Anda di halaman 1dari 16

Chemical Engineering Science 60 (2005) 5862 – 5877

www.elsevier.com/locate/ces

Computer modelling and numerical analysis of hydrodynamics and heat


transfer in non-porous catalytic reactor for the decomposition of ammonia
A.N. Waghode∗ , N.S. Hanspal, I.M.T.A. Shigidi, V. Nassehi, K. Hellgardt
Advanced Separation Technologies Group, Department of Chemical Engineering, Loughborough University, Loughborough, Leicestershire LE11 3TU, UK

Received 27 July 2004; received in revised form 19 April 2005; accepted 5 May 2005
Available online 1 July 2005

Abstract
Chemical reactors exhibit very complex behaviours such as multiple steady states, oscillations, etc. resulting from complex linkage
between the transport processes and the non-linear chemical reaction kinetics. Ammonia is a potential hydrogen source for a number of
fuel cell applications for small scale power generation useful for portable equipments. In the present work, we analyse the fluid dynamics
and heat transfer in catalytic microreactor systems for the decomposition of ammonia over a monolayer Ni non-porous catalyst. The overall
model for this convective–diffusive–reactive system consists of a flow model, a mass transport model, an energy conservation model and
a reaction kinetics model for ammonia decomposition. The flow model is described by the Stokes equation for a creeping flow regime.
The mass transport and energy conservation models are based on convective–diffusion equations. The rate of ammonia decomposition
can be measured as a function of the catalyst activity and ammonia concentration. A standard Galerkin finite element technique has
been applied for the solution of the flow equations. A slightly perturbed form of the mass continuity equation is used to satisfy the
Ladyzhenskaya–Babuška–Brezzi stability criterion. For the solution of convection–diffusion equations, a streamline inconsistent upwind
finite element scheme has been chosen to avoid any spurious oscillations. C0 -continuous 9-noded Lagrangian biquadratic isoparametric
finite elements are used for the approximation of the field variables. A second-order Taylor–Galerkin time-stepping scheme has been chosen
for the temporal discretisation of the flow equations whilst an implicit theta method has been used for convection–diffusion equations. The
results are presented in the form of velocity vectors and concentration, temperature contours and are examined for stability, convergence
and theoretical consistency.
䉷 2005 Elsevier Ltd. All rights reserved.

Keywords: Ammonia decomposition; Catalytic microreactor; Finite element modelling; Hydrodynamics; Heat transfer

1. Introduction converted to different forms of nitrogen oxides NOx when it


is combusted to produce power. The increasing environmen-
Ammonia decomposition is an important reaction for its tal concern have resulted in the impositions of strict regula-
involvement in a number of industrial applications such tions on reducing the emission levels of gaseous pollutants
as biomass gasification (Wang et al., 1999), which is a such as COx and NOx . A massive search for an alternative to
very attractive source of power generation that reduces re- conventional fuels have driven the interest in fuel cells. Fuel
liance on fossil fuel. About 60–80% of fuel nitrogen in the cell technology has been delivered significant attention in
biomass gets converted to ammonia in gasification. Am- recent years as a concomitant way for small scale electrical
monia is known as one of the main contaminants in this power generation through chemical transformations without
process because it forms fuel-nitrogen bonds that can be emission of environmental pollutants. The smooth opera-
tion of these fuel cells essentially depends on continuous
supply of clean hydrogen gas free from contaminant such
∗ Corresponding author. Tel.: +44 1509222546; fax: +44 1509223923. as carbon monoxide to avoid poisoning of the anode cata-
E-mail address: A.Waghode@lboro.ac.uk (A.N. Waghode). lyst. Due to these constraints, the fuel cell processes using
0009-2509/$ - see front matter 䉷 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2005.05.019
A.N. Waghode et al. / Chemical Engineering Science 60 (2005) 5862 – 5877 5863

hydrocarbon fuels requires secondary small-scale units such Ru and the effects of various supports such as H-ZSM-5,
as desulphurisation and oxidation for reduction of CO lev- HY, Al2 O3 and SiO2 on the reaction were analysed. Ammo-
els. Hydrogen generated from single step decomposition of nia decomposition approached 100% conversion when high
ammonia approves to be an attractive alternative to hydro- Ni loading was employed. They concluded that ammonia
carbons for fuel cells. Ammonia possesses high hydrogen decomposition decreases in the order Ru > Ir > Ni for the
content (17% by weight) and more importantly, provides a same nominal metal loading. The metal-support interaction
promising mode of storage for hydrogen for its on-site gen- plays a crucial role in determining the apparent activation
eration. energy for the ammonia decomposition. The ammonia de-
Hydrogen generated from ammonia decomposition is composition activity per metal site was found to be greater
a straightforward process with CO-free product gases N2 for silica support compared with alumina.
and H2 . Chellapa et al. (2002) carried out an experimental in-
Decomposition of ammonia is an endothermic reaction vestigation of the decomposition of ammonia as a prime
expressed as, source of hydrogen for PEM fuel cells over Ni.Pt/Al2 O3
catalyst. Intrinsic rate data was obtained between tempera-
2NH3 → N2 + 3H2 H = 11 kcal/mol
tures of 520 and 620 ◦ C and at ammonia pressures between
Though the commercial ammonia synthesis has reached its 50–780 Torr, generally applied in typical fuel cell applica-
level of saturation for any future development, the reverse tions. For all ammonia concentrations, the reaction was ob-
reaction of ammonia decomposition has been an interest of served to be of the first order with respect to ammonia in
intense research since last couple of decades. The attempts the operating temperature range. The activation energy for
reported in the literature before 1990 were mainly concerned the reaction was observed to be specific to that particular
with gaining a fair idea of the reaction kinetics of the ammo- catalyst employed to enhance the reaction. Using Ni-based
nia decomposition (Löffler and Schmidt, 1976) and in the catalyst, they presented a rate expression for this gas phase
last decade, the core focus was on investigating the nitriding reaction as,
processes (Djèga-Mariadassou et al., 1999) and NH3 abate-
ment (Mojtahedi and Abbasian, 1995). Yin et al. (2004) pre- r = kf pNH3 , (1)
sented a concise review of the reaction kinetics and mech-
anism of ammonia decomposition in conjunction with var- where,
ious catalyst developments for on-site generation of hydro-  
−49229
gen for end-use in fuel cell application. Their kinetic analy- kf = 1.309×1012 exp g mol NH3 /(g cat bar h).
RT
sis indicated that the recombinative desorption of adsorbed (2)
nitrogen atoms on active catalyst sites is the slowest hence-
forth, the rate determining step in ammonia decomposition. On the contrary, Papapolymerou and Bontozoglou (1997)
Despite these successful experimental investigations, a very have compared the unimolecular ammonia decomposi-
few predictive modelling attempts have been reported to ex- tion activity of various metallic polycrystalline wires. The
plain the decomposition chemistry (Deshmukh et al., 2004). unimolecular decomposition kinetics was expressed by
Wang et al. (1999) studied decomposition of ammonia at Langmuir–Hinshelwood model. They investigated that Ir is
high temperature and high pressure in a lab scale fixed bed highly catalytically active compared with Pd, Pt and Rh and
reactor over catalyst containing Fe, Co, Ru, Ir and Ni. The comparison of there experimental outcomes reported higher
effects of sulphur poisoning, carbon deposition and pres- decomposition rates on Ir than that obtained previously
ence of gases such as CH4 , CO and H2 were examined in with Ni.
the context of catalyst activity. Fresh Ni-based catalysts are Recently, Deshmukh et al. (2004) presented the decom-
shown to be the most efficient at low temperatures. The con- position chemistry of ammonia on Ru/Al2 O3 catalyst with
centration of hydrogen in the reaction gas was found to be a detailed microkinetic model describing the surface reac-
an important factor since higher ammonia conversions are tion mechanism. The mechanism consists of six reversible
reported at lower H2 partial pressures. elementary steps comprising adsorption–desorption of NH3 ,
The conventional processes for the production of hydro- N2 and H2 and hydrogen-abstraction steps from NHx inter-
gen such as steam reforming, autothermal reforming and mediates and their reverse.
partial oxidation of hydrocarbons and alcohols lead to pres- Though the microchemical reactors, micromachined re-
ence of undesirable by products such as COx . Removal of actors alias microreactors in short, have emerged accompa-
CO in ppm range, which is required for the Proton Exchange nied with developments in micro electrochemical systems
Membrane (PEM) fuel cells, makes these processes com- for portable appliances, they are found to be a sound alter-
plex and economically unfeasible. Such drawbacks restrict native for conventional complex, multi-unit pilot plants. Mi-
the practical implementation of these processes for vehicu- crochemical reactors offer inherent safety features through
lar and small-scale fuel cell applications. Choudhary et al. integrated control and sensor equipments in reactions deal-
(2001) systematically catalogued catalytic decomposition of ing with hazardous chemical intermediates with storage and
ammonia on supported metal catalysts such as Ni, Ir, and transportation constraints. These small-scale microchemical
5864 A.N. Waghode et al. / Chemical Engineering Science 60 (2005) 5862 – 5877

