International Journal of Thermal Sciences 111 (2017) 499e510

Contents lists available at ScienceDirect

International Journal of Thermal Sciences

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

A perturbative thermal analysis for an electro-osmotic flow in a slit

microchannel based on a Lubrication theory

ndez a, *
E.A. Ramos a, O. Bautista b, J.J. Lizardi c, F. Me
a
Departamento de Termofluidos, Facultad de Ingeniería, UNAM, M
exico, D. F. 04510, Mexico
b
ESIME Azcapotzalco, Instituto Polit

ecnico Nacional, Av. de las Granjas No. 682, Col. Santa Catarina, M

exico, D.F. 02250, Mexico
c
noma de la Ciudad de M

Colegio de Ciencia y Tecnología, Universidad Auto exico, Campus San Lorenzo Tezonco, Calle Prolongacio
n San Isidro 151, 09790,
Mexico

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

Article history: In this work, we develop a new thermal analysis for an electro-osmotic flow in a rectangular micro-
Received 2 March 2016 channel. The central idea is very simple: the Debye length that defines the length of the electrical double-
Received in revised form layer depends on temperature T. Therefore, if exists any reason to include variable temperature effects,
19 September 2016
the above length should be utilized with caution because it appears in any electro-osmotic mathematical
Accepted 23 September 2016
model. For instance, the presence of the Joule effect is a source that can generate important longitudinal
temperature gradients along the microchannel and the isothermal hypothesis is no longer valid. In this
manner, the Debye length is altered and as a consequence, new longitudinal temperature gradient terms
Keywords:
Microchannels
appear into the resulting governing equations. These terms are enough to change the electric potential
Electro-osmotic flow and the flow field. Taking into account the above comments, in the present study the momentum
Non-isothermal equations together with the energy and Poisson conservation equations are solved by using a regular
Joule heating perturbation technique. For this purpose, we introduce a dimensionless parameter a that measures the
temperature deviations of a reference temperature. For practical cases, this parameter is small compared
with unity and the theoretical predictions show; however, that for the used values of this parameter, the
volumetric flow rate decreases in comparison with the isothermal case.
© 2016 Elsevier Masson SAS. All rights reserved.

1. Introduction steady and transient conditions, determining the influence of the

Nowadays, the theoretical analysis of electroosmotic flows has
allowed the development of important applications such as drug
delivery, DNA analysis, biological/chemical agent detection sensors
on microchips [1], etc. In general, electroosmotic flows are capable
to induce fluid pumping and flow control using electric fields,
eliminating the need for mechanical pumps or valves with moving
components. Here, we are particularly interested in those physical
aspects that can modify the estimation of the volumetric flow rates.
For instance, the analysis of electro-osmotic flows in microchannels
can be drastically altered by imposing non-isothermal conditions.
In this direction, some analytical results for electro-osmotic flows
in microchannels with different geometries can be found in Maynes
and Webb [2], Su et al. [3] and Horiuchi and Dutta [4]. In parallel,
Tang et al. [5] and Tang et al. [6] developed numerical studies under
* Corresponding author.

ndez).
E-mail address: fmendez@unam.mx (F. Me
Joule effect on the electroosmotic flow. The numerical simulations
reveal that the influence of the Joule effect is fundamental to un-
derstand simultaneously the electroosmotic flow and the mass
species transport. Using triangle microchannels, similar results can
be also found in Liao et al. [7]. Therefore, the above studies show
the vital importance of the Joule heating effect. On the other hand,
uniform or variable thermal properties of the electrolyte solution
serves us to characterize more complex electroosmotic flow cases,
where conjugate effects can be important as was previously studied
by Sa
anchez et al. [8], Escando
n et al. [9] and Bautista et al. [10].
Other important cases where heat transfer processes have a special
importance on the electroosmotic flow can be found also in Vocale
et al. [11], Chakraborty et al. [12] and Escando
n et al. [13]. Addi-
tional efforts including non-isothermal conditions, slip velocities,
interfacial instabilities and concentration and mixed effects on
electroosmotic flows circulating in microchannels with different
geometries have been considered to clarify the importance of these
additional effects. The most relevant contributions to appreciate
the above aspects can be found in Matías et al. [14e16], Tandon and

