a
2
L
−3
−5
WeL = ꢀL · u · dP /ꢂ
5.43·10 . . .3.47·10
6.09·10 . . .5.48·10
hydrogen pressure between 10 bar and 30 bar. The reaction con-
ditions as well as characteristic dimensionless numbers regarding
hydrodynamics are summarized in Table 2. Isothermal operation
dh = d ·
Krischer–Kast
ꢀ
3
3
bed
2
hydraulic diameter
16ε /(9ꢃ(1 − εbed) )
Please cite this article in press as: A. Müller, et al., Evaluation of reactor concepts for the continuous production
of fine chemicals using the selective hydrogenation of cinnamaldehyde over palladium catalysts, Catal. Today (2014),
http://dx.doi.org/10.1016/j.cattod.2013.12.051

G Model
CATTOD-8842; No. of Pages7
ARTICLE IN PRESS
A. Müller et al. / Catalysis Today xxx (2014) xxx–xxx
3
Table 3
and after the reactor. The reactor consisted of a stainless steel tube
1.4301, ID 2.4 mm, OD 6 mm), the catalyst bed was fixed inside by
wire mesh.
For the MPSR a larger scale setup for concurrent upflow
operation was designed. Downflow operation was not possi-
ble due to the very high liquid demand to achieve slug flow.
Gas flow from compressed gas cylinders was adjusted by 4
Characteristic parameters for the reactor configurations.
(
Reactor configuration
PSR
MPSR
Catalyst
1.0 wt.% Pd/Al2O3
1.9
0.5 wt.% Pd/Al2O3
3.0
Mean particle diameter dP
−
3
[
10 m]
Channel cross-section
Round
2.4
Square
3.6
Channel characteristic length D
different mass flow controllers (Bronkhorst, Brooks, flow range
−
3
[
10 m]
3
8
.0–233,000 Ncm /min), liquid was pumped by a pulsation-free
D/dP ratio
Bed length Lbed [10 m]
Bed porosity εbed
Number of channels
Direction of flow
Total catalyst mass mcat [g]
1.26
500
0.57
1
Downflow
1.79
1.20
400
0.66
32
Upflow
92
460.0
0.121
−3
3
HPLC pump (Agilent, SD-1, flow range 0–800 cm /min). The con-
necting pipes and the reactor were heated electrically. A sampling
valve was installed behind the reactor to collect liquid samples. The
pressure was adjusted with two overflow valves after the reactor
outlet. After the passage, the two-phase flow was separated inside a
separation vessel. The system pressure was monitored by pressure
transmitters (Baumer, 0. . .160 bar) before and after the reactor.
The reactor consisted of a stainless steel tube (38 mm × 2.5 mm) of
−3
Total palladium mass mPd [10 g]
17.9
3.06·10−
3
Total geometric catalyst surface
2
[
m ]
Geometric specific surface area
3157.89
2000.0
2
3
[
m /m ]
6
00 mm length. At the bottom a gas distributor was installed (per-
forated plate, 37 holes) to generate a homogeneous bubble flow of
gas and liquid at the chosen velocities [24] (checked by CFD simula-
tions). The catalyst bed was fixed 11.5 cm above the gas distributor
(Fig. 2), i.e. more than three times the reactor diameter. Therefore a
fully developed homogeneous bubble flow was assumed according
to Wilkinson et al. [25]. Monolithic substrates (39 cpsi, cordierite,
IKTS Hermsdorf) were employed to form 32 parallel channels which
were filled with catalyst particles. 4 monolith segments (100 mm
length) were aligned and stacked on top of each other to form a
bed length of 400 mm. The catalyst particles were inserted man-
ually and fixed by wire mesh on both sides of the bed. In order
to avoid catalyst by-pass phenomena, empty monolith channels
were inactivated and the monolith segments were wrapped with a
PTFE bandage before being loaded in the reactor, thus reducing the
clearance between the monolith and the reactor wall.
