You are on page 1of 20

Journal of Colloid and Interface Science 275 (2004) 679–698

www.elsevier.com/locate/jcis

Frequency-dependent laminar electroosmotic flow in a closed-end


rectangular microchannel
Marcos,a C. Yang,a,∗ K.T. Ooi,a T.N. Wong,a and J.H. Masliyah b
a School of Mechanical and Production Engineering, Nanyang Technological University, Republic of Singapore 639798
b Department of Chemical and Materials Engineering, University of Alberta, Edmonton, AB, Canada T6G 2G6

Received 25 November 2003; accepted 4 March 2004


Available online 9 April 2004

Abstract
This article presents an analysis of the frequency- and time-dependent electroosmotic flow in a closed-end rectangular microchannel. An
exact solution to the modified Navier–Stokes equation governing the ac electroosmotic flow field is obtained by using the Green’s function
formulation in combination with a complex variable approach. An analytical expression for the induced backpressure gradient is derived.
With the Debye–Hückel approximation, the electrical double-layer potential distribution in the channel is obtained by analytically solving
the linearized two-dimensional Poisson–Boltzmann equation. Since the counterparts of the flow rate and the electrical current are shown to
be linearly proportional to the applied electric field and the pressure gradient, Onsager’s principle of reciprocity is demonstrated for transient
and ac electroosmotic flows. The time evolution of the electroosmotic flow and the effect of a frequency-dependent ac electric field on the
oscillating electroosmotic flow in a closed-end rectangular microchannel are examined. Specifically, the induced pressure gradient is analyzed
under effects of the channel dimension and the frequency of electric field. In addition, based on the Stokes second problem, the solution of
the slip velocity approximation is presented for comparison with the results obtained from the analytical scheme developed in this study.
 2004 Elsevier Inc. All rights reserved.

Keywords: Frequency-dependent electroosmosis; Closed-end rectangular microchannel; Onsager reciprocal relationship; Slip velocity approximation;
Green’s function method

1. Introduction accurately transport and manipulate the liquid samples in


microfluidic devices for chemical analyses and biomedical
Most solid surfaces would acquire a certain amount of diagnoses [3,4]. The electroosmotic pumping system works
electric surface charge when in contact with polar liquids without need of movable mechanical parts, thus increasing
such as electrolytes [1]. Due to electrostatic interactions, the long-term stability. In addition, electroosmosis allows pump-
presence of such surface charge can cause the redistribution ing liquids over a wide range of conductivity, which is essen-
of nearby counterions and co-ions in the liquid phase, lead- tial for biochemical applications [5].
ing to the formation of an electric double layer (EDL) [2]. Generally, use of electroosmosis involves direct current
In the diffuse layer of the EDL, the counterions predomi- (dc) electric field to generate a steady-state electroosmotic
nate over the co-ions to neutralize the surface charge. Thus flow. With an alternating current (ac) electric field, electroos-
the local net charge density is not zero. When an electric mosis still occurs, but the electroosmotic flow becomes time
field is applied tangentially along the charged surface, it will and frequency dependent. In recent years, ac electroosmo-
exert a coulumbic force on the ions within the EDL. The sis has drawn a great attention from researchers. In view of
migration of the mobile ions will carry the adjacent liquid the problem associated with electroosmosis in the conven-
phase by viscous shear stress, resulting in a bulk electroos- tional dc microelectrophoretic method (e.g., the difficulty
motic flow (EOF) [2]. The pressure-building ability of the in precisely locating the stationary level inside the micro-
electroosmosis has made it a powerful pumping means to electrophoretic cell), Minor et al. [6] proposed a new fast
method for measuring the electromobility of colloidal parti-
* Corresponding author. Fax: +65-6791-1859. cles using ac electric field. They showed that it is possible
E-mail address: mcyang@ntu.edu.sg (C. Yang). to apply an ac electric field with a specified frequency so
0021-9797/$ – see front matter  2004 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcis.2004.03.005
680 Marcos et al. / Journal of Colloid and Interface Science 275 (2004) 679–698

