Applied Mathematical Modelling 39 (2015) 4337–4350

Contents lists available at ScienceDirect

Applied Mathematical Modelling

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

Combined electroosmosis-pressure driven flow and mixing

in a microchannel with surface heterogeneity
S. Bhattacharyya ⇑, Subrata Bera
Department of Mathematics, Indian Institute of Technology, Kharagpur 721302, India

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

Article history: We investigate the combined pressure-driven electroosmotic flow near a wall roughness in
Received 6 October 2013 the form of a rectangular block mounted on one wall of an infinitely long microchannel.
Received in revised form 14 November 2014 The insulated rectangular block has a constant surface potential which is different from
Accepted 19 December 2014
the surface potential of the remaining part of the channel walls. The characteristics for
Available online 30 December 2014
the electrokinetic flow are obtained by numerically solving the Navier–Stokes equations
coupled with the Nernst–Planck equation for ion transport and the Poisson equation for
Keywords:
electric field. Vortical flow develops above the block when the surface potential of the
Debye layer
Potential patch
block is in opposite sign to that of the surface potential of the channel. The strength of
Nernst–Planck equation the re-circulating vortex grows as the surface potential of the block increases. Vortical flow
Microvortex also depends on the Debye length when it is in the order of the channel height. The com-
Surface roughness bined effects due to the geometrical modulation of the channel wall and heterogeneity in
surface potential is found to produce a stronger vortex and hence a stronger mixing, as
compared with the effect of either of these. The recirculating vortex, which appears on
the upper face of the block, grows and the average electroosmotic velocity increases with
the increase of the electrolyte concentration. The recirculating vortex does not appear
when EDL is thick and the effect of over-potential of the block on electroosmosis is negli-
gible. The loss of momentum near the obstacle is compensated by the electrostatic force
near the EDL, which prevents flow separation upstream or downstream of the obstacle.
The mixing performance of the present configuration is compared with several other cases
of surface modulation. The impact of the imposed pressure gradient on the vortical flow
due to surface heterogeneity is analyzed.
Ó 2014 Elsevier Inc. All rights reserved.

1. Introduction

One prominent difference between fluid motions in a micro/ nano-channel and the flow in a macrochannel is strong fluid-
wall interactions in the former case. As the channel size decreases, the surface-to-volume ratio increases. Various properties
of the walls, such as surface roughness of the order of nanometer and/ or heterogeneity in surface potential can therefore
greatly affect the fluid motions in microfluidics. An electric double layer, consisting of a Stern layer and a diffuse layer, is
formed near a charged wall. Under the action of an electric field, tangential to the solid surface, the surplus ions in the
diffused double layer experiences a Columbic force and starts moving towards the electrode, which in turn results in an

⇑ Corresponding author. Tel.: +91 3222283640; fax: +91 3222255303.

E-mail address: somnath@maths.iitkgp.ernet.in (S. Bhattacharyya).

http://dx.doi.org/10.1016/j.apm.2014.12.050
0307-904X/Ó 2014 Elsevier Inc. All rights reserved.
4338 S. Bhattacharyya, S. Bera / Applied Mathematical Modelling 39 (2015) 4337–4350

