Professional Documents
Culture Documents
∗guqiangshun@pi.ac.ae
Int. J. Comput. Methods Downloaded from www.worldscientific.com
†yfatt@pi.ac.ae
Li Hongying
Institute of High Performance Computing
A*STAR, 1 Fusionopolis Way, #16-16 Connexis
138632 Singapore, Singapore
lih@ihpc.a-star.edu.sg
Che Zhizhao
State Key Laboratory of Engines, Tianjin University
92 Weijin Rd, Nankai Qu, Tianjin Shi
Tianjin 300072, P. R. China
chezhizhao@tju.edu.cn
This paper presents a model for two-phase nanofluid-fluid flow and heat transfer. The
nonuniform nanoparticles are transported using Buongiorno model by convection, Brow-
nian diffusion and thermophoresis. This is the first attempt to employ Buongiorno model
for two-phase nanofluid-fluid flow. The moving interface between the nanofluid and the
immiscible fluid is captured using the level-set method. The model is first verified and
then demonstrated for coupled flow and heat transfer in (1) a water–alumina nanofluid-
filled cavity with a rising silicone oil drop and (2) stratified flow of water–alumina
nanofluid, pure water and silicone oil in a channel.
1. Introduction
With modern nanotechnology, highly uniform particles with an average size of
50 nm can be produced efficiently. These nanoparticles find very diverse engineering
† Corresponding author.
1850072-1
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
G. Qiangshun et al.
flow and heat transfer [Alloui et al. (2011); Chamkha and Abu-Nada (2012); Bachok
Int. J. Comput. Methods Downloaded from www.worldscientific.com
et al. (2012); Garoosi et al. (2013); Cianfrini et al. (2014); Turkyilmazoglu (2014)],
the assumption of having uniform concentration of nanoparticles in the base fluid
is frequently made. However, due to various transport processes, nonuniform con-
centration of nanoparticles in the base fluid can in fact arise due to various particle
transport mechanisms [Buongiorno (2005)]. As a result, the local thermodynamic
properties of nanofluid vary both spatially and temporally. This effectively alters
the transport processes in particular heat transfer in a fully coupled manner. Model-
ing of single-phase nanofluid flow with local variation of nanoparticle concentration
accounted for has been conducted using, among others, the Eulerian model (in which
the base fluid and the nanoparticles are assumed as separate continuous phase with
momentum and energy transfers) [Akgül and Pakdemirli (2016)] and Buongiorno
model (in which the nanofluid is assumed as a continuous phase with the dominant
Brownian and thermophoretic effects) [Kuznetsov and Nield (2010); Yang et al.
(2013); Malvandi and Ganji (2014); Corcione et al. (2015); Garoosi et al. (2015);
Elshehabey and Ahmed (2015); Mohyud-Din et al. (2017)].
Frequently in various applications, two-phase flow, where one phase is a
nanofluid and the remaining phase is an immiscible fluid, is encountered, for exam-
ple, nanofluid flow boiling in enhanced heat transfer [Lee and Mudawar (2007);
Cheng and Liu (2013); Duursma et al. (2015)] and drag reduction in slug two-phase
flow of air and water (nanofluid) [Pouranfard et al. (2015)] and fluid properties alter-
ation in enhanced oil recovery [Ogolo et al. (2012)]. However, numerical/analytical
modeling of these two-phase flows involving a nanofluid is very limited in exist-
ing literatures given the complex nature of coupled mass, momentum and energy
transports of the two phases at the evolving interface. Furthermore, the unknown
interface is required to be either tracked or captured.
Analytical modeling of fully developed stratified two-phase nanofluid-fluid flow
was considered [Van Gorder et al. (2012); Van Gorder (2013); Farooq and Lin
(2014)]. The nanoparticles are transported by both Brownian and thermophoretic
diffusions and therefore not uniform within the nanofluid. Steady-state developing
stratified two-phase nanofluid-fluid flow was modeled numerically in Abbasi and
1850072-2
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
coupled momentum and energy exchanges between phases accounted for. Thus, the
nanoparticles are transported by various mechanisms and therefore generally not
uniform in the nanofluid. Both VOF and CLSVOF models are named according to
the approach employed in capturing the interface. In these two models, the nanopar-
ticles are assumed uniformly distributed in the nanofluid. Therefore, these are reg-
ular two-phase flow models with modified uniform properties for the nanofluid.
It was highlighted in Wang et al. [2015] that, of the three models, the Eulerian-
Eulerian model is more appropriate for prediction of flow and heat transfer char-
acteristics as the model accounts for the heterogeneity of nanoparticle distribution
within the nanofluid. However, it should be noted that prediction of regular two-
phase flow models, e.g., VOF and CLSVOF, can be improved by incorporating
essential nanoparticle transport mechanisms leading to nonuniform nanoparticle
distribution within the nanofluid. This is attempted in this paper.
Therefore, this paper presents a two-phase flow model involving a nanofluid
interacting dynamically with another immiscible fluid in the presence of heat trans-
fer [Yap and Li (2015)]. The nanoparticles are transported by various mechanisms
using the model of Buongiorno [2005], leading to a nonuniform concentration and
therefore variable nanofluid properties. To the best knowledge of the authors, this
is the first attempt to employ Buongiorno model for two-phase nanofluid-fluid flow.
The transports of mass, momentum, nanoparticles and energy are fully coupled
to the interface between the two fluids. The interface between the two fluids is
captured using a level-set method. The model is applicable for low nanoparticle
concentration where there is negligible particle–particle interaction. Fortunately,
this is usually the case with nanofluid requiring generally low nanoparticle fraction
of within a few percent. Absorption (or desorption) of nanoparticles onto (out of)
interface is not considered. The governing equations are then solved using a finite
volume framework. The model is first verified by comparing to existing results of
a single-phase nanofluid flow with heat transfer. Finally, application of the model
to the case of a rising silicone oil drop in a water–alumina nanofluid-filled cavity is
demonstrated.