http://dx.doi.org/10.1016/j.ijthermalsci.2016.09.028
1290-0729/© 2016 Elsevier Masson SAS. All rights reserved.
500 E.A. Ramos et al. / International Journal of Thermal Sciences 111 (2017) 499e510
Kirby [17], Ge et al. [18] and Dey and Joo [19]. Special mention
deserve the works of Bahga et al. [20], Ghosh and Chakraborty [21]
and Yariv and Dorfman [22], where we can find special applications
of electroosmotic flows affected also by axially varying surface
conditions.
Although the previous specialized literature has reached an
important place to clarify the role of the Joule effect; however, all
previous contributions take into account that the ionic energy
parameter [1] given by the relationship aionic ¼ ezz=kB T is constant
if and only if the temperature is uniform. Here, the symbols e, z, z, kB
and T denote the electron charge, the valence, the zeta potential,
the Boltzmann constant and the temperature field of the electrolyte
solution, respectively. However, in the present case of study and
also in the references cited here, the temperature field of the
electroosmotic flow is not uniform due to the presence of impor-
tant longitudinal temperature gradients along the microchannel.
These temperature gradients are originated not only by the pres-
ence of Joule heating effect but they are also caused by the pressure
as will be shown later. In consequence, the above dimensionless
parameter aionic varies with the temperature and also the well-
known Debye length given by the relationship
lD ¼ ð2n∞ z2 e2 =εkB TÞ1=2 , where n∞ and ε denote the ionic number
concentration in the bulk solution and dielectric permittivity,
respectively.
Following the previous comments, the aim of this work is to
develop a new perturbative thermal analysis for a newtonian
electroosmotic flow, taking into account that the presence of Joule
heating effect and longitudinal pressure gradients can drastically
alter, among others, the volumetric flow rate and the induced
pressure field. For this purpose, we introduce a new dimensionless
parameter a that defines the deviations of the temperature with
respect to the isothermal case. For finite values of this parameter,
we show that the momentum conservation equations have new
terms which depend on temperature gradients and these terms are
caused in turn by new pressure gradients which depend on the
variable Debye length. However, in practical cases, this parameter a
assumes very small values. Therefore, the governing equations are
solved by applying a regular perturbation method for small values
of this parameter. The perturbative predictions reveal that even for
very small values of the parameter a, the volumetric flow rate
together with the velocity, pressure, temperature and electric fields
are accordingly modified.
2. Formulation and governing equations
In Fig. 1 we can appreciate a sketch of the physical model.
Owing to this geometry, a 2D rectangular coordinate system (x,y)
is adopted with the origin at the microchannel inlet and x-axis
along the microchannel centerline. Here, we assume that a new-
tonian laminar fluid is circulating into the rectangular micro-
channel. The slit microchannel is basically characterized by the
following geometrical dimensions: half height H and length L,
such that H≪L. In this manner, the above aspect ratio defines a
long rectangular microchannel and the thickness of the walls is
much smaller than any of the above geometrical lengths. We as-
sume, in addition, the microchannel walls are very good conduc-
tors of heat generated by the Joule heating effect. Therefore, we
accept also that the only mechanism for promoting the motion of
the fluid is carried out by employing electroosmotic forces, which
are induced by an electric field of strength E0 in the longitudinal
direction and given by E0¼f0/L, where f0 is the value of the
imposed electric potential at the microchannel inlet. In addition,
both ends of the microchannel are connected with two isothermal
liquid reservoirs, founding at temperature T0 and pressure P0
[23,24]. The heat losses at the interval 0xL for both lateral walls
of the microchannel are controlled by a convective heat transfer
process for which the heat transfer coefficient is known and given
lines below. The above heat losses cause the temperature of the
fluid can change in both longitudinal and transversal directions.
Because the surface of the walls is negatively charged, we assume
the existence of a thin electric double layer next to the walls and
the z-potential between the charge surface and the electrolyte
solution remains constant. Therefore, the net electric charge
contained in the thin electric double layer is the primary reason
for the electrokinetics effects, where positively charged ions can
be driven by an external electric field. Together with the previous
comments, we can add other well-known assumptions and hy-
pothesis that complete the present mathematical formulation and
they are basically the following [1]:
(a) the electric charge in the thin electric double layer obeys
the well-known Boltzmann distribution. The above is a conse-
quence for assuming a symmetric electrolyte of equal valence in
equilibrium with the charged surface, (b) we assume also that the
wall potential z is very small. In this manner, ez=kB T0 ≪1,
considering that in practical cases z < 25 mV. Therefore, the
Debye-Hückel approximation for the Poisson-Boltzmann equation
can be applied without any loss of generality; except for high
values of the zeta potential. In this last case, the mathematical
model develops here is no longer valid. In addition, the Debye-
Hückel approximation remains still valid for those cases when
temperature variations into the channel are important. This situ-
ation that corresponds to a non-isothermal electroosmotic flow,
the linearization process for the Poisson-Boltzmann equation must
take into account that the variable temperature effects, (c) the
streaming potential induced by the flow is smaller than the
external voltage, (d) the applied electric field is weak, i.e.,
f0 =L≪z=H. In this manner, the electric field is sligthly modified by
a negligible amount at the double layer [25], (e) we assume lD ≪H,
which means lD is a very thin layer, (f) the flow is laminar and the
corresponding Reynolds number is very small compared with
unity. In addition, because of H≪L, which is typical for micro-
fluidic systems driven by electrokinetic forces, the Lubrication
approximation can be used to obtain the flow field, (g) in order to
use the Boltzmann distribution, it is indispensable imposes that
the ionic Peclet number is small, which is equivalent to assume
equilibrium thermodynamics. The above can be appreciate from
the Nernst-Planck species conservation equations that establish a
species balance between ionic concentration gradients, electric
and convective terms. In a dimensionless version of the above
equations, the convective term can be neglected since it is
multiplied by the ionic Peclet number and for small values of this
parameter only prevails a balance between chemical and electric
forces, (h) the fluid properties do not vary with temperature.
However, this restriction can be eliminated as was widely studied
by Bautista et al. [10] for the case of an electro-osmotic flow in a
microchannel using the Phan-Thien and Tanner model, when the
viscosity is a dependent function on temperature and finally (i) we
assume that the total electric potential, F can be written as a
linear superposition of the externally applied potential, f and the
electrical potential due to the thin electric double layer at the
equilibrium state, j, i. e. F¼jþf, assumption which is only valid
for a thin electric double layer (lD ≪H). Under this leading order
consideration, therefore, the bulk of the fluid is completely free of
charge (electroneutrality condition) and in the absence of imposed
concentration gradients and under steady-state conditions, the
external potential f satisfies the Laplace equation.
Taking into account the previous assumptions and consider-
ations, the electroosmotic flow equations are the mass conserva-
tion equation,
 E.A. Ramos et al. / International Journal of Thermal Sciences 111 (2017) 499e510 501

Fig. 1. Schematic diagram of the studied physical model.

illustrated by Adjari [26,27]:

vu vv
þ ¼ 0; (1)
vx vy v2 f v2 f
þ ¼ 0: (7)
vx2 vy2
momentum equations for both directions,
In the above Equations (1), (2) And (6), the electric charge

 !
vu vu vp v u v u 2 2 density of the fluid is given by re ¼ 2z2 e2 n∞ j=kB Tðx; yÞ, where j
r u þv ¼ þh þ þ re Ex ; (2) is the electric potential which is generated into the thin electric
vx vy vx vx2 vy2
double layer, ninf is the ionic number concentration in the bulk
! solution, z is the absolute value of a (z:z) electrolyte solution


vv vv vp v2 v v2 v valency, e is the elementary charge, kB is the Boltzmann constant
r u þv ¼ þh þ þ re Ey ; (3) and T is the absolute temperature. Here, we must insist in the
vx vy vy vx2 vy2
following point: the above relationship for the charge density is,
among others, a function which depends on temperature field and
and the energy equation which can be written as,
this simple fact originates new temperature gradients in the gov-

 ! erning equations, as we will show lines below. However, we need
vT vT v2 T v2 T  
rCp u þv ¼k þ þ s Ex2 þ Ey2 : (4) first to complete the governing equations by defining also the
vx vy vx2 vy2 corresponding boundary conditions.
In Equations (1)e(4), u and v are the velocity components in
the x and y directions, respectively; T and p represent the tem- 2.1. Boundary conditions
perature of the electrolyte solution and the induced hydrodynamic
pressure. Also, Cp, k, h, s, r and re are the specific heat, the thermal For the present problem, the boundary conditions required to
conductivity, the dynamic viscosity, electrical conductivity, the solve the set of Eqs. (1)e(6) are the following:
density and the electric charge density of the electrolyte solution,
respectively. In the present work, the above properties are uðx; y ¼ ±HÞ ¼ vðx; y ¼ ±HÞ ¼ 0; (8)
assumed uniform, except the electric charge density, re, because it
depends on temperature, as we will show lines below. On the Pðx ¼ 0; yÞ ¼ Pðx ¼ L; yÞ ¼ P0 ; (9)
other hand, Ex and Ey are the electrical field strength in the x and y
directions and they can be evaluated from the following Tðx ¼ 0; yÞ ¼ Tðx ¼ L; yÞ ¼ T0 ; (10)
relationships,
f ðx ¼ 0Þ ¼ f0 ; f ðx ¼ LÞ ¼ 0; (11)
vFðx; yÞ vFðx; yÞ
Ex ¼  and Ey ¼  (5)
vx vy together with

where the variable Fðx; yÞ represents the total electric potential and vu vT vF
is governed by the well-known Poisson's equation, given by: y¼0: ¼ ¼ ¼ 0; (12)
vy vy vy

v2 F v2 F r s vT
þ 2 ¼ e; (6) y¼H: b ,VF ¼ 
n and k ¼ hðT  T∞ Þ; (13)
vx2 vy ε ε vy
while the external electric potential based on the previous b is a unit vector, perpendicular to the
In the last equation, n
assumption (i) is governed by the Laplace equation as was surface of the microchannel and s is the surface charge density for a
502 E.A. Ramos et al. / International Journal of Thermal Sciences 111 (2017) 499e510

non-metallic surface. The above equation demands that the surface

 !2
charge plus the charge of the ions in the whole double layer must vq vq v2 q v2 q 1 vj 2 vj vf
be zero. At the end of Appendix A, we have introduced specific εPe u þv ¼ ε2 2 þ 2 þ 2 þ
vc vY vc vY 2 vc 2 vc vc
details to transform that equation into an appropriate relationship