electroosmotic flow (EOF) in a channel. Thus, an alteration of the surface property can greatly affect the electric double layer
characteristics and thereby the electroosmosis.
In a macroscopic fluidic device, turbulent flow can be generated to enhance a mixing process with an acceptable mixing
time period. However, most microfluidic systems are inherently limited to a low Reynolds number which corresponds to a
laminar flow. Because of this, mixing of fluid in miniaturized bio-analytical instruments is dominated by the diffusion effect.
Techniques based on different actuation methods to enhance micro-mixing have been discussed by several authors. These
techniques can be broadly classified in two categories namely, active mixing and passive mixing [1,2]. Passive mixers typ-
ically use particular channel geometry configurations to increase the interfacial area between the liquids to be mixed. The
active mixers introduce moving parts inside microchannels or apply an AC-electric field.
The electroosmotic flow strongly depends on the surface charge density or f-potential. The enhancement of mixing in
microfluidics has been demonstrated by several authors [3–7] through surface potential modulation. The motivation behind
the surface potential heterogeneity is to induce a transverse flow, which in turn enhances the interfacial area of the species
to be mixed. Bhattacharyya and Nayak [6] and Chen and Conlisk [7] demonstrated a technique to enhance mixing in micro-
fluidics by introducing a potential patch on the wall of a micro/nano channel. There they have cited most of the recent stud-
ies on improved electrokinetic mixing through surface potential heterogeneity. Jain and Nandakumar [8] studied the optimal
heterogeneous surface pattern to enhance micromixing. The EOF in a heterogeneous microchannel under a periodic electric
field was considered by Lin and Chen [9]. Foregoing studies suggest that if the patch potential is of opposite sign to that of the
channel wall potential, the flow separates from the channel wall and a recirculation vortex develops above the patch. Most of
the studies on EOF with surface potential heterogeneity, except [6,7], is based on the free-slip model. In the free-slip model,
the electrolyte is considered to be electrically neutral out side the EDL and the ion distribution is governed by the equilib-
rium Boltzmann distribution. The non-uniformity in wall potential of the channel creates a non-uniform fluid flow and hence
induces a pressure gradient everywhere in the flow. The advection effect and non-neutrality of the electrolyte near a poten-
tial patch makes the equations for fluid flow and ion transport coupled. Thus, the assumption of Boltzmann distribution for
ions, which neglects the effects of fluid convection and electromigration due to imposed field, may not be valid.
The surface roughness of the order of a few angstroms can have a big impact on flows in microfluidic or nanofluidic chan-
nels [10]. Generally, microchannel surfaces exhibit certain degrees of roughness generated by the manufacturing techniques
or adhesion of biological particles from the liquids. The height of the surface roughness usually ranges from a few nanome-
ters to micrometers. The geometric modulation of a channel wall to increase the interfacial area between the liquids to be
mixed was studied by Ramirez and Conlisk [11]. Datta and Ghosal [12] made an asymptotic analysis to study the transport of
a solute in a straight microchannel of axially variable cross-sectional shape. Hu et al. [13] have analyzed the effect of surface
roughness and grooved microchannels on electroosmotic flow. The effect of undulation of a charged surface on the electro-
osmotic flow near the surface is analyzed by Lin [14,15]. It is found from those studies that the undulation of surface may
increase the magnitude of flow due to the increment of the total fixed charges on undulated surfaces. Bhattacharyya and
Nayak [16] found that the surface modulation involving geometric and surface potential heterogeneity causes a stronger
convection effect in nanochannels. There they have compared the results based on the non-linear Nernst–Planck model with
the results obtained by considering the equilibrium Poisson–Boltzmann model and found a large discrepancy near the sur-
face modulation. The groove-based microchannel to promote mixing was considered by Jain et al. [17].
Microvortices have an advantage in species mixing in microdevices. In some situations, the appearance of vortices need to
be suppressed so as to avoid aggregation of suspended particles leading to the eventual jamming of the device. We have
investigated the control of vortical flow through an imposed pressure-driven flow. The EOF may be accompanied by a pres-
sure driven flow due to a pressure difference along the longitudinal direction of the channel. The pressure gradient may arise
due to a leakage or a difference in the column height between the reservoirs connected by the channel [18]. Separation of
non-neutral electrolytes in a charged microchannel by using a pressure gradient in combination of electroosmotic flow is
made by Dutta [19]. A combined pressure-driven and electroosmotic flow is used to separate ionic species in a nanofluidic
channel by Gillespie and Pennathur [20]. In a recent paper, Song et al. [21] have studied analytically the convection–diffusion
of an analyte in a rectangular channel under EOF and pressure driven flow. There they have considered the effect of both
negative as well as a positive pressure gradient. The effect of fluid inertia on EOF in a microchannel with nonuniform surface
potential is analyzed by Bhattacharyya and Bera [22] through the Nernst–Planck model.
We focus on the combined pressure and electrokinetic effects on the formation of vortical flow in a microchannel in the
vicinity of a wall mounted rectangular block of height d, a fraction of the channel height. The f-potential along the block is
considered to be of opposite sign to that of the f-potential of the homogeneous parts of the wall. The geometrical modulation
and surface potential heterogeneity makes the equations for fluid flow, electric field and ion distribution coupled. The char-
acteristics for the electrokinetic flow are obtained by numerically solving the Poisson equation, the Nernst–Planck equation,
and the Navier–Stokes equations, simultaneously.

2. Mathematical Model

We consider a long rectangular channel of height h and width W with h  W filled with an incompressible Newtonian
electrolyte of uniform permittivity (e ) and viscosity (l). An obstacle in the form of a rectangular block of length (l) and
height (d) is considered to be mounted along the lower wall of the channel (Fig. 1). The external applied electric field is
 S. Bhattacharyya, S. Bera / Applied Mathematical Modelling 39 (2015) 4337–4350 4339

Fig. 1. Schematic diagram of the microchannel with wall mounted block of potential heterogeneity when h ¼ 10 lm and E0 ¼ 104 V/m.

generated by the electrodes placed at the inlet and the outlet of the channel. The number K ¼ E0 h=/0 measures the strength
of the external electric field in non-dimensional form. The non-dimensional equation for the distribution of external poten-
tial (w) is governed by the following Laplace equation

r2 w ¼ 0 ð1Þ
The wall and all sides of the block are electrically insulated i.e., rw:n ¼ 0, where n is the unit outward normal. Far upstream
ðx ! 1Þ and downstream ðx ! 1Þ of the block, w approaches a linear function of x i.e., w ¼ Kx.
The electric field Eð¼ Ex ; Ey ; Ez Þ is determined by the superposition of the external electric field along with the induced
electric field developed due to the redistribution of ions. The total electric potential U can be written as
U ¼ w ðx; y; zÞ þ / ðx; y; zÞ, where / is the induced electric potential. The dimensional variables are denoted by an asterisk
(). The charge density qe is related to the electric field as
X
r  ðe EÞ ¼ e r2 U ¼ qe ¼ zi eni ð2Þ
i

Here, zi and ni are respectively, the valance and ionic concentration of the i type ion, e is the elementary electric charge and
e ¼ 0 r , where 0 is the electric permittivity of vacuum and r , the dielectric constant of the solution. The non-dimensional
equation for induced electric potential is given by
2
@2/ @2/ @2/ ðjhÞ
B2 2
þ 2 þ C2 2 ¼  ðg  f Þ ð3Þ
@x @y @z 2
Here, B ¼ h=l and C ¼ h=W. We consider a symmetric electrolyte of valance zi ¼ 1. We denote the non-dimensional concen-
tration of cation by g and anion by f. We scale electric potential U by /0 ð¼ kB T=eÞ and ionic concentration ni by the bulk
ionic concentration n0 , cartesian coordinates by ðl; h; WÞ. Here, kB is the Boltzmann constant and T is the absolute tempera-
1=2
ture of the solution. The parameter j ¼ ½ð2e2 n0 Þ=ðe kB TÞ is reciprocal of the characteristic EDL thickness (kd ) and
jh ¼ h=kd .
Based on the conservation of momentum and ion flux, the governing equations in non-dimensional form can be expressed as
@u @ v @w
B þ þC ¼0 ð4Þ
@x @y @z
  2 
@u @u @u @u @p ðjhÞ @w @/
ReHS þ ReHS Bu þv þ Cw ¼ B  þ Bðg  f Þ þ r2 u ð5Þ
@t @x @y @z @x 2K @x @x
  2 
@v @v @v @v @p ðjhÞ @w @/
ReHS þ ReHS Bu þv þ Cw ¼  þ ðg  f Þ þ r2 v ð6Þ
@t @x @y @z @y 2K @y @y
  2 
@w @w @w @w @p ðjhÞ @w @/
ReHS þ ReHS Bu þv þ Cw ¼ C  þ C ðg  f Þ þ r2 w ð7Þ
@t @x @y @z @z 2K @z @z
" #    
@g @2g @2g @2g @g @g @g @g @w @g @w @g @w
Pe  B2 2 þ 2 þ C 2 2 þ Pe Bu þ v þ Cw  B2 þ þ C2
@t @x @y @z @x @y @z @x @x @y @y @z @z
  2
@g @/ @g @/ @g @/ ðjhÞ
 B2 þ þ C2 þ g ðg  f Þ ¼ 0 ð8Þ
@x @x @y @y @z @z 2
4340 S. Bhattacharyya, S. Bera / Applied Mathematical Modelling 39 (2015) 4337–4350

