International Communications in Heat and Mass Transfer 52 (2014) 132–139

Contents lists available at ScienceDirect

International Communications in Heat and Mass Transfer

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

Analysis of turbulent double-diffusive free convection in porous media

using the two-energy equation model☆
Marcelo J.S. de Lemos
Departamento de Energia — IEME, Instituto Tecnológico de Aeronáutica — ITA, 12228-900 São José dos Campos, SP, Brazil

a r t i c l e i n f o a b s t r a c t

Available online 15 January 2014 This paper presents an analysis of macroscopic heat and mass transport for turbulent flow in permeable struc-
tures, which is based on the thermal non-equilibrium assumption between the porous matrix and the working
Keywords: fluid. Two driving mechanisms are here considered to contribute to the overall momentum transport, namely
Double-diffusion fluid-temperature driven and concentration driven mass fluxes. The fluid temperature, however, is also affected
Natural convection by the solid temperature distribution as the two phases exchange heat through their interfacial area. Essentially,
Thermal non-equilibrium
here the double-diffusive natural convection mechanism is investigated for the fluid phase in turbulent regime.
Turbulence modeling
Porous media
Equations are presented based on the double-decomposition concept, which considers both time fluctuations
Volume-average and spatial deviations about mean values. This work intends to demonstrate that additional transport mecha-
Time-average nisms are mathematically derived if velocity, fluid temperature and mass concentration simultaneously present
Mass transport time fluctuations and spatial deviations about average values. A modeled form for the entire set of transport
equations is presented where turbulent transfer is based on a macroscopic version of the k–ε model.
© 2014 Elsevier Ltd. All rights reserved.

1. Introduction with the so-called two-energy-equation model, or 2EEM for short,

which is based on the Local Thermal Non-equilibrium Hypothesis
Analyses of double-diffusive phenomena in free convection in (LTNE) meaning that the average temperature of the fluid is not equal
permeable media has many environmental and industrial applications, to the average temperature of the solid matrix However, in ref. [17] no
such as in oil and gas extraction, movement of gas concentration into double-diffusion was considered.
the ground, contaminant dispersion in soils, grain storage and drying, Therefore, the purpose of this contribution is to extend the work in
petrochemical processes, electrochemical processes, to mention a few ref. [16] on turbulent double-diffusion using only one energy equation,
[1–9]. In some specific applications, the voids are large enough and assuming now the thermal non-equilibrium hypotheses in ref. [17],
the fluid mixture may become turbulent. In such instances, difficulties which requires an independent energy balance for each phase. As
arise in the proper mathematical modeling of the transport processes such, the expectation herein is that, by combining now such two models
under both temperature and concentration gradients. that were developed on separate, a larger number of physical processes
Usually, modeling of macroscopic transport for incompressible flows can now be more realistically tackled.
in rigid porous media has been based on the volume-average methodol-
ogy for either heat or mass transfer [10–14]. If fluctuations in time are 2. Local instantaneous transport equation
also of concern due the existence of turbulence in the intra-pore space,
a variety of mathematical models have been published in the literature The steady-state local (microscopic) instantaneous transport equa-
in the last decade. One of such views, which entails simultaneous applica- tions for an incompressible binary fluid mixture with constant properties
tion of both time and volume averaging operators to all governing flowing in an inert heterogeneous medium are given in details elsewhere
equations, has been organized and published in a book [15] that describes, and for that, they will be just repeated here. They read:
in detail, an idea known in the literature as the double-decomposition
concept (see chapter 3, pgs. 27–32 in ref. [15] for details). within the fluid:
In an earlier work [16], double-diffusive effects in porous media have
been treated considering thermal equilibrium between the porous matrix Continuity∇
u ¼ 0 ð1Þ
and the permeating fluid. Or say, in ref. [16] the fluid temperate was as- 2
sumed to be the same of that of the solid when analyzing double- Momentumρ∇
ðuuÞ ¼ −∇p þ μ∇ u þ ρ g ð2Þ
diffusive mechanisms. Later [17], buoyancy-free flows were investigated
 n
o

Energy‐fluid phase ρcp ∇
uT f ¼ ∇
k f ∇T f þ S f : ð3Þ
f
☆ Communicated by W.J. Minkowycz.
E-mail address: delemos@ita.br. Mass concentration ρ∇
ðu mℓ þ Jℓ Þ ¼ ρ Rℓ ð4Þ

0735-1933/$ – see front matter © 2014 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.icheatmasstransfer.2014.01.017
 M.J.S. de Lemos / International Communications in Heat and Mass Transfer 52 (2014) 132–139 133