2 !2
for the zeta potential at the wall. vf 1 vj
þ þ 2 2 ; (18)
vc ε 2 vc

2.2. Dimensionless governing equations

v2 f v2 f v2 j v2 j k2
2ε2 þ 2 2 þ ε2 2 þ 2 ¼ j; (19)
For a weak electric field and long microchannels, i.e. H≪L, we vc 2 vY vc vY ð1 þ aqÞ
have anticipated that the total potential F can be decomposed as
F≡jþf. The above is a direct consequence of the assumption (i). v2 f v2 f
However, in the present non-isothermal case is inevitable to as- ε2 þ ¼ 0; (20)
vc2 vY 2
sume that the electric potential into the thin double layer j is
function of both coordinates x and y in contraposition with previ-
where the last equation governs the external electric field and
ous works. Accordingly, the corresponding linearized Poisson-
represents just the Laplace equation as was illustrated by Pascall
Boltzmann equation must be modified to consider the non-
and Squires [28]. It is important to note that the above is only valid
isothermal case. In this manner, the set of equations (1)e(7) can
for a thin electric double layer according with the assumption (i)
be transformed with the aid of this last change.
given previously. In addition, the sum of the first two terms of the
Taking into account that F¼jþf as was previously commented,
right-hand side of Eq. (19) is equal to zero taking into account Eq.
we can introduce the following dimensionless variables that per-
(20). In the above equations, the dimensionless parameters are
mits us to simplify considerably the governing equations,
defined as:
x y u vL rUHS H rCp UHS H
c¼ ; Y¼ ; u¼ ; v¼ ; (14) ε¼
H
; Re ¼ ; Pe ¼ ; k ¼ kH;
L H UHS UHS H L h k
 0 
T  T0 p  p00 H2 f j 2z2 e2 n∞ DTc sE02 H2 fo
q¼ ;P ¼ ;f ¼ ;j ¼ ; k2 ¼ ;a ¼ ¼ 2¼ :
DTc hUHS L f0 z εkB T0 T0 kT0
and
z
where DTc ¼ sE02 H 2 =k represents the characteristic temperature On the other hand, the boundary conditions given by Eqs. 8e13
rise for this problem and was obtained by performing an order of can be transformed in dimensionless form as,
magnitude analysis of the energy equation and the details are
omitted for simplicity; UHS is the well-known Helmholtz-Smo- uðc; Y ¼ ±1Þ ¼ vðc; Y ¼ ±1Þ ¼ Pðc ¼ 0; YÞ ¼ Pðc ¼ 1; YÞ
¼ qðc ¼ 0; YÞ ¼ qðc ¼ 1; YÞ ¼ 0;
luchowski electroosmotic velocity and given by UHS ¼ εzE0 =h . In
addition, the modified pressure, p0 , is related with the pressure (21)
field p of Eqs. (2)e(3), through the relationship p0 ¼pεk2j2/2,
where k2 ¼ 2z2 e2 n∞ =εkB Tðx; yÞ. This change of variable was fðc ¼ 0Þ ¼ 1; fðc ¼ 1Þ ¼ 0; (22)
introduced [26] to transform the momemtum equations in such
manner that always can exist a relationship between the velocity vu vq vf vj
field and the external electric field. However, in the present case Y¼0: ¼ ¼ ¼ ¼ 0; (23)
vY vY vY vY
this change of variable originates additionally new temperature
gradients in both momemtum equations, as we can see lines
vj H s vf vq
below. Therefore, in terms of the previous dimensionless variables, Y¼1: ¼ ; ¼0 and ¼ Lð1 þ bqÞ:
the governing equations (1)e(4) and (6) and (7) take the following vY zε vY vY
form, (24)
The last two boundary conditions in Eq. (23) for the electric
vu vv
þ ¼0 (15) potentials follow from the symmetry condition, while the first two
vc vY
boundary conditions of Eq. (24) follow directly from Eq. (13),
accepting the fact that the microchannel surface is not metallic. In

 2
vu vu vP v2 u v2 u a j vq addition, we use the previous assumption (b), i. e. the wall potential
εRe u þv ¼ þ ε2 2 þ 2  k2 z is very small. In this manner, ez=kB T0 ≪1, considering that in
vc vY vc vc vY 22 ð1 þ aqÞ2 vc
practical cases z<25 mV. Therefore, the first boundary condition of
j vf Eq. (24) can be written as
 k2 ;
ð1 þ aqÞ vc
(16) Y¼1: j ¼ 1: (25)

! The details for deriving the above equation are given at the end

 of Appendix A. On the other hand, b ¼ sE02 H 2 =kðT0  T∞ Þ and
3 vv vv vP v2 v v2 v
ε Re u þv ¼ þ ε2 ε2 2 þ 2 L ¼ hðT0  T∞ Þ=sE02 H. In the convective boundary condition given
vc vY vY vc vY
by Eq. (24), L is a dimensionless parameter that measures the
2
a j vq competition between the heat losses at the surface of the micro-
 k2 ; (17)
22 ð1 þ aqÞ2 vY channel wall and the heat generated due to the Joule heating in the
fluid. In addition, k ¼ kH, with k denoting the Debye-Hückel
parameter which is related to the Debye length lD, through the
 E.A. Ramos et al. / International Journal of Thermal Sciences 111 (2017) 499e510 503

relationship k1 ¼ lD ¼ ð2n∞ z2 e2 =εkB T0 Þ1=2 , [29]. The system of Table 2

Eqs. 15e24 defines now, in dimensionless form, the non-isothermal Dimensionless parameters for the present
analysis.
electroosmotic flow in a slit microchannel. In the following section,
we use Lubrication theory to reduce the governing equations by Parameter Value
performing an asymptotic analysis for small values of the dimen- k 40
sionless parameter a. In practical cases, this parameter can assume 2 1000
values of the order of 0.004; therefore, the asymptotic analysis is ε 0.01
Re 0.1
just appropriate for this limit.
Pe 0.05
L 0.4
a 0.004
3. Lubrication theory b 0.008

