Computational Fluid Dynamics for Microreactors Used in Catalytic

Oxidation of Propane

S. Odiba; M. Olea*; S. Hodgson; A. Adgar

Teesside University, School of Science and Engineering, Borough Road, Middlesbrough TS1 3BA, UK
*[email protected]

Abstract: The optimal design of a microreactor explosive or toxic components, compared to

for a specific application can be explored based conventional chemical reactors. Due to the small
on simulation results obtained by using CFD channel dimensions, microreactors have a high
package of COMSOL Multiphysics. This surface-to-volume ratio for kinetically controlled
research deals with the design of suitable reactions and high heat and mass transfer rates
microreactors for the catalytic oxidation of (Ehrfeld et al, 1999). The small dimensions of
volatile organic compound (VOCs), using the microchannels allow short radial diffusion
propane as a model molecule. The microreactor time leading to a narrow residence time
considered consists of eleven parallel channels, distribution (RTD) (Rouge et al, 2000). What is
in which an Au/Cr/-Al2O3-catalyzed more, the scale-up can be easily performed just
combustion reaction takes place. Each channel is by multiplying the number (numbering-up) of
0.5 mm diameter and 100 mm long. the microreactors units without changing the
The catalytic microreactor was simulated for channels geometry. Catalyst material is usually
temperatures between 563 K and 663 K and inlet deposited on the wall of the reacting channels for
flow velocities between 0.05 and 1 m/s. As catalytic microreactor (Zanfir et al., 2003).
expected, the propane conversion increased The objective of this project is to develop a
quickly with increasing temperature for a fixed suitable microreactor model for VOC catalytic
inlet flow velocity. For a fixed temperature, the oxidation using the CFD package of COMSOL
propane conversion increased as the inlet flow Multiphysics. Propane oxidation over a catalyst
velocity decreased. The simulated results have (Au/Cr/-Al2O3) was chosen as a model reaction.
revealed that higher conversion rates of propane
can be achieved by decreasing inlet gas velocity C3 H 8 5O2 3CO2 4H 2 O (1)
from 1 m/s to 0.01 m/s and increasing
temperature from 593 K to 663 K The microreactor considered consists of
Optimum inlet gas velocity for propane eleven parallel channels, in which an Au/Cr/-
conversion of 99.99 % was between 0.05 m/s Al2O3-catalyzed oxidation reaction takes place.
and 0.1 m/s. Each channel is 0.5 mm diameter and 100 mm
long. The first design step was the selection of
Keywords: Volatile organic compound the most suitable microreactor geometry. This
oxidation, Catalytic microreactor design, was based on cold flow (without reaction) and
modeling, propane, CFD. the simulation results were obtained by using the
CFD module of COMSOL Multiphysics. Four
1. Introduction different geometries had been proposed (see
Figure 1) and simulated in 2D (see Figure 2) to
Microreactors are devices containing open evaluate the fluid behaviour in the
paths for fluids with characteristic dimensions in microchannels. The design of a microreactor
the range of 100 500 m and have multiple relies on having uniform flow distribution in all
channels where the chemical transformations parallel microchannels for optimum yield and
occur (W. Ehrfeld, et al, 1999). Different product selectivity. Relative standard deviation
reactions have been used to illustrate the was used as a measure of the quality of the flow
advantages of performing chemical reactions in distribution in the microchannels. As defined by
microreactors, which are particularly suited for (Delsman., et al, 2004):
highly exothermic, endothermic and fast n

