Memorias del XXXIV Encuentro Nacional y III Congreso Internacional de la AMIDIQ

7 al 10 de mayo de 2013, Mazatlán, Sinaloa, México

AN APPROXIMATE APPROACH FOR UPSCALING NONLINEAR REACTIVE TRANSPORT

IN POROUS MEDIA

Helen D. Lugo-Méndez, Francisco J. Valdés-Parada, J. Alberto Ochoa-Tapia

Departamento de Ingeniería de Procesos e Hidráulica. Universidad Autónoma Metropolitana-Iztapalapa,
Av. San Rafael Atlixco No.186, Col. Vicentina, 09340, México D.F., México. [email protected]

Abstract
In this work we use the method of volume averaging to upscale the process of diffusion and reaction of a
chemical species undergoing a homogeneous chemical reaction that obeys non-linear kinetics. If no
length- or time -scale constraints are imposed, the resulting macroscopic model is nonlocal, nonlinear
and strongly coupled with the microscale transport. Therefore, by imposing a convenient set of
constraints and assumptions (i.e., scaling postulates), we were able to derive a local macroscale model
that is nonlinear both in the reactive and diffusive terms. A highlight of our approach is that the closure
problem associated to the macroscale model can be written in a linear form by using Taylor series
expansions about the volume averaged concentration. The results from this work provide a justification
for upscaling nonlinear problems and it provides a threshold of applicability of the macroscale model in
terms of the Thiele modulus and the characteristic lengths and time-scales involved.

Introduction
Many systems of hierarchical nature involve transport and reaction in which the reaction rate is
nonlinear. The derivation of the equations valid at certain levels of scale usually involves assuming that
the nonlinear term exhibits a vis-à-vis resemblance with its counterpart at an inferior level of scale.
However, this approach involves assumptions that have not been acknowledged in the literature. In this
work, we use the method of volume averaging [1] to upscale the process of diffusive transport in a
porous medium that is saturated by only one fluid phase carrying a solute (species A ) undergoing a
homogeneous and nonlinear reaction. The results from this study are useful to clearly identify the
situations under which vis-à vis models are applicable. Furthermore, by allowing the effective diffusivity
to depend of the reaction rate, we were able to obtain models that are less limited than those previously
reported in the literature [2].

Upscaling
Consider a rigid and homogeneous porous medium, saturated by only one fluid phase (the  -phase) that
carries a solute (species A ) forming a dilute solution. In this work, only diffusive mass transport and a
homogeneous reaction in the  -phase are considered, however extensions to other situations are
straightforward. In this way, the governing equation for the concentration of species A at the pore-scale
is
cA
    D cA   R  cA  , in the  -phase
t (1)
As a first approach we assume that the solid phase (the -phase) is impermeable to mass transfer. In the
method of volume averaging, for a piecewise smooth function defined everywhere in the  -phase, the
intrinsic averaging operator over the averaging domain is introduced as
1
dV  y 
| V | y V  x  x y
     
x
(2)

329
© 2013 Academia Mexicana de Investigación y Docencia en Ingeniería Química (AMIDIQ)
ISBN: 978-607-95593-1-1
 Memorias del XXXIV Encuentro Nacional y III Congreso Internacional de la AMIDIQ
7 al 10 de mayo de 2013, Mazatlán, Sinaloa, México

The notation will be relaxed by using V | V | . Following the steps established in the method of
volume averaging [1]; and using the decomposition of the point concentration, cA  cA    cA [3]; the
local unclosed upscaled equation valid everywhere in the bulk of the porous medium results as

 cA   
 
 1 

V yA  ( x)
   D   cA  n c
 A dA( y)    R c
   A
t
  (3)
This equation is constrained to  r0 L , where L is the smallest large length-scale associated with
the problem under consideration and r0 is the characteristic length of the averaging domain, V .

