Chemical Engineering Science 138 (2015) 465–472

Contents lists available at ScienceDirect

Chemical Engineering Science

journal homepage: www.elsevier.com/locate/ces

Design and optimization of a catalytic membrane reactor for the direct

synthesis of propylene oxide
E. Kertalli a, M.F. Neira d'Angelo a, J.C. Schouten a,n, T.A. Nijhuis b
a
Eindhoven University of Technology, Department of Chemical Engineering and Chemistry, Laboratory of Chemical Reactor Engineering, P.O. Box 513, 5600
MB Eindhoven, The Netherlands
b
Sabic, Geleen, The Netherlands

H I G H L I G H T S

Membrane reactor design for propylene oxide direct synthesis in liquid phase.

Numerical simulation for describing the reactor concept.

In situ combination of two consecutive reactions (hydrogen peroxide synthesis and propylene epoxidation).

Optimization of the membrane reactor main properties.

Scaling up of the membrane reactor.

art ic l e i nf o a b s t r a c t

Article history: Using numerical simulations, a new membrane reactor is proposed for the direct synthesis of propylene
Received 2 June 2015 oxide (PO) in liquid phase. The reactor is a combination of two consecutive catalytic reactor units, one for
Received in revised form the hydrogen peroxide (H2O2) synthesis on a Pd/SiO2 catalytic membrane layer, and the second for the
25 August 2015
conversion of hydrogen peroxide with propylene (C3H6) to PO on a titanium silicalite-1 (TS-1) catalytic
Accepted 27 August 2015
Available online 6 September 2015
layer. The membrane reactor is described numerically by a set of kinetic-diffusion mass balance equa-
tions. The optimization of the reactor design is achieved by determining membrane pore size, thickness
Keywords: and gas pressures which provide conversion and selectivity performance comparable to the industrial
Hydrogen peroxide requirements. An optimal pore size of 0.2–0.4 μm was found for the Pd/SiO2 membrane layer. The results
Propylene oxide
show that a Pd/SiO2 membrane thickness of 250 μm and a TS-1 layer of 100 μm are necessary to ensure
Catalytic membrane
conversion and selectivity performance of the catalytic membrane reactor comparable to the industrial
Reactor design and optimization
Scaling up ones. Calculated these optimized dimensions of the membrane reactor, a total membrane area of
84,000 m2 is required for the production of 300 kton/year of propylene oxide.
& 2015 Elsevier Ltd. All rights reserved.

1. Introduction
Propylene oxide (PO) is an important compound due to its
versatility in the chemical industry. Since the main traditional
processes for PO production are becoming less and less attractive
due to economic and environmental concerns (Nijhuis et al.,
2006), possible alternatives have been searched in the last dec-
ades. The successful implementation of a new process on a large
scale lies on its capacity to provide safety, economic, environ-
mental and performance benefits (Cavani and Teles, 2009). The
first step towards meeting industrial requirements (Cavani, 2010;
n
Corresponding author
E-mail address: j.c.schouten@tue.nl (J.C. Schouten).
Sinha et al., 2004) for alternative PO production routes has been
achieved with the discovery of titanium silicalite-1 catalyst (Tar-
amasso et al., 1983). A successful application of this zeolitic
material is achieved in the HPPO process, where the epoxidation of
propylene with H2O2 leads to PO and water formation. As a result,
two large scale plants are currently operating in Belgium and
Thailand with a capacity of 300 and 390 kton/year of PO, respec-
tively (Bassler and Weidenbach, 2010; The Dow Chemical Com-
pany, 2014). However, the high cost of H2O2 production for a large
part determines the cost of the PO synthesis process (Huang and
Haruta, 2011). Therefore, the direct synthesis of propylene oxide
from hydrogen, oxygen and propylene was proposed as a green
and economically advantageous alternative (Monnier, 2001). Two
main streams have been distinguished in the research approach:
gas vs liquid phase reaction. The gas phase reaction is mainly