reactions. Their small reaction volume allows 1 ( f F )2

n 1
( 2)
handling of dangerous reactions such as with F n 1
 in cold flow (without reaction) at normal
where f is the flow rate per channels, F the temperature and pressure. This approach
mean flow rate, n is the number of channels, and calculates the flow field using the incompressible
is the relative standard deviation. form of the Navier-Stokes and continuity
Consequently, a larger value of the relative equations. Owing to the intense computational
standard deviation will correspond to a less demands, 2D model was use to accurately study
uniformly distributed flow over the channels. the flow behavior in the microchannels. A 3D
The reason is that non-uniformity of flow model was use to study the catalytic oxidation at
distribution of fluid among the channels steady state and atmospheric pressure. No slip
introduces dispersion, leading to loss in boundary condition is imposed on all internal
conversion and selectivity in catalytic walls. Au/Cr/-Al2O3 catalyst was assumed to
microreactor (Commenge, et al, 2002). It also uniformly be deposited onto the wall layer of the
leads to increase in pressure drop over the microchannels. Meshing was performed by the
microreactor. default triangular and tetrahedral meshing
algorithm available in COMSOL Multiphysics
with a mesh size of 0.01mm. The flow rates in
each channel were determined using COMSOL
post processing. This model shows the catalytic
oxidation of propane (VOCs) in a microreactor.
The reactant (propane) is transported from the
fluid bulk to the catalytic surfaces at the reactor
wall, where it reacts. A space-independent model
is set up to analyze the reaction using the
reaction engineering interface. The known
reaction kinetics is incorporated to solve the
space-dependent model of the microreactor
because all the interfaces with chemical
engineering module include fields for defining
arbitrary reaction kinetics.
The physics user interface supports
simulation of transport by convection and
diffusion in 1D, 2D and 3D. The dependent
variable is the molar concentration; c.
Type A1 Figure 1 and 2 shows the four proposed
Type A2 Type A3 Type A4
geometries created using CFD package of
Figure 1. Four proposed geometries
COMSOL Multiphysics in 2D.
The so called A2 geometry gave the
optimum Flow distribution from laminar to
3. Governing Equations
transitional regime (the lowest relative standard
The flow in the microchannel is laminar and
deviation). The second step involved the
given by the Navier-Stokes equations. The
simulation of the selected microreactor geometry
Navier-Stoke equations, solved by default in all
for propane oxidation. The A2 geometry was
the single phase flow interfaces are the
modified in a 3D model to further investigate the
compressible formulation of the momentum
catalytic oxidation of the model VOC. The
equation:
catalytic reaction takes place at the surface of
microchannel wall.
u 2
u u p ( (u (u)T ) ( u) I ) F
t 3
2. Use of COMSOL Multiphysics
. (3)
The four proposed microreactors geometries
And the continuity equations:
were developed in COMSOL v4.3b and a single
phase laminar compressible flow was simulated
 diluted species is used to solve for the
( u ) 0 (4) convection and reaction of the chemical species.
t The inlet concentrations and flow density is
derived from the partial pressures. The calculated
Where is the solutions density (kg/m3), flow field serves as input to the transport of
dynamic viscosity of the fluid (Ns/m2), u is the dilute species interface, to describe the
velocity vector (m/s), I denote the identity convective mass transport. The use of parameters
matrix, F is the volume force vector (N/m3), T is and variables helps to define a more complex
the absolute temperature (K) and p is pressure rate expression in COMSOL.
(Pa). This equation is applicable for The kinetics of propane oxidation for the catalyst
incompressible as well as compressible flows systems Au/Cr/-Al2O3 using partial pressures
where the density varies. The flow in this can be represented by Equation 1.
microreactor is assumed compressible because
the Mach number (Ma) is less than 0.3. k O2 k C3 H 8 PC3 H 8 PO2
rC3 H 8 (5)
The boundary conditions imposed on the k O2 PO2 5k C3 H 8 PC3 H 8
momentum balance are no slip at the reactor
walls: Equation 5 was used by (Lin, Jiunn-Nan and
Ben-Zu, Wan, 2004), when they investigated the
u = U0 inlet combustion of propane at different temperatures.
It was derived using the Mas-Van Krevelen
u=0 walls mechanism for reduction-oxidation.
No slip boundary condition for a stationary solid Where k O2 and k C3 H 8 are rate constants in the
wall, that is, that the fluid at the wall is not reduction-oxidation mechanism, PO2 and
moving. The outlet boundary conditions
describing fluid flow at an outlet are the PC3 H 8 are the partial pressures of oxygen and
pressure, and no viscous stress. The pressure
condition prescribes only a Dirichlet condition
propane, and rC3 H 8 is reaction rate of propane.
for the pressure: The Arrhenius expression for the temperature
p = p0 outlet dependence of k O2 and rC3 H 8 is:

This boundary condition is physically equivalent aO2

to a boundary that is adjacent to a large container k O2 O2 exp
(6)
(inlets) or exiting into a large container (outlets). RT
The flow in the microreactor is driven by a
pressure drop. The pressure at the inlet, p is aC3 H 8
slightly higher than that at the outlet, where the k C3 H 8 C3 H 8 exp
(7)
pressure is set to atmospheric pressure. The No RT
Viscous Stress condition specifies vanishing
viscous stress on the outlet. Where A is the Arrhenius constant, Ea. is the
activation energy, R is the gas constant and T is
2
(u (u ) ) ( u ) I ) n 0 the temperature in K.