defined as mℓ = ρℓ/ρ, ρℓ is the mass density of species ℓ (mass of ℓ

Nomenclature
over total mixture volume), ρ is the bulk density of the mixture
 
cF Forchheimer coefficient ρ ¼ ∑ ρℓ , p is the pressure, μ is the fluid mixture viscosity, g is the
ℓ
Cℓ volumetric molar concentration
gravity acceleration vector, cp is the specific heat, the subscripts f and s
cp specific heat
refer to fluid and solid phases, respectively, Tf and Ts are the fluid and
Dℓ diffusion coefficient
solid temperature, kf and ks are the fluid and solid thermal conductivi-
Ddisp Mass dispersion
ties and S is the heat generation term. If there is no heat generation ei-
Ddisp,t turbulent mass dispersion
ther in the solid or in the fluid, one has further Sf = Ss = 0. The
Dt turbulent mass flux
generation rate of species ℓ per unit of mixture mass is given in
g gravity acceleration vector
Eq. (4) by Rℓ. Also, as pointed out in ref. [16], an alternative way of writ-
I unity tensor
ing the mass transport equation is using the volumetric molar concen-
Jℓ mass diffusion coefficient
tration Cℓ (mol of ℓ over total mixture volume), the molar weight Mℓ
k turbulent kinetic energy per unit mass, k ¼ u′
u′ =2
(g/mol of ℓ) and the molar generation/destruction rate R ∗ℓ (mol of ℓ/
bkNi intrinsic (fluid) average of k
total mixture volume), giving:
K permeability
ℓ chemical species 
Mℓ ∇
ðu C ℓ þ Jℓ Þ ¼ M ℓ Rℓ : ð6Þ
mℓ mass fraction of component ℓ
Mℓ molar weight of component ℓ
p pressure Further, the mass diffusion flux Jℓ (mass of ℓ per unit area per unit
Prt turbulent Prandtl number time) in Eq. (4) or (6) is due to the velocity slip of species ℓ,
Sct turbulent Schmidt number
T temperature J ¼ ρℓ ðuℓ −uÞ ¼ −ρℓ Dℓ ∇mℓ ¼ −M ℓ Dℓ ∇C ℓ ð7Þ
u mass-averaged velocity of the mixture
uD Darcy velocity vector where Dℓ is the diffusion coefficient of species ℓ into the mixture. The
uℓ velocity of species ℓ second equality in Eq. (7) is known as Fick's Law, which is a constitutive
equation strictly valid for binary mixtures under the absence of any
additional driving mechanisms for mass transfer [10]. Therefore, no
Greek symbols Soret or Dufour effects are here considered.
β thermal expansion coefficient Rearranging Eq. (6) for an inert species, dividing it by Mℓ and
βC salute expansion coefficient dropping the index ℓ for a simple binary mixture, one has,
βϕ macroscopic thermal expansion coefficient
βCϕ macroscopic salute expansion coefficient ∇
ðu C Þ ¼ ∇
ðD∇C Þ: ð8Þ
λ fluid thermal conductivity
μ fluid mixture viscosity If one considers that the density in the last term of Eq. (2) varies with
μt Turbulent viscosity fluid temperature and concentration, for natural convection flow, the
μtϕ macroscopic turbulent viscosity Boussinesq hypothesis reads, after renaming this density ρT,
ε dissipation rate of k
h

i
〈ε〉i intrinsic (fluid) average of ε ρT ≅ ρ 1−β T f −T ref −βC C−C ref ð9Þ
ρ bulk density of the mixture
ρℓ mass density of species ℓ
ϕ porosity where the subscript ref indicates a reference value and β and βC are the
thermal and salute expansion coefficients, respectively, defined by,
 
1 ∂ρ  1 ∂ρ 
Subscripts 
β¼−  ; βC ¼ −  : ð10Þ
β buoyancy ρ ∂T f  ρ ∂C 
p;C p;T f
ℓ chemical species
t turbulent
Here, it is interesting to point out that in ref. [16] the temperature
ϕ macroscopic
used in Eq. (9) was the same as that of the solid, T = Tf = Ts. Further,
C concentration
it is important to note that, as it is going to be shown below, after
volume averaging Eqs. (3) and (5), Tf is going to be related to Ts due to
the exchange of heat between the two phases across the interstitial
Superscripts
area. Also, Eq. (9) is an approximation of Eq. (10) and shows how
i intrinsic (fluid) average
density varies with the fluid temperature and mass concentration in
v volume (fluid + solid) average
the body force term of the momentum equation.Substituting now
k turbulent kinetic energy
Eq. (9) into Eq. (2), one has,
s,f solid, fluid
h

i
2
ρ∇
ðuuÞ ¼ −∇p þ μ∇ u þ ρ g 1−β T f −T ref −βC C−C ref : ð11Þ

within the solid: Thus, the momentum equation becomes after some rearrangement,

Energy‐solid shase ðporous matrixÞ0 ¼ ∇
ðks ∇T s Þ þ Ss : ð5Þ h

i

 2
ρ∇
ðuuÞ ¼ −ð∇pÞ þ μ∇ u−ρ g ðβ T f −T ref þ βC C−C ref ð12Þ
where u is the mass-averaged velocity of the mixture, u ¼ ∑ mℓ uℓ, uℓ
ℓ
is the velocity of species ℓ, mℓ is the mass fraction of component ℓ, where (∇p)⁎ = ∇p − ρg is a modified pressure gradient.
134 M.J.S. de Lemos / International Communications in Heat and Mass Transfer 52 (2014) 132–139

3. Time and volume average operators — The double Further, the local volume average theorem can be expressed as
decomposition concept [18–20]:

For the sake of completeness, although the information below has
 Z

v i 1
been given in detail in a number of articles and books, including [15] h∇φi ¼ ∇ ϕhφi þ ni φ dS
ΔV
for example, it is convenient to recall the definition of time average Ai