1850072-3
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
G. Qiangshun et al.
(2015)] and flow boiling [Fang et al. (2015)]. In these examples, the nanoparticle
Int. J. Comput. Methods Downloaded from www.worldscientific.com
2. Problem Description
Figure 1 shows a domain consisting of two immiscible fluids, i.e., the −ve fluid and
the +ve nanofluid with each occupying the Ω− and the Ω+ regions, respectively.
The +ve nanofluid is a dilute mixture of nanoparticles and a base fluid. Due to the
presence of nanoparticles, it generally has better heat transfer properties compared
with those of the base fluid [Yu et al. (2008); Taylor et al. (2013)]. The Ω− and the
Ω+ regions are separated by an interface Γ, which can be highly irregular and evolves
over time. The level-set function [Osher and Sethian (1988)] is used to represent
the interface mathematically as
−d, if x ∈ Ω−
φ = 0, if x ∈ Γ (1)
+d, if x ∈ Ω (nanofluid),
+
where d is the shortest distance from the interface. The two fluids interact dynam-
ically via appropriate interfacial conditions at the interface. Transport processes
Fig. 1. Domain of interest with two fluids: the –ve fluid and the +ve nanofluid.
1850072-4
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
involving mass, momentum and energy between these fluids are therefore intimately
coupled.
3. Mathematical Formulation
3.1. Governing transport equations
A one-fluid formulation is employed in this paper. Basically, the domain of interest
is visualized as to consist of a single special fluid. This fluid is special in the sense
that its properties at a given time and location are set to the properties of either
the −ve fluid or the +ve nanofluid depending on whichever occupies that particular
location at that particular time. Any property of the special fluid can be determined
conveniently from the level-set function, as will be presented in Sec. 3.2. Within a
by UNIVERSITY OF NEW ENGLAND on 01/07/18. For personal use only.
1850072-5
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
G. Qiangshun et al.
both DB and DT are zero. Nanoparticle absorption onto and desorption from the
interface are not considered.
3.2. Properties
The level-set function provides a convenient means to evaluate the properties of the
special fluid. For this purpose, a smoothed Heaviside function H(φ) over a band of
width 2ε is defined:
0,
if φ < −ε
φ+ε 1 πφ
H(φ) = + sin , if |φ| ≤ ε (7)
2ε 2π ε
1, if φ > +ε.
by UNIVERSITY OF NEW ENGLAND on 01/07/18. For personal use only.
Int. J. Comput. Methods Downloaded from www.worldscientific.com
Then, any property of interest for the special fluid α can be expressed as either an
arithmetic mean of the form
α(φ) = (1 − H)α− + Hα+ (8a)
or a harmonic mean of the form
1 1−H H
= + , (8b)
α(φ) α− α+
where the subscripts − and + refer to quantity associated with the −ve fluid and
the +ve nanofluid, respectively. The density and specific heat capacity are evaluated
using the arithmetic mean. The viscosity, thermal conductivity, Brownian diffusion
coefficient and thermophoretic diffusion coefficient are evaluated using the harmonic
mean.
For the +ve nanofluid, the density, viscosity [Khanafer and Vafai (2011)], spe-
cific heat capacity and thermal conductivity [Khanafer and Vafai (2011)] are given,
respectively, by
ρ+ = ϕρp + (1 − ϕ)ρbf , (9a)
µbf
µ+ = , (9b)
1 − 34.87(dp /dbf )−0.3 ϕ1.03
(ρc)+ = ϕ(ρc)p + (1 − ϕ)(ρc)bf , (9c)
10 0.03
T kp
k+ = kbf 1 + 4.4Re0.4 0.66
p Prbf ϕ0.66 , (9d)
Tr,bf kbf
where the subscripts p and bf refer to quantity associated with the nanoparticles
and the base fluid, respectively.
In the evaluation of nanofluid viscosity via Eq. (9b), dp and dbf are, respectively,
the nanoparticle diameter and the equivalent diameter of the base fluid molecule.
In particular, dbf can be evaluated as
1/3
6M
dbf = 0.1 , (9e)
N πρ0,bf
1850072-6
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
where M , N and ρ0,bf are, respectively, the molar mass of the base fluid, the Avo-
gadro number and the mass density of the base fluid at T0 = 293 K. To calculate
the nanofluid thermal conductivity using Eq. (9d), Tr,bf is the freezing point of the
base fluid. The nanoparticle Reynolds number Rep and Prandtl number of the base
fluid Prbf are given by
2ρbf kB T
Rep = , (9f)
πµ2bf dp
cbf µbf
Prbf = , (9g)
kbf
where kB = 1.3806 × 10−23 J/K is the Boltzmann’s constant. Do note that there
are many theoretical relations and empirical correlations developed for viscosity
by UNIVERSITY OF NEW ENGLAND on 01/07/18. For personal use only.
and thermal conductivity, see, for example, discussed in Khanafer and Vafai [2011]
Int. J. Comput. Methods Downloaded from www.worldscientific.com
and Corcione [2011]. If desired, these can also be employed in place of Eqs. (9b)
and (9d).
Finally, the Brownian diffusion coefficient DB and thermal diffusion coefficient
DT of the nanofluid can be calculated, respectively, as
kB T
DB = , (9h)
3πµbf dp
kbf µbf ϕ
DT = 0.26 . (9i)
2kbf + kp ρbf
1850072-7
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
G. Qiangshun et al.
reference fluid.
Int. J. Comput. Methods Downloaded from www.worldscientific.com
For stratified flows, if required, a local mass correction approach [Yap et al.
(2005)] is employed in place of global mass correction to alleviate mass loss/gain
problem of the level-set method. Basically, the flowrate at any given cross-section
instead of mass as in the global mass correction approach is “conserved” by intro-
ducing a correction to the level-set function.