3 The transport of diluted species user interface

assumes chemical species transport through
In the 3D microreactor model, reactions diffusion and convection and implements the
occur at the catalytic surfaces located at the mass balance equation. The interface assumes
walls. The channel walls are however coated that all species present are dilute, and that their
with Au/Cr/-Al2O3 catalyst. So that as propane concentration is small compared to a solvent
gas flows through the reactor, oxidation reactions fluid or solid.
occur at the catalytic surfaces. A space- The mass transfer in the reactor domain is given
dependent model is generated and flux is applied by the stationary convection and diffusion
for the surface reaction to the catalytic surface. equation:
For an ideal gas, laminar flow and transport of
 c the bulk. The fluid flow rates inside the channels
u c ( Dc) R (8) were calculated by integrating across the normal
t component of the exit velocity. Since the flow is
stationary and compressible, it then guarantees
Where D denotes the diffusion coefficient (m2/s), that the flow into a channel and out is equivalent.
c is the species concentration (mol/m3), u is the This accounts for the fact that the velocity is
velocity vector (m/s) and R is the reaction rate non-uniform (laminar profile) across the cross-
expression for the species (mol/(m3.s)). The first section. To derive the flow rate in each channel I
term on the left-hand side of Equation 8 have used a Derived Values => line integration
represent the accumulation or consumption of of velocity field, y component in COMSOL.
the species. The second term on the left hand Solving the fluid flow at higher velocities (10
side accounts for the convective transport due to m/s - 100 m/s) the computation becomes highly
a velocity field u. This field can be obtained non-linear, and the non-linearity of the
from coupling the transport of diluted species computation can be thought of as a function of
user interface to one that describes fluid flow the Reynolds number. This can be solved by
(CFD or momentum balance). The first term on using a continuation sweep to incrementally
the right-hand side of Equation 8 accounts for increase the Reynolds number. In the 3D model,
the diffusion transport (interaction between the a good resolution of the flow is obtained by
dilute species and solvent). Finally, the second meshing the inlet faces of the channels using a
term on the right-hand side of Equation 8 free triangular mesh with suitably fine size
represents a source, or sinks term, usually due to settings to get a good resolution of the flow
a chemical reaction. across section. A swept mesh on the channel
In the microchannel, the boundary condition for domains and a free tetrahedral mesh on the inlet
an inflow is set as (concentrations): and outlet domains (manifolds). For the flow
study to resolve the laminar profile within the
c cin (9) channels, a suitable custom maximum element
size of 0.01m was used on the planes in the free
Resolving a parabolic or laminar flow profile triangular mesh for the 3D model.
for the 2D model requires at least more than two Propane fractional conversion, X is an
elements across each channel. The 2D model important characteristic for the catalytic
was resolved with a refined mesh with five oxidation and is defined as:
elements across each channel. The default C( C 3 H 8 ) in C( C 3 H 8 ) Ou t (10)

element type for laminar flow is a linear element, C( C 3 H 8 ) in
so a number of linear segments are resolved due Where C is the concentration in mol/m3
to the curved nature of the profile. Mesh is built
to optimally resolve the laminar flow in the
channels. First by modifying the geometry to
4. Results
introduce an extra rectangle (channel), in order
to introduce a boundary edge along the top and
bottom faces of each channel. Then an edge
mesh is built for the channel inlets with a
maximum element size of 1.25 x10-5 m per
channel. Mapped mesh is then used in the
channel domains to achieve a rectangular
element that have a long aspect ratio since flow
is likely to have a similar profile down the
channels. This uses a default mesh size of fluid
dynamics and coarse that is inherited from the
global size node of COMSOL Multiphysics.
Mesh is completed using a free triangular mesh
on the remaining domains with fluid dynamics
and extra-fine on the boundaries and coarse in Type A1 Type A2 Type A3 Type A4
Figure 2. CFD simulation of the four proposed
geometries, showing the contour plot of velocity
distribution in 2D

4.1. Effect of inlet velocity

This was tested by calculating a relative standard

deviation of the flow rates at various velocities
for each geometry as seen in Figure 3. Type
A1geometry showed different results than those
of the other geometries. For type A2, A3, and
A4, for inlet velocities higher than 30 m/s, it was
found that the deviation became constant. For
type A1 the deviation continued to increase until
approximately 90 m/s and then became constant.