devices generally have high heat and mass transfer coeffi- Accuracy and convergence of the model was proved with
cients, enhanced surface to volume ratio for reactions and the known analytical solutions. It was concluded that the ap-
smallest occupied volumes for convenient end-use in small plication of simple finite difference technique for temporal
scale applications. discretisation with finite element space discretisation could
Ganley et al. (2004a) developed a structured alumina- be very effective when dealing with complicate geometries
anodised alumina microreactor for decomposition of anhy- and physical properties variations incurred in coupled reac-
drous ammonia on Ru catalyst in the moderate tempera- tion and mass transfer phenomenon. Trotta and Del Giudice
ture range. Modifications in reactor geometry such as chan- (1985) extended the previously developed finite element
nelling, varying catalyst development procedures along with procedure to obtain solutions for convective heat and mass
addition of promoters exhibited moderate impacts on the transfer problems with series surface reactions in fluid–solid
overall conversion of ammonia in varying degrees. With systems with solids having active and inactive catalyst
these amendments, the overall conversion of ammonia was sites. The numerical scheme is found to be reliable and
achievable to the maximum extent of 99% at 600 ◦ C reac- general to investigate the effects of geometric and hydro-
tion temperature to permit hydrogen production equivalent dynamic parameters on the effectiveness factor in catalytic
to 60 W for satisfactory use in portable and mobile appli- cavity.
ances such as notebook computers. Park (1995) developed a nonlinear pseudo-homogeneous
Deshmukh et al. (2004) modelled ammonia decomposi- catalytic reactor model in 2-dimensional coordinate sys-
tion for the production of hydrogen in a complex, post mi- tem. Dankwert’s boundary conditions were imposed on
croreactor geometry. Using their own microkinetic model 2-dimensional axisymmetric domain of fixed bed cat-
for the reaction, they analysed flow behaviours in ideal plug alytic reactor. Streamline-Upwind Petrov Galerkin (SUPG)
flow reactors, CSTRs and simulated fluid dynamics in 2- finite element method is applied to achieve stable and
dimensional tubular reactors with axisymmetric flow and accurate results to flow systems with convection, diffu-
mass transfer characteristics. The simulated data lies in close sion and chemical reaction. It was concluded that the
vicinity of the experimental data reported by Ganley et al. uniform or laminar flow velocity distribution plays impor-
(2004b). The authors justified the microreactor device as a tant roles on the transient variations in concentration and
benign choice for a small scale hydrogen production because thermal fields along with the steady state solutions. Mass
of low pressure drop, sufficiently reasonable surface area and thermal diffusion coefficients or their combinations
per unit volumes and increased efficiency due to no moving were identified as one of the key parameters to affect the
parts. steady-state solutions. SUPG finite element method and
More recently, SZ´rensen et al. (2005) presented a ver- Streamline Upwind Based (SUB) finite difference methods
satile synthesis protocol for supported metal catalyst to be were found to deliver results with comparable stability and
employed for decomposition of ammonia in microreactor accuracy.
for COx -free production of hydrogen. The microreactor is To provide physically justifiable simulations of fluid flow,
formulated in silicon wafer by a deep reactive ion etching energy distribution and mass transport in T-shaped microre-
technique. By the use of supported Ru catalyst promoted actors, Hsing et al. (2000) proposed a finite element based
with barium, sufficient hydrogen was obtainable for 1 W fuel numerical algorithm for Pt-catalysed ammonia oxidation.
cell corresponding to maximum consumption by common The numerical scheme includes simultaneous solutions of
cellular phones. momentum balance, energy conservation and mass transport
The chemical reactor for decomposition of ammo- equations to asses the strong interactions between energy
nia is a peculiar example of a coupled system of transport and chemical reaction kinetics. Higher order in-
convection–reaction–diffusion arising in many chemical and terpolation functions were used for approximating velocity,
biological settings. Due to increasing awareness among sci- temperature and concentration fields than the pressure vari-
entists and engineers about hydrodynamics of reactors apart ables. To circumvent oscillations in results, the convection
from chemical kinetics and the reaction mechanism, various dominated flow conditions were simulated using SUPG fi-
attempts are being made towards investigating the complex nite element scheme. It was concluded that the average con-
and non-linear interactions between convection–diffusion version of ammonia decreased with increasing flow rates
and reaction. The formulation of the mathematical model of the reactant species. The simulated results were in good
comprehensive of all the reaction and flow situations is of agreement with the experimental data and the comparison
considerable complexity and the governing model equations suggested that the transport phenomena in microreactors
can only be solved by rigorous numerical procedures, even can be reliably predicted and easily understood compared
more if the geometry of the domain is highly irregular and to macroscopic reactor systems. However, the simulations
complex. are insensitive in analysing the reaction mechanism and no
Del Giudice and Trotta (1978) introduced a standard microkinetic or kinetic data can be extracted from the nu-
Galerkin finite element procedure for the solution of merical results. The small dimensions of microchemical re-
non-linear equation representing transient diffusion ac- actor systems reveal that the transport processes are strongly
companied by nth-order irreversible chemical reaction. dominated by the diffusion mechanism.
A.N. Waghode et al. / Chemical Engineering Science 60 (2005) 5862 – 5877 5865