 Z
and volume average. v i 1
h∇
φi ¼ ∇
ϕhφi þ ni
φ dS ð18Þ
The time average of a general quantity φ is defined as: ΔV
Ai
 v Z
Z tþΔt ∂φ ∂
i
 1
1 ¼ ϕhφi − ni
ðui φÞdS
φ¼ φ dt ð13Þ ∂t ∂t ΔV
Δt t Ai

where the time interval Δt is small compared to the fluctuations of the

where ni is the unit vector normal to the fluid–solid interface, pointing
average value, φ, but large enough to capture turbulent fluctuations of
from the fluid towards the solid phase, Ai is the fluid–solid interface
φ. Time decomposition can then be written as,
area within the REV. It is important to emphasize that Ai should not be
confused with the surface area surrounding volume ΔV. In ref. [15] it
φ ¼ φ þ φ′ ð14Þ
is shown that for a rigid, homogeneous porous medium saturated
with an incompressible fluid, the following relationships apply:
with φ′ ¼ 0. Here, φ′ is the time fluctuation of φ around its average φ.
Further, the volume average of φ taken over a Representative
i
Elementary Volume (REV, Fig. 1) in a porous medium can be written hφii ¼ hφi
as [18–20]: ð19Þ
φ ¼ iφ
Z ′
i i′
v 1 φ ¼ hφi :
hφi ¼ φ dV: ð15Þ
ΔV
ΔV
Therefore, a general quantity φ can be expressed by either,
v
The value 〈φ〉 is defined for any point x surrounded by a REV of size
ΔV. This average is related to the intrinsic average for the fluid phase as: i′ i
φ ¼ hφii þ hφi þ i φ þ φ′ ð20Þ
D Ev D Ei
φf ¼ ϕ φf ð16Þ or

where ϕ = ΔVf / ΔV is the medium porosity and ΔVf is the volume i

i i
φ ¼ hφi þ i φ þ φ′ þ φ′ : ð21Þ
occupied by the fluid in a REV. Furthermore, one can write:

i i Expressions (20) and (21) encompass what is recalled in the litera-

φ ¼ hφi þ φ ð17Þ
ture as the “double decomposition” concept where iφ′ can be under-
stood as either the time fluctuation of the spatial deviation or the spatial
with 〈iφ〉i = 0. In Eq. (17), iφ is the spatial deviation of φ with respect to D Ei
the intrinsic average 〈φ〉i. deviation of the time fluctuation. Also, i φ′ ¼ i φ′ ¼ 0.

Fig. 1. Representative elementary volume (R.E.V.), intrinsic average; space and time fluctuations (see ref. [15]).
 M.J.S. de Lemos / International Communications in Heat and Mass Transfer 52 (2014) 132–139 135

4. Time averaged transport equations 5. Macroscopic equations for buoyancy free flows

In order to apply the time average operator to Eqs. (1), (3), (5), (8) 5.1. Mean continuity equation
and (12), one considers,
When the average operators (13)–(15) are simultaneously applied
′ ′ ′ ′
u ¼ u þ u ; T f ¼ T f þ T f ; Ts ¼ Ts þ Ts; C ¼ C þ C ; p ¼ p þ p :
0
ð22Þ over Eqs. (1)–(2), macroscopic equations for turbulent flow are obtained.
Volume integration is performed over a Representative Elementary
Volume (REV) shown in Fig. 1 resulting in,
Substituting Eq. (22) into the governing equations and considering
constant properties for both the fluid and the solid,
∇
uD ¼ 0 : ð33Þ

∇
u¼0 ð23Þ
where, uD ¼ ϕhuii and huii identifies the intrinsic (liquid) average of the

 time-averaged velocity vector u.
 2
ρ∇
ðuuÞ ¼ −ð∇pÞ þ μ∇ u þ ∇
−ρ u′ u′ ð24Þ For non-buoyant flows, macroscopic equations considering turbu-
h

i
− ρ g β T f −T ref þ βC C−C ref lence have been already derived in detail for momentum, heat, and
mass transfer [15] and for this reason their derivation need not to be




 repeated here. They are read as follows.
ρcp ∇
uT f ¼ ∇
k∇T f þ ∇
−ρ cp u′ T ′f ð25Þ
f
6. Mean momentum transport
    
0 ¼ ∇
ks ∇T s : ð26Þ uD uD
 D Ei 
i 2
ρ∇
¼ −∇ ϕhpi þ μ∇ uD þ ∇
−ρϕ u′ u′ ð34Þ
ϕ



 
μϕ c ϕρj u ju
∇
u C ¼ −∇
D∇C þ ∇
−u′ C ′ ð27Þ − uD þ F pffiffiffiffiD D
K K

For clear fluid, the use of the eddy-diffusivity concept for expressing D Ei
v 2 i
the stress–strain rate relationship for the Reynolds stress appearing in −ρϕ u′ u′ ¼ μ t ϕ 2 D − ϕρhki I ð35Þ
3
Eq. (24) gives,

v 1 n
i
 h

i
iT o
2 D ¼ ∇ ϕhui þ ∇ ϕhui
−ρu0 u0 ¼ μ t 2D− ρk I ð28Þ 2
3 D Ei ð36Þ
i
h i hki ¼ u′
u′ =2
where D ¼ ∇u þ ð∇uÞT =2 is the mean deformation tensor, k ¼ u0
u0 =
2 is the turbulent kinetic energy per unit mass, μ t is the turbulent viscosity 2
i
and I is the unity tensor. Similarly, for the turbulent heat flux on the r.h.s. hki
μ t φ ¼ ρ cμ f μ ð37Þ
of Eqs. (25) and (27) the eddy diffusivity concept reads, hεii

μt μ where cμ is a constant and fμ is damping function to be presented later.