that electroosmosis is suppressed, whereas particles are still finitely extended rectangular microchannels has been analyt-
able to follow the field according to their dc mobility. Ramos ically studied by Yang et al. [16], Erickson and Li [17], and
et al. [7] experimentally studied the forces acting on parti- Marcos et al. [18]. Specifically, Marcos et al. [18] presented
cles of macromolecular sizes in microelectrode arrays. They a comparison of the results of ac electroosmosis between the
performed calculations for various forces affecting the ac no-slip condition and the slip velocity approximation using
electrokinetic behavior of particles. In a subsequent work, the Smoluchowski equation.
Ramos et al. [8] presented a simple model describing the This paper presents an analysis of the frequency-depen-
frequency dependence of the ac electroosmotic flow veloc- dent ac electroosmotic flow in a closed-end rectangular mi-
ity in planar microelectrodes. Barragán and Bauzá [9] stud- crochannel, which not only is relevant to the development
ied the influence of an ac sinusoidal perturbation on the of new microelectrophoretic methods [6], but also may pro-
electroosmotic flow through a cation-exchange membrane. vide insight into the development of novel electrokinetic
Their experimental results showed that the presence of the micropumps and microactuators [19]. Since the microchan-
ac perturbation affects the electroosmotic flow through the nel is closed, the ac electroosmosis should have a differ-
frequency of the ac signal and the stirring condition. Con- ent frequency response due to the superposition of the vis-
sidering that slow mixing processes occur in microfluidic de- cous waves generated by the electroosmotic flow and the
vices due to low Reynolds number-dominated laminar flows, induced counterflow. Further, the underlying mathematical
Oddy et al. [10] demonstrated an electrokinetic chaotic mi- modeling and solution pose more challenge and complex-
cromixer using fluctuating electric field. In their experiment, ity. The structure of the paper is organized as follows: Sec-
instability of electroosmotic flow driven by sinusoidally al- tion 2 presents the governing equations for ac electroosmo-
ternating electric fields was observed. sis. The modified Navier–Stokes equation is solved analyti-
In the literature, several theoretical studies have been re- cally using the Green’s function formulation. The exact solu-
ported on the ac electroosmotic flow in microcapillaries of tions for the frequency responses of electroosmotic flow and
various geometrical domains. Using slip velocity approach, the induced pressure gradient are obtained. Section 3 pro-
Minor et al. [6] carried out an analysis of the dynamic as- vides the EDL potential distribution by solving the Poisson–
pects of electroosmosis and electrophoresis in a parallel slit. Boltzmann equation with the Debye–Hückel approximation.
Also in a parallel slit, Dutta and Beskok [11] presented an In Sections 4 and 5, we derive an analytical expression for
analytical model for the time-periodic EOF, and compared the double integration and demonstrate the validity of On-
their analytical solution of electroosmotic flow field with the sager’s principle of reciprocity. Section 6 presents the results
Stokes second problem. The slip velocity model was first and discussion. In Appendix A, the solution under a slip ve-
proposed by Overbeek [12], who showed that for microchan- locity approximation based on the Stokes second problem is
nels with relatively thin EDL thickness, the flow field out- also provided for comparison.
side the EDL is an irrotational flow with a slip velocity
boundary condition determined by the well-known Smolu-
chowski equation. Recently, the problem associated with ac 2. Dynamics of electroosmotic flow
electroosmotic flow in cylindrical microcapillaries was ad-
dressed by Reppert and Morgan [13], Kang et al. [14], and Fig. 1 shows a rectangular microchannel of height 2a
Bhattacharrya et al. [15]. Reppert and Morgan [13] artifi- and width 2b and a cartesian coordinate system for analysis
cially divided the flow into two regimes: near-wall regime of the electroosmotic flow. The liquid in the microchannel
and bulk-fluid regime. In the near-wall regime, their analy- is assumed to be an incompressible, newtonian, symmetric
sis neglected the fluid inertial effect; while in the bulk- electrolyte of constant density, ρ, and viscosity, µ. The chan-
fluid regime, the slip velocity approach was used. Using nel wall is uniformly charged with a zeta potential, ζ . Based
the Green’s function method, Kang et al. [14] developed a on the assumption of a laminar, fully developed flow, the mo-
theoretical model for ac electroosmosis. They used an an- tion of the electrolyte under the application of an external
alytical scheme to solve the complete Poisson–Boltzmann time- and frequency-dependent electric field, E(t), through
equation without assuming the Debye–Hückel approxima- the infinitely extended rectangular microchannel can be de-
tion and thin EDL. More recently, the ac laminar EOF in in- scribed by a modified Navier–Stokes equation, which takes