1850072-8
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
end of a given time step, the nanoparticles diffused into the −ve fluid CV (i, j) are
distributed back into the neighboring +ve nanofluid CVs of (i − 1, j + 1), (i, j + 1),
(i + 1, j + 1) and (i + 1, j). This can be achieved by setting the nanofluid volume
fraction in these neighboring CVs as:
ϕi,j ∆Vi,j
ϕin,jn = ϕin,jn + fin,jn , (13a)
∆Vin,jn
where
Sin,jn
fin,jn = in=i+1,jn=j+1 , (13b)
in=i−1,jn=j−1 Sin,jn
1, if φi,j = −ve, φin,jn = +ve
Sin,jn = (13c)
by UNIVERSITY OF NEW ENGLAND on 01/07/18. For personal use only.
0, otherwise.
Int. J. Comput. Methods Downloaded from www.worldscientific.com
To complete the procedure, the nanoparticle volume fraction φ for CV (i, j) is set
to φ = 0.
4. Solution Procedure
4.1. Numerical method
The transport equations Eqs. (2)–(5) can be written in the form of a generic tran-
sient convection–diffusion equation:
∂(ρ̃Φ)
+ ∇ · (ρ̃uΦ) = ∇ · (Γ̃∇Φ) + S, (14)
∂t
where ρ̃, Γ̃ and S are the appropriate “density”, “diffusion coefficient” and source
term. The source term contains all other terms that cannot be fitted neatly into the
convection or diffusion terms. This generic equation is solved using a finite volume
method [Patankar (1980); Versteeg and Malalasekera (2007)] on a staggered mesh
arrangement. Scalar variables are defined at the node of the CVs. The staggered
velocity components are defined at the surface of the CVs. The convection term is
modeled using a second-order upwind scheme with SUPERBEE limiter [Roe (1983)]
implemented via a deferred correction manner. A fully implicit scheme is used for
time integration. The velocity–pressure coupling of the Navier–Stokes equations is
handled with the SIMPLER algorithm.
To capture the evolving interface accurately, the level-set method requires higher
order numerical schemes. The evolution of the level-set function (Eq. (10)) and
its redistancing (Eq. (11)) are spatially discretized with WENO5 [Jiang and Peng
(2000)] and integrated using TVD-RK2 [Shu and Osher (1988)]. These schemes are
computationally intensive. To reduce the computational effort, the level-set method
is implemented in a narrow-band procedure [Peng et al. (1999)], in which the level-
set function is solved only within a band of certain thickness from the interface.
This reduces one order of computational effort.
1850072-9
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
G. Qiangshun et al.
µ− = 0.081 Pa s, (15b)
1850072-10
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
(a) ρ+ (b) µ+
Int. J. Comput. Methods Downloaded from www.worldscientific.com
(e) DB (f) DT
1850072-11
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
G. Qiangshun et al.
Alumina nanoparticle:
dp = 25 nm, (17a)
(17d)
In a pure water-silicon oil system, interfacial tension depends weakly on temper-
ature [El-Hamouz (2007); Peters and Arabali (2013)]. In El-Hamouz [2007], the
measurement shows an interfacial tension decrease of merely 2% when temperature
varies from 25◦ C to 80◦ C, i.e., ≈0.04%/◦C. In Peters and Peters and Arabali [2013],
it was found to interestingly slightly increase 4% ovet T from 18◦ C to 35◦ C, i.e.,
≈0.2%/◦ C. In a separate interfacial tension measurement for a different system,
i.e., water-titanium oxide nanofluid and mineral oil [Sohel Murshed et al. (2008)],
a decrease of roughly 6% is found over the range of 30◦ C to 45◦ C, i.e., ≈ 0.4%/◦ C.
For problems considered in this paper, the temperature ranges from 32◦ C to 42◦ C.
Variation of interfacial tension due to temperature is therefore expected to be small,
in the range of a few percent. In the absence of complete interfacial tension data
for water–alumina nanofluid and silicone oil system at different temperatures and
nanoparticle volume fraction, a constant value of σ = 0.03 N/m is used. With this,
interfacial tension gradient-induced Marangoni effect is not considered. Neverthe-
less, demonstration including Marangoni effect will also be presented. For the tem-
perature range of 32◦ C (305 K) to 42◦ C (315 K), interfacial tension is assumed to
decrease 5% linearly as
σ = σo − 0.005(T − 305)σo , (18)
where σo = 0.03 N/m and T in K. To account for the Marangoni effect due
to this variation of interfacial tension, an additional source term of (n̂ ×
(dσ/dT )∇T ) × n̂δ(φ) is included into the RHS of the momentum equation (Eq. (3)).
5.2. Verification
To the best knowledge of the authors, the current work is the first attempt to
model two-phase nanofluid-fluid flow with Buongiorno model for nanofluid. There-
fore, there is no existing result of which direct verification is possible. The cur-
rent model consists of two essential components: (1) level-set approach to capture
1850072-12
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
the interface of a two-phase flow and (2) nanofluid flow with Buongiorno model.
The approach taken in the current paper is to verify these two components sep-
arately. After both components (1) and (2) are verified, the approach is deemed
indirectly verified for two-phase nanofluid-fluid flow. Component (1) was verified in
the authors’ earlier work [Yap and Chai (2012); Li et al. (2015)] and therefore will
not be repeated here. To verify component (2), the limiting case of a single-phase
nanofluid flow with Buongiorno model is attempted. Specifically, verification is con-
ducted for natural convection of a single-phase water–alumina nanofluid in a square
cavity with prescribed temperature gradient (Fig. 4). This problem is selected as
there exists excellent results in Corcione et al. [2013] to be compared with. The left
and right walls are maintained at a temperature of, respectively, Th and Tc . Both
the top and bottom walls are insulated. The flow is essentially buoyancy-driven.
It should be noted that as there is only one phase in the domain, the interface
capturing procedure of the level-set method is temporarily disabled.
Solutions for two cases with different uniform initial nanoparticle volume frac-
tions, i.e., ϕo = 0 and ϕo = 0.04, are obtained. Of course, the case of ϕo = 0 cor-
responds to the situation without alumina nanoparticles, i.e., pure water. Figure 5
shows the present steady-state solutions: streamlines for the case of ϕo = 0.04
(Fig. 5(a)), vertical velocity component v along the horizontal midplane of the
cavity (Fig. 5(b)), dimensionless temperature θ = TTh−T c
−Tc for the case of ϕo = 0.04
(Fig. 5(c)) and temperature T along the horizontal midplane of the cavity (Fig. 5(d))
superimposed on the results of Corcione et al. [2013]. The present solutions, for both
cases of ϕo = 0 and ϕo = 0.04, agree well with those of Corcione et al. [2013]. The
effect of alumina nanoparticles can be carefully observed in Figs. 5(b) and 5(d).