Alhumaizi et al. (2003) mathematically analysed the is quite justifiable on the basis that the employed catalyst is
one-dimensional reaction–diffusion–convection system us- non-porous which promotes the surface reactions. The gov-
ing standard reduction techniques such as finite difference, erning equations are in macroscopic form applicable to the
finite element and orthogonal collocation based on method geometric scale of reactor. On the contrary, the reaction ki-
of lines. The numerical solution of the governing equa- netics is expressed in a microscopic form in terms of active
tions can be stiff in temporal direction due to diffusive sites present on the catalyst surface. The appropriateness of
terms and steep in spatial direction due to convective terms. the inclusion or projection of such a microscopic term to
They analysed both steady state and transient behaviour macroscopic equations derived over the bigger scale of re-
of a homogeneous tubular reactor with an autocatalytic actor domain has been examined. However, the developed
reaction. They concluded that methods like essentially non- algorithm holds the capability of dealing with any complex
oscillatory method and total variation diminishing method and non-linear reaction kinetics for homogenous or hetero-
which are a special version of high resolution finite dif- geneous catalysed reaction.
ference technique enhance the accuracy eliminating the The important aspect in the present study is analysing the
oscillations but at the expense of increased computational influence of surface reaction on fluid flow and heat distri-
costs. bution characteristics in a microreactor. The form of surface
The present article is a preliminary outcome of an on- reaction kinetics equation may vary depending on the nature
going research on characterising the operation of ammonia of reaction or the physical states of constituting components
based fuel cells. It is an attempt to simulate the fluid dy- (macroscopic) or the catalyst activity (microscopic). Com-
namics and heat transfer in a microchemical reactor for mercial software may not offer flexibility in adopting such
decomposition of ammonia whose foundation is firmly various forms of rate of reaction expressions. In addition,
lying on basics of chemical reaction engineering, mass the stability of solutions depends on the type of the time-
transfer and computational fluid dynamics. Two microre- stepping scheme employed for the temporal discretisation of
actor systems were modelled to simulate hydrodynamics the governing equations. If the advection term is the domi-
and heat transfer characteristics in two distinct possible nating mode of transport of chemical species and energy, an
cases improper choice of time stepping scheme may lead to un-
stable and oscillatory results and higher order time-stepping
(a) the decomposition of ammonia in a channel with cat- schemes may be requisite.
alytically active wall for generation of ammonia: pro- In case of geometrically complex domains, the accuracy
duction for subsequent use in a fuel cell; of approximations can be enhanced by mesh refinement ei-
(b) the decomposition of ammonia over a catalytically ac- ther throughout the domain or in the vicinity of curved parts
tive cylinder which represents the external surface of a of the domain where the variations in field unknowns is as-
fuel cell for direct operation with ammonia. sumed to be significant. Most of the commercial software
exhibit their shortcomings in dealing with excessive mesh
In both the cases, the non-porous monolayer Ni metal refinements and their number crunching sub-routines cannot
surface film has been employed as a catalyst. The stan- cope with large number of equations to be solved simulta-
dard Galerkin finite element method has been employed neously. The objective behind the numerical analysis is just
for the solution of equations of conservation of mass and limited to have a preliminary idea about performance of the
momentum. The energy and concentration equations are reactor in support for any ongoing experimentation and op-
convection–diffusion equations which may require special timal design is not a point to be accessed. In such cases,
treatment to avoid spurious oscillations in the simulated the developed algorithm is expected to provide considerable
results if advection is the dominating transport mechanism flexibility that any commercial software devoted to this field
for chemical species and/or heat. A streamline inconsistent may not offer. The simulated results are examined for accu-
upwinding scheme has been used for solution of the mass racy, stability and theoretical consistency.
transport and energy conservation equations. The physical
domains of the microreactors have been tessellated with 9-
nodded quadrilateral Lagrangian isoparametric elements to
compensate for the curved geometrical features of the do- 2. Mathematical statement of the problem
mains. Second order biquadratic interpolation functions are
used for approximating all the field variables i.e., velocity, Since in one of the microreactors considered here, the
pressure, concentration and temperature. reaction is occurring on the walls, the assumption of constant
If the reaction kinetics is non-linear, it poses additional temperature and concentration in lateral direction associated
constraints on selection of the iterative procedures for the with 1-D model is inappropriate. The 3-D models may be
solution of the algebraic equations derived from the standard computationally flexible but may not be probably justified
Galerkin finite element formulation. As a result, to avoid in terms of computational economy. Therefore, a 2-D model
complexity arising in numerical analysis, the intrinsic rate of would be a good compromise between these two alternatives
reaction is expressed in terms of the catalyst activity which for microreactor simulations.
5866 A.N. Waghode et al. / Chemical Engineering Science 60 (2005) 5862 – 5877

2.1. Conservation of mass 2.3. Mass transport equation

The multicomponent gas mixture consists of ammonia, Mass transport in the present case can be expressed by
nitrogen, hydrogen and argon and it is assumed to be an ideal the convection–diffusion equation as,
and purely Newtonian incompressible homogeneous fluid.
Incompressibility is a reasonable assumption since ammonia jci j ci jci
+ vx + vy
decomposition occurs at low pressures. jt jx jy
   
The continuity of mass is mathematically represented by, j jc i j jc i
= Dxi + Dyi + Sci , (5)
jx jx jy jy
1 jp jv x jv y
+ + = 0, (3)
cs jt
2 jx jy i = 1, . . . N,

where,  is the fluid density, vx and vy are the components of where, c is the concentration of the gas component and Dxi
the velocity vector, t is the time variable, p is the hydrostatic and Dyi are the mass diffusion coefficients in the x and y di-
pressure and cs is the speed of sound in the gaseous fluid. rections, respectively. N is the total number of components
Eq. (3) is also referred to as the perturbed form of the in gaseous reaction system. The last term Sc represent the
continuity equation which has been employed in order to source/sink term introduced by the first order ammonia de-
satisfy the Ladyzhenskaya–Babuška–Brezzi (LBB) stability composition reaction explained in the successive sections.
criterion in the modelling of incompressible flow problems In this equation, the mass diffusivity for each constitutive
(Reddy, 1993). In physical sense, the inclusion of the term component is assumed to be identical and constant through-
1 jp out the simulations.
cs2 jt
in the continuity equation amounts to considering
the fluid to be slightly compressible (Zienkiewicz and Wu,
1991). The physical significance of this term can be referred 2.4. Energy conservation equation
from the work of Zienkiewicz et al. (1999).
In order to model the thermal field inside the non-porous
2.2. Conservation of momentum catalytic microreactors, a 2-dimensional transient energy
balance equation is considered in a Cartesian coordinate sys-
The conservation of momentum in an incompressible tem. The solid walls of the domain are considered to be insu-
flow system is generally represented by the Navier–Stokes lated so that there is no heat exchange with the surroundings.
equation, which consists of terms of convective as well With the assumptions of negligible effects of boiling and
as diffusive transport. The contribution of convective and condensation, the final energy conservation equation takes
diffusive mechanisms to momentum transport can be eval- the form as,
uated by the value of Reynolds number, which is a ratio of  
jT jT jT
inertial to viscous forces. In the present studies, the value cp + cp vx + vy
of Reynolds numbers in both the reactors is calculated to jt jx jy
   