" #    
@f @2f @2f @2f @f @f @f @f @w @f @w @f @w
Pe  B2 2 þ 2 þ C 2 2 þ Pe Bu þ v þ Cw þ B2 þ þ C2
@t @x @y @z @x @y @z @x @x @y @y @z @z
  2
@f @/ @f @/ @f @/ ðjhÞ
þ B2 þ þ C2  f ðg  f Þ ¼ 0 ð9Þ
@x @x @y @y @z @z 2
2 2 2 @ 2
where r2 ¼ B2 @x@
2 þ @y2 þ C @z2 . We scale the velocity field by the Helmholtz–Smoluchowski velocity U HS ð¼ e E0 /0 =l) and
@


time t by h=U HS . Here, pressure is scaled by lU HS =h. The Reynolds number based on U HS is defined as ReHS ¼ U HS h=m, Schmidt
number Sc ¼ m=Di, Peclet number Pe ¼ ReHS Sc. Here, Di is the diffusivity of the i type species and kinematic viscosity m ¼ l=q.
The Reynolds number (Re) is 0:15  102 for pure EOF case ðG ¼ 0Þ when f ¼ 1:0 and h ¼ 10 lm.
The width (W) of the channel is considered to be on the order of the length of the channel. So, C ¼ h=W  1. Thus, all
gradients with respect to z can be neglected and the flow can be treated as two-dimensional. The present analysis can also
be interpreted as EOF within a channel of width h and height W h, as the gravitational and buoyancy effects are neglected
here.
Along the solid walls of the channel we assume a no-slip condition and ion impermeability i.e.,
u ¼ v ¼ 0; / ¼ f; Ni  n ¼ 0 ð10Þ
where n is the unit outward normal vector. Walls are assumed to be at constant f-potential with f ¼ fp along the surface of
the block. The over-potential of the block ð0:5 6 x
0:5; 0 6 y 6 0:2Þ is defined as /p ¼ fp  f. The wide microchannel is
assumed to be sufficiently long upstream and downstream of the block and the flow is assumed to be fully developed
EOF at the far upstream and downstream of the block. In the far upstream and downstream of the block, the conditions
are governed by the fully developed 1-D combined pressure-electroosmotically driven flow i.e.,
in
u ¼ uin ; v ¼ 0; / ¼ /in ; g ¼ g in ; f ¼f ð11Þ
in
The form of uin ; /in ; g in and f are governed by the combined effect of a constant pressure gradient G and constant surface
potential f. The inlet and outlet conditions can be derived in a similar manner as described in Bera and Bhattacharyya
[23]. In determining the inlet and outlet boundary conditions, the pressure gradient is assumed to be constant and is deter-
mined by the non-dimensional quantity,
2
,
1 dp h
G¼ U HS ð12Þ
l dx 2
The average non-dimensional speed due to a pressure driven flow is G=6 corresponding to an imposed pressure gradient G.
We introduce the effective Reynolds number Re, based on the average incoming flow as Re ¼ ð1 þ G=6ÞReHS .

2.1. Mass Transport equation

The electrokinetic transport of uncharged sample species is considered in the present analysis. In absence of no chemical
reaction or absorption of species on the wall, the transport of species are governed by convection and diffusion. The govern-
ing equation for species transport in non dimensional form is
@c 1 2
þ ðq:rÞc ¼ r c ð13Þ
@t Pes
where the advection speed is q. Here c is the dimensionless species concentration, scaled by the reference concentration c0
and Ds is the diffusion coefficient of the species. In this case, the Peclet number is based on sample diffusivity Ds and is
defined as Pes ¼ U HS h=Ds . No mass flux is assumed along the channel wall. Thus, the boundary conditions of C are as follows:
@c @c
c ¼ cin at x ¼ L; ¼ 0 at x ¼ L and ¼ 0 at y ¼ 0; 1
@x @y
where the value for cin is specified as cin ¼ 1 on the lower inlet (0 6 y 6 0:5) and cin ¼ 0 on the upper inlet (0:5 < y 6 1) of the
channel. The non-dimensional channel length L is taken to be sufficiently long so as to establish the sample concentration
independent of x.