This exercise verifies all components of the present model except the interface
capturing procedure of the level-set method. The implemented interface capturing
procedure of the level-set method has been verified in the authors’ earlier work [Yap
and Chai (2012); Li et al. (2015)] and therefore will not be repeated here.
1850072-13
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
G. Qiangshun et al.
by UNIVERSITY OF NEW ENGLAND on 01/07/18. For personal use only.
1850072-14
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
nanoparticles will not be uniformly distributed and therefore justifies the use of
Eq. (5) in determining the temporal and spatial distribution of nanoparticles. Of
course, heat is transferred simultaneously from the hot nanofluid and silicone oil to
pure water.
Given the symmetry of the problem, computations are made for only left half
of the domain. The following initial and boundary conditions apply:
Initial conditions:
u = 0, in whole domain, (19a)
Th , in nanofluid and silicone oil
T = (19b)
Tc , in pure water,
ϕo , in nanofluid
ϕ= (19c)
0, otherwise.
Boundary conditions:
u = 0, ∂v = 0, at symmetric plane
∂x (19d)
u = 0, at walls,
1850072-15
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
G. Qiangshun et al.
these plots are the velocity, temperature and concentration fields at t = 0, 0.09375,
0.1875, 0.2813, 0.375 and 0.45 s. Note that in order to avoid overcrowding the figure,
only one in every eight velocity vectors is plotted. The first plot in Fig. 7 shows
the grid-independent test conducted. The superimposed solutions were obtained on
two consecutively refined meshes of 100 × 400 CVs with ∆t = 1.875 × 10−4 s and
150 × 600 CVs with ∆t = 9.375 × 10−5 s. A mesh of 100 × 400 CVs with ∆t =
1.875 × 10−4 s is sufficient to capture all the essential features of the solution. This
is also verified for the cases of ϕo = 0.02 and 0.04. Therefore, the same mesh size is
used in these cases.
The effect of variable interfacial tension (Eq. (18))-induced Marangoni effect is
shown in Fig. 8. The interfacial profile and temperature field are almost exactly
identical to those of a constant interfacial tension, except small difference in
by UNIVERSITY OF NEW ENGLAND on 01/07/18. For personal use only.
1850072-16
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
1850072-17
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
G. Qiangshun et al.
by UNIVERSITY OF NEW ENGLAND on 01/07/18. For personal use only.
Int. J. Comput. Methods Downloaded from www.worldscientific.com
Fig. 8. Solutions for constant and variable σ at t = 0.1875, 0.2813, 0.375 and 0.45 s.
fields reveals the close association of energy and nanoparticle transport to these
flow structures. In fact, these flow structures drive dominantly both energy and
nanoparticle transport.
The very sharp features in temperature and nanoparticle volume fraction fields
suggest, at least for the time and length scales considered, a minimal role of other
transport mechanisms, i.e., diffusion in energy transport and Brownian diffusion
and thermophoresis in nanoparticle transport. This is however not unexpected given
the very small coefficients associated with these transport mechanisms, for example,
both DB and DT are in the order of 10−11 m2 /s.
Generally, for these cases, from the perspective of modification in thermal con-
ductivity and heat capacity, the presence of nanoparticles does not help much in heat
transfer enhancement from the hotter lower region to the upper cooler region. The
1850072-18
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
Fig. 9. Solutions for ϕo = 0.02 at t = 0, 0.09375, 0.1875, 0.2813, 0.375 and 0.45 s.
1850072-19
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
G. Qiangshun et al.
by UNIVERSITY OF NEW ENGLAND on 01/07/18. For personal use only.
Int. J. Comput. Methods Downloaded from www.worldscientific.com
Fig. 10. Solutions for ϕo = 0.04 at t = 0, 0.09375, 0.1875, 0.2813, 0.375 and 0.45 s.
Shown in Fig. 11 is the interesting case of replacing the pure water in the upper
region with nanofluid. Now the enclosure contains only nanofluid encapsulating the
silicone oil drop. The volume fraction of nanoparticle is uniform with ϕo = 0.02.
Both the flow and temperature fields have very similar characteristics to that of the
case with pure water encapsulating a silicone oil drop (ϕo = 0, Fig. 7). The vortical
1850072-20
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
Fig. 11. Solutions for ϕo = 0.02 (nanofluid only with silicone oil drop) at t = 0, 0.09375, 0.1875,
0.2813, 0.375 and 0.45 s.
structure trailing the rising silicone drop is larger though. The drop rises faster than
any cases considered as now the drop is always surrounded by a denser nanofluid
and therefore driven by a larger buoyancy force. Region of low nanoparticle fraction
forms near the trailing edge of the silicone drop at t = 0.2813 s, see the fourth plot
in Fig. 11. Once this region forms, it extends gradually into the flow under the
1850072-21
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
G. Qiangshun et al.
by UNIVERSITY OF NEW ENGLAND on 01/07/18. For personal use only.
Int. J. Comput. Methods Downloaded from www.worldscientific.com
Fig. 12. Solutions for Ro = 0.15Lo and ϕo = 0.02 at t = 0, 0.09375, 0.1875, 0.2813, 0.375 and
0.45 s.
influence of the vortical flow structure. This also attributes to the dominance of
convective transport with minimal diffusion effect.
The effect of drop radius is now investigated. With drop radius increased from
Ro = 0.15Lo (Fig. 12), 0.25Lo (Fig. 9) to 0.35Lo (Fig. 13), the larger drop is closer
to the wall and therefore experiences a larger wall effect. Increasingly constrained
1850072-22
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
Fig. 13. Solutions for Ro = 0.35Lo and ϕo = 0.02 at t = 0, 0.09375, 0.1875, 0.2813, 0.375 and
0.45 s.
by the wall, the drop undergoes a larger deformation leading to a higher drag. As a
result, the drop rises slowly. The flow still dominantly drives energy and nanopar-
ticle transport. Both temperature and nanoparticle distribution are least affected
for the case of the smallest drop, i.e., Ro = 0.15Lo, as its motion only generates a
weak vortical flow.