be less than 0.5. By theory and experimental investigations, j jT j jT
= kx + ky + Sh, (6)
for the values of Reynolds numbers less than 8, the Stokes jx jx jy jy
law gives more representative account of flow process ne-
glecting the convective terms. However, it should be noted where, T is the temperature, cp is the specific heat capacity
that the inclusion of convective terms in the equation of of the gas mixture at constant pressure, kx and ky are the ther-
motion is straightforward and does not require any spe- mal conductivity of the gaseous fluid in x and y directions,
cific modifications in the numerical scheme used in this respectively. The last term Sh corresponds to source/sink
work. terms introduced by the heat liberated/absorbed during the
In absence of body forces, the conservation of momen- course of reaction.
tum for the creeping incompressible Newtonian fluid flow
is expressed by the Stokes equation. In 2-dimensional coor-
2.5. Specific heat capacity equation
dinate system, the Stokes equation takes the form as (Bird
et al., 2002),
Specific heat capacity of an ideal gas at constant pressure
      can be measured as a function of temperature as,
  jvx = − jp + j 2 jvx + j jv jv
jt jx jx  jyx + jxy ,
 jx  jy
   (4) cpi = a + bT + cT 2 + dT 3 , (7)
  jvy = − jp + j 2 jvy + j jv jv
 jxy + jyx ,
jt jy jy jy jx
where, i = 1, . . . , N and a, b, c & d are empirical constants
where,  is the fluid viscosity. found in the literature (Reid et al., 1977). (see Appendix A)
A.N. Waghode et al. / Chemical Engineering Science 60 (2005) 5862 – 5877 5867

The heat capacity of a multicomponent ideal gas mixture Schmidt, 1976; Ertl and Huber, 1980; Djèga-Mariadassou
can be calculated as, et al., 1999). In most of the studies, the rate of reaction is
N
considered to be of the first order with respect to ammonia

partial pressure.
cpi = nxi cpi , (8)
In the present study, the rate of ammonia decomposition
i=1
is measured in terms of catalyst turnover frequency. We con-
where, nx is the mole fraction of an individual component sider a cylindrical surface impregnated by nonporous mono-
in the gaseous mixture. layer Ni catalyst.
Assume a unit length of catalyst cylinder of 1 cm diameter.
2.6. Thermal conductivity equation Find the circumference of the pellet to be, LC = .D =
3.1416 cm. If the diameter of a Ni atom is DNi = 1.38 A0 =
The thermal conductivity of a polyatomic gas at low den- 1.38 × 10−10 m. Assuming close packing and monolayer of
sity is reliably calculated by the Eucken equation (Bird Ni atoms, the number of Ni atoms that can fit around the
et al., 2002): circumference of the cylinder is NNi = 227.65 × 106 .
  For ammonia decomposition, the turnover number of am-
5 R
ki = cpi + i , (9) monia at 873 K (the reaction temperature) is equal to 0.55
4 Mi (molecules of NH3 ) site−1 s−1 (Choudhary et al., 2001). If
where, i = 1, . . . , N, we assume that every Ni atom is catalytically active, then the
R is the universal gas constant and M is the molecular monoatomar turnover frequency of the circumference of the
weight of the component gas. The thermal conductivity of cylinder is, TOFC =125.21×106 (molecules of NH3 ) s−1 as
gaseous mixture at low pressure can be calculated by, in terms of moles, TOFC = 2.0789 × 10−16 mol of NH3 s−1 .
This number can be multiplied by the number of slices
N

nxi ki in the direction of length of the cylinder. For a unit length
k x = ky = N . (10) (1 cm) of the cylinder, this number would be 72.46 × 106 .
i=1 j =1 nxi ij
Hence, the turnover number of the Ni catalyst pellet for
The coefficients ij are calculated by the following equa- the decomposition of ammonia would be,
tion:
 2 TOFC = 1.506 × 10−8 moles of NH3 s−1 .
   1/2  
1 Mi −1/2  i Mj 1/4
 .
ij = √ 1 + 1+ In the similar way, the total amount of ammonia reacting on
8 Mj j Mi the catalytically active surface of the channel microreactor
(11) has been determined.

2.7. Equation for viscosity 2.9. Boundary conditions

Viscosity of an ideal gas at low pressures can be expressed As discussed in the subsequent section, the flow domains
as a polynomial function of temperature as, considered in the present work are of two different geomet-
rical characteristics. The above-described governing equa-
i =a+bT +cT 2 , (12)
tions are solved subject to the following boundary condi-
where, a, b and c are empirical constants obtained from tions. At the inlet of the catalytic microreactors, a plug flow
literature. (www.cheric.org) (see Appendix B). velocity field is specified. On the impermeable boundaries
The effective viscosity of a multicomponent ideal gas mix- of the microreactors, no slip-wall velocity boundary condi-
ture is calculated by the constitutive equation as, tions are imposed. Zero pressure datum has been specified at
the outlet. The concentration of entering gases is specified at
N

nxi i the inlet of the reactor. In the case of the microreactor with
= N . (13) centrally located catalytically active cylinder, no-slip bound-
i=1 j =1 nxi ij
ary condition is specified on the periphery of the cylinder.
The coefficient ij is calculated by Eq. (11). The gases entering into this microreactor are preheated to
the temperature of 873 K. In case of the channel microre-
2.8. Rate of reaction actor, the catalytically active section wall is maintained at
constant temperature of 873 K.
Decomposition of ammonia is a gas phase reaction and
usually, the rate expression can be expressed in terms of par- 2.10. Finite element formulations
tial pressures of reacting components. Many attempts can
be found in literature solely devoted to the chemical ki- Three different types of governing equations viz, conser-
netics of the ammonia decomposition reaction. (Löffler and vation of momentum, mass transport and energy balance
5868 A.N. Waghode et al. / Chemical Engineering Science 60 (2005) 5862 – 5877

equations are discretised to ordinary linear algebraic equa- where 0  1. The functions given at time-level n + t
tions using the finite element technique. can be evaluated at other time levels using the interpolation,
A|n+t = A|n+1 + (1 − )A|n . (15)
2.10.1. Working equations of the conservation of momentum
The finite element method used for the solution of conser- The final form of the finite element working equations for
vation of momentum equation is based on the UVP scheme the mass transport convection–diffusion equation can be rep-
in which velocity components and pressure are calculated resented as,
simaltaneously. A finite element formulation is developed for
the two dimensional time dependent incompressible Stokes [Kij ]n+1 {cj }n+1 =[Kij ]n {cj }n +{Fj }n +{Bj }n+t , (16)
jp
equations. The inclusion of the term 1c2 jt in the continu- where,
s
ity equation allows utilisation of equal order interpolation     
jNj jN j
models for the velocity and pressure and hence significantly Kijn+1
= Ni Nj + tW i u +v
e jx jy
increases the flexibility of the developed solution scheme  
jNi jNj jN i jN j
(Nassehi, 2002). + t Dx + Dy dx dy,
In this scheme, the spatial variables are left continuous jx jx jy jy
and the governing equation is discretized in time only us- (17)
ing second order Taylor–Galerkin method (Donea, 1984).  
The equation is then discretized in space using the stan- Fj = tN i SCi dx dy, (18)
e
dard Galerkin method. The accuracy of the time stepping   
method is dependent on the highest order of the time deriva- jcie jcie
Bj = tN i Dx nx + D y ny j. (19)
tive remaining in the expansion after its truncation (Nassehi, e jx jy
2002). The final working equations of the Stokes model for
In an analogous fashion, corresponding coefficients for the
the flow dynamics of the free flow regime are represented
matrix [K] and load vector {C} valid for the mass trans-
in the work of Hanspal et al. (2005).
port model can be obtained using the interpolation at time
level n.
2.10.2. Working equations of mass transport and energy In the similar fashion, the final working equations for the
conservation energy conservation model can be written in matrix form as,
As discussed earlier, the mass transport and energy con-
servation equations consist of convective–diffusive terms. If [Eij ]n+1 {Tj }n+1 =[Eij ]n {Tj }n +{Mj }n +{Lj }n+t , (20)
the mass/thermal diffusion coefficient is small, it results in a
high Peclet number. In such cases, the convective–diffusive where,
   