As was previously mentioned, we propose an asymptotic anal-

ysis for the dimensionless governing equations. From the physical
and nondimensionless parameters given on Tables 1 and 2, we can
readily deduce that terms of order εRe, ε2, ε3Re, 1/2, 1/22 and 1/(2ε)2
are very small compared with a and in turn, the latter is also small
compared with unity. In addition, we assume that those terms k2 =2
are small or order unity. Therefore, we propose first to deduce the
governing equations by taking into account the Lubrication limit,
where terms of order εRe, ε2, ε3Re, 1/2, 1/22 and 1/(2ε)2 can be
neglected. After that, we apply a regular perturbation scheme for
small values of the dimensionless parameter a. It means that in a
first approximation, we can discard those terms of Eqs. 15e20 and
therefore, we suggest first for mass and momentum conservation
equations, the following regular expansion for each one of the
variables u; v; P; q; j and f:
 
J ¼ J0 þ O ε2 ; εRe; ε3 Re; … þ …;
velocity field is necessary to known, together with the external
electric field, also the temperature field. This is the reason to
include the energy equation for the present problem. Although that
the energy equation given by Eq. (18) requires for this section a
different treatment, we prefer a brief discussion lines below. On the
other hand, taking a¼0 in Eqs. (28) and (29), which corresponds to
an isothermal electroosmotic flow, we recover the well-known
motion of a newtonian fluid governed only by a purely electroos-
motic flow, which is independent of the temperature and whose
solution can be found elsewhere [10], [30]. This last case reveals
now the importance of the parameter a: even for very small values
of this parameter, as we will show later, the temperature deviations
of the isothermal hypothesis induce strong changes in the pressure
and velocity fields. However, we need to understand first the
asymptotic structure of governing equations related with the
(26) temperature and electric potentials fields.

3.1. Energy equation

such that the leading order of Eqs. 15e17, retaining obviously terms
of order a, can be written in the following form,
For the energy equation (18), both diffusive and convective
vu0 vv0 terms must be retained. Otherwise, the heat transport would be
þ ¼0 (27) only dominated by diffusive and Joule effects, a situation which is
vc vY
impossible since the fluid is moving. The details of the above
considerations to reduce the energy equation can be found else-
2
vP0 v2 u0 a j0 vq0 j0 vf0 where [10], and obviously, we adopt here the same procedure.
¼  k2  k2 (28)
vc v Y2 22 ð1 þ aq0 Þ2 vc ð1 þ aq0 Þ vc Furthermore, Eq. (18) admits a first integral in the transverse di-
rection and after applying the boundary conditions (23) and (24),
2
we obtain that:
vP0 a j0 vq0
¼ k2 : (29)
vY 22 ð1 þ aq0 Þ2 vY Z1 Z1
vq vq1
εPe 0 3
u0 dY þ ε Pe v0 dY
Equations 27e29 represent the motion of a newtonian fluid vc vY
0 0
governed by pressure, viscosity, and external electric forces. This
2
situation is clearly indicated by Eq. (28) and represents the limit of v q0
2
vf0
¼ ε2  Lf1 þ bq g þ (30)
the Lubrication theory. However in the present case, the second vc2
0
vc
term of the right-hand side of Eq. (28) shows the strong matching
The above equation, together with Eqs. 27e29 can easily solved.
between velocity and temperature fields: in order to predict the
However, Eq. (30) requires, such as Eqs. 27e29, that f0 is known.
For this reason, we provide in the following subsection, the
Table 1
Material transport properties and geometrical parameters.
simplified governing equations for the electric fields f0 and j0 .

Parameter Value Unit 3.2. Poisson-Boltzmann equation

cp 3760 J/kg K
E0 ~103 V/m Under the Lubrication approximation, Eqs. (19) and (20) can be
h 20e100 W/m2 K
written as,
H 50e200 mm
k 0.61e0.7 W/m K
L ~102 m v2 j0 k2
¼ j (31)
ε 0.01e0.10 e vY 2 ð1 þ aq0 Þ 0
ε ~1010 C/V m
z ~102 V
h ~103e101 kg/m s
and
lD 1e10 nm
r ~103 kg/m3 v2 f0 v2 f0 v2 f
s ε2 þ þ ε2 21 ¼ 0: (32)
0.1e10 S/m vc 2 vY 2 vY
504 E.A. Ramos et al. / International Journal of Thermal Sciences 111 (2017) 499e510

For the last equation, we must retain up to terms of order ε2. The
reason to make it is the following: if we only consider terms of d2 f0;0 v2 f1;0
þ ¼ 0: (44)
order ε0, then we obtain from Eq. (20) that, dc2 vY 2

v2 f0 The last equation can be transversally integrated from Y ¼ 0 to 1

¼ 0; (33) by considering that at,
vY 2

however, the above equation is not sufficient to know the depen- vf1;0 vf1;0
Y¼0: ¼0 and Y ¼1: ¼ 0: (45)
dence of the potential f0 on the dimensionless longitudinal coor- vY vY
dinate c. Therefore, the asymptotic analysis of the above equation The above boundary conditions were deducted from Eqs. (23)
requires at least retain higher-order terms in ε2. A similar case was and (24). Therefore, we obtain that the external electric potential
developed by Lin et al. [31], studying a depth-averaged electroki- is governed by the equation,
netic flow with the aid of a three-dimensional model for micro-
channels. Here, we have adopted the same procedure and from Eq. d2 f0;0
(32), we obtain up to terms of order ε2, the following equation, ¼ 0; (46)
dc2
v2 f0 v2 f1
þ ¼ 0: (34) the above equation has a trivial solution given as f0;0 ¼ 1  c. A
vc2 vY 2 similar procedure permits us to obtain that f0;1 ¼ 0. The remaining
The above equation can be transversally integrated by using system of Eqs. 40e43 up to terms of zero-order in a, represents the
appropriate boundary conditions. However, we prefer to carry out well-known isothermal and purely electroosmotic flow, and except
the above integral in the following subsection taking into account that the velocity component u0,0 is dependent of the transverse
small values of the thermal parameter a. coordinate Y, the other dependent variables are only functions of
the longitudinal coordinate c.
Similarly, for terms of order a1, we can write:
4. Solution in the limit of a/0
vu0;1 vv0;1
þ ¼ 0; (47)
As we anticipate, Eqs. 27e32 can be solved with the aid of a reg- vc vY
ular perturbation scheme. For this purpose, we suggest the following
expansions for the dependent variables u0 , v0 , P0, q0 and j0 , taking v2 u0;1 k2 2 dq0;0   dP
0;1
into account that the zero-order solution corresponds to a purely ¼ j0;0 þ k2 j0;0 q0;0  j0;1 þ (48)
vY 2 22 dc dc
electroosmotic flow without pressure gradients and temperature
effects. Therefore, with the aid of the asymptotic expansions, and

u0 ¼ u0;0 þ a u0;1 þ …; v0 ¼ a v0;1 þ …; (35) v2 j0;1  

¼ k2 j0;1  j0;0 q0;0 : (49)
vY 2
P0 ¼ a P0;1 þ …; q0 ¼ q0;0 þ a q0;1 þ …; (36)
It should be noted that the energy equation up to terms of order
a has not been included since it is not necessary for determining the
f0 ¼ f0;0 þ af0;1 þ …; (37) induced pressure P0;1 , as we show lines below. The governing
equations 40e49 must be solved with the following boundary
f1 ¼ f1;0 þ af1;1 þ …; (38) conditions:

and u0;i ðc; Y ¼ 1Þ ¼ 0 ; v0;i ðY ¼ ±1Þ ¼ 0; for i ¼ 0; 1; …

(50)
P0;i ðc ¼ 0; 1Þ ¼ q0;i ðc ¼ 0; 1Þ ¼ 0 for i ¼ 0; 1; …
j0 ¼ j0;0 þ aj0;1 þ …; (39)

then eqs. 27e31 can be transformed into the following system. Up

4.1. Zero-order solution in a
to terms of order a0, we have:

du0;0 Therefore, the zero-order solution of Eqs. 40e43 can be easily

¼0 (40) solved and after some algebraic manipulations, we have that,
dc
coshðkYÞ coshðkYÞ
d2 u0;0
u0;0 ¼ 1  and j0;0 ¼ : (51)
coshðkÞ coshðkÞ
¼ k2 j0;0 (41)
dY 2
02 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi3 1 02