3. Numerical methods

We solved the coupled set of governing non-linear equations for fluid flow and ionic species concentration through a
finite volume method on a staggered grid system. In the staggered grid system, the scalar quantities are evaluated at each
cell center and the velocity components are evaluated at the midpoint of the cell sides to which they are normal. The dis-
cretized form of the governing equations is obtained by integrating the governing equations over each control volumes. Dif-
ferent control volumes are used to integrate different equations. Near the block, both the imposed and induced electric
potential undergo a sharp change. In order to capture the sharp change in variable values accurately, we use the higher-order
 S. Bhattacharyya, S. Bera / Applied Mathematical Modelling 39 (2015) 4337–4350 4341

upwind scheme, QUICK (Quadratic Upwind Interpolation Convective Kinematics) to discretize the convective and electromi-
gration terms in both concentration and Navier Stokes equations. The QUICK scheme uses a quadratic interpolation/ exter-
polation between the three nodal values of variables to estimate its value at the interface of the control volume. The upwind
scheme imparts stability to the numerical solution in the region where a steep gradient in variables occur. An implicit first-
order scheme is used for discretising the time derivative terms. The resulting discretized equations are solved iteratively
through the pressure correction based iterative algorithm SIMPLE [24]. The iteration starts by assuming the induced electric
potential / at every cell center.
We considered a non-uniform grid spacing along y-direction and uniform grids along the other axis (Fig. 2(b)) and dt was
taken as 0:001. To check the effects of grid spacing, computations have been performed for three different form of grids
namely, Grid 1: 200  240, Grid 2: 400  240 and Grid 3: 400  500 for EOF with-out pressure gradient (G ¼ 0) and com-
pared with the results due to Mirbozorgi et al. [25]. In Grid 1 and Grid 2, we considered dy to vary between 0:005 to 0:01
with dx is either 0:02 (for Grid 1) or dx ¼ 0:01 (for Grid 2). In Grid 3, we considered dx ¼ 0:01 and 0:0025 6 dy 6 0:005.
Fig. 2(a) and (b) suggests that the results obtained by Grid 2 and Grid 3 agree fairly well with each other and these results
are in close agreement with the result due to [25]. Thus, we find that Grid 2 is optimal. We made several test runs to place the
upstream and downstream boundaries and found that the variables become independent of x and approaches the fully
developed EOF at a distance 4l from the center of the block. This distance is found to be larger when two side-by-side
blocks are considered.

4. Results and discussions

The height of the rectangular block could be any fraction of the channel height. We present results in which the length is
the same as the channel height i.e., h ¼ l. The f-potential along the surface of the block is considered to be of opposite sign to
that of the f-potential on the homogeneous walls of the channel. The imposed electric field is E0 ¼ 104 V/m and the channel
height 10 lm. The ionic concentration of the electrolyte is varied from 107 M to 105 M so that the Debye–Huckel param-
eter jh varies between 10:38 to 103:8. Beyond this range of jh, the EOF within a plane channel becomes independent of the
Debye length. Several authors e.g., Jain et al. [26] restricted their attention to analyze the micromixing for jh ¼ 100 with
channel width h ¼ 10 lm. In the present analysis, we consider viscosity l ¼ 0:001 kg/m s, density q ¼ 1000 kg/m3 , Faraday
constant F ¼ 96,500 C mol1 and gas constant R ¼ 8:315 J/mol K at temperature T ¼ 300 K. We assumed the diffusion coef-
ficient of ions are the same as Dþ ¼ D ¼ 1:3  1010 m2 =s, thus Sc = 7692.31. The form of the imposed electric field and
potential distribution are presented in Fig. 3.

4.1. Effect of Debye-Hückel parameter on EOF with G = 0

We present the streamline patterns for various values of the Debye layer thickness i.e., jh ¼ 15; 40; 100 when h ¼ 10l in
absence of the imposed pressure gradient i.e., G ¼ 0 in Fig. 4(b)–(d). The f-potential along the block is considered to be 1 and
is 1 along the rest of the channel surface. For the sake of comparison, we have included the streamlines (Fig. 4(a)) when the
f-potential of the block is the same as that of the channel walls i.e., fp ¼ 1. The flow field close to the block is two dimen-
sional. However, the streamlines shows a parallel flow at the far upstream or downstream of the block. We find that the

Fig. 2. Comparison of the present NP-model at far upstream and downstream of the block or patch i.e., (x ¼ 4l) with Mirbozorgi et al. [25] and the effects
of grid size for fully developed EOF with G ¼ 0 when channel height ðhÞ is 10 lm; kd ¼ 0:66 lm ði:e:; jh ¼ 15Þ; f ¼ 1 and Re = 0.02. (a) Velocity; (b) ionic
concentration.
4342 S. Bhattacharyya, S. Bera / Applied Mathematical Modelling 39 (2015) 4337–4350

Fig. 3. Distribution of the applied electric field and potential near the block within the microchannel when h ¼ 10 lm and E0 ¼ 104 V/m. (a) Electric field
lines; (b) potential lines.