Closure problem
To close Eq. (3), it is necessary to derive and formally solve the BVP for the concentration deviations,
c A , which can be obtained by subtracting Eq. (3) to Eq. (1). In the resulting expression, if the length-
scale associated with the pores,  , satisfies  Lc the non-local diffusion term can be discarded with
respect to the local diffusion term, however, despite these remarkable simplifications, the closure
problem requires to be solved in the entire macroscopic domain. To solve the closure in a more
convenient domain, the complex microscopic geometry of the representative region is modelled as a
spatially periodic porous medium generated by three non-unique lattice vectors l i (i  1, 2,3) defining a
representative unit cell,  . If the characteristic lengths, r0 and , of the representative region and unit
cell are constrained by  r0 L , then both cA   and cA   can be treated as position-invariant in
 and the boundary condition for c A at the entrances and exits of the macroscopic domain are no
longer necessary. Moreover, if the time-scale constraint 1  c 
1
D t * 2
 (where c  2
 R D cA ) is
satisfied the influence of the initial condition over the c A -fields becomes negligible, and the volume
averaged concentration and its derivatives can be assumed as constants in the time-scale of c A . These
scaling postulates lead to a quasi-steady, local and periodic closure problem, given by
 
   D cA   R cA  cA    0, in 
(4)

n  D cA  n  D cA  , at 
(5)
cA  y  li   cA  y  , i  1,2,3 (6)
cA    0 (7)
In the above equations,  represents the  -phase contained within the unit cell,  denotes the  - 
interface, and R are the reaction rate deviations, which can be a nonlinear function. The average
constraint given in Eq. (7) allows bounding the fields of the concentration deviations.

Linear closure problem

With the aim of obtaining an explicit solution for the closure problem, in this section we follow an
approach similar to the one used by Whitaker [see Problem 1-11 in ref. 1] in order to linearize the
reaction rate term. This is accomplished by
 Using a Taylor-series expansion of R around cA   , and
 Filtering the redundant information by carrying out an order of magnitude analysis based on the
time/length-scale constraints established in the scaling postulates previously stated.

330
© 2013 Academia Mexicana de Investigación y Docencia en Ingeniería Química (AMIDIQ)
ISBN: 978-607-95593-1-1
 Memorias del XXXIV Encuentro Nacional y III Congreso Internacional de la AMIDIQ
7 al 10 de mayo de 2013, Mazatlán, Sinaloa, México

If the reaction-rate function described by R  R(cA ) is an infinitely differentiable function in  , then

it can be locally defined by its convergent Taylor series. This means that there exists some neighborhood
r  0 such that for all cA   cA    r, cA    r  the reaction-rate function is given by [4]
dR 1 d2R
Rc
A |xy
 R c
A  |x

dcA
cA |x  y 
2! dcA2 
cA2 |x  y , 
cA  r c A    (8)
 
 cA  |x  cA  |x

Applying the intrinsic average operator to Eq. (8), and carrying out an order of magnitude estimate
based on the stated scaling postulates leads to the following simplified expressions for R cA   and
 R  c 
A

dR
R  cA |x  y and  R c   |x  R  c  |x
(9)
c A |xy dcA 
A A
 c A  |x

So that Eq. (4) can be written as

dR
   D cA   cA  0, in 
dcA 
 c A  |x
(10)
This equation is subject to the boundary conditions given by eqs. (5)-(7). The closure problem can be
therefore treated as local, periodic, quasi-steady, and lineal.

Formal solution of the simplified closure problem

The formal solution of the closure problem in  , based on Green's functions and taking into account the
fact that volume averaged quantities can be assumed constants within  , can be expressed as follows
dR
cA (x, t )    Gqs (x, y)cA  y, t  dV (y)  b cA   (11)
dcA
 cA  |x y