equation falls into the class of hyperbolic partial differen- jN j jN j
Eijn+1 = cp Ni Nj +t cp Wi u +v
tial equations since it is primarily dominated by the convec- e jx jy
tion terms. If ordinary Galerkin finite element formulation  
jNi jNj jN i jN j
is employed, spurious oscillations are observed in concen- + t kx +ky dx dy, (21)
jx jx jy jy
tration/temperature profiles. To achieve stability in results,
 
special techniques such as streamline upwind technique can
Mj = tN i Shi dx dy, (22)
be applied (Zienkiewicz and Taylor, 2002). e
In upwinding scheme, the weight function used in the   
jT e jT e
standard Galerkin finite element formulation is modified by Lj = tN i kx i nx + ky i ny j. (23)
applying a mathematical operator on it, multiplied by a nu- e jx jy
merical parameter. The modified weight function developed In order to cope with the curved geometrical features of
according to the upwinding formulation takes the form, the reactor domain, distorted elements in the computational
v̄i mesh need to be used. Therefore, the discretised working
WI = NI + NI,i , (14) equations are transformed into a local natural coordinate
|v|2
system using isoparametric mapping. The integrals in the
where, NI and NI,i are the shape functions and their deriva- elemental stiffness equations are hence calculated using a
tives, respectively, v is the magnitude of the velocity vec- Gauss–Legendre quadrature. The following steps outline the
tor and  is the upwinding coefficient calculated in terms of solution procedure:
Peclet Number. In the present study, inconsistent streamline
upwinding is used for both mass transport and energy con- 1. The entire domain of interest is discretized into a mesh
servation equations in which the modified weight function of finite elements.
is applied only to the convective terms. The temporal dis- 2. Initial values of the pressure, velocity, concentration and
cretization is carried out by implicit  time-stepping method. temperature fields (mostly zero everywhere) are stored
The discretization is carried out at a time-level n + t, except at the inlet as input arrays.
A.N. Waghode et al. / Chemical Engineering Science 60 (2005) 5862 – 5877 5869

3. The initial temperature in the reactor is assumed to be to 360 numbers of elements. The finite element mesh of the
the reference temperature. reactor domain is shown in Fig. 2.
4. The viscosity of the inlet gaseous mixture is calculated The time step of 1 s and the value of time-stepping param-
at the reference temperature. eter equal to 0.75 is used in the simulations. Velocity vectors
5. The flow equations are solved assuming constant density and pressure along with concentration contours are plotted
and the velocity & the pressure values are obtained. after 20 time steps. Responding to the imposed boundary
6. The mass transport equations are solved assuming con- conditions, the simulated flow field over the reactor domain
stant mass diffusivity and density resulting into concen- obtained after 20 time steps is shown in Fig. 3. The flow
tration values over the domain for the each individual throughout the domain is observed to be fully developed
gaseous component. which is evident from the parabolic velocity profile. Near
7. Using the set of auxiliary equations heat capacity and the catalytic cylinder, low-pressure wakes are formed and
thermal conductivity values are calculated as a function a diverted flow is observed around the cylinder. The veloc-
of temperature. ity values in the narrow gap between the cylinder and the
8. The energy conservation equation is solved to yield the domain walls are maximum due the channelling effect.
temperature field over the reactor domain. The resulting pressure field distribution over the domain is
9. Steps 3–8 are iterated until convergence is reached. Con- shown in Fig. 4. All the pressure values are in Pa. The pres-
vergence is checked using a calculated ratio of the Eu- sure values are gradually decreasing in the direction of flow
clidean norm (Lapidius and Pinder, 1982) between suc- reaching the zero value at the exit. The wake formation near
cessive iterations to the norm, a solution that satisfies the catalytic cylinder, observed from the velocity profile, is
 also reflected in the pressure contours. The total hydrostatic
 N
 i=1 |X r+1 − Xi |2 fluid pressure drop over the entire domain is 0.002 Pa.
 i  (24) The simulated ammonia concentration profile in the mi-
N r+1 2
i=1 |Xi | croreactor domain is shown in Fig. 5 with all the concen-
tration values mentioned in mol m−3 . Though the flow has
is sought where i is the number of iteration cycle, N is attained fully developed status, the concentration profile has
the total number of degrees of freedom and is a pre- not reached steady state which is evident from the varying
selected convergence tolerance value. ammonia concentration in the flow direction. This reduction
The time variable is incremented by t. Using the con- in ammonia concentration is attributed mainly to the fluid
centration and temperature values of the previous iteration, flow and consumption of ammonia due to reaction on the
density and viscosity are updated and the computations are catalytic surface of the cylinder. The effect of diverted flow
repeated for the next time step. The simulation algorithm is caused by the solid catalytic cylinder can be visualised from
developed in FORTRAN environment and the finite element Fig. 5 in the form of a tail of ammonia concentration in the
mesh generation is carried out in COSMOS interface. direction reverse to that of flow indicating some sort of back
diffusion.
The concentration profile for the product hydrogen gas is
3. Computational results and discussions shown in Fig. 6 with concentration values in mol m−3 . As the
hydrogen is being produced only on the surface of catalytic
Results are presented for two types of non-porous catalytic cylinder, its concentration is maximum around the periph-
microreactors with different geometrical characteristics. In ery of the cylinder. Due to the wake formation around the
all the simulations, the fluid is a purely Newtonian gaseous cylinder, hydrogen is found to be accumulating due to lower
fluid composed of ideal gases at all temperatures encoun- velocity values around this specific region. In the first half
tered in the simulations. The density of inlet gas mixture is section of the microreactor domain, the hydrogen concentra-
0.77 kg m−3 which changes as the composition of gaseous tion is ignorable which again gradually decreases from the
mixture changes. The mass diffusivity of individual gaseous catalytic cylinder towards the exit. Exactly the similar form
component in the mixture is equal to 1.28 × 10−5 m2 s−1 of variation could be replicated in the nitrogen concentra-
which is constant throughout the simulations. tion.
In the first case, the microreactor domain is consisting of a The heat dissipated over the microreactor domain can be
centrally located non-porous catalytically active cylinder as conveniently represented by the temperature variation shown
shown in Fig. 1. The length of the reactor is 0.295 m whereas in Fig. 7. All the temperature values are expressed in K and
its width measures 0.02 m. The diameter of non-porous cat- plotted at the steady state attained after 60 successive itera-
alytic cylinder is 0.01 m. The boundary conditions imposed tions. The gases entering the catalytic reactor are preheated
on the microreactor domain walls and exterior boundary of to the temperature of 873 K, which is supposed to be within
catalytic cylinder are shown in Fig. 1. The microreactor do- the feasible reaction temperature range. The temperature be-
main is discretised into finite element mesh using 9-noded fore the catalytic section is nearly constant at the predefined
quadrilateral C0 -continuous biquadratic Lagrange elements. value of 873 K. In the very close front of the catalytic cylin-
The total number of nodes counts to be 1536 corresponding der, the profile shows a gradual change with 5–6 K drop in
5870 A.N. Waghode et al. / Chemical Engineering Science 60 (2005) 5862 – 5877

vx = vy = 0

vx = 0.01 m/s
vy = 0.0

CNH3 = 0.005 mol/m3 vx = vy = 0


CAr = 0.005 mol/m3
CN2 = 0.0 Catalyst Pellet P=0
CH2 = 0.0

T = 8730 K
vx = vy = 0

Fig. 1. Boundary conditions on microreactor domain with centrally located non-porous catalytic cylinder.