2
!2 Pek Pek L b Pek1
Z1 q0;0 ¼ a1 exp@4 1
þ 1
þ 2 5cA þ b1 exp@4
dq0;0 d2 q0;0   df0;0 2ε 2ε ε 2ε
εPe u0;0 dY ¼ ε2  L 1 þ bq0;0 þ
dc dc2 dc sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi3 1



0
Pek1 2 Lb5 A 1 1
(42)  þ 2 c þ 1 ;
2ε ε b L

d2 j0;0 (52)
¼ k2 j0;0 (43)
dY 2 where the constants a1 and b1 are defined as,
 E.A. Ramos et al. / International Journal of Thermal Sciences 111 (2017) 499e510 505

2 ffi3
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

2
Lb5

 1  exp4Pek
2ε 
1 Pek1
2ε þ ε2
1 1
a1 ¼  1 2 ffi3
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 ffi3
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
b L
2
2
Lb5 b
exp4Pek
2ε
1
þ Pek1
2ε þ 2  exp4
ε
Pek 1
 2ε
Pek 1
þ 25
2ε
L
ε

and

2 ffi3
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

 2
8 9

>
> 1 exp4Pek
2ε
1
 Pek1
2ε þ Lε2b5 >
>
1 1 < =
b1 ¼ 1 2 ffi3
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 ffi3  1
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
b L >
>
2
2 >
>
: b b ;
exp4 1 þ
Pek
2ε
Pek 1
2ε þ 2 5  exp4 1 
L
ε
Pek Pek 1
2ε þ 25L
2ε ε

respectively. In the above equation, the factor k1 given by the in- continuity equation, we can integrate the resulting equation in the
R1
tegral k1 ¼ 0 u0;0 dY, can be evaluated with the aid of Eq. (51) and transverse direction, from Y¼0 to Y¼1 and applying appropriate
the result is, boundary conditions, we obtain the following equation

tanhðkÞ d2 P0;1 d2 q0;0 dq0;0

k1 ¼ 1  : ¼ c1  d1
k dc 2 dc 2 dc
(55)

where the constants c1 and d1 are given by,

4.2. First-order solution in a c1 ¼

3 1 1
½senhð2 kÞ  senhð2 kÞ 
4 2
k
162 cosh2 ðkÞ 4k 3
In this case, we need to solve the system of Eqs. 47e49. There-
fore, we proceed in the following manner: we replace the zero- and
order solution given by Eq. (51) into Eq. (49). Te resulting equa-  

3 1 1
tion can be integrated twice with respect to the transverse coor- d1 ¼ tanhðkÞ þ ½senhðkÞ  senhðkÞ  2coshðkÞ
vj1;0 4 coshðkÞ k
dinate Y and applying the boundary conditions at Y¼0, vY ¼ 0 and
at Y¼1, j1,0¼0, we obtain that:
respectively.
k q0;0 Equation (55) can be integrated directly and the result is given
j0;1 ¼ ðtanhðkÞcoshðkYÞ  YsenhðkYÞÞ: (53) by,
2 coshðkÞ
On the other hand, the solution up to terms of order a for the P0;1 ¼ c1 q0;0  d1 G þ d1 fGð1Þ  Gð0Þgc þ d1 Gð0Þ; (56)
velocity component u0;1 can be easily determined by replacing the
corresponding temperature and electric profiles into Eq. (48). The where
resulting equation can be integrated again twice and we obtain, 02 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi3 1 02

2
Pek Pek L b Pek1
dq0;0  dP1
0;1 GðcÞ ¼ f1 exp@4 1
þ 1
þ 2 5cA þ f2 exp@4
u0;1 ¼ U1 ðYÞ  U2 ðYÞq0;0 þ Y2  1 ; (54) 2ε 2ε ε 2ε
dc 2 dc
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi3 1



where the functions U1(Y) and U2(Y) are defined as, Pek1 2 Lb5 A 1 1
 þ 2 c þ  1 c;
h
2ε ε b L
1 1 i 1
U1 ðYÞ ¼ k2
Y 2
 1 þ ½coshð2 kYÞ  coshð2 kÞ (57)
82 cosh2 ðkÞ 2
and the constants f1 and f2 are given as,
and
a1
f1 ¼ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
k coshðkYÞ

U2 ðYÞ ¼ ftanhðkÞ  Y tanhðkYÞg: 2
2 coshðkÞ Pek1
2ε þ Pek1
2ε þ Lε2b
We must emphasize at this point that the induce pressure P0;1
remains unknown. Substituting the flow velocity u0;1 in the and
506 E.A. Ramos et al. / International Journal of Thermal Sciences 111 (2017) 499e510

b1
f2 ¼ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

 ffi:
2
Pek1
2ε  Pek1
2ε þ Lε2b

4.3. Nondimensionless volumetric flow rate

Finally, the dimensionless volumetric flow rate into the micro-

channel can be found with the aid of the velocity expansion
u0 ¼ u0;0 þ a u0;1 þ …, recognizing that u0;0 and u0;1 are given by
R1
Eqs. (51) and (54) and the following relationship Q ¼ 0 u0 dY. In
this manner, the dimensionless volumetric flow rate is given by


tanhðkÞ 1 dq0;0 dP0;1
Q ¼1 þ a g1  h1 q0;0  ; (58)
k 3 dx dx

where the constants g1 and h2 are given as,

  Fig. 2. Asymptotically computed dimensionless velocity component u0 for a¼0, 0.004,
3 1 senhð2kÞ 4 2 0.006, 0.008 and 0.01. The values of the dimensionless parameters are: k ¼ 40, 2¼1000,
g1 ¼  coshð2kÞ  k
162 cosh2 ðkÞ 2k 3 ε¼0.01, Pe¼0.05, L¼0.4 and b¼0.008.