−ρ cp u′ T ′f ¼ cp ∇T f ; −ρ u′ C ′ ¼ t ∇C ð29Þ
Prt Sct
7. Macroscopic turbulence field
where Prt and Sct are known as the turbulent Prandtl and Schmidt
The intrinsic turbulent kinetic energy per unit mass and its dissipa-
numbers, respectively.
tion rate are governed by the following equations,
Further, a transport equation for the turbulent kinetic energy is
obtained by multiplying first, by u′, the difference between the instanta-   μt 

∂
i


i i
D Ei
neous and the time-averaged momentum equations. Thus, applying ρ ϕhki þ ∇
uD hki ¼ ∇
μ þ ϕ ∇ ϕhki −ρ u′ u′ : ∇uD
∂t σk
further the time average operator to the resulting product, one has, ð38Þ
ϕhkii j uD j i
þ ck ρ pffiffiffiffi −ρϕhε i
K
 ′ 
p 2
ρ∇
ðukÞ ¼ −ρ∇
u′ þ q þ μ∇ k þ P þ GT þ GC − ρε ð30Þ
∂

  μ tφ 
  D Ei  i
hε i
ρ ρ
i
ϕhε i þ ∇
uD hε i
i
¼∇
μ þ ∇ ϕhε i
i
þ c1 −ρ u′ u′ : ∇uD
∂t σε hkii
i i2
ϕ hε i j uD j hε i
where P ¼ −ρu′ u′ : ∇u is the generation rate of k due to gradients of the þ c2 ck ρ pffiffiffiffi −c2 f 2 ρϕ
K hkii
mean velocity and
ð39Þ
μt
GT ¼ −ρ β g
u′ T ′f ¼ β g
∇T f ð31Þ where the c's are constants and f2 is a another damping function. Usual-
Prt
ly, two forms of the k–ε model are employed, namely the High Reynolds
(Launder and Spalding [21]) and Low Reynolds number (Abe et al. [22])
μt turbulence models. The constants and formulae used as damping func-
GC ¼ −ρ βC g
u′ C ′ ¼ βC g
∇C ð32Þ
Sct tions are showed in Table 1.

are the thermal and concentration generation rates of k due to tempera- 8. Two-energy equation model (2EEM)
ture and concentration fluctuations, respectively. Also, q = u′ ∙ u′/2
and, on the right of Eqs. (31) and (32), the models in Eq. (29) have Similarly, macroscopic energy equations are obtained for both fluid
been applied. and solid phases by applying time and volume average operators to
136 M.J.S. de Lemos / International Communications in Heat and Mass Transfer 52 (2014) 132–139

Table 1
Damping functions and constants for turbulence models.

High Reynolds turbulence model proposed Low Reynolds turbulence model proposed by
by Launder and Spalding [21] Abe et al. [22]
( "  2 #)
fμ 1.0 n h io2
Þ0:25 y ðk2 =νεÞ
1− exp − ðνε14ν 1 þ 2 5 0:75 exp − 200
ðk =νεÞ
( "  2 # )
f2 1.0 n h io2
0:25
ðk2 =νεÞ
1− exp − ðνε3:1ν
Þ y
1−0:3 exp − 6:5

σk 1.0 1.4
σε 1.33 1.3
c1 1.44 1.5
c2 1.92 1.9

2 3
Eqs. (3) and (5). As in the flow case, volume integration is performed Z
6 1 7
i
over a Representative Elementary Volume (REV), resulting in, ∇
4 ni k f T f dA5 ¼ K f ;s
∇ T s
ΔV
Ai
8 0 19
>
> >
> Local conduction : 2 3 : ð46Þ
>
>
> B C>>
> Z D Ei

 < B D Ei D Ei D
′
i ′ i E D E C =
B i i C 6 1 7
ρ cp ∇
ϕ Bhui T f þ i i
uT þ u Tf þ i ′i ′
u T C ¼ −∇
4 ni ks T s dA5 ¼ Ks; f
∇ Tf
f >
>
>
B
@ |fflfflfflffl{zfflfflfflffl} |fflfflfflfflfflfflffl{zfflfflfflfflfflfflffl} |fflfflfflfflfflffl{zfflfflfflfflfflffl} A>
C >
> ΔV
> >
>
: thermal disperson turbulent heat turbulent thermal > ; Ai

2 3 flux disperson
 D E Z Z
6 i 1 7 1 Finally, Eqs. (40) and (42) can be rewritten using the concept of
∇
4k f ∇ ϕ T f þ ni k f T f dA5 þ ni
k f ∇T f dA
ΔV ΔV overall effective conduction in the form,
A A
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
i
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
i
 D E 
conduction interfacial heat transfer
 i
D Ei  1
Z
ρ cp ∇
uD T f ¼ ∇
Keff ; f
∇ T f þ ni
k f ∇T f dA
ð40Þ f ∇V
Ai

ð47Þ
where the expansion,

* +i n o Z
D Ei

 D Ei 

D Ei D Ei i
0 ¼ ∇
Keff ;s
∇hT s i −
1
ni
ks ∇T s dA ð48Þ
u′ T ′f ¼ u′ i þ i u′ T ′f þ i T ′ ¼ u′ i T ′f þ i u′ i T ′ ΔV
Ai