Fig. 2. Finite element mesh of a microreactor with centrally located non- porous catalytic cylinder.

Fig. 3. Simulated flow field in a microreactor with centrally located non-porous catalytic cylinder.

0.0022

0.002

0.0018

0.0016

0.0014

0.0012

0.001

0.0008

0.0006

0.0004

0.0002

Fig. 4. Simulated pressure distribution over a microreactor domain with centrally located non-porous catalytic cylinder.
A.N. Waghode et al. / Chemical Engineering Science 60 (2005) 5862 – 5877 5871

0.005

0.0046

0.0042

0.0038

0.0034

0.003

0.0026

0.0022

0.0018

0.0014

0.001

0.0006

0.0002

Fig. 5. Variations in ammonia concentration in microreactor with centrally located non-porous catalytic cylinder.

Fig. 6. Variations in hydrogen concentration in microreactor with centrally located non-porous catalytic cylinder.

the temperature due to the heat sink corresponding to the verged solutions, successive mesh refinement is carried out.
endothermic nature of the reaction. Although the heat ab- The mesh is first refined in the flow direction to duplicate
sorbed during the course of the reaction is very low com- the total number of elements that were present in the parent
pared to the enthalpy of the inlet gases, it is expected to mesh. No deviations are found in concentration and temper-
slightly influence the temperature field around the catalyst. ature profiles compared to those obtained before the mesh
In the mathematical model, the heat sink term is included refinement. When the mesh is successively refined to trip-
in the equations corresponding to the nodes located on the licate the number of elements, the same performance is re-
catalyst surface that inevitably ignores the heat sink that oc- peated confirming that the solutions obtained with the orig-
curs all over the catalyst surface. To compensate the cumu- inal mesh were already converged. However, a significant
lative effect, the total heat sink is redistributed to the nodes rise in computational time is reported which for the densest
present on the catalyst surface. To make the energy balance mesh is several hours.
computationally accurate, the value of the heat sink given at Depending on the nature of the governing differential
the catalyst surface nodes is such that the total is equal to equations, boundary conditions are needed to be enforced
the heat sink integrated over the catalyst surface. The tem- along the whole boundary or parts of the boundary enclos-
perature in the reactor at the rear of catalytic section again ing the domain. These boundary conditions emerge from the
attains a steady state value. physics underlying the process. In the presented results, the
The accuracy of every finite element solution relies to simulated variations of the field unknowns are found to be
certain extent on selection of appropriate meshes. To have perfectly symmetrical along the axis of domain parallel to
an idea about the optimum mesh density, generating con- the direction of flow. This can be regarded as an evidence
5872 A.N. Waghode et al. / Chemical Engineering Science 60 (2005) 5862 – 5877

872

870

868

866

864

862

860

858

856

Fig. 7. Simulated thermal field in microreactor with centrally located non-porous catalytic cylinder.

for the accuracy of the model. However, for any future sim- 6.62
ulations on the physical domain, only half of the geometri- 6.61
cal domain is sufficient which should result in a reduction 6.6
of computational cost. 6.59
% Conversion

The governing mode of transport of heat and mass is de- 6.58


cided by the dimensionless Peclet number (Pe) defined as 6.57
the ratio of convective to diffusive transport. In the present 6.56
case, the value of the Peclet numbers for both heat and 6.55
mass transport are calculated to be around 5, which strongly 6.54
signifies that the transport is not primarily by convection 6.53

mechanism. To provide mathematical evidence for the above 6.52


1.00E-05 1.20E-05 1.40E-05 1.60E-05 1.80E-05 2.00E-05 2.20E-05
conclusion, simulations were carried out without any up-
winding in the standard Galerkin scheme. As per the pre- Diffusion Coefficient (m2/s)
sumption, no paracitic oscillations were observed in the re- Fig. 8. Effect of diffusion coefficient on conversion of ammonia.
sults confirming that convection is not the principle mode
for the transport of gases and energy in the present case.
Since, the mass transport is not found to be dominated by diffusion mechanism is having a mild influence on the extent
convection; the ammonia concentration will need more re- of ammonia decomposition.
actor length to attain steady state concentration. On the con- Simulations are carried out in the context of the ex-
trary, concentration of hydrogen is almost negligible in the perimental data reported by Choudhary et al. (2001) re-
first half of the reactor and the maximum hydrogen concen- garding the activity of 65% Ni/SiO2 /Al2 O3 dispersed
tration is observed near the catalyst. catalyst for decomposition of ammonia. From their ex-
The effect of diffusion on the conversion of ammonia perimental analysis, the turnover frequency for the cata-
was studied by varying the diffusion coefficient value over a lyst at the reaction temperature of 873 K is found to be
certain interval. The conversion in the catalytic reactor can 0.55 molecules site−1 s−1 . Simulations are performed sub-
be calculated by the formula stituting these values in the reaction source/sink term of the
convection–diffusion mass transport equation. The overall
% conversion conversion of ammonia over the microreactor is calculated
Inlet concentration to the reactor−outlet concentration to be 88.7% which is reasonably close to the experimental
= .
Inlet concentration to the reactor value of 79.5% obtained by Choudhary et al. (2001) at
(25) the identical reaction temperature using a continuous-flow
reactor system consisting of conventional fixed bed reactor,
The conversion is plotted against the mass diffusion coef- a quartz reactor along with a furnace. As the fluid flow
ficient in Fig. 8. The % conversion of ammonia increases patterns in this fixed bed reactor is likely to be different
slightly with increase in diffusion coefficient and the to that occurring in the microreactor considered here, the
A.N. Waghode et al. / Chemical Engineering Science 60 (2005) 5862 – 5877 5873

vx = vy = 0
vx = 0.01 m/s
vy= 0

Catalyst (T = 873˚ K)
CNH3 = 0.005 mol/m3 P=0
CAr = 0.005 mol/m3
CN2 = 0.0
CH2 = 0.0
vx = vy = 0

Fig. 9. Boundary conditions on microreactor domain with catalytically active wall.