Fig. 1. Geometry of the rectangular channel.


Marcos et al. / Journal of Colloid and Interface Science 275 (2004) 679–698 681

the form [2] An analytical solution for the dynamic electroosmotic flow
in a closed-end rectangular microchannel under a time-
∂u(x, y, t) dP (t)
ρ =− + E(t)ρe (x, y) dependent electric field will be sought by using the Green’s
∂t dz function approach. According to the Green’s function meth-
 2 
∂ u(x, y, t) ∂ 2 u(x, y, t) od [20], the solution of Eq. (5) subjected to the initial and
+µ + , (1)
∂x 2 ∂y 2 boundary conditions given in Eqs. (6a)–(6c) can be explic-
itly expressed as
where u(x, y, t) is the transient velocity field, P (t) is the in-
duced pressure to generate a backflow in a closed-end chan-   t¯  h a/D
b/D  h
nel, and ρe (x, y) is the local net charge density due to the ū(X, Y, t¯ ) = REAL − dτ G(X, Y, t¯|X , Y  , τ )
presence of the EDL. In combination with the Boltzmann
τ =0 X  =0 Y  =0
distribution, ρe (x, y) is expressed as [2]   
d P0 ωDh2 τ
ρe (x, y) = zv e0 (n+ − n− ) × exp i +φ
  dZ ν
zv e0  
= −2zv e0 n0 sinh ψ(x, y) , (2) iωDh2 τ
kb T  
+ GE0 exp
ν
where zv is the electrolyte valence, e0 is the elementary  
charge, n0 is the ionic number concentration in the bulk
solution, kb is the Boltzmann constant, T is the absolute tem- × sinh Ψ (X , Y  ) dX dY  . (7)
perature, and ψ(x, y) is the local EDL potential.
The external sinusoidal electric field and the induced si- Here, the Green’s function G(X, Y, t¯|X , Y  , τ ) can be ob-
nusoidal backpressure can be expressed as tained by using the separation of variables method. The ex-
pression for G(X, Y, t¯|X , Y  , τ ) is given by [21,22]
E(t) = E0 eiωt , (3a)
P (t) = P0 e i(ωt +φ)
, (3b) G(X, Y, t¯|X , Y  , τ )
∞ ∞ 
2 2
D2 D π (2m − 1)2
where E0 is the amplitude of the applied electric field =4 h exp − h
strength, ω is the alternating frequency, i is the unit of the ab 4 b2
m=1 n=1
imaginary number, P0 is the magnitude of the pressure, and  
(2n − 1)2
φ is the phase lag due to fluid inertia. + (t¯ − τ )
a2
Introducing the dimensionless parameters
 
 
2m − 1 Dh 2m − 1 Dh 
u µ x y × cos πX cos πX
ū = , t¯ = t, X= , Y= , (4a) 2 b 2 b
U ρDh2 Dh Dh
 