ð41Þ
where

has been used in light of the double decomposition concept given by h i

Eqs. (19)–(21) (see ref. [15] for details). For the solid phase, one has, Keff ; f ¼ ϕ k f I þ K f ;s þ Kt þ Kdisp þ Kdisp;t ð49Þ

8 9
>
< h Z >
= Z

i i 1 1 Keff ;s ¼ ½ð1−ϕÞ ks  I þ Ks; f ð50Þ
0 ¼ ∇
ks ∇ ð1−φÞ T s − ni ks T s dA − ni
ks ∇T s dA :
>
: ΔV >
; ΔV
Ai Ai
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} interfacial heat transfer
conduction The turbulent heat flux and turbulent thermal dispersion terms, Kt
ð42Þ and Kdisp,t, are here modeled through the Eddy diffusivity concept as
[17]:

i
i
In Eqs. (40) and (42), T s and T f denote the intrinsic time-
 νt
average temperature of solid and fluid phases, respectively. These equa- Kt þ Kdisp;t ¼ ϕ ρ cp
φ
I ð51Þ
f Prt
tions are the macroscopic energy balances for the fluid and the porous φ

matrix (solid), respectively.

Also, in order to use Eqs. (40) and

i (42), the
underscored terms have where Prtϕ is the macroscopic turbulent Prandtl number for the fluid en-
i
to be modeled as a function of T s and T f . To accomplish this, a ergy equation.
gradient type diffusion model is used for all terms not involving the
interfacial heat transfer, in the form, 9. Interfacial heat transfer, hi

 
D Ei  D Ei In Eqs. (40) and (42) the heat transferred between the two phases
Turbulent heat flux : − ρ cp ϕ u′ i T ′f ¼ Kt
∇ T f ð43Þ can be modeled by means of a film coefficient hi such that,
f

  D Ei  D Ei  D Ei  Z Z
Thermal dispersion : − ρ cp ϕ i ui T ¼ Kdisp
∇ T f ð44Þ i 1 1
hi ai hT s i − T f ¼ ni
k f ∇T f dA ¼ ni
ks ∇T s dA ð52Þ
f ∇V ΔV
Ai Ai

  D Ei  D Ei
i ′i ′
Turbulent thermal dispersion : − ρ cp ϕ u T ¼ Kdisp;t
∇ T f
f
where ai = Ai / ΔV is the interfacial area per unit volume. In porous
ð45Þ media, the high values of ai make them attractive for transferring
 M.J.S. de Lemos / International Communications in Heat and Mass Transfer 52 (2014) 132–139 137

thermal energy via conduction through the solid followed by Coefficients Ddisp, Dt and Ddisp,t in Eq. (58) appear due to the non-
convection to a fluid stream. linearity of the convection term. They come from the modeling of the
 D E  following mechanisms:

 i D Ei  

i D Ei

ρ cp ∇
uD T f ¼ ∇
Keff ; f
∇ T f þ hi ai T s − T f ; D Ei D Ei
f
• Mass dispersion : − i ui C ¼ Ddisp
∇ C ð62Þ
ð53Þ

n   D Ei

i o
i D Ei

′ ′
0 ¼ ∇
Keff ;s
∇ T s −hi ai T s − T f : ð54Þ • Turbulent mass flux : − u′ i C ′ i ¼ −huii hC ii ¼ Dt
∇ C ð63Þ

Wakao et al. [23]. proposed a correlation for hi for closely packed bed D Ei D Ei
and compared results with their experimental data. This correlation • Turbulent mass dispersion; − i u′ i C ′ ¼ Ddisp;t
∇ C : ð64Þ
reads,

hi D 0:6 1=3 Here also mechanisms (63) and (64) are added up as [16];
¼ 2 þ 1:1ReD Pr : ð55Þ
kf
D Ei 1 μ t D Ei D Ei
 D Ei
i
Kuwahara et al. [24] also obtained the interfacial convective heat − u′ C ′ ¼ ϕ
∇ C ¼ hDt i ∇ C ¼ Dt þ Ddisp;t
∇ C : ð65Þ
ρ Sctϕ
transfer coefficient for laminar flow, as follows,
  11. Double-diffusion effects with a two-energy equation model
hi D 4ð1−ϕÞ 1 1=2 1=3
¼ 1þ þ ð1−ϕÞ ReD Pr ; valid for 0:2 b ϕ b 0:9;
kf ϕ 2
11.1. Mean flow
ð56Þ
Focusing now attention to buoyancy effects only, application of the
Eq. (56) is based on porosity dependency and is valid for packed
volume average procedure to the last term of (24) leads to,
beds of particle diameter D.
Following this same methodology, in which the porous medium is
D h

iEv
considered an infinite number of solid square rods, Saito and de ρ g β T f −T ref þ βC C−C ref
Lemos [25] proposed a correlation for obtaining the interfacial heat
Z h

i
transfer coefficient for turbulent flow as, ΔV f 1
¼ ρ g β T f −T ref þ βC C−C ref dV: ð66Þ
 