To characterize the reactor performance the conversion of CAL
XCAL was calculated according to Eq. (4) and the selectivity to HCAL
SHCAL according to Eq. (5), assuming volume constancy.
c
CAL
XCAL = 1 − ꢁ
(4)
(5)
ci
c
HCAL
SHCAL
=
c
+ cHCOL
HCAL
The reaction rate can be considered as limited by external and
internal mass transfer effects. Calculations of the Thiele modu-
lus based on experiments in a batch stirred tank reactor revealed
a catalyst efficiency of less than 50%, caused by the rather high
thickness of the active catalyst layer. In order to compare both
reactor concepts the surface specific reaction rate rS was calcu-
lated based on the geometric surface area of the catalyst particles
(
SP), the conversion and the initial concentration of CAL cCAL,0
3. Results and discussion
according to Eq. (6). A comparison based on the mass of palladium
would lead to deceptive results, as only part of the active metal is
utilized.
3.1. Pellet string reactor (PSR)
3
.1.1. Hydrodynamic flow observations
In order to achieve high mass transfer rates and a narrow resi-
˙
cCAL,0 · X
· VL
CAL
rS
=
.
(6)
SP
dence time distribution, it has been proven to be beneficial to work
in the slug flow regime [10,13]. Therefore experimental conditions
were chosen to produce a gas–liquid slug flow in the empty channel
before the catalytic packing. Some selected examples at different
flow conditions are shown in Fig. 3. It can be seen that the slug
flow leads to a good wetting of the catalyst bed and a periodical
exposition of the catalyst to the gas phase. Sometimes gas bubble
coalescence was observed in the empty channel when slugs became
too thin and especially at low liquid velocity and gas holdup, lead-
ing to a slightly irregular flow pattern. Furthermore, in most cases
coalescence and break-up of bubbles inside the packing occurred
caused by the changing acceleration between the spheres. The open
cross-section in the packed bed ranges from 4.1 mm2 to 1.7 mm ,
thus leading to an acceleration of the interstitial velocity by fac-
tor 2.5. That is why flow irregularities in the empty channel are
assumed to have only minor impact on the reactor performance.
2
.3. Reactor setups
Two different experimental setups were designed and installed
to test the reactor concepts. The main parameters are summa-
rized in Table 3. The first setup in laboratory scale for the PSR was
designed similarly to the setup proposed by Haase et al. [23] for con-
current downflow (Fig. 2). Gas flow from compressed gas cylinders
was adjusted by 4 different mass flow controllers (Bronkhorst, flow
range 2.0–100,000 Ncm /min), liquid was pumped by a pulsation-
free HPLC pump (Agilent, SD-1, flow range 0–200 cm /min). The
connecting pipes and the reactor were heated electrically to main-
tain the reaction temperature and temperatures were recorded by
several thermocouples (type K). Pressure resistant glass channels
3
3
2
(
Bohemia Crystal) were installed before and after the reactor to
observe the flow regime. Gas and liquid were mixed by injecting
the gas into the liquid to produce regular slug flow. The occur-
ring two-phase flow was recorded by a high-speed video camera
system (VDS-Vosskuehler, HCC1000, illumination: multiLED, GS
Vitec). A sampling valve was installed behind the reactor to col-
lect liquid samples. After the passage, the two-phase flow was
separated inside a separation vessel. A manual pressure controller
3.1.2. Variation of liquid velocity
Reaction studies on the effect of the liquid velocity were con-
ducted at constant hydrogen pressure (10 bar) and gas holdup (0.5).