As a matter of convenience, we defined the closure variable b as

b  
y
 Gqs (x, y)n D dA(y)
(12)
Notice that b solves the BVP given by eqs. (1.4-58) in ref. [1], which is independent of the reaction
rate. In eqs. (11) and (12), Gqs (x, y) is the passive Green's function that solves the following BVP
  D Gqs    (x  y), in 
(13)
n  D Gqs  0, at 
(14)
Gqs  y  li   Gqs  y  , i  1,2,3 (15)
The result given by Eq. (11) clearly separates the reactive and diffusive contributions to the deviations.
However, it remains an implicit solution. To overcome this issue, we formally solve the linear closure
problem given by eqs. (10) and (5)-(7), using integral equation formulations based on Green's functions,
the result can be expressed as
 
cA (x)     Grx (x, y)n D dA(y)  cA   (16)
 y 
  

331
© 2013 Academia Mexicana de Investigación y Docencia en Ingeniería Química (AMIDIQ)
ISBN: 978-607-95593-1-1
 Memorias del XXXIV Encuentro Nacional y III Congreso Internacional de la AMIDIQ
7 al 10 de mayo de 2013, Mazatlán, Sinaloa, México

Where Grx is the reactive-diffusive Green's function, which solves the following BVP
dR
D 2Grx  y; y0   Grx  y; y0     y  y0  , in 
dcA
 cA  |x
(17)
n  D Grx  y; y0   0, at 
(18)
Grx  y  i ; y0   Grx  y; y0  , i  1,2,3 (19)
Grx    0 (20)
Indeed, Eq. (16) is an explicit solution of the closure problem and can be used to close the macroscale
model. However, the Green's function Grx couples reactive and diffusive transport as recently pointed
out in the literature for diffusive and dispersive reactive transport [5, 6]. With the aim of separating both
contributions, we substitute Eq. (16) into Eq. (11), the result is given by
dR
cA (x)  brx cA    b cA   (21)
dcA
 cA  |x

For convenience, we defined the closure variable, brx , as

brx   
y y0 
Gqs (x, y)Grx (y, y0 )n D dA(y) dV (y0 )
(22)
Certainly, we may rewrite Eq. (21) in the more compact form,
cA  b cA   (23)
where the closure variable b is given by
dR
b brx  b (24)
dcA
 c A  |x

The BVPs for b and brx are obtained by substituting eqs. (21)and (23) into Eq. (10) and are coupled
since b is the source term in the differential equation for brx.,

Closed upscaled model

To develop the closed form of the upscaled reaction-diffusion equation in terms of effective transport
coefficients, eqs. (21) and (9) are substituted into Eq. (3). The resulting equation is
cA  
   Ddiff-rx cA     R  cA    (25)
t
where Ddiff-rx is the effective reaction-diffusion coefficient composed by the effective diffusivity tensor,
Deff , and a diffusion-like coefficient that is dependent of the reaction rate, Drx . These effective
coefficients can be written in terms of the closure variables as follows,
 1 
Ddiff-rx  c A     D I    n  y  b   y  dA( y ) 
 V yA  
Deff

D dR

V dc A   
n   y  b rx y, c A   dA(y)
 c A  |x yA
 

Drx
332
© 2013 Academia Mexicana de Investigación y Docencia en Ingeniería Química (AMIDIQ)
ISBN: 978-607-95593-1-1
  1 
Memorias del XXXIV
Ddiff-rx  
c A  Encuentro
V yA 
 D I Nacional ny  y
III  b 
Congreso
 y  dA( y ) 
Internacional de la AMIDIQ
7 al 10 de mayo 
 de 2013, Mazatlán, Sinaloa, México 
Deff

D dR

V dc A   
n   y  b rx y, c A   dA(y)
 c A  |x yA 

Drx
(26)
Results
Computation of the effective diffusivity
The numerical solution of the closure problems can be carried out by numerical or analytical means. In
the first case, one can solve the BVPs for b and b using any numerical method. In this work we used
the software Comsol Multiphysics to numerically solve the closure problem using finite elements. In
order to derive the analytical closure problem solution, we use a Chang's unit cell [7, 8], which is a non-
periodic domain conformed by two homothetic concentric regions. Since Chang's unit cell is not
periodic, homogeneous Dirichlet-type boundary conditions are enforced at the outer boundary of the cell,
xx
thus abandoning the periodic boundary condition, Eq. (6). Figure 1 presents the values for Deff,rxn / D as
a function of the closure Thiele modulus 2  R  cA   / c0  2
/ D , for different porosity values. This
figure shows the good agreement between the predictions obtained using Chang's unit cell and those
obtained using a spatially periodic model for   1 and   50 , demonstrating the periodic boundary
conditions do not play a particularly important role over these range. For 1    50 , the differences
between the predicted values of the coefficient using 2D periodic and non-periodic systems are higher to
those obtained using 3D models; however, these differences are still within a 10% threshold with respect
to the numerical solution.