and
represents a maximum bound for which the longitudinal temper-
3 tankðkÞ
h1 ¼ tanh2 ðkÞ þ 1 ;
2 k
in the above equation, we have taken into account that the
dimensionless volumetric flow rate is defined as Q ¼ Q =Qc , where
Qc¼2HUHS and represents the characteristic volumetric flow rate
per unit width of the microchannel.
It is important to note that the induced pressure P0 ¼ a P0;1 þ
…; depends only on the longitudinal coordinate c. However, the
solution for the longitudinal velocity, u0 , depends on both longi-
tudinal and transverse coordinates and v0 will not be zero, in
general. We should keep in mind that the dimensionless parameter
a defines the temperature deviations from the isothermal case
generating new longitudinal pressure and temperature gradients.
5. Results and discussion
ature gradients can seriously disturb the velocity profile. In our
case, a assumes smaller values (the maximum value of a in the
present analysis is a¼0.01). The reason to consider in our analysis
values of a0.01, is due to that for larger values, the velocity be-
comes negative, situation which physically is inadmissible. In other
works, the perturbation solution is only valid for the range of
0a0.01. For larger values of a outside of this range, the system of
equations 27e32 should be numerically solved. However, the above
range is a very typical case since the physical properties and pa-
rameters needed to evaluate a are very representative.
Fig. 2 shows the dimensionless component of the velocity u0 as a
function of the dimensionless transverse coordinate Y for five
values of the parameter a and selected values of the dimensionless
parameters involved in the present work. Here, we can appreciate
classical distributions for the velocity profiles: near to the wall, the
velocity component suddenly goes to zero and far away from the

In this work, some typical transport properties and character-

istic values of geometric parameters for the microchannel are given
on Table 1, which was previously given in Section 3. In addition,
Table 2 shows the corresponding dimensionless parameters.
Taking into account that the thermal parameter a represents the
ratio between the characteristic temperature rise (~ sE20H2/k) along
the microchannel caused by Joule heating effect and the reference
temperature T0, we can anticipate for which values of a, the lon-
gitudinal temperature gradient may or may not have significant
effects on the velocity, pressure and electric fields. The procedure is
the following: when we derive Eq. (16), we have used the trans-
formation p0 ¼pεk2j2/2. If the above transformation is introduced
into Eq. (2), we generate then an additional term of the form
z2 e2 n∞ 2 vT
kB T 2
j vx . The above term appears in dimensionless form just as
the fourth on the right-hand side of Eq. (16). Therefore, for the latter
term involving longitudinal temperature gradients is relevant in Eq.
(2), then this term should be of the same order of magnitude as the
viscous term, which in order of magnitude is equivalent to write
hUHS 2 2

H2
 z ke Tn2∞ , where all symbols and characteristic scales were
B
Fig. 3. Dimensionless velocity profiles as function of the dimensionless transverse
defined in Section 2.2. Therefore, the above order relationship can coordinate Y and different values of the longitudinal coordinate c¼0.1, 0.5, 0.7 and 0.9.
be written in terms of the thermal parameter as. Using the data of The values of the dimensionless parameters are: k ¼ 40, 2¼1000, ε¼0.01, Pe¼0.05,
the physical properties of Table 1, we found that a~101. This value L¼0.4, a¼0.004 and b¼0.008.
 E.A. Ramos et al. / International Journal of Thermal Sciences 111 (2017) 499e510 507

wall, the velocity reaches the well-known Helmholtz-Smo-

luchowski electroosmotic velocity. However, for each value of the
coordinate Y and just in regions far from the wall, the velocity in-
creases sligthly in comparison with the isothermal case, meanwhile
the velocity decreases sligthly for zones next to the wall, which
reveals then the influence of the thermal parameter a. The above
can be seen from the two boxes included in this figure. In this
manner, the concavity of these curves is directly connected with the
values of the parameter a. This point will be discussed in the
Conclusions section, where appropriate comments related to the
presence of the induced pressure and temperature gradients have
been included.
On Fig. 3, dimensionless velocity profiles are plotted, again, as
function of the dimensionless transverse coordinate Y, for four
different values of the longitudinal coordinate c¼0.1, 0.5, 0.7 and
0.9 and similar values of dimensionless parameters of the above
figure. Now, we can see that the parabolic character of the velocity
profile is practically maintained the same; however, we include
additional boxes into the main figure, where the concavity of the
curves can change sligthly depending of the assumed values of the
Fig. 5. Dimensionless induced pressure P0 as function of the dimensionless coordi-
longitudinal coordinate. For instance, for c¼0.1, we have in this case nate c, for three values of the parameter b¼0.004, 0.007 and 0.009. The values of the
the smallest values for the velocity distribution and the curve is dimensionless parameters are: k ¼ 40, 2¼1000, ε¼0.01, Pe¼0.05, L¼0.4 and a¼0.004.
convex; after for c¼0.5 and 0.7, the velocity profiles reach
maximum values with practically the same profile, but the in-
crements of the velocities transform the distribution towards a The above behavior is repeated if the value of a changes; however,
concave curve and for c¼0.9, the velocity profile again returns to be the pressure distributions are amplified for increasing values of this
a smooth convex curve. We believe that this change of concavity is thermal parameter. Due to these favorable and unfavorable pres-
due to induced pressure gradients along the longitudinal coordi- sure distributions along the longitudinal coordinate, is that velocity
nate, as show in the next two Figures. profiles can change their shape.
On Fig. 4 the pressure profile is shown as function of the On Fig. 5 we show again the pressure profiles as function of the
dimensionless coordinate c for five different values of the thermal longitudinal coordinate c. However in the present case, we assume
parameter a(¼0,0.004,0.006,0.008 and 0.01), which defines the for a fixed value of the parameter a¼0.004, different values of the
non-isothermal case. Clearly, the case of a¼0 represents the dimensionless parameter b. This dimensionless parameter mea-
isothermal case where pressure gradients are absent. This corre- sures the importance or competition between the characteristic
sponds a purely electroosmotic flow, Bautista et al. [10]. However, temperature rise causes by the Joule heating effect and character-
for values of a differents to zero, the pressure distributions change istic temperature drop with the environment outside the micro-
drastically: the pressure first is increased up to a value of c¼0.25, channel. For increasing values of this parameter, the pressure
where the pressure reaches a different maximum value for each profiles decrease following a similar behavior of the previous
value of a. After this point, the pressure begins to decrease until the figure.
value zero is reached practically in c¼0.65. After that, the pressure The dimensionless temperature profiles, q, as function of the
changes then its behavior modifying the concavity of the profiles. longitudinal coordinate c are depicted on Fig. 6. In this case, we