The observed cinnamaldehyde conversion for different flow condi-
tions inside the PSR is depicted in Fig. 4 and reveals two different
effects. Between 0.02 m/s and 0.08 m/s the conversion decreases
with increasing velocity as the residence time in the reactor is
reduced. In contrast, at the lowest velocity 0.01 m/s the conversion
(
Tescom, type 26) was installed in the outlet gas line to regulate
the system pressure over a wide range. The system pressure was
monitored by pressure transmitters (Baumer, 0. . .250 bar) before
Please cite this article in press as: A. Müller, et al., Evaluation of reactor concepts for the continuous production
of fine chemicals using the selective hydrogenation of cinnamaldehyde over palladium catalysts, Catal. Today (2014),
http://dx.doi.org/10.1016/j.cattod.2013.12.051

G Model
CATTOD-8842; No. of Pages7
ARTICLE IN PRESS
4
A. Müller et al. / Catalysis Today xxx (2014) xxx–xxx
Fig. 2. Schematic drawing of the PSR experimental setup (left) and the MPSR reactor design (right).
is considerably lower compared to 0.02 m/s, although the residence
time is even higher. While the three points follow a linear depen-
dence with a slope proportional to the reaction rate, the point
at 0.01 m/s differs significantly. This effect could be explained by
strong external mass transfer limitations or a reduced catalyst wet-
ting. As proven by several authors [13,14,26], external mass transfer
in packed mini-channels is considerably lower at a small two-phase
velocity. Based on a mass transfer correlation for gas–liquid–solid
reactions in a PSR proposed by Langsch et al. [14], an overall mass
During the experiments at low velocity, stagnant gas bubbles in
the packing were observed by the high-speed video camera. This
results in an incomplete use of the active catalyst surface as the
transport of reaction mixture to the catalyst is hindered and there-
fore only a low conversion can be achieved. Based on these findings,
an optimization between good wetting and a high residence time
should be done.
3.1.3. Variation of gas holdup and hydrogen pressure
−
1
−1
transfer coefficient of 0.4 s
up to 1.81 s was estimated for
A variation of gas holdup at one specific liquid phase veloc-
ity was generated by increasing the gas feed flow rate. This also
increases the two-phase velocity, thus causing a shorter residence
time in the reactor. In turn, a lower conversion would be expected.
On the other hand, higher gas holdup also improves the external
mass transfer, which has been observed in a PSR for example by
Kallinikos et al. [26], Haase et al. [13] and Langsch et al. [15] as
well as in traditional fixed bed reactors in general [27]. An increase
of superficial gas velocity is supposed to improve mass transfer
based on two different phenomena. The thickness of the laminar
boundary layers is reduced due to enhanced turbulences created
the given reaction conditions. As the coefficient is almost linearly
dependent on the liquid velocity, the decreasing mass transfer rate
does not serve alone to explain the very low conversion at 0.1 m/s.
Fig. 3. Observed flow regimes above and inside the packing for different operating
conditions.
Fig. 4. Determined CAL conversion for different liquid velocities in the PSR (PH2
10 bar, HG = 0.5).
=
Please cite this article in press as: A. Müller, et al., Evaluation of reactor concepts for the continuous production
of fine chemicals using the selective hydrogenation of cinnamaldehyde over palladium catalysts, Catal. Today (2014),
http://dx.doi.org/10.1016/j.cattod.2013.12.051

G Model
CATTOD-8842; No. of Pages7
ARTICLE IN PRESS
A. Müller et al. / Catalysis Today xxx (2014) xxx–xxx
5
Fig. 5. Determined CAL conversion for different gas holdups in the PSR with pressure
effect (uL,s = 0.02 m/s).
Fig. 7. Selectivity toward HCAL in the PSR at two different pressures (ꢀ 10 bar,
᭹ 30 bar, uL,s = 0.02 m/s) and in the MPSR at two different liquid velocities (ꢁ
−
3
−3
uL,s = 0.8 × 10 m/s, ꢄ uL,s = 1.6 × 10 m/s, PH2 = 10 bar), lines indicating trends.