1850072-23
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
G. Qiangshun et al.
Initial conditions:
Boundary conditions:
u = uo , v = 0, at inlet
∂u
= 0, v = 0, at outlet (20d)
∂x
u = 0, at walls,
T = Tc , at inlet
∂T
= 0, at outlet (20e)
∂x
T = Th , at walls,
1850072-24
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
ϕo , in nanofluid
ϕ = , at inlet
0, otherwise
(20f)
∇T
n · DB ∇ϕ + DT = 0, at walls and outlet.
T
For demonstration purpose, the governing parameters are set to Lo = 0.0001 m,
uo = 1 m/s, Th = 315 K, Tc = 305 K and ϕo = 0.04. This gives a Reynolds number
of roughly
2ρbf uo Lo
Re = ∼ 200,
µbf
by UNIVERSITY OF NEW ENGLAND on 01/07/18. For personal use only.
well within the limit of a laminar flow. Note that this definition of Re is used only
Int. J. Comput. Methods Downloaded from www.worldscientific.com
as a reference and would be even lower if the properties of silicone oil are used
instead. Four scenarios differ in the location of the nanofluid layer at the inlet are
considered; these are
0.0 ≤ y/Lo ≤ 0.0
0.0 ≤ y/Lo ≤ 0.3
ϕo = 0.04 for . (21)
0.0 ≤ y/Lo ≤ 0.6
0.3 ≤ y/Lo ≤ 0.6
Obviously, the first scenario of φo = 0.04 for 0.0 ≤ y/Lo ≤ 0.0 corresponds to the
situation without nanofluid at the inlet, i.e., only pure water as the lower layer and
silicone oil as the upper layer. The last scenario of ϕo = 0.04 for 0.3 ≤ y/Lo ≤ 0.6
corresponds to the situation shown in Fig. 14 in which the water–alumina nanofluid
flows in the inlet as the middle layer ranging from 0.3 ≤ y/Lo ≤ 0.6.
Plotted in Fig. 15 are the steady-state solutions for the scenario of ϕo = 0.04
for 0.0 ≤ y/Lo ≤ 0.0 (without nanofluid, i.e., pure water and silicone oil only). The
interfacial profiles obtained on two consecutively refined meshes of 320 × 80 CVs
with ∆t = 6.25 × 10−7 s and 640 × 160 CVs with ∆t = 3.125 × 10−7 s are super-
imposed for the purpose of grid-independent study. A mesh of 320 × 80 CVs with
∆t = 6.25 × 10−7 s is sufficient to resolve the solution. The interface quickly flattens
not too far away from the inlet. Computations were also made using a longer chan-
nel with a length of 8Lo . There is no difference observed in the solution. Therefore,
the channel length of 4Lo is sufficient for the cases considered. Also shown in Fig. 15
are the velocity and temperature fields, both with the interfacial profile superim-
posed. Generally, a steeper transverse temperature gradient occurs adjacent to the
lower wall (in contact with pure water), compared with that of the upper wall (in
contact with the silicone oil). Further coupled with a larger thermal conductivity
of pure water in comparison to silicone oil, heat transfer into the water layer at the
lower wall is therefore larger than that into the silicone oil layer at the upper wall.
From the temperature field in Fig. 15, the temperature around the interface along
1850072-25
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
G. Qiangshun et al.
by UNIVERSITY OF NEW ENGLAND on 01/07/18. For personal use only.
Int. J. Comput. Methods Downloaded from www.worldscientific.com
Fig. 15. Solutions for ϕo = 0.04 for 0.0 ≤ y/Lo ≤ 0.0 (without nanofluid, i.e., pure water and
silicone oil only).
the channel varies very little. The resulted Marangoni effect is verified to be minute
and can be safely ignored.
Upon careful examination of the interfacial profiles for the scenarios considered
(Eq. (21)), the presence of nanoparticles does not alter noticeably the interfacial
profile and therefore hydrodynamic behavior. These plots for different scenarios are
therefore not presented. For stratified two-phase flow, the interfacial profile is dom-
inantly affected by the fluids’ viscosities. Even though the viscosity of the combined
miscible nanofluid–water layer (i.e., nanofluid and pure water layers) nominally
increases in the presence of nanoparticles, the viscosity of this layer (of the order
of 10−3 –10−4 Pa s) is still at least one order of magnitude smaller than that of the
silicone oil layer (0.08 Pa s). The effect of an increased viscosity of the combined
nanofluid–water layer is therefore negligible. The interfacial profile is still dictated
primarily by the silicone oil viscosity. This leads to very similar interfacial profiles
for all these scenarios.
The nanoparticle distribution for different scenarios is shown in Fig. 16. The-
oretically, nanoparticles can be transported transversely from the nanofluid layer
into the pure water layer. This transverse transport relies on three mechanisms:
convection, Brownian diffusion and thermophoresis. For convection, note that from
1850072-26
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
Fig. 15, the velocity field is almost parallel within the pure water layer (or com-
bined nanofluid–water in other scenarios). Therefore, convection does not transport
much nanoparticles from the nanofluid layer into the pure water layer. Directed
against nanoparticle concentration gradient, Brownian diffusion drives nanoparticles
from high nanoparticle volume fraction region (nanofluid layer) to low nanoparticle
volume fraction region (pure water layer). Finally, directed against temperature gra-
dient, thermophoresis drives nanoparticles from the high temperature region near
the lower wall toward the low temperature region at the channel center. In the
scenario of ϕo = 0.04 for 0.0 ≤ y/Lo ≤ 0.3, thermophoresis tends to drive the
nanoparticles from the lower nanofluid layer into the middle pure water layer. In
contrast, in the scenario of ϕo = 0.04 for 0.3 ≤ y/Lo ≤ 0.6, thermophoresis tends to
drive the nanoparticles back into the middle nanofluid layer. However, with similar
physical conditions as those prevail in Sec. 5.3, transport due to both Brownian
diffusion and thermophoresis is minimal. Therefore, nanoparticles mostly remain
1850072-27
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
G. Qiangshun et al.
within the nanofluid layer while they are transported downstream dominantly by
convection.