http://dx.doi.org/10.1016/j.ces.2015.08.034
0009-2509/& 2015 Elsevier Ltd. All rights reserved.
466 E. Kertalli et al. / Chemical Engineering Science 138 (2015) 465–472
focused on the bi-functional Au–titanium based catalyst (i.e., TiO2,
Ti/SiO2 and TS-1) (Nijhuis et al., 2005, 2010; Yap et al., 2004;
Bravo-Suárez et al., 2008). Despite the high PO selectivity of the
Au–Ti catalyst, several limitations such as low propylene conver-
sion and hydrogen efficiency, safety concerns and catalyst pro-
ductivity make the gas phase PO synthesis still far from being
operated on a large scale (Sinha et al., 2004). The second option is
the liquid phase PO synthesis on Pd or Pd–Pt/TS-1 catalyst (Laufer,
1999; Meiers and Hölderich, 1999; Jenzer et al., 2001). The reaction
occurs in two steps where H2O2 is first in situ generated on Pd
particles and then utilized on titanium silicalite-1 catalyst. Several
limitations are encountered in the direct PO liquid phase synth-
esis. The explosive nature of the reactants, the instability of H2O2
(Samanta, 2008) and the undesired hydrogenation of propylene to
propane (Jenzer et al., 2001) have been reported as obstacles to the
high PO selectivity and productivity. The limited PO yields are
related to the low selectivity of H2O2, which is the rate limiting
step for the direct synthesis of propylene oxide (Monnier, 2001).
Furthermore, the palladium nanoparticles, responsible for H2O2
formation, are also active in its hydrogenation and decomposition
to water (Samanta, 2008) lowering its selectivity. Therefore, the
controlled and separated feed of hydrogen in the system is a key
parameter to control in order to improve the performance of the
direct synthesis of propylene oxide by improving the H2O2 and PO
selectivity and safety concerns. This can be achieved with an
engineering approach where the reactor design allows this opti-
mization of the process. A possible reactor device to implement in
the direct PO synthesis is the membrane reactor where controlled
and separated feed of the reactants can easily be obtained.
In the present work we present a new catalytic membrane
reactor concept for the direct propylene oxide synthesis in the
liquid phase. This reactor design is the coupling of two consecutive
reactions, H2O2 synthesis and propylene epoxidation. The reactor
design allows one to feed hydrogen and oxygen (in excess)
through the membrane separately from the propylene. This con-
figuration limits the hydrogenation of H2O2 and propylene and
improves the safety operation of the process. In this system, the
PO production is optimized by maximizing separately the selec-
tivity of H2O2 (rate limiting step) and epoxidation. This optimiza-
tion of the reactor concept consists in ensuring that H2O2 synthesis
step reaches selectivity targets of 50–60%, for hydrogen conver-
sions higher then 10% (García-Serna et al., 2014) and PO selectivity
of 50%, based on H2O2, and conversions of 90% (Cavani and Teles,
2009; Chowdhury et al., 2006). Moreover, the two reaction steps
are then combined in a single reactor unit allowing the utilization
of H2O2 in situ produced immediately lowering its concentration
in the reactor. This limits the decomposition of H2O2 on the Pd
catalyst improving even further the selectivity of H2O2, the rate
limiting step of the overall process (Paunovic et al., 2015). The
optimization of the reactor design was achieved with the imple-
mentation of a mathematical model which describes the two step
reactor concept. Key parameters such as the membrane pore size,
the membrane thicknesses and the gas concentrations are eval-
uated in order to ensure conversion and selectivity values which
are competitive with existing processes (García-Serna et al., 2014;
Cavani and Teles, 2009; Chowdhury et al., 2006). The optimized
reactor design was then scaled up to a plant capacity of 300 kton/
year of PO.
2. Reactor concept
A schematic representation of the catalytic membrane reactor
is shown in Fig. 1. The reactor consists of two separated units
where two consecutive reactions occur. The first reactor unit is a
membrane contactor layer of Pd/SiO2 catalyst where hydrogen and
Fig. 1. Longitudinal section of a tubular membrane reactor and catalytic layers.
oxygen react to form H2O2. The concentrations of hydrogen and
oxygen are kept outside the explosive limits. The second reactor
unit is a titanium silicalite-1 (TS-1) catalytic layer. The H2O2 pro-
duced in the first step reacts with propylene to PO on TS-1. The
first requirement for both reaction steps to occur is the presence of
a liquid phase (Cavani and Teles, 2009). The consensus in the lit-
erature is that methanol is the best solvent for the direct synthesis
of PO due to high reaction rates (Russo et al., 2013) and high
hydrogen solubility (Edwards et al., 2005). Furthermore, the pre-
sence of methanol improves the H2O2 synthesis by removing the
strongly adsorbed hydrogen peroxide from the catalytic sites.
Methanol flow is also used to sweep the PO produced in the
membrane reactor. Therefore, both catalytic layers are solvent fil-
led. Oxygen and hydrogen, fed to the membrane contactor in the
gas phase, dissolve and diffuse in the liquid filled pores of Pd/SiO2
catalytic layer. These gases react on the Pd catalyst to form H2O2.
The H2O2 is either reduced to water in the presence of hydrogen
(on Pd catalyst) or diffuses to the interface of the catalytic layers.
Here, H2O2 diffuses in the second layer (TS-1) and immediately
reacts with propylene, present in the methanol, to form PO.
The optimization of the reactor design is achieved by deter-
mining key parameters such as the membrane pore size and
thickness and the concentrations of the reactants. The membrane
reactor concept, chosen for this study, presents a tubular geo-
metry. Compared to the planar geometry, the tubular one is more
advantageous for industrial applications due to the higher area
provided and easier sealing of the edges. However, the present
model applies also to planar membranes with similar operation
conditions as the tubular one (reactants residence time and
transport phenomena through the membrane contactor).
3. Model reactor approach
3.1. Modeling equations
The model consists of a set of mass balance equations which
describes the concentrations profile of all the components through
the reactor. A schematic representation of these profiles is shown
in Fig. 2. The results from the simulation are used to calculate the
conversion, the selectivity and the flux of the different compounds.
Two reactor units have been considered:

Pd/SiO2 catalytic membrane layer (unit 1),

TS-1 layer catalytic layer (unit 2).
(a) Unit 1: A well mixed gas phase (hydrogen and oxygen) is
considered to diffuse into the pores of the Pd/SiO2 catalytic
membrane layer filled with a stagnant liquid film (methanol). The
 E. Kertalli et al. / Chemical Engineering Science 138 (2015) 465–472 467

Deff−i d2CH2O2
2
⁎ ± ri = 0
rTS −1 dx2 (8)

where i¼ PO, H2O2, C3H6.