recirculation zone appears above the block for higher values of jh and this zone expands with the increase of jh. Due to the
geometric modulation, the flow accelerates in the region above the block and an induced pressure gradient develops due to
the non-uniformity of the fluid velocity. When a potential patch along the block is considered, an EDL of opposite nature will
develop in the vicinity of the surface of the block. As the fluid approaches the EDL on the block, it experiences an electrostatic
force which is along the opposite direction to the incoming flow. This electrostatic force is strong within the EDL where
charge density is high. The negative body force adjacent to the block induces a flow separation and a recirculation zone
on the upper face of the block. The electrostatic force grows with the rise of ionic concentration of electrolyte i.e., increase
of jh. For sufficiently large jhð 1Þ i.e., thin EDL, the vortical flow becomes independent of jh as the fluid outside the thin
EDL becomes electrically neutral. The retarding electrostatic force for lower values of jh is not strong enough to induce a
flow separation from the upper surface of the block. Our computation suggests that for jh < 10, no flow separation above
the block occurs. The streamline patterns in Fig. 4 shows a vortical flow near the block when block f-potential is of opposite
sign to that of the homogeneous parts of the channel wall. It may be noted that in a pressure driven flow, separation occurs
from the lower wall of the channel downstream of the block at any values of Reynolds number. In EOF, the electrostatic body
force compensate the momentum loss near the block and prevents the flow separation.
The geometrical modulation of channel wall and surface potential heterogeneity creates an induced pressure field. We
present the distribution of cross-sectional averaged pressure (pav g ) near the block for different values of EDL thickness in
Fig. 5. We find that as the fluid approaches the block, it experiences an adverse pressure gradient and the flow separation
near the upstream top corner of the block occurs. A large pressure drop across the block is observed for all the cases. A favor-
able pressure gradient above the block induces an acceleration in fluid. The pressure remains constant at far upstream and
down stream of the block.
The size of the vortex, which develops above the block, as a function of jh is illustrated in Fig. 6(a). The size of the vortex
at different fp based on the free-slip model is also indicated in Fig. 6(a). The vortex size grows with the increase of jh and
achieves a constant value at large jh. Our result shows that the EOF in the present case depends on the EDL thickness for
finite values of jhð< 40Þ. At sufficiently low jh, the charge density within the EDL is not strong enough to produce a retard-
ing electrostatic force to overcome the fluid inertia and create a flow separation. On the other hand, at large jhðP 40Þ the
flow field outside the EDL is electrically neutral and the form of the bulk flow becomes independent of EDL. For this, we find
that the vortical flow becomes independent of jh when jh P 40. We find from Fig. 6(b) that the vortex size expands with
the increase of over-potential of the block. When the over-potential of the block is increased, more number of counter ions
accumulate in the EDL and hence the charge density of the EDL grows. Thus the increase in fp induces a stronger vortex for
any values of the EDL thickness (Fig. 6(a)). Fig. 6(b) shows that at large jh, the vortex size increases linearly with the increase
of over-potential of the patch. However, the relation between the vortex size and patch potential is not linear when EDL
effect on the bulk flow is not negligible i.e, jh ¼ 15. The size of the vortex at different fp based on the free-slip model is also
indicated in Fig. 6(a).
When we consider a thin EDL i.e., at large jh, the net charge density can be considered to be zero outside the EDL. In this
case, the electrokinetic interactions outside the EDL is negligible and the bulk flow is governed by the viscous stress and
induced pressure gradient. Ions convect parallel to the wall i.e., along the direction of the imposed electric field. A slip model
based on the Helmholtz–Smollochwoski slip velocity at the edge of the EDL is valid to analyze the EOF. However, for finite
values of EDL thickness, the double layer influences the ion transport in the bulk region and the electrostatic forces are non-
negligible outside the EDL. The convection effects on ion distribution above the block for jh < 100 is evident from the
 S. Bhattacharyya, S. Bera / Applied Mathematical Modelling 39 (2015) 4337–4350 4343

Fig. 4. Streamlines near the block (a) with homogeneous f-potential /p ¼ 0 (i.e., fp ¼ 1) when jh ¼ 100; (b) Heterogeneous f-potential i.e., /p ¼ 2
(fp ¼ 1) when jh ¼ 15; (c) /p ¼ 2 (fp ¼ 1) when jh ¼ 40; (d) /p ¼ 2 (fp ¼ 1) when jh ¼ 100; Here, G ¼ 0; h ¼ 10 lm, f ¼ 1 and E0 ¼ 104 V/m.

Fig. 5. Distribution of cross-sectional averaged pressure along the channel length for different values of jhð¼ 15; 40; 100Þ for EOF with G ¼ 0. Here,
h ¼ 10 lm, f ¼ 1; fp ¼ 1 (i.e., /p ¼ 2) and E0 ¼ 104 V/m. Here, the characteristic pressure is p0 ¼ lU HS =h ¼ 0:018 Pa.

results. This shows that the equilibrium Boltzmann distribution of ions for moderate values of EDL thickness is not valid. A
slight non-zero value of the net charge density i.e., ðg  f Þ may lead to a big change in the induced potential ð/Þ by virtue of
Eq. (3). We present the distribution of induced potential ð/Þ for two values of jh ¼ ð15; 100Þ in Fig. 7(a) and (b). The
4344 S. Bhattacharyya, S. Bera / Applied Mathematical Modelling 39 (2015) 4337–4350

Fig. 6. Variation of the height of vortex center (hp ) as a function of (a) jh at different patch potential (fp ¼ 1; 2). The dashed lines in represents the result
based on free-slip model; (b) fp when jh ¼ 15 and 100 for EOF with G ¼ 0. Here, h ¼ 10 lm, f ¼ 1 and E0 ¼ 104 V/m.

formation of a double layer on the solid walls of the channel is evident from these results. The equipotential lines are dis-
torted in the region above the block. This distribution is large for thick EDL (low jh) cases, where the region is electrically
non-neutral. At large jhð¼ 100Þ, the equi-potential lines are almost parallel to the x-axis, which implies that the induced
electric field is along the direction of normal to the wall.

4.2. Comparison with free-slip model

We now compare our computed solutions for different values of the Debye layer thickness with the solution obtained by
the free-slip model. In the free-slip model, it is assumed that the Debye layer is thin and the fluid is electrically neutral out-
side of it. A slip velocity condition based on the Smoluchowski velocity is imposed at the outer edge of the Debye layer. The
electric field lines are assumed to be tangential to the EDL. We have imposed u =U HS ¼ 1 on the homogeneous parts of the
channel wall and along the surface of the block (0:5 6 x
0:5; 0 6 y 6 0:2) we use, q  t ¼ fp E  t. where t is tangent along
the surface of the block and q is the velocity vector. The Navier–Stokes equations are solved with the above boundary con-
ditions. Comparison of the velocity profiles at x ¼ 0 is presented in Fig. 8(a) and (b). We find that there exists a difference
between the results based on the present model and the free-slip model, particularly for a thick EDL. The free-slip model
over-predicts the result based on the present non-linear model. The result for jh ¼ 100 is close to the free-slip model with
a maximum percentage difference between the streamwise velocity being 16:5%. We have already seen that the EOF far