In the scenario of ϕo = 0.04 for 0.0 ≤ y/Lo ≤ 0.3 (Fig. 16(a)), the lower
nanofluid layer becomes thinner along the channel as it is squeezed by both the
middle pure water layer and the more viscous upper silicone oil layer. The aver-
age nanoparticle volume fraction within the nanofluid layer is higher than that in
the same layer at the inlet. Similar higher average nanoparticle volume fraction
within the nanofluid layer is also observed to occur in the scenario of ϕo = 0.04
for 0.3 ≤ y/Lo ≤ 0.6 (Fig. 16(b)). For the scenarios of 0.0 ≤ y/Lo ≤ 0.6 and
0.3 ≤ y/Lo ≤ 0.6, there is a thin region adjacent to the interface with higher
nanoparticle volume fraction. This accumulation of nanoparticles is driven by trans-
verse convection pushing the interface downwards as the more viscous silicone oil
by UNIVERSITY OF NEW ENGLAND on 01/07/18. For personal use only.
layer expands during the transient flow development. This thin region of high
Int. J. Comput. Methods Downloaded from www.worldscientific.com
1850072-28
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
for 0.3 ≤ y/Lo ≤ 0.6) scenarios are almost identical. However, with the presence
of nanoparticles adjacent to the lower wall (and similar concentration near the
lower wall, see Figs. 16(a) and 16(b)), NuL numbers for the second (ϕo = 0.04
for 0.0 ≤ y/Lo ≤ 0.3) and third (ϕo = 0.04 for 0.0 ≤ y/Lo ≤ 0.6) scenarios are
almost identical and most importantly lower than those of the first (ϕo = 0.04
for 0.0 ≤ y/Lo ≤ 0.0) and fourth (ϕo = 0.04 for 0.3 ≤ y/Lo ≤ 0.6) scenarios.
Note that Nusselt number can be construed as the ratio of convective to conduction
heat transfer. With increased thermal conductivity attributed to nanoparticles, a
higher increase in conduction heat transfer from the wall to the combined nanofluid–
water layer relative to convective heat transfer occurs. In fact, a similar reduction
in Nusselt number is also observed under conditions considered in the work of
Nimmagadda and Venkatasubbaiah [2015].
For the upper wall, NuU does not seem to be affected much in the presence
of nanoparticles in the lower combined nanofluid–water layer. It remains almost
identical for all scenarios. This is not unexpected as the flows in the upper silicone
oil layer are hydrodynamically very similar, and constant thermophysical silicone
oil properties dictating heat transfer at the upper wall are employed resulting in
similar heat transfer characteristics. Generally, NuU decreases rapidly at the inlet
and then flattens, i.e., a profile similar to that of a typical Graetz problem.
The bulk temperature Tb can be used to quantify heat transferred into the fluids.
It increases with the amount of heat absorbed. As the cumulative heat absorbed
from the wall along the channel increases, Tb increases along the channel for all
scenarios. The bulk temperature along the channel is almost identical for the first
(0.0 ≤ y/Lo ≤ 0.0) and the fourth scenarios (0.3 ≤ y/Lo ≤ 0.6). Comparing to
the first scenario (0.0 ≤ y/Lo ≤ 0.0), the presence of nanoparticles in the middle
layer in the fourth scenario (0.3 ≤ y/Lo ≤ 0.6) does not alter the heat transfer
performance. This suggests that augmentation of thermal properties attributed to
1850072-29
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
G. Qiangshun et al.
by UNIVERSITY OF NEW ENGLAND on 01/07/18. For personal use only.
Int. J. Comput. Methods Downloaded from www.worldscientific.com
1850072-30
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
ϕo = 0.02 for 0.0 ≤ y/Lo ≤ 0.0 is exactly identical to that of ϕo = 0.04 for
0.0 ≤ y/Lo ≤ 0.0, as the nanofluid layer is absent and therefore there are only the
pure water and silicone oil layers.
The variation of NuU along the channel is almost identical for all scenarios
considered. From Fig. 18(a), NuL for the scenario of ϕo = 0.02 for 0.3 ≤ y/Lo ≤ 0.6
is identical to the corresponding scenario of ϕo = 0.04 for 0.3 ≤ y/Lo ≤ 0.6. This
reinforces the fact that nanoparticles only minimally enhance heat transfer within
the nanofluid itself. This enhancement is negligible in a comparison with pure water
(base fluid) as the baseline. For the scenarios of ϕo = 0.02 for 0.0 ≤ y/Lo ≤ 0.3
and 0.0 ≤ y/Lo ≤ 0.6, thermal conductivity augmentation is less compared with
the corresponding scenarios of ϕo = 0.04. Therefore, on a relative basis, conduction
heat transfer is weaker leading to a higher NuL .
by UNIVERSITY OF NEW ENGLAND on 01/07/18. For personal use only.
Along the channel, Tb for the scenarios of ϕo = 0.02 for 0.0 ≤ y/Lo ≤ 0.3 and
Int. J. Comput. Methods Downloaded from www.worldscientific.com
0.0 ≤ y/Lo ≤ 0.6 is slightly lower than that of the corresponding scenarios with
ϕo = 0.04. The bulk temperature differences ∆Tb at the outlet between the scenarios
with (0.0 ≤ y/Lo ≤ 0.3 and 0.0 ≤ y/Lo ≤ 0.6) and without (0.0 ≤ y/Lo ≤ 0.0 and
0.3 ≤ y/Lo ≤ 0.6) nanoparticles adjacent to the lower wall are roughly 0.1 K and
0.05 K, respectively, for the case of ϕo = 0.04 and ϕo = 0.02. Lowering the amount
of nanoparticles near the lower wall indeed reduces heat transfer performance.