 
2n − 1 Dh 2n − 1 Dh 
z P0 × cos πY cos πY .
Z= , P0 = , 0 = E0 Dh Re0 ,
E 2 a 2 a
Dh Re0 ρU 2 ζ
zv e0 ψ (8)
Ψ= ,  = 2zv e0 n0 ζ ,
G (4b)
kb T ρU 2 Substituting Eq. (8) into Eq. (7) and carrying out the integra-
tion, we can get the nondimensional fluid velocity distribu-
where U is the reference velocity, Dh = 4ab/(a + b) is the tion in the microchannel
hydraulic diameter of the rectangular channel, ν = µ/ρ is  ∞ ∞
the liquid kinematic viscosity, and Re0 = Dh U/ν is the ref- 64 d P0
erence Reynolds number. ū(X, Y, t¯ ) = REAL − (−1)m+n
π 2 Dh2 dZ m=1 n=1
Substituting Eqs. (2) and (3) into Eq. (1), we can obtain
ωD 2 t¯  D2 
  exp i ν h + φ − exp − 4h Rmn t¯ + iφ
∂ ū ∂ 2 ū ∂ 2 ū d P0 ωDh2 t¯ × 
= + − exp i + φ (2m − 1)(2n − 1) Rmn + i 4ω
∂ t¯ ∂X2 ∂Y 2 dZ ν
 
ν
  2m − 1 Dh
2
iωDh t¯ × cos
−G E 0 exp sinh Ψ (X, Y ). (5) 2 b
πX
ν
 
2n − 1 Dh
The initial and boundary conditions applicable to Eq. (5) × cos πY
are 2 a
∞ ∞
 16  
ū(X, Y, t¯ )t¯=0 = 0, (6a) − G E0 Cmn
  ab
∂ ū  ∂ ū  m=1 n=1
= 0, = 0, (6b) iωD 2  D2 
∂X  X=0 ∂Y  Y =0 exp ν h t¯ − exp − 4h Rmn t¯
× 
ū|X=b/Dh = 0, ū|Y =a/Dh = 0. (6c) Rmn + i 4ω
ν
682 Marcos et al. / Journal of Colloid and Interface Science 275 (2004) 679–698


 
2m − 1 Dh
× cos πX Hence, we can show the expression for the induced pressure
2 b gradient is

  
2n − 1 Dh d P0 E
π 2 Dh2 G 0
× cos πY , (9) =−
2 a dZ 4ab
  D2 
∞ ∞ m+n
1−exp − 4h Rmn t¯
where m=1 n=1 (−1) Cmn (2m−1)(2n−1)Rmn
×   D2  . (13)
 h a/D
b/D  h
  ∞ ∞ 1−exp − 4h Rmn t¯
2m − 1 Dh 
Cmn = cos πX m=1 n=1 (2m−1)2 (2n−1)2 Rmn
2 b D2 
 
X =0 Y =0

  (i) For a small time elapsed, i.e., t¯  1, exp − 4h Rmn t¯
2n − 1 Dh D2 
× cos πY  ≈ 1 − 4h Rmn t¯ , the velocity distribution can therefore be
2 a expressed as
× sinh Ψ (X , Y ) dX dY  ,
 
(10a) ∞ ∞
  16 d P0 (−1)m+n
(2m − 1)2 (2n − 1)2 ū(X, Y, t¯ ) = − ¯
t
Rmn = π 2 + . (10b)
2
π dZ (2m − 1)(2n − 1)
m=1 n=1
b2 a2
 
2m − 1 Dh
Eq. (9) shows that the flow in a closed-end rectangular mi- × cos πX
2 b
crochannel actuated by an applied oscillating electric field
 
2n − 1 Dh
consists of simple linear superposition of two parts: (i) the × cos πY
2 a
flow due to the applied electric field, and (ii) the flow due to
2 ∞

the induced backpressure gradient. In this study, two cases 4Dh
− E
G 0 t¯ Cmn
are being discussed: ab
m=1 n=1

 
2m − 1 Dh
Case 1: transient electroosmotic flow field under a dc × cos πX
2 b
electric field (ω = 0)
 