1.5E+00
Type A4
Normalized flow rate

1.0E+00
Figure 3. Relative standard deviation of the flow 100 m/s
distribution as a function of the inlet manifold velocity 5.0E-01 60 m/s
as calculated for the 4 geometries 10 m/s
1 m/s
4.2. Effect of inlet velocity on flow 0.0E+00
distribution inside the microchannels 0 2 4 6 8 10 12
Channel number
This was tested by calculating the normalized Figure 4. Flow distribution over the channels,
flow rates and plotting the normalized flow rate calculated at an inlet velocity of 1 m/s, 10 m/s, 60 m/s,
against the channel number as seen in Figure 4. and 100 m/s
Normalized flow rate is calculated by dividing
channels compared to the inner channels near the
the actual flow rate per channel by the mean flow
inlet. The highest flow rates occurred for the
rate. For all geometries there was an effect on the
centre channels. This was particularly
outer channels located furthest away from the
pronounced for type A1 which has a much
manifold inlet. The flow was less for these
narrowed active inlet area compared to type A2,
A3, and A4 respectively. This produced a more
steep parabolic profile in the centre channels
compared to type A2, A3, and A4 as seen in
Figure 4. At inlet velocities below 10 m/s, a
constant flow rate is found, which means that the
flow distribution is independent of flow rate. All
the geometries (Type A1, A2, A3, and A4) gave
similar results in this regime. In this region
where the flow distribution is independent of the
flow rate, the flow field is determined by wall
friction. At an inlet velocity of 60 m/s and above,
inertial forces become important.
 simulation result shows that propane
4.2. Catalytic oxidation of propane concentration is decreasing by increasing gas
inlet velocity.
The next step was to modify geometry A2
into a 3D model to further investigate the
catalytic oxidation of VOCs. Simulation results
are shown in Figure 5, 6 and 7.

Figure 8. Propane concentration gradient along a

channel at a velocity of 0.01m/s

Figure 5. Concentration gradient of propane oxidation

over Au/Cr/-Al2O3 in the microreactor at inlet
velocity of 0.01 m/s

Figure 9. Propane concentration gradient along a

channel at a velocity of 0.05m/s

Figure 6. Concentration gradient of propane oxidation

over Au/Cr/-Al2O3 in the microreactor at inlet velocity of
0.1 m/s

Figure 10. Propane concentration gradient along a

channel at a velocity of 0.1m/s
Figure 7. Velocity profile of the laminar flow field in
the microreactor at temperature of 593 K

4.3. Effect of gas inlet velocity on propane

concentration

When the reactant inlet velocity is increase, it

means that the residence time of reactants is
decreased, and the total propane conversion
decreases as shown in Figure 7, 8, 9 and 10.
Residence time is important parameters for Figure 11. Propane concentration gradient along a
catalytic oxidation of propane, because it channel at a velocity of 1m/s
determines the reactor productivity. The 3D
4.4. Effect of temperature on propane Figure 12. Propane concentration and conversion
conversion against temperature at a velocity of 0.01m/s

Since both reaction rate and species diffusivities

depend on temperature, increasing the
temperature from 593 K to 663 K has a
significant influence on the conversion of
propane as shown in Figure 7 to 16. This
increase in temperature also affects the flow in
the microchannels.
The combustion of the propane is a strongly Figure 13. Propane concentration and conversion
exothermic reaction with a high negative value against temperature at a velocity of 0.05m/s
of standard Gibbs free energy over a significant
temperature range. In standard terms to increase
the equilibrium fractional conversion of propane
the reaction temperature should be reduced. For
the exothermic reaction, if the temperature is
increased the equilibrium conversion should be
reduced. For this reaction (equation 1) the
standard Gibbs free energy does reduce as
temperature increases but even at higher
temperature (593 K - 663 K) the equilibrium Fugure 14. Propane concentration and conversion
fractional conversion can be calculated as > against temperature at a velocity of 0.1m/s
99.9%.
An examination of the Arrhenius equations
(equation 5 and 6) indicates that the rate
constants increase with temperature. The effect
of this should be that as the reaction temperature
increases the reaction rate should increase and
the rate of approach to the equilibrium fractional
conversion should increase. These predictions
are consistent with the behaviour shown in
(Figures 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, and Figure 15. Propane concentration and conversion
16). The rate at which the equilibrium against temperature at a velocity of 1m/s
conversion is approached consistently increases
as the temperature increases. The increasing
reactant velocity decreases the residence time in
the reactor as indicated previously .This has the
effect of increasing the point in the reactor where
the equilibrium fractional conversion is
approached but all the results obtained in the
simulation in terms of velocity and temperature
changes are consistent with the effects described.
Figure 16. Propane conversion and conversion against
temperature at a velocity of 0.01, 0.05, 0.1, and 1 m/s