The reaction term is positive for PO and negative for H2O2 and
propylene due to their respective formation and consumption in
the epoxidation reaction.
The reaction rate for PO synthesis from H2O2 and propylene is
adapted from the work of Wu et al. (2013) where the propylene
epoxidation was performed in a fixed bed reactor on a supported
TS-1 catalyst:
⎛ kPO ⁎RT ⁎KH2O2 ⁎CH2O2 ⁎CC3H6 ⁎ϕ CH3OH ⁎HE,CH3OH ⎞
r PO = ⎜⎜ ⎟⎟
⎝ 1 + K H2O2 ⁎ C H2O2 + K C 3H 6 ⁎ ϕ CH3OH ⁎ H E,CH3OH ⁎ RT + KPO ⁎ CPO ⎠

Fig. 2. Schematic illustration of the concentrations profile along the catalytic (9)
⁎a
membrane reactor studied in the mathematical model.
The components which do not participate in the reaction on
synthesis of H2O2 (Eq. (1)) and its further reduction to water (Eq. the TS-1 catalyst are assumed to simply diffuse along this catalytic
(2)) are considered to take place on the Pd catalyst at rates rH2O2 layer. In this case, the mass balance contains only a diffusion term:
and rred, respectively:
Deff−i d2Ci
⁎ 2 =0
H2 + O2 = H2 O2 (1) 2
rTS −1 dx (10)

where i¼ O2, H2, H2O.

H2 O2 + H2 = 2H2 O (2)
The system is operated at steady state and isothermal condi-
The mass balance of the components involved in the reactions tions. The set of ordinary differential equations previously
above is described as follows: described is solved with Athena Visual Studio software. The
simulation results were used to calculate key parameters such as
Deff−i d2Ci conversion (Xi), selectivity (Si) and fluxes (Ji). These parameters
2
⁎ 2 ± ri = 0
rmembr dx (3) were calculated as shown below:
where i¼O2, H2, H2O2, H2O. The reaction term ri is positive for Ji−in − Ji−out
Xi = ⁎100
H2O2 and water formed in this catalytic layer and negative for Ji−in (11)
oxygen and hydrogen. The equation for oxygen contains a single
reaction term due to the consumption in the direct H2O2 synthesis. Ciproduced
The reaction rate expression for hydrogen is described utilizing Si = ⁎100
Creacted compound (12)
two negative reaction terms due to its consumption in the H2O2
synthesis and in the reduction of H2O2 to water. dCi dC
The kinetic rate expressions published by Voloshin et al. Ji = Deff |in − Deff i |out
dx dx (13)
(Voloshin et al., 2007; Voloshin and Lawal, 2009),which provide
complete information on the reactions involved, were used in this
study: 3.2. Boundary conditions

k H2O2 ⁎CPd ⁎(RT )2⁎KH2 ⁎KO2 ⁎CH2 ⁎CO2 Fig. 3 shows a schematic representation of the boundaries of
r H2O2 =
(1 + RT ⁎KH2 ⁎CH2 + RT ⁎KO2 ⁎CO2 )2 (4) the catalytic membrane reactor.
Center refers to the interface between the two reactor units.
Right and left refer to the two extremes of the catalytic membrane
k red ⁎CPd ⁎RT ⁎KH2 ⁎KH2O2 ⁎CH2 ⁎CH2O2 reactor where hydrogen–oxygen (left) and propylene–methanol
rred =
(1 + (RT ⁎KH2 ⁎CH2 )0.5 + (KH2O2 ⁎CH2O2 )0.5)4 (5) (right) are respectively fed.
Center: For the boundary condition in the center we assume
The components which do not participate in the reactions on the that the flux of each component that leaves one of the reactor
Pd catalyst are assumed to simply diffuse along the Pd/SiO2 units equals the flux that enters the other reactor unit. Moreover,
membrane layer. The mass balance of these components (i.e., the
unreacted compounds in the epoxidation reaction) contains only a
diffusion term:

Deff−i d2Ci
2
⁎ 2 =0
rmembr dx (6)

where i¼C3H6, PO.