Fig. 10. Simulated flow field over microreactor domain with catalytically active wall.

discrepancy between the values of conversion is sufficiently to that of reactor in the perpendicular direction. The catalyst
justified. In addition, the catalyst synthesized by Choudhary wall is heated by an external heat source to the temperature
et al. (2001) is in finely powered form with approximately of 873 K. The wall reactor domain is discretized into a finite
8% dispersion of Ni metal on the SiO2 /Al2 O3 support. In element mesh comprised of 9-noded rectangular bi-quadratic
this case, approximately 4.8 × 1018 active sites are exposed Lagrange elements. The total number of elements counts to
to ammonia and in our case 1.7 × 1016 catalytically active 300 with 1281 number of nodes.
sites are considered. However, supported catalysts exhibit Responding to the prescribed boundary conditions, the
mass transfer limitations mainly due to pore-diffusion which flow field developed over the entire computational domain
reduces conversion. In the present work, the catalyst under of the non-porous catalytic wall reactor can be visualised
implementation is Ni metal with monoatomar dispersion on from the velocity vectors plotted in Fig. 10. The flow is ob-
the support surface with each Ni atom catalytically active served to be fully developed resulting in a parabolic profile
therefore resulting in higher value of conversion compared throughout the domain, which is a characteristics of incom-
to the experimental one. pressible viscous flow. The changes in fluid viscosity, den-
The model has been theoretically validated by numerical sity and temperature are not showing any drastic influence
evaluations based on mass balance calculations over the en- on the velocity profile.
tire reactor domain. The discrepancy between the inlet and The hydrostatic pressure profile corresponding to fully
outlet masses is bounded to the satisfactory limit of 1%. developed velocity field is shown in Fig. 11 in the form of
In the second case, a rectangular channel wall reactor contours and the values indicated are in Pa.
is considered as a domain under study. Some intermediate Due to the zero pressure datum applied at the exit bound-
portion of the reactor wall is catalytically active and the ary, the pressure values gradually decrease to zero in the
width of the catalytic section is equal to the width of the direction of flow. The overall pressure drop across the flow
reactor in z-direction. The geometrical features of the reactor domain is calculated to be 3.6 × 10−4 Pa.
along with the imposed boundary conditions are shown in The solution of the convection–diffusion–reaction mass
Fig. 9. The length of this catalytic non-porous reactor is transport equation gives the concentration values of the re-
equal to 0.015 m with 0.005 m in width. The reactor extends actant gas ammonia along with the product gases hydro-
to 0.05 m in width in the z-direction. The catalytically active gen, nitrogen and the inert carrier argon. The concentration
region of the walls is 0.005 m in length and has a width equal profile of the reactant ammonia is shown in Fig. 12 in the
5874 A.N. Waghode et al. / Chemical Engineering Science 60 (2005) 5862 – 5877

3.80E-004
3.60E-004
3.20E-004
2.80E-004
2.40E-004
2.00E-004
1.60E-004
1.20E-004
8.00E-005
4.00E-005
0.00E+000

Fig. 11. Simulated pressure distribution over microreactor domain with catalytically active wall.

0.004998
0.004996
0.004994
0.004992
0.00499
0.004988
0.004986
0.004984
0.004982
0.00498
0.004978
0.004976
0.004974
0.004972
0.00497
0.004968
0.004966
0004964

Fig. 12. Variations in ammonia concentration in microreactor with catalytically active wall.

form of coloured contours with the scale provided in the to be at reference temperature of 273 K. The external heat
unit of mol m−3 to compare the relative values. The results is only applied to catalytic sections of the domain walls and
presented in Fig. 12 are after 30 time steps. A reduction the maximum temperature is observed in the vicinity of that
in ammonia concentration is observed near the walls of the section. The applied heat gets dissipated to the central sec-
domain where reaction is taking place. tion of the reactor from the domain walls. Though the heat
The nitrogen gas concentration in mol m−3 is shown in consumed during the reaction is small, it shows its influence
Fig. 13 in the form of colour contours. The concentration of on the temperature variation in the close vicinity of catalytic
nitrogen is maximum near the catalytic section of the do- sections. The applied heat is observed to get propagated in
main walls. Near the entrance of the reactor, the concentra- the flow direction towards the exit. Though the reaction is
tion value is minimum. This concentration profile is exactly endothermic, the heat absorbed by the reaction is very less
opposite to that with ammonia where concentration value showing no significant influence on the overall energy dis-
is less near the catalytic walls. An exactly similar profile is tribution in the reactor.
obtained for the other product gas hydrogen with all the pro- Like the previous case, the optimum mesh has been found
files found to be symmetrical along direction of the flow of by successive mesh refinements. It results into no further
gases. improvement in accuracy of results but caused unnecessary
The energy distribution over the reactor domain can be increase in computational time. Since, the Peclet number
explained with the help of temperature variation in the reac- for both mass transport and heat transport is very small, no
tor in Fig. 14. The temperature values indicated on the scale oscillatory behaviour is observed even when no upwinding
are in K. The gases entering the reactor domain are assumed technique is applied to Galerkin finite element formulation
A.N. Waghode et al. / Chemical Engineering Science 60 (2005) 5862 – 5877 5875

Fig. 13. Variations in nitrogen concentration in microreactor with catalytically active wall.

880
840
800
760
720
680
640
600
560
520
480
440
400
360
320
280

Fig. 14. Simulated thermal field in microreactor with catalytically active wall.

of the convective term. The continuity of the mass trans- solutions except that the computational time has increased
ferred throughout the domain is examined and the discrep- drastically for calculations using small time-step values.
ancy between inlet and outlet masses is found to be less than
0.7% confirming the conservation of mass.
A permissible range of time step values are considered in 4. Conclusions
simulations of both these microreactors to have an idea about
variations in field variable on time scale. However, no visible The fluid flow behaviour of gaseous fluid has been
change has been observed in the stability and accuracy of the analysed in two different types of non-porous catalytic
5876 A.N. Waghode et al. / Chemical Engineering Science 60 (2005) 5862 – 5877

microreactors. In both these microreactors, the flow field is ny component of the unit normal vector in the y
observed to be fully developed when a sufficient channel direction
length for flow is provided. In the first microreactor with Ni elemental shape function/weight function
centrally located catalytically active cylinder, diverted flow p pressure
has been observed in the vicinity of the pellet due to sudden p̃ interpolated form of pressure
low-pressure wake formation. P normalised pressure
Ammonia is getting consumed near the catalytically ac- P pressure drop
tive surface. In the same way, product gases nitrogen and Sc mass source/sink term
hydrogen are found to be formed and accumulated near the Sh energy source/sink term
reaction surface. The concentration of the inert carrier Ar- t time variable
gon remains unchanged throughout the reactor. T temperature
In the case of the microreactor with centrally located cat- TOF turnover number
alytically active cylinder, the external heat is supplied by U normalised x component of the velocity
preheating the reacting gaseous mixture. The applied heat is vn fluid velocity normal to the boundary
found to be dissipated in the domain primarily in the direc- ṽx interpolated form of the x component of the
tion of flow due to diffusive transport mechanism. Though velocity vector
the endothermic heat of reaction has a very low value, it ṽy interpolated form of the y component of the
seems to affect the energy distribution near the catalytic velocity vector
section with no significant effect on the global tempera- vx velocity in the x direction
ture variation in the reactor. Since, the sink term due to vy velocity in the y direction
the reaction is measured on a microscopic level in terms of V normalised y component of the velocity
turnover frequency of the catalyst, it should be corrected by
a pre-estimated numerical value when incorporated in the Greek letters
macroscopic global energy balance equation for the reactor
system. ˙ shear rate
In the case of the channel microreactor, the external heat  boundary of the solution domain
is supplied at the catalytic section of the wall thereby nul- e elemental boundary
lifying the effect of microscopic heat of reaction sink term  viscosity of fluid
on macroscopic level of reactor. The maximum temperature  time stepping parameter
is observed to be in the vicinity of the catalytic section as  fluid density
predicted.  solution domain
The conversion values of ammonia in previously reported e solution domain
experiments in the literature and in our simulations exhibit
appreciable match when considered in reaction kinetics point
of view However, this comparison does not seem to be ra-
tional when considered the macroscopic phenomena of fluid Appendix A.
dynamics and mass transport on the geometric scale of re-
actor systems. Empirical constants in the heat capacity equation cp =
In both the cases, the convection shows a negligible a + b · T + c · T 2 + d · T 3 , cal mol−1 K −1 are given in the
impact on both transport mechanisms viz. heat and mass following table (Table 1).
transfer.