ΔV ΔV f
hi D ReD 0:8 1=3 ΔV f
4 Re 7
¼ 0:08 Pr ; for 1:0x10 b D b 2:0x10 ; valid for 0:2 b ϕ b 0:9;
kf ϕ ϕ
ð57Þ Expanding the left hand side of Eq. (66) in light of Eq. (17), the buoyancy
term becomes,
Table 2 shows three variant correlations for the fluid to solid heat
transfer coefficient hi and the specific surface area of the porous medium D h

iEv
ai, which appears in both energy equations. ρ g β T f −T ref þ βC C−C ref
"   #
10. Mass transport D Ei

i
¼ ρ g ϕ βϕ T f −T ref þ βC ϕ 〈 C 〉 −C ref
ð67Þ
 D E  D E
i i
i
i
∇
uD C ¼ ∇
Deff
∇ ϕ C ð58Þ þ ρ gβϕ i T þ ρ gβC ϕ i C
|fflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflffl}
¼0 ¼0

Deff ¼ Ddisp þ Ddiff þ Dt þ Ddisp;t ð59Þ

where the third and forth terms on the r.h.s. are null since 〈iφ〉i = 0. Here,
coefficients βϕ and βCϕ are the macroscopic thermal and salute expansion
i 1 μϕ coefficients, respectively. Assuming that gravity is constant over the REV,
Ddiff ¼ hDi I ¼ I ð60Þ
ρ Sc expressions for them based on Eq. (67) are given as,

1 μ tϕ D
Ev D
Ev
Dt þ Ddisp;t ¼ I ð61Þ ρ β T f −T ref ρ βC C−C ref
ρ Sctϕ
βφ ¼ D E  ; βC ¼ D E  : ð68Þ
i φ i
ρ ϕ T f −T ref ρ ϕ C −C ref
where Sctϕ is a macroscopic turbulent Schmidt number.

Table 2
Correlations for heat transfer coefficient and fluid-to-solid specific area ai.

Reference Correlation eqn. ai Flow regime

6ð1−ϕÞ
Wakao et al. [21]. hi D
¼2þ 1:1Re0:6 1=3 Laminar
kf D Pr (55) D

 4ð1−ϕÞ
Kuwahara et al. [24]. hi D
¼ 1 þ 4ð1−ϕ Þ
þ 12 ð1−ϕÞ1=2 ReD Pr1=3 (56) D Laminar
kf ϕ
Saito and de Lemos [25]
0:8 4ð1−ϕÞ
Turbulent
hi D
k
¼ 0:08 ReϕD Pr1=3 (57) D
f
138 M.J.S. de Lemos / International Communications in Heat and Mass Transfer 52 (2014) 132–139

Including Eq. (67) into Eq. (34), the macroscopic time-mean Navier– Further, expanding the r.h.s. of Eq. (71) in light of Eqs. (17) and
Stokes (NS) equation for an incompressible fluid with constant properties (19), one has
is given as,
D Ei
k
−ρ βC ϕ ϕ g
u′ C ′
 


uD uD i 2 i


ρ∇
¼ −∇ ϕhpi þ μ∇ uD þ ∇
−ρϕ 〈 u′ u′ 〉

〈 〉
k i
ϕ "  # ¼ −ρ βC ϕ ϕ g
u′ i þ i u′ Þð C ′ i þ i C ′
D Ei 

D 
′
i ′
i Ei Di ′ i ′ Ei D ′
i i ′ Ei Di ′ ′
i Ei
i
− ρ gϕ β ϕ T f −T ref þ βC ϕ 〈 C 〉 −C ref ð69Þ k
¼ −ρ βC ϕ ϕ g
u C þ u C þ u C þ u C