(b) Unit 2: The epoxidation of propylene with the in-situ gen-
erated H2O2 takes place on the titanium silicalite-1 catalytic layer
at a rate rPO:
H2 O2 + C3 H6 = PO + H2 O (7)
Fig. 3. Schematic representation of the boundary conditions for the membrane
The mass balance of the components involved in the epoxida- reactor. left = GO2− H2/Lmethanol interface. right = C3H6 -methanol/TS-1 interface.
tion reaction is the following equation: center¼ interface between the two catalytic layers.
468 E. Kertalli et al. / Chemical Engineering Science 138 (2015) 465–472

the concentration of each component in both catalytic layers is the Table 1

same in the center: Constant values utilized in the model approach

Deff−i dCi−membr D dC Name Value Units

⁎ + eff−i ⁎ i−TS−1 = 0
rmembr dx r TS−1 dx (14)
Deff 1–10 m2/s
kH2O2 0.015 molH2 O2/gPd s
kH2O 0.0774 molH2 O2/gPd s
Ci−membr − Ci−TS−1 = 0 (15)
kPO 0.179 molPO/ gTS − 1 m3 s
kH2 0.567 1/atm
where i refers to all the components.
kO2 0.367 1/atm
Right and left: Eq. (16) describes the boundary condition in the
two extremes of the catalytic units. Here, the concentration of the kH2O2 0.00241 m3/mol

reactants is known being it the inlet concentration of each reac- kC3H6 0.00072 m3/mol

tant: KPO 0.00174 m3/mol

CPd 6.0E þ4 gPd/m3
Ci−membr = Ci−in−membr (16) HeH2 0.78 mol/m3 atm
HeO2 1.3 mol/m3 atm
where i ¼O2, H2, C3H6. HeC3H6 260 mol/m3 atm
The inlet concentrations of the components in the liquid phase ϕC3H6 1 –
have been calculated from partial pressures and solubility data of a 5.73 n 2.35 –
the respective gases in methanol according to Eq. (17). Henry's
constants were taken from literature data (Battino and Collo-
boration, 1981; Young and Colloboration, 1981; Shin and Chad- the present work are 0.35 and 2, respectively (Li, 2008). The values
wick, 2010): of the constants utilized in the model reactor approach are sum-
marized in Table 1.
Ci = He ⁎Pi (17)