2n − 1 Dh
× cos πY , (14a)
In this case, the phase lag is also zero (φ = 0). Hence the 2 a
fluid velocity distribution is expressed as indicating that the velocity is proportional to time at an initial
moment. Also the corresponding pressure is
∞ ∞ ∞ ∞ (−1)m+n Cmn
64 d P0 E 0
ū(X, Y, t¯ ) = − (−1)m+n d P0 π 2 Dh2 G m=1 n=1 (2m−1)(2n−1)
π 2 Dh2 dZ m=1 n=1 =− ∞ ∞ 1
. (14b)
dZ 4ab m=1 n=1 2 2
(2m−1) (2n−1)
D2 
1 − exp − 4h Rmn t¯ D2 
× (ii) For a large time, i.e., t¯ → ∞, exp − 4h Rmn t¯ ≈ 0,
(2m − 1)(2n − 1)Rmn the velocity distribution becomes

 
2m − 1 Dh ∞ ∞
× cos πX 64 d P0 (−1)m+n
2 b ū(X, Y, t¯ ) = −

  π 2 Dh2 dZ m=1 n=1 (2m − 1)(2n − 1)Rmn
2n − 1 Dh
 
× cos πY 2m − 1 Dh
2 a × cos πX
D2  2 b
∞ ∞
 
16   1 − exp − 4h Rmn t¯ 2n − 1 Dh
− GE 0 Cmn × cos πY
ab Rmn 2 a
m=1 n=1

  ∞ ∞
2m − 1 Dh 16   Cmn
× cos πX − G E0
2 b ab Rmn
m=1 n=1

 
 
2n − 1 Dh 2m − 1 Dh
× cos πY . (11) × cos πX
2 a 2 b

 
For a closed-end rectangular microchannel, the net flow rate 2n − 1 Dh
× cos πY , (15a)
equals zero. Mathematically, it fulfills the condition 2 a
and the corresponding induced pressure gradient becomes
 h
b/D  h
a/D
∞ ∞ (−1)m+nCmn
d P0 E
π 2 Dh2 G 0
ū(X, Y, t¯ ) dX dY = 0. (12) =−
m=1
∞ ∞
n=1 (2m−1)(2n−1)Rmn
. (15b)
dZ 4ab 1
−b/Dh −a/Dh m=1 n=1 2 2
(2m−1) (2n−1) Rmn
Marcos et al. / Journal of Colloid and Interface Science 275 (2004) 679–698 683

Eqs. (15a) and (15b), respectively, give the flow field and the to convert the electrical energy, zv e0 n0 ζ into the fluid kinetic
induced pressure gradient in a closed-end rectangular mi- energy, ρU 2 /2. Furthermore, it is interesting to discuss an-
crochannel under the steady-state situation [22]. Also it is in- other limiting situation that as ω → ∞, Eq. (18) is reduced
teresting to note that the mathematical formulation given in to Eq. (14b), suggesting the induced pressure gradient un-
Eq. (15a) is identical to the steady-state velocity distribution der extremely high applied frequencies is identical to that at
for pressure-driven liquid flow in a rectangular microchannel the initial moment under a dc electric field because in both
under electroviscous effect [23,24]. For electrokinetic flows, cases, the flow occurs only within the EDL region.
there is a link between streaming potential flow due to pres-
sure gradient and electroosmotic flow due to electric field,
and its discussion is presented later. 3. Electrical potential distribution

Case 2: steadily oscillating electroosmotic flow field under Kang et al. [14] have shown that for microchannels with
an ac electric field fully developed electroosmotic flows under a time- and
frequency-dependent electric field, the EDL potential dis-
In this situation, the exponential term on the right-hand tribution can still be described by the well-known Poisson–
side of Eq. (9) decays, and hence Eq. (9) is reduced to Boltzmann equation provided that the frequency of the elec-
 tric field is not very high, say less than 1 MHz. The Poisson–
64 d P(t¯ ) Boltzmann equation in a rectangular domain is expressed as
ū(X, Y, t¯ ) = REAL −
π 2 Dh2 dZ  
∂ 2ψ ∂ 2ψ 2zv e0 n0 zv e0 ψ