μϕ c ϕρju ju 0 1
− uD þ F pffiffiffiffiD D
K K B D Ei



C
′ ′
¼ −ρ βC ϕ ϕ g
B huii hC ii þ i u′ i C ′ þ u′ i i C ′ þ i u′ C ′ i C
k i i
@|fflfflfflfflffl
ffl{zfflfflfflfflfflffl} |fflfflfflfflfflffl{zfflfflfflfflfflffl} |fflfflfflfflfflfflffl{zfflfflfflfflfflfflffl} |fflfflfflfflfflfflffl{zfflfflfflfflfflfflffl} A
I
II ¼0 ¼0
where the superscript * on the pressure gradient that would appear
in Eq. (69) by the volume-average of Eq. (24), has been dropped. ð72Þ
As pointed out by [16], it is interesting to comment on role of
coefficients βϕ and βC ϕ on the overall mixture density value. Here, The last two terms on the right of Eq. (72) are null since 〈iC′〉i = 0
only fluids that became less dense with increasing temperature and 〈iu′〉i = 0. In addition, the following physical significance can be
are considered. However, two situations might occur when increas- inferred to the two remaining terms, which were fully commented
D Ei upon in ref. [16] and for that they will be just listed:
ing C , namely the mixture might become less dense with the ad-
dition of a lighter solute, or else, a denser fluid may result by mixing I. Generation/destruction rate of turbulence energy due to macroscopic
a heavier component to it. Implications of that on the stability of concentration fluctuations,
the entire fluid system were discussed in ref. [16] where more II. Generation/destruction rate due to turbulent concentration
details can be found. dispersion.
A model for Eq. (72) is still needed in order to solve an equation for
11.2. Turbulent field 〈k〉i, which is a necessary information when computing μ t ϕ using
Eq. (37). Consequently, terms I and II above have to be modeled as a
As mentioned, this work extends and combines earlier develop- D Ei
ments for turbulent double-diffusion using the thermal equilibrium function of average concentration, C . To accomplish this, a gradient
model [16] with the hypothesis of thermal non-equilibrium [17]. For type diffusion model is used, in the form,
clear fluid, the buoyancy contribution to the k equation is given by
Eqs. (31) and (32). • Buoyancy generation of 〈k〉i due to turbulent salute fluctuations:
Volume averaging Eq. (31) in reference [16] has resulted in the term,  D Ei 
k ′ ′ k
−ρ βC ϕ ϕ g
huii hC ii ¼ ρ βC ϕ ϕ g
Dt
∇ C : ð73Þ
D Ev D Ei
v i 0 ′ k 0 ′
hGT i ¼ Gβ ¼ −ρβ g
u T f ¼ −ρβϕ ϕ g
uT f
μ tϕ D Ei
k • Buoyancy generation of 〈k〉i due to turbulent salute dispersion:
¼ βϕ ϕ g
∇ Tf ð70Þ
Prt ϕ
D Ei  D Ei 
k i ′i ′ k
−ρ βC ϕ ϕ g
u C ¼ ρ βC ϕ ϕ g
Ddisp;t
∇ C : ð74Þ
where the model in Eq. in (29) has been applied. Eq. (70) represents an
additional macroscopic generation/destruction rate of 〈k〉i due to
temperature variation in porous media, where βkϕ is a macroscopic
D Ev The buoyancy concentration coefficients seen above, namely Dt and
β u0 T ′ f Ddisp,t, were used before in Eqs. (63) and (64), respectively. Note that the
coefficient given by β kϕ ¼ D Ei . In reference [16], coefficients β
terms given by Eqs. (73) and (74) arise only if the flow is turbulent and
ϕ u0 T ′ f
if buoyancy is of importance.
(Eq. (10)), βϕ (Eq. (68)) and βkϕ (Eq. (70)) were all assumed to be Using then Eq. (65) the macroscopic buoyancy generation of k due to
equal, for simplicity. Also, in ref. [16] the temperature used in Eq. (70) concentration fluctuations can be modeled as,
was the same regardless of the phase. Here, however, it is the gradient
D Ei D Ei
of the intrinsic fluid temperature T f that is considered to promote i k
GβC ¼ −ρ βC ϕ ϕ g
u0 C ′
the driving mechanism to generate/destroy turbulence.
 μt ð75Þ
k i k i
In order to add the effect of concentration variation within the fluid, ¼ ρ βC ϕ ϕ g
½ Dt þ Ddisp;t
∇ 〈 C 〉  ¼ βC ϕ ϕ ϕ g
∇ 〈 C 〉
Sctϕ
one applies the volume average operator to Eq. (32) such that,

where μ t φ , Sctϕ and the two coefficients Dt and Ddisp,t have been defined
D Ev D Ei
v
hGC i ¼
i
GβC ¼ −ρβ C g
u C 0 ′
¼
k
−ρβC ϕ ϕg
u C 0 ′
ð71Þ before. D Ei
i
Final transport equations for hki ¼ u′
u′ =2 and hεii ¼ μ
D Ei
∇u′ :ð∇u′ ÞT =ρ, in their so-called High Reynolds number form can
where the coefficient β kC φ , for a constant value of g within the REV, is

v now include the buoyancy generation terms due to temperature and
β u0 C ′
given by βkC ϕ ¼ C 0 ′
i , which, in turn, is not necessarily equal to βC ϕ concentration fluctuations as,
ϕ uC

given by Eq. (68). However, for the sake of simplicity and in the
  μt 

i i i i i i i
absence of better information, one can use a similar argument as in ρ∇
uD hki ¼ ∇
μ þ ϕ ∇ ϕhki þ P þ G þ Gβ þ GβC −ρϕhε i
σk
reference [16] and make use of the assumption βC ¼ βC ϕ ¼ βkC ϕ . ð76Þ
 M.J.S. de Lemos / International Communications in Heat and Mass Transfer 52 (2014) 132–139 139