Figure 1.- Effect of closure Thiele modulus  , and porosity on the effective reaction-diffusion
coefficient predicted in 2D and 3D unit cells of a spatially periodic array of squares and cubes, and in
cylindrical and spherical Chang's unit cells.

In order to determine the range of applicability of the macroscale model, we solved Eq. (25) in a one-
dimensional domain subject to the following boundary conditions

 d cA
x  0, cA  c0 ; x  L, 0 (27)
dx

333
© 2013 Academia Mexicana de Investigación y Docencia en Ingeniería Química (AMIDIQ)
ISBN: 978-607-95593-1-1
 Memorias del XXXIV Encuentro Nacional y III Congreso Internacional de la AMIDIQ
7 al 10 de mayo de 2013, Mazatlán, Sinaloa, México

The concentration fields are compared with the averages of those arising from the solution of the
governing equations at the pore scale in a two-dimensional domain (i.e., direct numerical simulations or
DNS). The results are presented in Figure 2 for two porosity values and considering three expressions
for the reaction rate. From these results, it is observed that for situations in which the reaction rate, at the
macroscale, overcomes the diffusive transport, it is more difficult for the upscaled model to reproduce
the results from the DNS regardless of the periodic or non-periodic nature of the unit cell. Despite this
limitation, the methodology used in this work has a wide range of applicability in many systems of
hierarchical nature involving transport and nonlinear sources.

Figure 2.- Macroscopic concentration profiles for three reaction rates predicted from direct numerical simulations (DNS)
and with the upscaled model using periodic and non-periodic unit cells for the closure problem solution.

Conclusions
In this work we presented a methodology to upscale transport and reaction processes involving nonlinear
terms. Our approach was based on the linearization of the reaction rate about the average concentration
value. The length-scale constraints supporting this approximation were identified and our comparison
with direct numerical simulations suggests that this approach is reliable. The derivations presented in
this work can certainly be extended to other situations such as dispersive and reactive transport,
heterogeneous reactions, etc.

References
1. Whitaker S. The Method of Volume Averaging. Kluwer Academic Publishers, 1999.
2. Wood B.D. and Whitaker S. “Diffusion  and  reaction  in  biofilms.” Chem. Eng. Sci. Vol. 52, No 3, p. 397-425, 1998.
3. Gray W. G. “A derivation of the equations for multiphase transport.” Chem. Engng. Sci. Vol. 30 , p. 229-233, 1975.
4. Arfken G. B, Weber H. J. “Mathematical Methods for Physicists.” Seventh Edition. Academic Press, 2013.
5. Valdés-Parada F. J., Alvarez-Ramirez J. “On the effective diffusivity under chemical reaction in porous media.” Chem.
Eng. Sc., Vol. 6, No 13, p. 4100-4104, 2010.
6. Valdés-Parada F. J., Aguilar-Madera C. J., Álvarez-Ramírez J. “On diffusion, dispersion  and  reaction  in  porous  media.”
Chem. Eng. Sc. Vol. 66, No. 10, p. 2177–2190, 2010.
7. Chang H. C. “Multi-scale analysis of effective transport in periodic  heterogeneous  media.” Chem. Eng. Comm. Vol. 15,
p. 83-91, 1982.
8. Ochoa-Tapia J.   A.,   Stroeve   P.,   Whitaker   S.   “Diffusive transport in two-phase media: spatially periodic models and
Maxwell’s  theory  for  isotropic  and  anisotropic systems.” Chem. Eng. Sci. Vol. 49, p. 709-726, 1994.

334
© 2013 Academia Mexicana de Investigación y Docencia en Ingeniería Química (AMIDIQ)
ISBN: 978-607-95593-1-1