Eq. (18) refers to the right boundary condition. In this position, the
amount of each component that exits the reactor equals that in the 4. Results and discussion
methanol solution:
Deff−i dCi−TS−1 The optimization of the catalytic membrane reactor in terms of
⁎ ⁎A − Vmeth ⁎Ci−TS−1 = 0 pore size and thickness is achieved by studying separately the two
r TS−1 dx (18)
steps of the reaction and combining them in the final reactor
where i ¼O2, H2, H2O2, PO, H2O. design. The effect of methanol flow rate and gas partial pressures
As left boundary condition, Eq. (19) describes the mass balance has also been included in order to observe the best operating
of propylene at the interface between unit 1 and the gas phase. conditions for the direct propylene oxide synthesis.
The flux of propylene at the exit of unit 1 equals the amount of
propylene in the gas phase, which is in equilibrium with the liquid
4.1. Optimization of the Pd/SiO2 membrane pore size
phase. m is the partition coefficient of propylene between gas and
liquid phase:
The presence of methanol is essential for the direct synthesis of
Deff−C3H6 dCC3H6 −membr PO. The pore size of the membrane plays an important role in
⁎ ⁎A − Vsweep ⁎CC3H6 −membr ⁎m = 0
rmembr dx (19) ensuring complete wettability of the membrane reactor with the
solvent. The membrane morphology should also prevent the liquid
m is defined as
to exit the reactor. This is achieved by controlling the pressure
Cgas−C3H6 inside the pores according to the following equation (Li, 2007):
m=
Cliq−C3H6 (20) Pgas − P liq < Pcapillary (23)
Eq. (21) describes the left boundary condition for low-volatility Therefore, the pore size of the Pd/SiO2 membrane should be
compounds at the interface between unit 1 and the gas phase. The such that the Young–Laplace equation is satisfied (Li, 2007):
fluxes of these compounds are zero:
2⁎γ ⁎ cos θ
dCi−membr Pcapillary =
=0 porerad (24)
dx (21)
This equation assumes that the pores of the membrane are
where i ¼H2O2, PO, H2O
cylindrical, which is largely accepted assumption (Bhattacharyya
and Butterfield, 2003).
3.3. Diffusion coefficients
Therefore, for a given hydrophilic material (i.e., silica) and a
liquid medium (i.e., methanol), the pore size of the membrane can
The effective diffusion coefficients are defined as follows:
be directly correlated with the capillary pressure as shown
ϵ
Deff = DAB ⁎ in Fig. 4. The contact angle for hydrophilic materials is smaller
τ (22)
then 90o (30o).
Knudsen diffusion is excluded due to the presence of methanol Mild differential pressures are desired to ensure mechanical
in the pores of the Pd/SiO2 (Satterfield, 1969). Therefore, bulk stability of the Pd/SiO2 catalytic layer (Hsieh, 1996).
diffusion of the gas in the liquid is the main type of diffusion. In addition, limiting the differential pressure to 1.5 atm ensures
Reasonable values for the diffusion coefficients of a gas in a liquid that the viscous flow of the gases through the membrane will not
phase are between [1–10]n10  9 m2/s (Satterfield, 1969). The bulk take place (Li, 2007). Therefore, an estimation of the pressure
diffusion coefficient is corrected with the porosity and tortuosity window in the system can be obtained according to the following
of the membrane. The values of porosity and tortuosity assumed in equation:
 E. Kertalli et al. / Chemical Engineering Science 138 (2015) 465–472
Fig. 4. Capillary pressure dependence on the pore size of the Pd/SiO2 catalytic
membrane layer.
Fig. 5. H2 flux and conversion as a function of the Pd/SiO2 catalytic membrane
thickness. CH2 = 6 mol/m3, CO2 = 10 mol/m3, CC3H6 = 2000 mol/m3, methanol ¼
20 ml/h. TS-1 thickness: 40 μm
Pgas − P liq = 1.5 atm
(25)
Thus, a Pd/SiO2 membrane layer with a pore size within 0.2-
0.4 μm is enough to guarantee the correct operation of the
membrane reactor.
4.2. Optimization of the Pd/SiO2 membrane thickness
As shown in Figs. 5 and 6, the membrane thickness influences
the H2O2 synthesis. To study and optimize the Pd/SiO2 membrane
thickness, parameters such as gas partial pressure (Pi), TS-1
thickness and methanol flow rate were kept constant. The effect of
these parameters has been studied in more detailed further on in
this work.
Fig. 5 shows the dependence of the hydrogen flux at the exit of
Pd/SiO2 membrane and the hydrogen conversion on the membrane
thickness. The hydrogen flux across the membrane decreases with the
membrane thickness as expected from Fick's law. However, the non-
linear shape of the curve clearly indicates the contribution of both
phenomena: diffusion and reaction. The optimal design of the mem-
brane reactor would require the complete consumption of the
hydrogen in the Pd/SiO2 layer resulting in no hydrogen flux exiting
the membrane. This can be obtained for membranes thicker than
350 μm. However, Fig. 5 also shows the dependence of the hydrogen
469
Fig. 6. H2 utilization as a function of the Pd/SiO2 catalytic membrane thickness.
CH2 = 6 mol/m3, CO2 = 10 mol/m3, CC3H6 = 2000 mol/m3, methanol¼ 20 ml/h. TS-1
thickness: 40 μm.
Fig. 7. H2O2 flux and conversion as a function of the TS-1 catalytic membrane
thickness. CH2 = 6 mol/m3, CO2 = 10 mol/m3, CC3H6 = 2000 mol/m3, methanol=
20 ml/h . Pd/SiO2 thickness: 250 μm.
conversion on the membrane thickness. Industrial requirements for
hydrogen conversion of 50% are already reached with a membrane
thickness of 150 μm. However, high hydrogen conversions are directly
related to the H2O2 selectivity (Samanta, 2008). High concentrations of
H2O2, in the presence of hydrogen, favour the formation of water.
Therefore, the hydrogen utilization becomes an important parameter
in order to reach the industrial selectivity targets of 60% (García-Serna
et al., 2014). Fig. 6 shows the hydrogen utilization for different
membrane thicknesses. For thinner membranes, the hydrogen is
mainly utilized in the H2O2 synthesis reaction. The formation of water
is slower than that of the H2O2. However, if the membrane thickness
becomes larger than 250 μm, the increase in H2O2 concentration
favours its reduction to water which then becomes significant. A
membrane thickness of 250 μm ensures a H2O2 selectivity of 70%,
which is within the industrial requirements.
The hydrogen liquid loss represents the amount of unreacted
hydrogen which diffuses through the TS-1 catalytic layer.
4.3. Optimization of the TS-1 thickness layer
Fig. 7 shows the influence of the TS-1 thickness layer on the
flux and conversion of H2O2, which was previously synthesized in
the optimized 250 μm Pd/SiO2 thickness and 0.2 μm pore size.
The H2O2 flux decreases with the thickness of the catalytic
layer as expected from Fick's diffusion law. A TS-1 layer of 70 μm is
470 E. Kertalli et al. / Chemical Engineering Science 138 (2015) 465–472

Fig. 8. H2O2 concentration and selectivity dependence on the TS-1 thickness layer
Fig. 9. The dependence of H2O2 conversion on the methanol flow rate for two
at the interface between Pd/SiO2 membrane–TS-1 catalyst. CH2 = 6 mol/m3,
different TS-1 thickness layers. CH2 = 6 mol/m3, CO2 = 10 mol/m3, CC3H6 =
CO2 = 10 mol/m3, CC3H6 = 2000 mol/m3, methanol ¼20 ml/h. Pd/SiO2 thickness:
2000 mol/m3 , Pd/SiO2 thickness¼ 250 μm
250 μm.