  μt 
 References
i i
ρ∇
uD hεi ¼ ∇
μ þ ϕ ∇ ϕhε i
σε [1] P. Nithiarasu, T. Sundararajan, K.N. Seetharamu, Double-diffusive natural convection
hεi h i
 i in a fluid saturated porous cavity with a freely convecting wall, Int. Commun. Heat
i
i i i i
þ i c1 P þ c2 G þ c1 Gβ þ GβC −c2 ρϕhεi ð77Þ Mass Transfer 24 (8) (1997) 1121.
hki [2] M. Mamou, P. Vasseur, E. Bilgen, Multiple solutions for double-diffusive convection
in a vertical porous enclosure, Int. J. Heat Mass Transf. 38 (10) (1995) 1787.
[3] B. Goyeau, J.-P. Songbe, D. Gobin, Numerical study of double-diffusive natural
convection in a porous cavity using the Darcy–Brinkman formulation, Int. J. Heat
where, σk = 1, σε = 1.3, c1 = 1.44, c2 = 1.92, cμ = 0.09 and ck = 0.28 Mass Transf. 39 (7) (1996) 1363.
are non-dimensional constants (see ref. [15]). The production terms have [4] A.A. Mohamad, R. Bennacer, Double diffusion natural convection in an enclosure
filled with saturated porous medium subjected to cross gradients; stably stratified
the following physical significance: fluid, Int. J. Heat Mass Transf. 45 (18) (2002) 3725.
D Ei [5] R. Bennacer, H. Beji, A.A. Mohamad, Double diffusive convection in a vertical enclo-
1. P i ¼ −ρ u′ u′ : ∇uD is the production rate of 〈k〉i due to gradients sure inserted with two saturated porous layers confining a fluid layer, Int. J. Therm.
Sci. 42 (2) (2003) 141.
of u D̅ ; [6] M. Mamou, M. Hasnaoui, A. Amahmida, P. Vasseur, Stability analysis of double diffu-
i
2. Gi ¼ ck ρ ϕ hkpi ffiffiKjffiuD j is the generation rate of the intrinsic average of 〈k〉i sive convection in a vertical brinkman porous enclosure, Int. Commun. Heat Mass
Transfer 25 (4) (1998) 491.
due to the action of the porous matrix; [7] R. Bennacer, A. Tobbal, H. Beji, P. Vasseur, Double diffusive convection in a vertical
μt
i enclosure filled with anisotropic porous media, Int. J. Therm. Sci. 40 (1) (2001) 30.
3. Giβ ¼ βkϕ ϕ Prtϕ g
∇ T f is the generation of 〈k〉i due to mean temper- [8] F. Moukalled, M. Darwish, Double diffusive natural convection in a porous rhombic
ϕ
annulus, Numer. Heat Transfer, Part A 64 (5) (2013) 378–399.
ature variation within
μt E fluid, and
D the i [9] A. Khadiri, A. Amahmid, M. Hasnaoui, A. Rtibi, I Soret effect on double-diffusive
4. GiβC ¼ βkC ϕ ϕ Sctϕ g
∇ C is the generation of 〈k〉i due to concentra- convection in a square porous cavity heated and salted, from below, Numer. Heat
ϕ
Transfer, Part A 57 (1 I) (2010) 848–868.
tion gradients. [10] C.T. Hsu, P. Cheng, Thermal dispersion in a porous medium, Int. J. Heat Mass Transfer
33 (1990) 1587.
[11] J. Bear, Dynamics of Fluids in Porous Media, Dover, New York, 1972.
[12] J. Bear, Y. Bachmat, A Generalized Theory on Hydrodynamic Dispersion in Porous
12. Conclusions Media, I.A.S.H. Symp. Artificial Recharge and Management of Aquifers, Haifa,
Israel, P.N., 72, I.A.S.H., 1967. 7–16.
In this work, equations were derived for turbulent double-diffusive [13] S. Whitaker, Equations of motion in porous media, Chem. Eng. Sci. 21 (1966) 291.
[14] S. Whitaker, Diffusion and dispersion in porous media, J. Amer. Inst. Chem. Eng 3
natural convection in porous media. Derivations were carried out (13) (1967) 420.
under the light of the double decomposition concept [15]. Extra terms [15] M.J.S. de Lemos, Turbulence in Porous Media: Modeling and Applications, 2nd edi-

tion Elsevier, Amsterdam, 2012.
appearing in the equations needed to be modeled in terms of uD , T f
D E [16] M.J.S. de Lemos, L.A. Tofaneli, Modeling of double-diffusive turbulent natural con-
and C . Here, two different models were combined in order to broaden vection in porous media, Int. J. Heat Mass Transfer 47 (19–20) (2004) 4221–4231.
[17] M.B. Saito, M.J.S. de Lemos, A macroscopic two-energy equation model for turbulent
the ability to analyze more complex flow systems. The first model dealt flow and heat transfer in highly porous media, Int. J. Heat Mass Transfer 53 (11–12)
with characterizing turbulent double-diffusive mechanism but was lim- (2010) 2424–2433.
[18] J.C. Slattery, Flow of viscoelastic fluids through porous media, A.I.Ch.E.J. 13 (1967)
ited to situations were the so-called thermal equilibrium between
1066.
phases applied [16]. In addition, the second description of turbulent [19] S. Whitaker, Advances in theory of fluid motion in porous media, Ind. Eng. Chem. 61
flow in porous media made no consideration about buoyancy effects (1969) 14.
but was able to handle situations where the difference in both the [20] W.G. Gray, P.C.Y. Lee, On the theorems for local volume averaging of multiphase
system, Int. J. Multiphase Flow 3 (1977) 333.
fluid and the solid material was considerable [17]. By combining the [21] B.E. Launder, D.B. Spalding, The numerical computation of turbulent flows, Comput.
two models in one single mathematical characterization, the work Methods Appl. Mech. Eng. 3 (1974) 269–289.
herein aims at extending the tool described in detail in ref. [15] to [22] K. Abe, Y. Nagano, T. Kondoh, An improve k–ε model for prediction of turbulent
flows with separation and reattachment, Trans. JSME 58 (1992) 3003–3010.
solve an ever-broader range of practical problems in engineering. [23] N. Wakao, S. Kaguei, T. Funazkri, Effect of fluid dispersion coefficients on
particle-to-fluid heat transfer coefficients in packed bed, Chem. Eng. Sci. 34
(1979) 325–336.
Acknowledgments [24] F. Kuwahara, M. Shirota, A. Nakayama, A numerical study of interfacial convective
heat transfer coefficient in two-energy equation model for convection in porous
The author is indebted to the research funding agencies in Brazil, media, Int. J. Heat Mass Transfer 44 (2001) 1153–1159.
[25] M.B. Saito, M.J.S. de Lemos, A correlation for interfacial heat transfer coefficient
namely CNPq and CAPES, for their continuous support for more than for turbulent flow over an array of square rods, J. Heat Transf. 128 (2006)
two decades. 444–452.