By maintaining ϕo = 0.04 and all other governing parameters fixed, the inlet
velocity is now reduced to uo = 0.1 m/s. This reduces Re and Peclet numbers for
both heat and mass transfers. Diffusion transport becomes relatively more impor-
tant if compared with the original scenarios of uo = 1 m/s. Plotted in Fig. 19 are
the solutions of uo = 0.1 m/s with ϕo = 0.04 for 0.0 ≤ y/Lo ≤ 0.3. The increase in
the strength of transverse diffusion transport can easily be identified for nanopar-
ticles (resulting in a more uniform nanoparticle distribution within the nanofluid
layer particularly near the inlet in comparison with Fig. 16(a)) and energy (affect-
ing the fluids temperature well into the center of the channel in comparison with
Fig. 15). As a result, NuL is lower than the corresponding scenarios with uo = 1 m/s
(Fig. 20). Similar observation of having NuL higher for the scenarios of ϕo = 0.04
for 0.0 ≤ y/Lo ≤ 0.0 and 0.3 ≤ y/Lo ≤ 0.6 than those for the scenarios of ϕo = 0.04
for 0.0 ≤ y/Lo ≤ 0.3 and 0.0 ≤ y/Lo ≤ 0.6 can be seen from Fig. 20.
NuU is however interestingly different from the scenarios with uo = 1 m/s. With
uo = 0.1 m/s, NuU for the scenarios of ϕo = 0.04 for 0.0 ≤ y/Lo ≤ 0.3 and
0.0 ≤ y/Lo ≤ 0.6 is almost identical but higher than those of ϕo = 0.04 for 0.0 ≤
y/Lo ≤ 0.0 and 0.3 ≤ y/Lo ≤ 0.6 (which are themselves almost identical). For
the scenarios of ϕo = 0.04 for 0.0 ≤ y/Lo ≤ 0.3 and 0.0 ≤ y/Lo ≤ 0.6, the
fluids absorb comparatively more heat from the lower wall due to the presence of
nanoparticles near the lower wall. At any given location, the fluids temperature
is generally higher resulting in a smaller effective transverse temperature gradient.
Heat diffusion from the upper wall to upper silicone oil layer decreases. With a lower
diffusion heat transport relatively to the convective heat transport, NuU increases.
With a reduced uo , the resident time for fluids in the channel is higher (roughly
1850072-31
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
G. Qiangshun et al.
by UNIVERSITY OF NEW ENGLAND on 01/07/18. For personal use only.
Int. J. Comput. Methods Downloaded from www.worldscientific.com
Fig. 19. Solutions for ϕo = 0.04 for 0.0 ≤ y/Lo ≤ 0.3 for uo = 0.1 m/s.
Fig. 20. Heat transfer performance for ϕo = 0.04 and uo = 0.1 m/s.
1850072-32
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
10 times) compared with the corresponding scenarios with uo = 1 m/s. The fluids
absorb more heat in the flow process. Therefore, Tb along the channel and at the
outlet is much higher (Fig. 20).
6. Concluding Remarks
This paper presents a model for two-phase nanofluid-fluid flow with heat transfer.
Nanofluid is modeled using Buongiorno’s approach. The model is verified against
solution for natural convection of nanofluid in a cavity. For demonstration, solu-
tions for flow and transport in a water–alumina nanofluid-filled cavity with a rising
silicone oil drop and stratified flow of water–alumina nanofluid, water and silicone
oil in a channel are presented. In this study, water–alumina nanofluid is employed.
by UNIVERSITY OF NEW ENGLAND on 01/07/18. For personal use only.
Nevertheless, the model works regardless of the type of base fluid and nanoparticles,
Int. J. Comput. Methods Downloaded from www.worldscientific.com
References
Abbasi, M. and Baniamerian, Z. [2014] “Analytical simulation of flow and heat transfer of
two-phase nanofluid (stratified flow regime),” Int. J. Chem. Eng. 2014, 474865.
Akgül, M. B. and Pakdemirli, M. [2016] “Numerical analysis of mixed convection of
nanofluids inside a vertical channel,” Int. J. Comput. Meth. 13, 1650012.
Alloui, Z., Vasseur, P. and Reggio, M. [2011] “Natural convection of nanofluids in a shallow
cavity heated from below,” Int. J. Therm. Sci. 50, 385–393.
Bachok, N., Ishak, A. and Pop, I. [2012] “Flow and heat transfer characteristics on a
moving plate in a nanofluid,” Int. J. Heat Mass Transf. 55, 642–648.
Brackbill, J. U., Kothe, D. B. and Zemach, C. [1992] “A continuum method for modelling
surface tension,” J. Comput. Phys. 100, 335–354.
Buongiorno, J. [2005] “Convective transport in nanofluids,” J. Heat Transf. 128, 240–250.
Chamkha, A. J. and Abu-Nada, E. [2012] “Mixed convection flow in single- and double-lid
driven square cavities filled with water–Al2 O3 nanofluid: Effect of viscosity models,”
Eur. J. Mech. B-Fluids 36, 82–96.
Cheng, L. and Liu, L. [2013] “Boiling and two-phase flow phenomena of refrigerant-based
nanofluids: Fundamentals, applications and challenges,” Int. J. Refrig. 36, 421–446.
Cianfrini, C., Corcione, M., Habib, E. and Quintino, A. [2014] “Buoyancy-induced con-
vection in Al2 O3 /water nanofluids from an enclosed heater,” Eur. J. Mech. B-Fluids
48, 123–134.
Ciloglu, D. and Bolukbasi, A. [2015] “A comprehensive review on pool boiling of nanoflu-
ids,” Appl. Therm. Eng. 84, 45–63.
1850072-33
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
G. Qiangshun et al.
Corcione, M. [2011] “Empirical correlating equations for predicting the effective thermal
conductivity and dynamic viscosity of nanofluids,” Energ. Convers. Manage. 52, 789–
793.