5. Conclusions

The use of a steady-state COMSOL

Multiphysics model for the simulation of
propane catalytic oxidation in a microreactor was
demonstrated in this paper. The effect of three
important parameters namely the manifold
shape, the inlet velocity, and temperature for a 1. Ehrfeld, W; Haverkamp,V.; Hessel, V.
cold flow (non-reaction) on fluid flow uniformity Microreactors, in Ullmans Encyclopedia of
and with reaction was investigated for an 11 industrial Chemistry, Wiley-VCH, Weinheim,
channel microreactor. Distributions of fluid flow 1999
over the microchannels with 4 different manifold 2. Rouge, A.; Spoetzl, B; Gebauer, H; Schenk,
geometries have been calculated using a 2D CFD R, and Renken, A. Microchannel reactors for fast
module of COMSOL Multiphysics. periodic operation: the catalytic dehydrogenation
The flow behaviour has been assessed based of isopropanol, Chem Eng Sci, 56: 1419-1427,
on the relative standard deviation of the flow (2000).
rates from a mean. Simulation results indicate 3. Zanfir, M; Gavriilidis, A. Catalytic
two flow regimes depending on the inlet velocity combustion assisted methane steam reforming in
for the 4 geometries. a catalytic plate reactor, Chem Eng Sci, 58, 3947-
For all geometries at low flow rates (below 3960 (2003)
10 m/s), the flow distribution appears to be 4. Commenge, J. M., Falk, L., Corriou, J. P.,
completely determined by wall friction. In this Matlosz, M, Optimal design for flow uniformity
regime all 4 geometries yielded similar results. in microchannel reactor', American Institute of
For operation below this velocity the flow Chemical Engineers Journal, 48(2), 345-358
distribution appears to be independent of flow (2002).
rate (as seen in Figure 4 for an inlet velocity of 1 5. Lin, Jiunn-Nan, and Wan, Ben-Zu, Reaction
m/s). At flow rates above a transitional velocity Kinetics of propane Catalytic Combustion on
(60 m/s for type A2, A3, and A4 and 90 m/s for Cr/-Al2O3, Co/Cr/-Al2O3, and Au/Cr/-Al2O3,
type A1), inertial effects start to influence the J. Chin. Inst. Chem. Engrs, 35(2), 149-159
flow distribution with a constant relative (2004).
standard deviation. Geometries A2, A3, and A4 6. COMSOL Multiphisics training course hand
show a similar trends; for type A1 a different book 2012 Publications division Home page.
effect is observed, the relative standard deviation http://comsol.com (accessed June, 2013)
continues to increase until it appears to level off 7. John D. Anderson, Computational fluid
at a higher velocity. It appears that the observed dynamics: The basics with applications,
effect is linked to the nature of the flow regime McGraw-Hill, Inc. London, 1995.
in the manifold. A plot of relative standard
deviation of flow rates against inlet manifold
velocity, showed a significant difference
between geometries A1, and A2, A3, A4. The
relative standard deviations are in order of
magnitude, less for type A2, A3, and A4. Type
A2 gave the optimum flow distribution from
laminar to transitional with the lowest relative
standard deviation.
The 3D simulation results showed that the
propane concentration decreases by increasing
the gas inlet velocity. By increasing the
temperature from 593 K to 663 K increase the
conversion rate. The simulated results have
revealed that higher conversion rates of propane
can be achieved by decreasing inlet gas velocity
from 1 m/s to 0.01 m/s and increasing
temperature from 593 K to 663 K
Optimum inlet gas velocity for propane
conversion of 99.99 % was between 0.05 m/s
and 0.1 m/s.

6. References