3
.3. Residence time distribution
by the higher interstitial velocities which results in higher mass
transfer coefficients. Secondly, increasing gas holdup also affects
the interfacial gas–solid and gas–liquid areas positively. As Langsch
et al. [15] proposed, the gas–liquid area is further enlarged because
smaller gas bubbles are produced due to bubble breakup at higher
velocity. As shown in Fig. 5, the positive effect of enhanced mass
transfer prevails in the system, leading to a significant increase
in conversion with higher gas holdup. Furthermore, the conver-
sion of CAL benefits from an increased hydrogen pressure. This can
be explained by an increased hydrogen solubility in the solvent
Orienting residence time distribution analyses were undertaken
in the PSR at uL,s = 0.04 m/s and a gas holdup of 0.5 and in the
−
3
MPSR at the lowest liquid velocity (0.8·10 m/s) and a gas holdup
of 0.45. The results show a very broad distribution for the MPSR,
which serves to explain the low selectivity at this flow conditions.
A Péclet number of approx. 2 can be estimated for an axial disper-
sion model. On the one hand that might be caused by backmixing
effects due to the slow upflow. On the other hand bypass flow might
occur as the channels in the MPSR have a square cross section, thus
liquid can pass by the catalyst particles easier at the corners com-
pared to a channel with round cross section. In the literature, it was
already shown that the residence time distribution for a single PSR
is supposed to be rather narrow for two-phase slug flow [10,12].
An example for a residence time distribution for the PSR is shown
in Fig. 8. The results obtained from step response studies indicate a
cell number of N = 16 when a cell model is applied or a Péclet num-
ber of 30 for an axial dispersion model. This corresponds to the good
selectivity that was achieved for the PSR in the reaction studies and
matches well with the correlation for the Péclet number proposed
by Kallinikos et al. [10]. Further residence time studies to verify this
behavior for the PSR are currently under preparation.
3
3
toluene from 37.1 mol/m (10 bar) to 111.2 mol/m (30 bar) accord-
ing to the correlation of Jáuregui-Haza et al. [28]. The positive effect
of higher gas holdup was observed similarly for both pressures.
3.2. Multi-channel reactor
Reaction studies on the influence of gas holdup and liquid veloc-
ity were also performed in the scale-up concept, the multi-channel
reactor. For the gas holdup similar trends compared to the PSR were
observed (Fig. 6), proving the applicability of the concept in general.
At the lowest liquid velocity, the impact of gas holdup on the con-
version is even more pronounced. In contrast, the effect of liquid
velocity gives interesting results. At the lowest gas holdup (0.45)
liquid velocity hardly affects the conversion, whereas it has a sig-
nificant impact at higher gas holdups. Also the selectivity shows
different trends for the lowest velocity compared to higher values
3.4. Comparison of reactor concepts
Most interesting for a commercial application is a comparison
of the reactor concepts regarding the achievable selectivity and
(Fig. 7). This indicates very strong mass transfer limitations and
deviation from plug flow caused by backmixing in the low velocity
range and shows the necessity of a thorough reactor characteriza-
tion during scale-up.
Fig. 8. Measured residence time distribution in the PSR (uL,s = 0.04 m/s, HG = 0.5) and
in the MPSR (uL,s = 0.8 × 10−3 m/s, HG = 0.45) with model curves ( cell model, · · ·· · ·
Fig. 6. Determined CAL conversion for different gas holdup and liquid velocities in
the MPSR (PH2 = 10 bar).
axial dispersion model).
Please cite this article in press as: A. Müller, et al., Evaluation of reactor concepts for the continuous production
of fine chemicals using the selective hydrogenation of cinnamaldehyde over palladium catalysts, Catal. Today (2014),
http://dx.doi.org/10.1016/j.cattod.2013.12.051

G Model
CATTOD-8842; No. of Pages7
ARTICLE IN PRESS
A. Müller et al. / Catalysis Today xxx (2014) xxx–xxx
6
residence time in the reactor. Also as expected, the reaction rate is
slightly increased with higher pressure.
As a result of the low space velocity in the MPSR also the reaction
rate is considerably lower. On the other hand, it shows the same
trend as the single PSR, thus, a comparable reaction rate can be
expected for operation at higher space velocity. In addition also the
selectivity would be influenced positively with rising space velocity
as already shown in Section 3.2.