Fig. 4. Dimensionless induced pressure P0 as function of the dimensionless coordi-

nate c, for five values of the parameter a¼0, 0.004, 0.006, 0.008 and 0.01. The values of Fig. 6. Dimensionless temperature profiles as function of the longitudinal coordinate
the dimensionless parameters are: k ¼ 40, 2¼1000, ε¼0.01, Pe¼0.05, L¼0.4 and c, for three values of the Pe
clet number, Pe. The values of the dimensionless param-
b¼0.008. eters are: k ¼ 40, 2¼1000, ε¼0.01, L¼0.4 and b¼0.008.
508 E.A. Ramos et al. / International Journal of Thermal Sciences 111 (2017) 499e510
Fig. 7. Dimensionless temperature profiles as function of the longitudinal coordinate
c, for three values of the dimensionless parameter, b¼0.004, 0.007 and 0.009. The
values of the dimensionless parameters are: k ¼ 40, 2¼1000, ε¼0.01, Pe¼0.05 and
L¼0.4.
show the influence of the Pe
clet number, Pe. For increasing values
of this parameter, the dimensionless temperature distributions
assume smaller values since the fluid will reach a better convective
heat transport and therefore fluid tends to remain colder. In this
manner, the central region of the microchannel is where the most
relevant increments of the temperature are reached because of the
ends of the microchannel are maintained at uniform temperature.
Therefore, we can see that the combination of the parameters Pe
and b determine clearly the influence of the Joule effect on the
temperature rise. In the following figure, we can appreciate in more
detail the above aspect.
Fig. 7 shows again the variation of the dimensionless tempera-
ture as function of the dimensionless coordinate c, for three
different values of the parameter b. In this figure, for increasing
Fig. 8. Dimensionless temperature profiles as function of the longitudinal coordinate
c, for three values of the dimensionless parameter L¼0.2, 0.5 and 0.8. The values of the
dimensionless parameters are: k ¼ 40, 2¼1000, ε¼0.01, Pe¼0.05 and b¼0.004.
values of b, the temperature profile drops, which seems to be a
contradiction; however, we must have care with the use of the
dimensionless parameters b and L. This last parameter is defined as
L ¼ hðT0  T∞ Þ=sE02 H. Therefore, for the fixed value of L¼0.4 used
in this figure (and also for any value of L), means that
hðT0  T∞ Þ  E02 H. Using this last condition into the definition of b,
we see easily that b~Hh/k, which corresponds to the usual Biot
number. Therefore, for increasing values of b, we obtain a stronger
convective effect and thus, the temperature and pressure profiles
both decay, as can be appreciate in Figs. 5 and 7, respectively. In
contraposition, the case of decreasing values of b with fixed values
of L¼0.4 corresponds to a stronger Joule effect.
Fig. 8 presents the dimensionless temperature as function of the
longitudinal coordinate c for different values of the dimensionless
parameter L(¼0.2,0.5,0.8). As the parameter L increases, the
dimensionless temperature profiles decrease, which means among
others, a greater heat dissipation through the microchannel wall,
and therefore lower values on the dimensionless temperature are
achieved when the condition of heat flux at the external surface of
the microchannel dominates on the Joule heating effect. These re-
sults are in agreement with those obtained by Ref. [10], who carried
out numerical and asymptotic analyses of the Joule heating effect
with variable properties on an electroosmotic flow of a non-
Newtonian fluid.
Finally, on Fig. 9, the dimensionless volumetric flow rate as
function of the parameter L is shown as function of five values of
the thermal parameter a. In all cases, the influence of the parameter
a is the same: for increasing values of that parameter, the corre-
sponding volumetric flow rates decrease for fixed values of L and
therefore, the influence of both parameters a and L regulate
simultaneously the reduction of the volumetric flow rates.
Although it is really difficult to compare directly the results of the
present work with other experimental or theoretical results,
because of the influence of the parameter a has not been previously
reported, we can appreciate that other authors have considered the
influence of other longitudinal effects on the electroosmotic flow
obtaining similar results. Special mention deserve the works of
Bahga et al. [20], Ghosh and Chakraborty [21] and Yariv and Dorf-
man [22], where we can find special applications of electroosmotic
flows affected also by axially varying surface conditions.
Fig. 9. Dimensionless volumetric flow rate as a function of the dimensionless
parameter L, for five values of the thermal parameter a¼0, 0.004, 0.006, 0.008 and
0.01. The values of the dimensionless parameters are: k ¼ 40, 2¼1000, ε¼0.01, Pe¼0.05
and b¼0.008.
 E.A. Ramos et al. / International Journal of Thermal Sciences 111 (2017) 499e510 509

6. Conclusions
In the present work, we have developed a regular perturbation
analysis to describe the non-isothermal electro-osmotic flow
circulating into a slit microchannel. The use of the dimensionless
thermal parameter a, which measures deviations of the tempera-
ture with respect to the isothermal case, is introduced. In practical
cases, this parameter normally assumes smaller values than unity.
However, and even for small values of this parameter, the de-
viations for the pressure and temperature fields are important
enough to have to be taken into account. Therefore, new longitu-
dinal temperature gradients appear into the momentum conser-
vation equations modifying drastically the velocity, pressure and
electric fields. In this manner, the explicit influence of this
parameter a determines that the volumetric flow rate also be
modified. However, these new temperature gradients are not suf-
ficient to regulate the volumetric flow rates because of two addi-
tional parameters can also modulate the intensity of these
gradients: L and b help us to be careful about the factors affecting
the determination of volumetric flow rate. In fact, these last pa-
vj
; (A.6)
rameters control the competition between Joule heating effect and
vY
the heat losses at the walls of the microchannel, such that both
longitudinal temperature gradients as transverse compete simul-
taneously for estimating the prediction of the main variables of
interest.
Acknowledgments

ndez acknowledge the support
A. Ramos, O. Bautista and F. Me
from PAPIIT-UNAM, under contract number IN112215.
Appendix A
We explain in brevity the origin of the fourth term of the right-
hand side of Eq. (16) and the last term of the right-hand side of Eq.
(17). From the relationship
p0 ¼ p  εk2 j2 =2 ¼ p  ε½ð2n∞ z2 e2 =εkB Tðx; yÞÞ1=2 2 j2 =2 and taking
into account that k depends directly on temperature, the pressure
gradients which appear into the momentum conservation equa-
tions (2)e(3) with the aid of above transformation, can be written
as:
obtain directly the set of Equations (16) And (17). In this manner, we
have illustrated the procedure to obtain these body forces arising
from temperature variations.
In this Appendix, we derive also the boundary condition given
by Eq. (24) for the surface charge. The central idea is that surface
charge s can be related with the Zeta potential z as was shown by
Grahame [32] by demanding that surface charge plus the charge of
the ions in the whole double layer, must be zero. The total charge in
RH
the double layer is 0 re dy; where the electric charge density of the
fluid is given by re ¼ 2z2 e2 n∞ j=kB Tðx; yÞ. Therefore, using the
above sum and the relationship for re, we get in dimensionless form
that,
Z1
Hs v2 j
¼ dY; (A.5)
εz vY 2
0
the above definite integral can easily be evaluated obtaining that,
Hs
¼ 
εz
Y¼1
where we have used the symmetric condition. In addition, we can
use the asymptotic expansions for j (under the Lubrication
approximation and small values of a), for obtaining that up to terms
of order a0 and with the aid of the zero order solution for j0;0,
Hs
¼ k: (A.7)
εz
In the above equation, we have assumed low potential at the
wall and obviously the linear Poisson-Boltzmann equation. The
details can be found in Grahame [32]. Therefore, a linear relation-
ship between the surface charge s and z, is established. In this
manner, we can use a zeta potential as a boundary condition at the
wall because for small values of the zeta potential, the relationship
between potential and surface charge is linear. Therefore, for a
given value of the surface charge, the above relationship guarantees
that the zeta potential is also given at the wall.
References