Notation Appendix B.

c concentration of gaseous component Empirical constants in the viscosity equation  = a + b ·


cp specific heat capacity at constant pressure T + c · T 2 , cP are given in the following table (Table 2).
cs speed of sound in the fluid
Dx mass diffusivity of gaseous component in the
x-direction Table 1
Dy mass diffusivity of gaseous component in the Component/constants a b c d
y-direction
kx thermal conductivity in x-direction Ammonia 6.524 5.693E-03 4.078E-06 −2.83-09
Nitrogen 7.440 −0.324E-03 6.400E-06 −2.79E-09
ky thermal conductivity in y-direction Hydrogen 6.483 2.215E-03 −3.298E-06 1.826E-09
nx component of the unit normal vector in the x Argon 4.969 −0.767E-05 1.234E-08 0.0
direction
A.N. Waghode et al. / Chemical Engineering Science 60 (2005) 5862 – 5877 5877

Table 2 Hsing, I.M., Srinivasan, R., Harold, M.P., Jensen, K., Schmidt, M.A., 2000.
Simulation of micromachined chemical reactors for heterogeneous
Component/constants a b c
partial oxidation reactions. Chemical Engineering Science 55, 3–13.
Ammonia −0.0009372 3.899E-05 −4.405E-09 Lapidius, L., Pinder, G.F., 1982. Numerical Solution of Partial Differential
Nitrogen 0.003043 4.989E-05 −1.093E-08 Equations in Science and Engineering. Wiley, New York.
Hydrogen 0.002187 0.0000222 −3.751E-09 Löffler, D.G., Schmidt, L.D., 1976. Kinetics of ammonia decomposition
Argon 0.004387 6.399E-05 −1.284E-08 on polycrystalline Pt. Journal of Catalysis 41, 440–454.
Mojtahedi, W., Abbasian, J., 1995. Catalytic decomposition of ammonia
in fuel gas at high temperature and pressure. Fuel 74 (11), 1698–1710.
Nassehi, V., 2002. Practical Aspects of Finite Element Modelling of
References Polymer Processing. Wiley, Oxford.
Papapolymerou, G., Bontozoglou, V., 1997. Decomposition of NH3 on
Alhumaizi, K., Henda, R., Soliman, M., 2003. Numerical analysis of Pd and Ir comparison with Pt and Rh. Journal of Molecular Catalysys
a reaction–diffusion–convection system. Computers and Chemical A: Chemical 120, 167–171.
Engineering 27, 579–594. Park, S.-K., 1995. Application of Petrov–Galerkin finite element method
Bird, R.B., Stewart, W.E., Lightfoot, E.N., 2002. Transport Phenomena. to nonlinear transient diffusion–convection–reaction system. Chemical
second ed.. Wiley, New York. Engineering Communications 139, 159–200.
Chellapa, A.S., Fisher, C.M., Thomson, W.J., 2002. Ammonia Reddy, J.N., 1993. An Introduction to Finite Element Method. second ed.
decomposition kinetics over Ni–Pt/Al2 O3 for PEM fuel cell McGraw-Hill Inc, New York.
applications. Applied Catalysis A: General 227, 231–240. Reid, R.C., Prausnitz, J.M., Sherwood, T.K., 1977. The Properties of
Choudhary, T.V., Sivadinarayana, C., Goodman, D.W., 2001. Catalytic Gases and Liquids. third ed. McGraw Hill International, New York.
ammonia decomposition: COx -free hydrogen production for fuel cell SZ´rensen, R.Z., Nielsen, L.J.E., Jensen, S., Hansen, O., Johannsen,
applications. Catalysis Letters 72 (3–4), 197–201. T., Quaade, U., Christiansen, C.H., 2005. Catalytic ammonia
Del Giudice, S., Trotta, A., 1978. Transient diffusion with n-order decomposition: miniatured production of COx -free hydrogen for fuel
irreversible chemical reaction: a finite element approach. Chemical cells. Catalysis Communications 6, 229–232.
Engineering Science 33, 697–702. Trotta, A., Del Giudice, S., 1985. Finite element solution of simultaneous
Deshmukh, S.R., Mhadeshwar, A.B., Vlachos, D.G., 2004. Microreactor convection diffusion and reaction on catalytic surface. Computers and
modelling for hydrogen production from ammonia decomposition Chemical Engineering 9 (2), 167–173.
on Rubidium. Industrial and Engineering Chemistry Research 43, Wang, W., Padban, N., Ye, Z., Andersson, A., Bjerle, I., 1999. Kinetics of
2986–2999. ammonia decomposition in hot gas cleaning. Industrial and Engineering
Djèga-Mariadassou, G., Shin, C.-H., Bugli, G., 1999. Tamaru’s model for Chemistry Research 38 (11), 4175–4182.
ammonia decomposition over titanium oxynitride. Journal of Molecular Yin, S.F., Xu, B.Q., Zhou, X.P., Au, C.A., 2004. A mini-review on
Catalysis A: Chemical 141, 263–267. ammonia decomposition catalysts for on-site generation of hydrogen
Donea, J.A., 1984. Taylor-Galerkin method for convective transport for fuel cell applications. Applied Catalysis A: General 277, 1–9.
problems. International Journal for Numerical Methods in Engineering Zienkiewicz, O.C., Wu, J., 1991. Incompressibility without tears—how
20, 101–119. to avoid restrictions on mixed formulation. International Journal of
Ertl, G., Huber, M., 1980. Mechanism and kinetics of ammonia Numerical Methods in Engineering 32, 1189–1203.
decomposition on Iron. Journal of Catalysis 61, 537–547. Zienkiewicz, O.C., Nithiarasu, P., Codina, R., Vázquez, M., Ortiz, P.,
Ganley, J.C., Seebauer, E.G., Masel, R.I., 2004a. Development of a 1999. The charecteristic-based-split procedure: an efficient and accurate
microreactor for the production of hydrogen from ammonia. Journal algorithm for fluid problems. International Journal for Numerical
of Power Sources 137, 53–61. Methods in Fluids 31, 359–392.
Ganley, J.C., Seebauer, E.G., Masel, R.I., 2004b. Porous anodic alumina Zienkiewicz, O.C., Taylor, R.L., 2002. The Finite Element Method. fifth
posts as a catalyst support in microreactors for production of hydrogen ed. vol. 3. Butterworth-Heinemann, Oxford.
from ammonia. A.I.Ch.E. Journal 50 (4), 829–850. www.cheric.org.
Hanspal, N.S., Waghode, A.N., Nassehi, V., Wakeman, R.J., 2005.
Numerical analysis of coupled Stokes/Darcy flows in industrial
filtrations. Transport in Porous Media, submitted for publication.

Anda mungkin juga menyukai