4. Conclusions
In this contribution two reactor concepts based on a pellet string
reactor were evaluated for the continuous selective hydrogenation
of cinnamaldehyde to hydrocinnamaldehyde. The PSR can present
an interesting concept for the production of fine chemicals com-
pared to conventional batch reactors. As a mini-scale fixed bed
reactor it offers easy catalyst handling without separation steps and
provides a high heat transfer capacity. By applying a defined flow
regime (slug flow) in each reaction channel, very efficient catalyst
wetting and a close residence time distribution can be achieved.
Therefore a high selectivity can be reached, that is comparable to
the performance of a batch reactor. Scale-up can be easily done by
parallelization of several channels (numbering-up). Based on opti-
mization studies in a single channel, the optimum conditions for
production can be identified with less time and resources. During
scale-up special care has to be taken of the hydrodynamics. It was
shown that it is possible to reach the selectivity from batch exper-
iments also in the multi-channel PSR as long as the liquid velocity
is sufficiently high.
A direct comparison of the reactor concepts is difficult at this
stage, as the liquid velocity differs in approximately one order of
magnitude and the systems were operated in different flow modes.
Furthermore, effects of liquid maldistribution, external mass trans-
fer and by-passing on the reactor performance should be evaluated
quantitatively in more detail before a final conclusion about the
simplicity of scale-up can be drawn. Nevertheless, first experiments
indicate that the results generated in a single channel can be trans-
ferred to a larger scale reactor with multiple parallel channels while
still achieving comparable reaction rate and selectivity. Therefore
the potential of this reactor concept will be further investigated so
it can be eventually applied also for commercial syntheses.
Future work will focus on the application of a gas and liquid
distributor for the MPSR to generate a constant slug flow also in
downflow operation in order to further study the effect of residence
time distribution on the reactor performance. Moreover, additional
reaction studies should be carried out in a random fixed bed reactor
to evaluate the impact of the structured bed.
Fig. 9. Comparison of the selectivity toward HCAL in batch and continuous reactors
at a CAL conversion of approx. 16% (apart from value at conversion X = 60%).
reaction rate. Results for the selectivity are presented in Fig. 7. In
general a high selectivity of around 90% toward HCAL was achieved
in the single PSR. It was found, that the selectivity is relatively
independent of the gas holdup and only moderately affected by
the hydrogen pressure, whereas elevated pressure reduces the
selectivity. In contrast, the MPSR achieves a significantly poorer
selectivity, depending on the liquid velocity and gas holdup. Differ-
ent effects can be observed for the different liquid velocities. The
−
3
selectivity is worst at the lowest liquid velocity (0.8·10 m/s) and
is further reduced by increasing gas holdup. In contrast, at higher
liquid velocity, increasing the gas holdup has a positive effect on
the selectivity. For the latter case, the selectivity reaches almost
the values of the single PSR. This shows again the strong impact of
the flow conditions on the reactor performance. The low selectivity
indicates strong backmixing in the MPSR at low liquid throughput
which is further impaired by a high gas holdup. A very low Péclet
number of only around 2 was estimated by residence time distri-
bution studies, as discussed in Section 3.3. With increasing liquid
velocity flow conditions should resemble plug flow, thus leading to
an improved performance.
For further comparison batch experiments were performed with
the same catalysts assuming the absence of external mass transfer
limitations and ideal mixing. The achieved selectivity at a conver-
sion of 16% is about the same as in the single PSR (92.8%), as shown
in Fig. 9. For the MPSR the selectivity of the batch experiments is
almost achieved for the higher liquid velocities, whereas for the
lowest liquid velocity the selectivity is certainly lower (around
7
3%).