Fig. 7. Distribution of induced potential ð/Þ in the vicinity of the block when (a) jh ¼ 15 and (b) jh ¼ 100 for EOF with G ¼ 0. Here, h ¼ 10 lm,
f ¼ 1; fp ¼ 1 (i.e., /p ¼ 2) and E0 ¼ 104 V/m.
 S. Bhattacharyya, S. Bera / Applied Mathematical Modelling 39 (2015) 4337–4350 4345

downstream and upstream of the block resembles the slip model. In the vicinity of the block, the convection effect and the
effect of induced field become prominent in the ion distribution. These effects are neglected in the free-slip model. In
Fig. 6(a) we have shown the comparison of vortex size which sets in above the block for different values of the EDL thickness
with the vortex size estimated via the free-slip model. The free-slip model over-predicts the vortex size. However, at large
values of jh, the computed values based on the present model approaches the vortex size based on the free-slip model.
The electric current density (j) is defined as
X       
ni ni n
j¼F zi Ni ¼ j0 Rzi r þ Rz2i rU þ Pe qRzi i ð14Þ
i
n0 n0 n0

where j0 ð¼ FDi n0 =h) is the characteristic electric current density and q is the non-dimensional velocity. The average current
density is
ZZ
1
J¼ jx dxdy ð15Þ
A A

where jx is the x-component of the vector j and A is the area of the portion of the channel considered in the present com-
putation. The average current density based on a free-slip model is defined as J Slip . Fig. 9 shows that the average current den-
sity in this microchannel deviates from the current density based on the free-slip model for moderate values of jh. At large
values of jhð> 40Þ, the average current density approaches the value which is obtained by the free-slip model.

4.3. Effect of Reynolds number

We now consider the effect of an imposed pressure gradient on the EOF within a microchannel with surface roughness
and heterogeneity in wall potential. In presence of a favorable pressure gradient, the flow separation and formation of eddy
above the block gets delayed with respect to the patch potential. The extra momentum generated through the imposed pres-
sure gradient will oppose the negative electroostatic force along the top face of the block. Fig. 10 depicts the critical value of
the f-potential for which the recirculation zone appears as a function of the Reynolds number. Here the Reynolds number is
varied by varying the imposed axial pressure gradient (G). The critical value of the patch potential is found to be higher than
the case in which the EOF is considered over a potential patch (absence of block i.e., d ¼ 0). The strong induced favorable
pressure gradient above the block compensates the momentum loss due to the opposing EDL. For this, at a fixed Reynolds
number, the over-potential of the patch for the onset of vortex becomes higher than the case where geometric modulation is
neglected.
Fig. 11(a)–(c) shows the streamline patterns for different choices of the imposed pressure gradient ðG ¼ 2; 0; 2Þ. We find
that the vortex expands when the imposed pressure gradient acts in the adverse direction of the EOF. It shrinks in presence
of a favorable pressure gradient. We find that a small amount of imposed pressure gradient has a big impact on the EOF near
the block. At moderate values of Re, the mixed EOF will be dominated by the fluid inertia effect and the flow field resembles
the plane Poiseuille flow.
Fig. 12 provides an estimation for the height of the vortex center as a function of the imposed pressure gradient (G) for
different values of the EDL thickness. It is evident from this result that the vortex shrinks as the fluid Reynolds number
increases which means the increase of the imposed favorable pressure gradient. We find that the vortex size becomes

Fig. 8. Variation of (a) u-velocity and (b) v-velocity on the top of the block at x ¼ 0 for different values of jhð¼ 15; 40; 100Þ for EOF with G ¼ 0. Here
h ¼ 10 lm, f ¼ 1; fp ¼ 1 (i.e., /p ¼ 2) and E0 ¼ 104 V/m. Dotted lines corresponds to the result based on slip model. Arrow points the increasing order of
jh.
4346 S. Bhattacharyya, S. Bera / Applied Mathematical Modelling 39 (2015) 4337–4350

Fig. 9. Distribution of the scaled average current ðJ=JSlip Þ with jh for EOF with G ¼ 0 when h ¼ l ¼ 10 lm, f ¼ 1; fp ¼ 1 (i.e., /p ¼ 2) and E0 ¼ 104 V/m.

Fig. 10. Critical value of the patch potential ðfp Þ for the on-set of vortex at different Reynolds number (Re). Here h ¼ 10 lm, jh ¼ 15; f ¼ 1 and E0 ¼ 104 V/m.

independent of the Debye layer thickness when jh is bigger than 40. Our results show that a favorable pressure gradient of
0:036 kPa/m (G ¼ 1) produces 6:8% reduction of vortex size. We find that the vortex size at any values of Reynolds number
does not vary with the rise of jh when jh P 40. This implies that the imposed pressure gradient does not influence the elec-
tro-neutrality of the electrolyte for moderate values of jh i.e., when EDL is thin.