vp vp0 2z2 e2 n∞ vj z2 e2 n∞ 2 vT [1] Karniadakis G, Beskok A, Aluru N. Microflows and nanoflows. Springer; 2005.
¼ þ j  j (A.1) [2] Maynes D, Webb B. Fully developed electro-osmotic heat transfer in micro-
vx vx kB T vx kB T 2 vx channels. Int J Heat Mass Transf 2003;46:1359e69.
[3] Su J, Jian Y, Chang L. Thermally fully developed electroosmotic flow through a
and rectangular microchannel. Int J Heat Mass Transf 2012;55:6285e90.
[4] Horiuchi K, Dutta P. Joule heating effects in electroosmotically driven micro-
channel flows. Int J Heat Mass Transf 2004;47:3085e95.
vp vp0 2z2 e2 n∞ vj z2 e2 n∞ 2 vT
¼ þ j  j : (A.2) [5] Tang G, Yang C, Chai J, Gong H. Joule heating effect on electroosmotic flow and
vy vy kB T vy kB T 2 vy mass species transport in a microcapillary. Int J Heat Mass Transf 2004;47:
215e27.
Replacing the above equations into Eqs. (2) and (3), we obtain [6] Tang G, Yang C, Chai C, Gong H. Numerical analysis of the thermal effect on
that, electroosmotic flow and electrokinetic mass transport in microchannels. Anal
Chim Acta 2004;507:27e37.

 ! [7] Liao Q, Zhu X, Wen T. Thermal effects on mixed electro-osmotic and pressure-
vu vu vp0 z2 e2 n∞ 2 vT v2 u v2 u driven flows in triangle microchannels. Appl Therm Eng 2009;29:807e14.
r u þv ¼ þ j þh þ [8] S

anchez S, Me
ndez F, Martínez-Sua
stegui L, Bautista O. Asymptotic analysis
vx vy vx RT 2 vx vx2 vy2 for the conjugate heat transfer problem in an electro-osmotic flow with
temperature-dependent properties in a capillary. Int J Heat Mass Transf
2z2 e2 n∞ vf
þ j (A.3) 2012;55:8163e71.
RT vx [9] Escando
n J, Bautista O, Me
ndez F, Bautista E. Theoretical conjugate heat
transfer analysis in a parallel flat plate microchannel under electro-osmotic
and and pressure forces with a phan-thien-tanner fluid. Int J Therm Sci 2011;50:
1022e30.

 ! [10] Bautista O, Sa
nchez S, Arcos JC, Me
ndez F. Lubrication theory for electro-
vv vv vp0 v2 v v2 v z2 e2 n∞ 2 vT osmotic flow in a slit microchannel with the phan-thien and tanner model.
r u þv ¼ þh þ þ j (A.4)
vx vy vy vx2 vy2 RT 2 vy J Fluid Mech 2013;722:496e532.
[11] Vocale P, Geri M, Cattani L, Morini G, Spiga M. Electro-osmotic heat transfer in
elliptical microchannels under {H1} boundary condition. Int J Therm Sci
The above two equations when are written in dimensionless 2013;72:92e101.
form, with the aid of dimensionless variables (14), permits us to [12] Chakraborty R, Dey R, Chakraborty S. Thermal characteristics of
510 E.A. Ramos et al. / International Journal of Thermal Sciences 111 (2017) 499e510
electromagnetohydrodynamic flows in narrow channels with viscous dissi-
pation and joule heating under constant wall heat flux. Int J Heat Mass Transf
2013;67:1151e62.
[13] Escando
n J, Santiago F, Bautista O, Me
ndez F. Hydrodynamics and thermal
analysis of a mixed electromagnetohydrodynamic-pressure driven flow for
phanthientanner fluids in microchannel. Int J Therm Sci 2014;86:246e57.
[14] Matías A, Sa
nchez S, Me
ndez F, Bautista O. Influence of slip wall effect on a
non-isothermal electro-osmotic flow of a viscoelastic fluid. Int J Therm Sci
2015;98:352e63.
[15] Barrat J-L, Bocquet L. Influence of wetting properties on hydrodynamic
boundary conditions at a fluid/solid interface. Faraday Discuss 1999;112:
119e27.
[16] Bocquet L, Barrat J-L. Hydrodynamics within the electric double layer on
slipping surfaces. Phys Rev Lett 2004;93:257805.
[17] Tandon V, Kirby BJ. Zeta potential and electroosmotic mobility in microfluidic
devices fabricated from hydrophobic polymers: 2. slip and interfacial water
structure. Electrophoresis 2008;29:1102e14.
[18] Ge Z, Jin L, Yang C. Microfluidic concentration of sample solutes using joule
heating effects under a combined ac and dc electric field. Int J Heat Mass
Transf 2015;85:158165.
[19] Dey M, Joo SW. Joule heating induced interfacial instabilities in free-surface
electro-osmotic flows. Int J Heat Mass Transf 2015;87:295e302.
[20] Bahga SS, Vinogradova OI, Bazant MZ. Anisotropic electro-osmotic flow over
super-hydrophobic surfaces. J Fluid Mech 2010;644:245e55.
[21] Ghosh U, Chakraborty S. Electrokinetics over charge-modulated surfaces in
the presence of patterned wettability: role of the anisotropic streaming
potential. Phys Rev E 2013;88. 033001e1e033001e18.
[22] Yariv E, Dorfman KD. Electrophoretic transport through channels of periodi-
cally varying cross section. Phys Fluids 2007;19. 037101e1e037101e7.
[23] Xuan X, Li D. Analytical study of joule heating effects on electrokinetic
transportation in capillary electrophoresis. J Chromatogr A 2005;1064:
227e37.
[24] Xuan X, Sinton D, Li D. Thermal end effects on electroosmotic flow in a
capillary. Int J Heat Mass Transf 2004;47:3145e57.
[25] Saville DA. Electrokinetic effects with small particles. Ann Rev Fluid Mech
1977;9:321e37.
[26] Adjari A. Electro-osmosis on inhomogeneously charged surfaces. Phys Rev
Lett 1995;4:755e9.
[27] Adjari A. Transverse electrokinetic and microfluidic effects in micropatterned
channels: Lubrication analysis for slab geometries. Phys Rev E 2001;65:
016301.
[28] Pascall AJ, Squires TM. Electrokinetics at liquid/liquid interfaces. J Fluid Mech
2011;684:163e91.
[29] Masliyah J, Bhattacharjee S. Electrokinetic and colloid transport phenomena.
Wiley-Interscience; 2006.
[30] Rice CL, Whitehead R. Electrokinetic flow in a narrow cylindrical capillary.
J Phys Chem 1965;69:1922e35.
[31] Lin H, Storey BD, Santiago JG. A depth-average electrokinetic flow model for
shallow microchannels. J Fluid Mech 2008;608:43e70.
[32] Grahame DC. The electrical double layer and the theory of electrocapillarity.
Chem Rev 1947;41:441e501.