enough to ensure a H2O2 conversion of 90%, which is within

industrial targets. However, since the hydrogen peroxide con-
sumption on the TS-1 catalyst is a consecutive reaction of its
synthesis on Pd/SiO2 catalyst, the concentration and the selectivity
of H2O2 available at the unit 1–unit 2 interface are influenced by
the thickness of the TS-1 layer as shown in Fig. 8. This is attributed
to the change of the H2O2 concentration, driving force across the
catalytic layer.
An increase in the thickness of the TS-1 layer results in an
increase of the H2O2 concentration at the interface Pd/SiO2–TS-1
catalyst. Higher H2O2 concentrations are desired for obtaining a
higher PO production. However, higher concentrations of H2O2
also promote the H2O2 hydrogenation reaction, therefore the
selectivity of H2O2 at the catalysts interface decreases. In order to
guarantee high H2O2 concentration and selectivity values to
Fig. 10. H2 flux and H2O2 selectivity dependence on the H2 partial pressure.
respond to industrial requirements (70%), a TS-1 thickness of CO2 = 10 mol/m3, CC3H6 = 2000 mol/m3, methanol ¼ 20 ml/h, Pd/SiO2 thickness¼
100 μm is required. 250 μm, TS-1 thickness¼ 100 μm.

4.4. Optimization of the methanol flow rate

High methanol flow rates are beneficial to increase the gradient

of the product concentration across the membrane layers (i.e.,
driving force for diffusion) and for the removal of the heat released
during the exothermic reaction (Russo et al., 2013). However, the
presence of high methanol flow rates contributes to a dilution of
the final product, which is undesired for a large scale process. This
leads to the need of expensive separation units, thus limiting the
economic feasibility of the process.
Fig. 9 also shows the dependence of the H2O2 conversion on
methanol flow rate as a function of TS-1 thickness. Higher H2O2
conversion can be achieved by decreasing the methanol flow rate.
Lower methanol flow rates increase the concentration of H2O2 in
the system, which leads to higher H2O2 conversions and lower
selectivities. Notice that the H2O2 conversion depends on metha- Fig. 11. O2 flux and H2O2 selectivity dependence on the O2 partial pressure.
nol flow rate only at low flow rates values, whereas the H2O2 CH2 = 6 mol/m3, CC3H6 = 2000 mol/m3, methanol ¼20 ml/h, Pd/SiO2 thickness¼
conversion becomes independent of the methanol flow rate for 250 μm, TS-1 thickness¼ 100 μm.
higher methanol flow rates. For high methanol flow rates, H2O2
becomes too diluted and it does not influence the H2O2 conversion 4.5. Optimization of PO2 and PH2
any further. For a thicker TS-1 layer (100 μm), a methanol flow rate
of 2 ml/h is sufficient to use in the system. With lower methanol The influence of gas partial pressures on the hydrogen flux and
flow rates, thinner TS-1 catalytic layers (70 μm) can be utilized to H2O2 selectivity is shown in Figs. 10 and 11. In order to optimize
achieve conversion industrial targets. the gas partial pressures, the other parameters were kept constant.
 E. Kertalli et al. / Chemical Engineering Science 138 (2015) 465–472
Fig. 12. Dependence of the membrane area on the activity of the TS-1 catalyst.
The fluxes of hydrogen and oxygen, at the exit of the Pd/SiO2
membrane layer, increase with increasing the gas partial pressure
at the feed side. This is due to the increased driving force of the gas
across the catalytic membrane layer. On the other hand, lower
hydrogen partial pressures benefit the H2O2 selectivites by low-
ering the rate of hydrogenation reaction.
Fig. 11 shows that the oxygen partial pressure has nearly no
effect on the H2O2 selectivity. Due to solubility properties, oxygen
is always in excess compared to hydrogen, therefore oxygen is not
the limiting reactant. Even more, oxygen is involved only in the
H2O2 synthesis reaction, therefore it is not influencing the H2O2
selectivity. The handling of hydrogen:oxygen mixtures is strongly
limited by the explosive nature of the reactants. The combination
of safety and operation requirements leads to the final optimal
ratio of hydrogen:oxygen mixture of 4:96%. This combines the low
hydrogen concentration needed for optimal H2O2 selectivity with
the requirements of operating this mixture outside the explosive
regime. Similar to oxygen, the propylene partial pressure is
expected to have almost no effect on the H2O2 conversion and
selectivity. This is mainly due to the large excess of propylene in
the system. Due to its high solubility in methanol (2 orders of
magnitude higher than hydrogen and oxygen), propylene will
never be the limiting reactant in the system.