An additional comparison based on the surface specific reaction
rate and the space velocity is presented in Fig. 10. In general it was
found that the reaction rate increases with rising space velocity up
to a certain optimum due to the reduction of external mass trans-
fer limitations. For the PSR at 10 bar an opposite trend for higher
space velocities is already indicated, which is caused by the reduced
Acknowledgements
The authors thank the Federal Ministry of Research and Edu-
cation (BMBF) for funding parts of this work. A.M. gratefully
acknowledges the German Academic Exchange Service (DAAD) for
a travel scholarship to attend the conference CAFC10.
References
[
[
1] G.S. Calabrese, S. Pissavini, From batch to continuous flow processing in chem-
icals manufacturing, AIChE Journal 57 (2011) 828–834.
2] T. Bauer, S. Haase, Comparison of structured trickle-bed and monolithic reactors
in Pd-catalyzed hydrogenation of alpha-methylstyrene, Chemical Engineering
Journal 169 (2011) 263–269.
[
3] D.S. Scott, W. Lee, J. Papa, Measurement of transport coefficients in gas–solid
heterogeneous reactions, Chemical Engineering Science 29 (1974) 2155–2167.
Fig. 10. Surface specific reaction rates vs. space velocity in the PSR (ꢀ 10 bar, ᭹
[4] M.D. Lee, S.H. Shen, T.K. Chung, C.K. Kuei, Flow and dispersive behaviors of
cylindrical pellet string reactor, AIChE Journal 30 (1984) 639–641.
3
0 bar) and MPSR (ꢁ 10 bar).
Please cite this article in press as: A. Müller, et al., Evaluation of reactor concepts for the continuous production
of fine chemicals using the selective hydrogenation of cinnamaldehyde over palladium catalysts, Catal. Today (2014),
http://dx.doi.org/10.1016/j.cattod.2013.12.051

G Model
CATTOD-8842; No. of Pages7
ARTICLE IN PRESS
A. Müller et al. / Catalysis Today xxx (2014) xxx–xxx
7
[
[
[
[
[
5] R.K. Sharma, D.L. Cresswell, E.J. Newson, Effective diffusion coefficients and
tortuosity factors for commercial catalysts, Industrial & Engineering Chemistry
Research 30 (1991) 1428–1433.
6] O. Solcova, P. Schneider, Axial dispersion in single pellet string columns
packed with cylindrical particles, Chemical Engineering Science 49 (1994) 401–
[17] F.M. Dautzenberg, M. Mukherjee, Process intensification using multifunctional
reactors, Chemical Engineering Science 56 (2001) 251–267.
[18] M.N. Kashid, A. Gupta, A. Renken, L. Kiwi-Minsker, Numbering-up and mass
transfer studies of liquid–liquid two-phase microstructured reactors, Chemical
Engineering Journal 158 (2010) 233–240.
[19] A. Cabiac, T. Cacciaguerra, P. Trens, R. Durand, G. Delahay, A. Medevielle, D.
Plée, B. Coq, Influence of textural properties of activated carbons on Pd/carbon
catalysts synthesis for cinnamaldehyde hydrogenation, Applied Catalysis A:
General 340 (2008) 229–235.
[20] G.R. Cairns, R.J. Cross, D. Stirling, Hydrogenation of cinnamaldehyde using cat-
alysts prepared from supported palladium phosphine complexes, Journal of
Catalysis 166 (1997) 89–97.
[21] A. Hammoudeh, S. Mahmoud, Selective hydrogenation of cinnamaldehyde over
Pd/SiO2 catalysts: selectivity promotion by alloyed Sn, Journal of Molecular
Catalysis A: Chemical 203 (2003) 231–239.
[22] V. Satagopan, S.B. Chandalia, Selectivity aspects in the multiphase hydrogena-
tion of alpha,beta-unsaturated aldehydes over supported noble-metal catalysts
– I, Journal of Chemical Technology and Biotechnology 59 (1994) 257–263.
[23] S. Haase, T. Bauer, New method for simultaneous measurement of hydro-
dynamics and reaction rates in a mini-channel with Taylor flow, Chemical
Engineering Journal 176–177 (2011) 65–74.