4.4. Mixing of species

The mixing of solute in a plane microchannel is dominated by molecular diffusion which is a slow process and the
required mixing length therefore becomes long. The transverse EOF can enhance mixing performance by stretching and fold-
ing the fluids. In Fig. 13(a) we present the concentration profile along the cross-section which is at a distance 4l downstream
of the channel center (Fig. 1). The perfect mixing of solute should imply that concentration profile is close to 0:5. We present
the concentration for different EOF configurations. It is evident that the mixing of solute is high for the present configuration.
In Fig. 13(b) we present the estimated dimensional time required to achieve the steady concentration profile at far down-
stream. It is evident from the results that the surface modulation leads to a lower time required to achieve a steady-state
for solute distribution. Steady-state is achieved faster when the advection effect is raised.
We have quantified the mixing index g [8] as
2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 3
1
PN 2
ðc  c Þ
g¼6 7
N i¼1 i m
41:0  qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
PN 5  100% ð16Þ
1 2
N i¼1 0ðc  c m Þ

where the variable ci is the concentration at data point i, while c0 and cm are the concentration at each point if the solution
are unmixed and the concentration with perfect mixing (i.e., 0:5), respectively. Here, N is the total number of points in the
cross-section and the variable c0 takes on a value of 0 or 1 at any point across the channel cross-section.
 S. Bhattacharyya, S. Bera / Applied Mathematical Modelling 39 (2015) 4337–4350 4347

Fig. 11. Streamlines for different values of Re when h ¼ 10 lm, jh ¼ 40, f ¼ 1; /p ¼ 2 (i.e., fp ¼ 1) and E0 ¼ 104 V/m. (a) Re ¼ 0:09  102 ðG ¼ 2Þ; (b) for
EOF, Re ¼ 0:15  102 ðG ¼ 0Þ; (c) Re ¼ 0:21  102 ðG ¼ 2Þ.

The variation of mixing index along the channel length is shown in Fig. 14(a) for different flow configurations with G ¼ 0.
We demonstrate results for the present configuration i.e., a wall mounted block with heterogeneous potential and compared
it with the cases of a channel with potential patch, wall mounted block with homogeneous potential and side-by-side blocks.
We have also obtained results for mixing efficiency within a plane channel. The presence of a potential patch and =or surface
roughness have no impact on mixing at the upstream. However, mixing efficiency is enhanced when the combined effects of
step change in f-potential and surface roughness is considered. Our results show the present configuration i.e., a block with
surface potential heterogeneity produces a 10% mixing enhancement as compared to a plane channel and 4% enhancement is
found when the channel is patterned with a potential patch and wall mounted block with homogeneous potential (/p ¼ 0)
producing a 5% mixing efficiency. It is evident from Fig. 14(a) that the two side-by-side blocks produce a marginal enhance-
ment in mixing efficiency (1%) as compared to the present configuration. Thus, a wall mounted block with surface potential
heterogeneity is found to be an optimal choice for enhanced mixing. Fig. 14(a) shows that a 90% mixing index in a micro-
channel is possible only through the combined geometric and surface potential modulation. It is also clear from Fig. 14(a)
that the mixing index attains a constant value at a lower length when surface modulation is considered, which implies that
the mixing length reduces due to the surface modulation. In Fig. 14(b), we present the mixing efficiency of present model for
various values of the block height (d) with over-potential /p ¼ 4. The increase in block height produces an enhanced mixing.
4348 S. Bhattacharyya, S. Bera / Applied Mathematical Modelling 39 (2015) 4337–4350

Fig. 12. Variation of height of the vortex center from the lower wall of the channel with the imposed pressure gradient (G) for different values of EDL
thickness ðkd ¼ 0:6 lm; 0:3 lm; 0:1 lmÞ when h ¼ 10 lm, f ¼ 1:0; /p ¼ 2 (i.e., fp ¼ 1:0) and E0 ¼ 104 V/m.

Fig. 13. (a) Profiles for non dimensional solute concentration when G ¼ 0 for different configurations namely, plane channel (d ¼ 0; lb ¼ 0; /p ¼ 0), with
patch (d ¼ 0; lb ¼ 0; /p ¼ 4), single block (d ¼ 0:2; lb ¼ 1; /p ¼ 4), two blocks (d ¼ 0:2; lb ¼ 2; /p ¼ 4). Consult the online version for plots in color codes; (b)
Variation of transient time for solute distribution with (G) for wall mounted block and plane channel. Here, h ¼ 10 lm, jh ¼ 100, f ¼ 2 and E0 ¼ 104 V/m.

Fig. 14. Variation of mixing index ðgÞ when G ¼ 0 along the channel length for (a) different configurations (b) present configurations for different block
height (d ¼ 0:1  0:5) with /p ¼ 4. Here, jh ¼ 100, h ¼ 10 lm, f ¼ 2 and E0 ¼ 104 V/m.
 S. Bhattacharyya, S. Bera / Applied Mathematical Modelling 39 (2015) 4337–4350 4349

Fig. 15. Mixing index ðgÞ with pressure gradient (G) for different configurations when jh ¼ 100. Here, h ¼ 10 lm, f ¼ 2 and E0 ¼ 104 V/m. The
corresponding Reynolds number varies between 0:3  102 ðG ¼ 0Þ to 0:9  102 ðG ¼ 10Þ.

The effect of pressure gradient on the mixing efficiency far downstream is shown in Fig. 15 for several flow configurations.
The mixing efficiency decreases monotonically with the increase of an imposed pressure gradient. The advection effect in
solute transport rises with the increase in Reynolds number (increment in Peclet number), which leads to a reduced mixing
efficiency in all the cases considered.

5. Conclusions

The combined pressure-driven electroosmotic flow and mixing near a wall roughness in the form of a rectangular block
mounted on one wall of an infinitely long microchannel is studied. The insulated rectangular block has a constant surface
potential which is different from the surface potential of the remaining part of the channel walls. The characteristics for
the electrokinetic flow are obtained by numerically solving the Poisson equation, the Nernst–Planck equations, and the
Navier–Stokes equations in a coupled manner. The mixing performance is quantified through the dispersion of the solute
distribution.
A vortical flow developed above the block when the surface potential of the block is in opposite sign to that of the surface
potential of the channel. The strength of the re-circulating vortex grows as the surface potential of the block increases. Vor-
tical flow also depends on the Debye length when Debye layer thickness is in the order of the channel height. The recircu-
lating vortex, which appears on the upper face of the block grows and the average electroosmotic velocity increases with the
increase of the electrolyte concentration. We find that the results based on the free-slip model over-predicts the results due
to the present model for low to moderate range of Debye layer thickness, however, the two models merge for a sufficiently
small value of the Debye layer thickness. Control of vortical flow through an imposed pressure gradient is analyzed.
Our results suggest that the mixing performance in a microchannel is improved due to the combined effect of geometric
modification and surface potential heterogeneity. The enhancement in mixing index for this combined modification is found
to be almost double as compared to the cither of these surface modulation effects.