4.6. Scaling up of the catalytic membrane reactor
A scale up of the membrane reactor concept was considered for
a PO plant capacity of 300 kton/year. The catalytic area of the
membrane reactor needed for large scale production is depending
on the activity of the catalyst. Fig. 12 shows the dependence of the
membrane area on the TS-1 catalyst activity.
The industrial target for a TS-1 catalyst industrial application
requires a productivity of 1 kgPO/kg h. However, in the literature,
typical TS-1 catalysts show activities higher than 10 kgPO/kg h
(Oyama, 2008). Therefore, a productivity of 5 kgPO/kgcat h is
assumed as reasonable for the performance of the catalyst. This is
translated into a membrane area of 84,000 m2 needed for large
scale application of a membrane reactor for the direct PO synth-
esis. Considering tubular membranes of 1 m length and 2 mm
diameter as possible reactor configuration, 1.3n107 tubes are nee-
ded for the large production of PO. The high numbers of reactors
required for the application of this process raise concerns about
the feasibility of the direct synthesis of PO in a membrane reactor.
However, the membrane area needed for the PO production is
connected to the activity of the catalyst. Therefore, further
471
research in developing and improving the catalyst performance
would lead to a redaction of the number of tubes needed for a
large scale PO production making the feasibility of this process
more realistic.
5. Conclusions
A catalytic membrane reactor is proposed for the direct propylene
oxide synthesis in liquid phase. The reactor characteristics were
described with a mathematical model. The membrane reactor is
designed to perform industrial conversion and selectivity targets and
operates outside explosive regime. This new reactor concept allows
one to couple two consecutive reactions: hydrogen peroxide synth-
esis and its reaction with propylene to propylene oxide. The opti-
mization of the direct propylene oxide synthesis is obtained in two
steps. First, the hydrogen peroxide synthesis, which is the rate lim-
iting step of the process, is optimized by ensuring that the conversion
and the selectivity of hydrogen peroxide reach the values of 50 and
60%, respectively. Second, the epoxidation of propylene with hydro-
gen peroxide is optimized reaching a propylene oxide selectivity
higher than 50%. The large scale application of the catalytic mem-
brane reactor is considered for a plant of 300 ktonPO /year capacity. A
reactor area of 84,000 m2 is necessary to obtain this industrial
capacity. The feasibility of the large scale application of the propylene
oxide synthesis on a membrane reactor is related to the improve-
ment of the catalyst activity which would lead to a reduced mem-
brane area needed for the process.
Nomenclature
in inlet
out outlet
i ordinal number of components
Deff−i effective diffusion of component i
ki kinetic constant of component i
Ki adsorption equilibrium constant of component i
Ci concentration of component i
Hei Henry constant of component i
ϕC3H6 fraction of solvent in the system
Vsweep gas volumetric flow rate
Vmeth liquid volumetric flow rate
CPd palladium concentration
a 5.73n2.35 ( gcat /fraction of changing liters of reactor to
m3)
Acknowledgements
The research leading to these results has received funding from
the European Community's Seventh Framework Programme
through the Collaborative Project INCAS, under agreement n
245988. The author would like to thank Mart de Croon for helpful
discussions.
References
Bassler, P., Weidenbach, M., 2010. The New HPPO Process for Propylene Oxide:
From Joint Development to Worldscale Production. Technical Report.
Battino, R., International Union of Pure and Applied Chemistry, 1981. Oxygen and
Ozone. Pergamon, Oxford, New York.
Bhattacharyya, D., Butterfield, D.A., 2003. New Insights into Membrane Science and
Technology: Polymeric and Biofunctional Membranes. Elsevier, Amsterdam,
Boston.
472 E. Kertalli et al. / Chemical Engineering Science 138 (2015) 465–472
Bravo-Suárez, J.J., Bando, K.K., Akita, T., Fujitani, T., Fuhrer, T.J., Oyama, S.T., 2008.
Propane reacts with O2 and H2 on gold supported TS-1 to form oxygenates with
high selectivity. Chem. Commun. 28, 3272–3274.
Cavani, F., 2010. Catalytic selective oxidation: the forefront in the challenge for a
more sustainable chemical industry. Catal. Today 157 (1–4), 8–15.
Cavani, F., Teles, J.H., 2009. Sustainability in catalytic oxidation: an alternative
approach or a structural evolution?. ChemSusChem 2 (6), 508–534.
Chowdhury, B., Bravo-Suárez, J.J., Daté, M., Tsubota, S., Haruta, M., 2006. Tri-