Corcione, M., Cianfrini, M. and Quintino, A. [2013] “Two-phase mixture modeling of nat-
ural convection of nanofluids with temperature-dependent properties,” Int. J. Therm.
Sci. 71, 182–195.
Corcione, M., Cianfrini, M. and Quintino, A. [2015] “Enhanced natural convection heat
transfer of nanofluids in enclosures with two adjacent walls heated and the two opposite
walls cooled,” Int. J. Heat Mass Transf. 88, 902–913.
Duursma, G., Sefiane, K., Dehaene, A., Harmand, S. and Wang, Y. [2015] “Flow and heat
transfer of single- and two-phase boiling of nanofluids in microchannels,” Heat Transf.
Eng. 36, 1252–1265.
El-Hamouz, A. [2007] “Effect of surfactant concentration and operating temperature on
the drop size distribution of silicon oil water dispersion,” J. Disper. Sci. Technol.
by UNIVERSITY OF NEW ENGLAND on 01/07/18. For personal use only.
28, 797–804.
Int. J. Comput. Methods Downloaded from www.worldscientific.com
1850072-34
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
Pouranfard, A. R., Mowla, D. and Esmaeilzadeh, F. [2015] “An experimental study of drag
reduction by nanofluids in slug two-phase flow of air and water through horizontal
pipes,” Chin. J. Chem. Eng. 23, 471–475.
Roe, P. L. [1983] “Some contributions to the modelling of discontinuous flows,” Proc. 15th
Summer Seminar on Applied Mathematics, 27 June–8 July, La Jolla, CA, pp. 23–34.
Ryzhkov, I. I. and Minakov, A. V. [2014] “The effect of nanoparticle diffusion and ther-
mophoresis on convective heat transfer of nanofluid in a circular tube,” Int. J. Heat
Mass Transf. 77, 956–969.
Saien, J. and Bamdadi, H. [2012] “Mass transfer from nanofluid single drops in
liquid−liquid extraction process,” Ind. Eng. Chem. Res. 51, 5157–5166.
Sert, I. O., Sezer-Uzol, N. and Kakaç, S. [2016] “Numerical approaches for convective
heat transfer with nanofluids,” Chapter 7, in Microscale and Nanoscale Heat Transfer :
Analysis, Design, and Application, eds. Rebay, M., Kakaç, S. and Cotta, R. M. (CRC
Press, Boca Raton), pp. 183–205.
Shu, C.-W. and Osher, S. [1988] “Efficient implementation of essentially non-oscillatory
shock capturing schemes,” J. Comput. Phys. 77, 439–471.
Sohel Murshed, S. M., Tan, S.-H. and Nguyen, N.-T. [2008] “Temperature dependence
of interfacial properties and viscosity of nanofluids for droplet-based microfluidics,”
J. Phys. D: Appl. Phys. 41, 085502.
Sussman, M., Smereka, P. and Osher, S. [1994] “A level set approach for computing solution
to incompressible two-phase flow,” J. Comput. Phys. 114, 146–159.
Taylor, R., Coulombe, S., Otanicar, T., Phelan, P., Gunawan, A., Lv, W., Rosengarten,
G., Prasher, R. and Tyagi, H. [2013] “Small particles, big impacts: A review of the
diverse applications of nanofluids,” J. Appl. Phys. 113, 011301.
Turkyilmazoglu, M. [2014] “Nanofluid flow and heat transfer due to a rotating disk,”
Comput. Fluids 94, 139–146.
Van Gorder, R. A. [2013] “Rare exact solution to a model of two-phase flow consisting of
nanofluid adjacent to a clear fluid,” Int. J. Heat Mass Transf. 61, 201–208.
Van Gorder, R. A., Prasad, K. V. and Vajravelu, K. [2012] “Convective heat transfer in the
vertical channel flow of a clear fluid adjacent to a nanofluid layer: A two-fluid model,”
Heat Mass Transf. 48, 1247–1255.
Versteeg, H. K. and Malalasekera, W. [2007] An Introduction to Computational Fluid
Dynamics: The Finite Volume Method, 2nd Edition (Prentice Education Limited,
England).
1850072-35
2nd Reading
January 3, 2018 14:9 WSPC/0219-8762 196-IJCM 1850072
G. Qiangshun et al.
Wang, P., Lv, J., Bai, M., Wang, Y., Hu, C. and Zhang, L. [2014] “Numerical simulation
on the flow and heat transfer process of nanofluids inside a piston cooling gallery,”
Numer. Heat Trans. A-Appl. 65, 378–400.
Wang, P., Lv, J., Bai, M., Li, G. and Zeng, K. [2015] “The reciprocating motion charac-
teristics of nanofluid inside the piston cooling gallery,” Powder Technol. 274, 402–417.
Yang, C., Li, W. and Nakayama, A. [2013] “Convective heat transfer of nanofluids in a
concentric annulus,” Int. J. Therm. Sci. 71, 249–257.
Yap, Y. F. and Chai, J. C. [2012] “Level-set method for multiphase flows,” Comput. Therm.
Sci. 4, 507–515.
Yap, Y. F. and Li, H. Y. [2015] “Modeling of two-phase nanofluid-fluid flow with heat
transfer,” in The 26th Int. Symp. Transport Phenomena, 27 September–1 October,
Leoben, Austria, Paper 118, pp. 1–10.
Yap, Y. F., Chai, J. C., Toh, K. C., Wong, T. N. and Lam, Y. C. [2005] “Numerical
modeling of unidirectional stratified flow with and without phase change,” Int. J. Heat
by UNIVERSITY OF NEW ENGLAND on 01/07/18. For personal use only.
Yap, Y. F., Chai, J. C., Wong, T. N., Toh, K. C. and Zhang, H. Y. [2006] “A global mass
correction scheme for the level-set method,” Numer. Heat Tr. B-Fund 50, 455–472.
Yu, W., France, D. M., Routbort, J. L. and Choi, S. U. S. [2008] “Review and comparison
of nanofluid thermal conductivity and heat transfer enhancements,” Heat Transf. Eng.
29, 432–460.
1850072-36