[24] Y.T. Shah, B.G. Kelkar, S.P. Godbole, W.D. Deckwer, Design parameters estima-
tions for bubble column reactors, AIChE Journal 28 (1982) 353–379.
[25] P.M. Wilkinson, A.P. Spek, L.L. van Dierendonck, Design parameters estima-
tion for scale-up of high-pressure bubble columns, AIChE Journal 38 (1992)
544–554.
[26] L.E. Kallinikos, N.G. Papayannakos, Intensification of hydrodesulphurization
process with a structured bed spiral mini-reactor, Chemical Engineering and
Processing: Process Intensification 49 (2010) 1025–1030.
4
08.
7] O. Solcova, K. Soukup, P. Schneider, Diffusion coefficients and other transport
characteristics of peculiarly shaped porous materials in the single pellet-string
column, Microporous and Mesoporous Materials 91 (2006) 100–106.
8] K. Takacs, H.P. Calis, A.W. Gerritsen, C.M. VanDenBleek, The selective catalytic
reduction of nitric oxide in the bead string reactor, Chemical Engineering Sci-
ence 51 (1996) 1789–1798.
9] C.N. Satterfield, A.A. Pelossof, T.K. Sherwood, Mass transfer limitations in a
trickle-bed reactor, AIChE Journal 15 (1969) 226–234.
[
10] L.E. Kallinikos, N.G. Papayannakos, Fluid dynamic characteristics of a structured
bed spiral mini-reactor, Chemical Engineering Science 62 (2007) 5979–5988.
11] L.E. Kallinikos, N.G. Papayannakos, Operation of a miniscale string bed reactor in
spiral form at hydrotreatment conditions, Industrial & Engineering Chemistry
Research 46 (2007) 5531–5535.
12] A.I. Hipolito, M. Rolland, C. Boyer, C. de Bellefon, Single pellet string reactor
for intensification of catalyst testing in gas/liquid/solid configuration, Oil & Gas
Science and Technology – Revue d’IFP Energies nouvelles 65 (2010) 689–701.
13] S. Haase, M. Weiss, R. Langsch, T. Bauer, R. Lange, Hydrodynamics and mass
transfer in three-phase composite minichannel fixed-bed reactors, Chemical
Engineering Science 94 (2013) 224–236.
[
[
[
[
[
14] R. Langsch, S. Haase, R. Lange, Hydrodynamics and mass transfer in a pel-
let string reactor for gas–liquid–solid reactions, Chemie Ingenieur Technik 85
(
2013) 642–655.
15] R. Langsch, J. Zalucky, S. Haase, R. Lange, Investigation of a packed bed in a mini
channel with a low channel-to-particle diameter ratio: flow regimes and mass
transfer in gas–liquid operation, Chemical Engineering and Processing: Process
Intensification 75 (2014) 8–18.
[27] F. Larachi, M. Cassanello, A. Laurent, Gas–liquid interfacial mass transfer in
trickle-bed reactors at elevated pressures, Industrial & Engineering Chemistry
Research 37 (1998) 718–733.
[
16] F. Kapteijn, T.A. Nijhuis, J.J. Heiszwolf, J.A. Moulijn, New non-traditional mul-
[28] U.J. Jáuregui-Haza, E.J. Pardillo-Fontdevila, A.M. Wilhelm, H. Delmas, Solubility
of hydrogen and carbon monoxide in water and some organic solvents, Latin
American Applied Research 34 (2004) 71–74.
tiphase catalytic reactors based on monolithic structures, Catalysis Today 66
(
2001) 133–144.
Please cite this article in press as: A. Müller, et al., Evaluation of reactor concepts for the continuous production
of fine chemicals using the selective hydrogenation of cinnamaldehyde over palladium catalysts, Catal. Today (2014),
http://dx.doi.org/10.1016/j.cattod.2013.12.051