methylamine as a gas-phase promoter: highly efficient epoxidation of propy-
lene over supported gold catalysts. Angew. Chem. (International ed. in English)
45 (3), 412–415.
Edwards, J., Solsona, B., Landon, P., Carley, A., Herzing, A., Kiely, C., Hutchings, G.,
2005. Direct synthesis of hydrogen peroxide from H2 and O2 using
TiO2-supported Au–Pd catalysts. J. Catal. 236 (1), 69–79.
García-Serna, J., Moreno, T., Biasi, P., Cocero, M.J., Mikkola, J.-P., Salmi, T.O., 2014.
Engineering in direct synthesis of hydrogen peroxide: targets, reactors and
guidelines for operational conditions. Green Chem. 16 (5), 2320.
Hsieh, H.P., 1996. Inorganic Membranes for Separation and Reaction. Elsevier,
Amsterdam.
Huang, J., Haruta, M., 2011. Gas-phase propene epoxidation over coinage metal
catalysts. Res. Chem. Intermed. 38 (1), 1–24.
Jenzer, G., Mallat, T., Maciejewski, M., Eigenmann, F., Baiker, A., 2001. Continuous
epoxidation of propylene with oxygen and hydrogen on a Pd–Pt/TS-1 catalyst.
Appl. Catal. A: Gen. 208 (1–2), 125–133.
Laufer, W., 1999. Propylene epoxidation with hydrogen peroxide over palladium
containing titanium silicalite. J. Mol. Catal. A: Chem. 141 (1–3), 215–221.
Li, K., 2007. Ceramic Membranes for Separation and reaction. John Wiley,
Chichester, England, Hoboken, NJ.
Li, N., 2008. Advanced Membrane Technology and Applications. Wiley, Hoboken, NJ.
Meiers, R., Hölderich, W., 1999. Epoxidation of propylene and direct synthesis of
hydrogen peroxide by hydrogen and oxygen. Catal. Lett. 59, 161–163.
Monnier, J.R., 2001. The direct epoxidation of higher olefins using molecular oxy-
gen. Appl. Catal. A: Gen. 221 (1), 73–91.
Nijhuis, T.A., Visser, T., Weckhuysen, B.M., 2005. Mechanistic study into the direct
epoxidation of propene over gold/titania catalysts. J. Phys. Chem. B 109 (41),
19309–19319.
Nijhuis, T.A., Makkee, M., Moulijn, J.A., Weckhuysen, B.M., 2006. The production of
propene oxide: catalytic processes and recent developments. Ind. Eng. Chem.
Res. 45 (10), 3447–3459.
Nijhuis, T.A., Chen, J., Kriescher, S.M.A., Schouten, J.C., 2010. The direct epoxidation
of propene in the explosive regime in a microreactor—a study into the reaction
kinetics. Ind. Eng. Chem. Res. 49 (21), 10479–10485.
Oyama, S.T., 2008. Mechanisms in Homogeneous and Heterogeneous Epoxidation
Catalysis. Elsevier, Amsterdam.
Paunovic, V., Schouten, J.C., Nijhuis, T., 2015. Direct synthesis of hydrogen peroxide
in a wall-coated microchannel reactor over Au–Pd catalyst: a performance
study. Catal. Today 248, 160–168.
Russo, V., Tesser, R., Santacesaria, E., Di Serio, M., 2013. Chemical and technical
aspects of propene oxide production via hydrogen peroxide (HPPO Process).
Ind. Eng. Chem. Res. 52 (3), 1168–1178.
Samanta, C., 2008. Direct synthesis of hydrogen peroxide from hydrogen and
oxygen: an overview of recent developments in the process. Appl. Catal. A: Gen.
350 (2), 133–149.
Satterfield, C.N., 1969. Mass Transfer in Heterogeneous Catalysis. M.I.T. Press,
Cambridge, Mass.
Shin, S.B., Chadwick, D., 2010. Kinetics of heterogeneous catalytic epoxidation of
propene with hydrogen peroxide over titanium silicalite (TS-1). Ind. Eng. Chem.
Res. 49 (17), 8125–8134.
Sinha, A.K., Seelan, S., Tsubota, S., Haruta, M., 2004. A three-dimensional meso-
porous titanosilicate support for gold nanoparticles: vapor-phase epoxidation
of propene with high conversion. Angew. Chem. (International edition in
English) 43 (12), 1546–1548.
Taramasso, M., Perego, G., Notari, B., 1983. Preparation of Porous Crystalline Syn-
thetic Material Comprised of Silicon and Titanium Oxides.
The Dow Chemical Company, 2014. SCG-DOW Group Reaches Key Milestone toward
Thailand HPPO Plant Start-up News—News Center. Technical Report.
Voloshin, Y., Lawal, A., 2009. Kinetics of hydrogen peroxide reduction by hydrogen
in a microreactor. Appl. Catal. A: General 353 (1), 9–16.
Voloshin, Y., Halder, R., Lawal, A., 2007. Kinetics of hydrogen peroxide synthesis by
direct combination of H2 and O2 in a microreactor. Catal. Today 125 (1–2),
40–47.
Wu, G., Wang, Y., Wang, L., Feng, W., Shi, H., Lin, Y., Zhang, T., Jin, X., Wang, S., Wu,
X., Yao, P., 2013. Epoxidation of propylene with H2O2 catalyzed by supported
TS-1 catalyst in a fixed-bed reactor: experiments and kinetics. Chem. Eng. J.
215, 306–314.
Yap, N., Andres, R.P., Delgass, W.N., 2004. Reactivity and stability of Au in and on TS-
1 for epoxidation of propylene with H2 and O2. J. Catal. 226 (1), 156–170.
Young, C.L., International Union of Pure and Applied Chemistry, 1981. Hydrogen and
Deuterium. Pergamon Press, Oxford, New York.