Acknowledgement

The authors wish to thank the Council of Scientific & Industrial Research, Govt. India for providing the financial support
through the project grant 26(0202)/12/EMR-II.

References

[1] F. Tian, B. Li, D.Y. Kwok, Tradeoff between mixing and transport for electroosmotic flow in heterogeneous microchannels with nonuniform surface
potentials, Langmuir 21 (2005) 1126–1131.
[2] S. Liu, P.K. Dasgupta, Electroosmotically pumped capillary flow-injection analysis valve-based injection systems and sample throughput, Anal. Chim.
Acta 283 (1993) 739–745.
[3] A. Ajdri, Electro-osmosis on inhomogeniously charged surfaces, Phys. Rev. Lett. 75 (1995) 4.
[4] H.-Y. Wu, C.-H. Liu, A novel electrokinetic micromixer, Sensor Actuat. A–phys. 118 (2005) 107–115.
[5] A.E. Herr, J.I. Molho, J.G. Santiago, M.G. Mungal, T.W. Kenny, Electroosmotic capillary flow with nonuniform zeta potential, Anal. Chem. 72 (2000)
1053–1057.
[6] S. Bhattacharyya, A.K. Nayak, Electroosmotic flow in micro/nanochannels with surface potential heterogeneity: an analysis through the Nernst Planck
model with convection effect, Colloids Surf. A 339 (2009) 167–177.
[7] L. Chen, A.T. Conlisk, Effect of nonuniform surface potential on electroosmotic flow at large applied electric field strength, Biomed. Microdevices 11
(2009) 251–258.
4350 S. Bhattacharyya, S. Bera / Applied Mathematical Modelling 39 (2015) 4337–4350

[8] M. Jain, K. Nandakumar, Optimal patterning of heterogeneous surface charge for improved electrokinetic micromixing, Comput. Chem. Eng. 49 (2013)
18–24.
[9] T.-Y. Lin, C.-L. Chen, Analysis of electroosmotic flow with periodic electric and pressure fields via the lattice PoissonBoltzmann method, Appl. Math.
Model. 37 (2013) 2816–2869.
[10] H. Wätzig, S. Kaupp, M. Graf, Inner surface properties of capillaries for electrophoresis, Trends Anal. Chem. 22 (2003) 588–604.
[11] J.C. Ramirez, A.T. Conlisk, Formation of vortices near abrupt nano-channel height changes in electro-osmotic flow of aqueous solutions, Biomed.
Microdevices 8 (2006) 325330.
[12] S. Datta, S. Ghosal, Dispersion due to wall interaction in microfludic separation systems, Phys. Fluids 20 (2008) 012103.
[13] Y.D. Hu, C. Werner, D.Q. Li, Influence of the three-dimensional heterogeneous roughness on electrokinetic transport in microchannels, J. Colloid
Interface Sci. 280 (2004) 527–536.
[14] S.-H. Lin, Electrokinetic flow near an undulated, charged surface, Colloids Surf. B, Biointerfaces 81 (2010) 224–234.
[15] S.-H. Lin, Charged layer with undulated surface: configuration of electrical double layer, Colloids Surf. B, Biointerfaces 78 (2010) 127–139.
[16] S. Bhattacharyya, A.K. Nayak, Combined effect of surface roughness and heterogeneity of wall potential on electroosmosis in microfluidic/nanofuidic
channels, J. Fluid Eng. – ASME 132 (2010) 041103.
[17] M. Jain, A. Rao, K. Nandakumar, Numerical study on shape optimization of groove micromixers, Microfluid Nanofluid 15 (2013) 689–699.
[18] P. Dutta, A. Beskok, Analytical solution of combined electroosmotic/pressure driven flow in two dimensional straight channels: finite Debye Layer
effects, Anal. Chem. 73 (2001) 1979–1986.
[19] D. Dutta, A numerical analysis of nanofluidic charge based separations using a combination of electrokinetic and hydrodynamic flows, Chem. Eng. Sci.
93 (2013) 124–130.
[20] D. Gillespie, Separation of ions in nanofluidic channels with combined pressure-driven and electro-osmotic flow, Anal. Chem. 85 (2013) 2991–2998.
[21] H. Song, Yi. Wang, K. Pant, Scaling law for cross-stream diffusion in microchannels under combined electroosmotic and pressure driven flow,
Microfluid Nanofluid 14 (2013) 371–382.
[22] S. Bhattacharyya, S. Bera, Nonlinear electroosmosis pressure-driven flow in a wide microchannel with patchwise surface heterogeneity, J. Fluid Eng. –
ASME 135 (2013) 02130.
[23] S. Bera, S. Bhattacharyya, On mixed electroosmotic-pressure driven flow and mass transport in microchannels, Int. J. Eng. Sci. 62 (2013) 165–176.
[24] C.A.J. Fletcher, Computational Techniques for Fluid Dynamics, Springer Ser. Comput. Phys., second ed., vol. I–II, Springer-Berlin, Heidelberg, New York,
1991.
[25] S.A. Mirbozorgi, H. Niazmand, M. Renkrizbulut, Electroosmotic flow in reservoir-connected flat microchannels with non-uniform zeta potential, J. Fluid
Eng. – ASME. 128 (2006) 1133–1143.
[26] M. Jain, A. Yeung, K. Nandakumar, Efficient micromixing using induced-charge electroosmosis, J. Microelectromech. Syst. 18 (2009) 376–384.