∞ + = sinh . (19)
(−1)m+n ∂x 2 ∂y 2 εr ε0 kb T
× 
m=1 n=1
(2m − 1)(2n − 1) Rmn + i 4ων By using the previously defined dimensionless parameters,

  we can nondimensionalize Eq. (19) as
2m − 1 Dh
× cos πX
2 b ∂ 2Ψ ∂ 2Ψ

  + = K 2 sinh Ψ, (20)
2n − 1 Dh ∂X2 ∂Y 2
× cos πY
2 a where κ = (2zv2 e02 n0 /εr ε0 kb T )1/2 is the Debye–Hückel pa-
∞ ∞
16   Cmn rameter (1/κ is normally referred to as the EDL thickness),
− G E(t¯ )  and K = κDh . Assuming that the zeta potential is small and
ab Rmn + i 4ω


m=1 n=1

ν using the Debye–Hückel linear approximation, sinh Ψ ≈ Ψ ,
2m − 1 Dh we can linearize Eq. (20) to
× cos πX
2 b

   ∂ 2Ψ ∂ 2Ψ
2n − 1 Dh + = K 2 Ψ. (21)
× cos πY , (16) ∂X2 ∂Y 2
2 a Eq. (21) is subjected to the following boundary conditions:
 
where ∂Ψ  ∂Ψ 
  = 0, = 0, (22a)
d P(t¯ ) d P0 ωDh2 t¯ ∂X X=0 ∂Y Y =0
= exp i +φ , (17a)
dZ dZ ν Ψ |X=b/Dh = ζ̄ , Ψ |Y =a/Dh = ζ̄ , (22b)
 
iωDh2 where ζ̄ is the nondimensional zeta potential, defined as ζ̄ =
 t¯ ) = E
E( 0 exp t¯ . (17b)
ν zv e0 ζ /kb T .
Using the separation of variables method, we can solve
The induced pressure gradient expression for a closed-end
Eq. (21), and the solution of Eq. (21) is expressed as [21–23]
system can be obtained as
d P(t¯ ) E(
π 2 Dh2 G  t¯ ) Ψ (X, Y )
=− ∞
(2n−1)2 π 2 Dh2 1/2 
dZ 4ab (−1)n+1 cosh 1 + KY
∞ ∞ 4K 2 b2
= 4ζ̄
(−1)m+n Cmn
n=1 (2m−1)(2n−1) Rmn +i 4ω  (2n−1)2 π 2 Dh2 1/2 Ka 
m=1 n=1 (2n − 1)π cosh 1 +
× ∞ ∞ 1
ν
. (18)
4K 2 b2 Dh
m=1 n=1 (2m−1)2 (2n−1)2 Rmn +i 4ω 
(2n − 1)Dh
ν × cos πX
2b
Taking advantage of the exact solutions obtained in this
∞ (2m−1)2 π 2 Dh2 1/2 
study, we examine Eqs. (13) and (18), and find that the in- (−1)m+1 cosh 1 + KX
4K 2 a 2
+ 4ζ̄
duced pressure gradient for both dc and ac electric fields is (2m−1)2 π 2 Dh2 1/2 Kb 
proportional to the strength of applied electric field and the m=1 (2m − 1)π cosh 1 +
4K 2 a 2 Dh
 According to the definition for

dimensionless parameter G. (2m − 1)Dh
 in Eq. (4b), G
G  physically denotes a measure of the ability × cos πY . (23)
2a
684 Marcos et al. / Journal of Colloid and Interface Science 275 (2004) 679–698

Certainly, the Debye–Hückel linear approximation used for The flow rate Q(t) in the microchannel can be obtained
solving the Poisson–Boltzmann equation is valid only when by
the zeta potential is small (usually less than 25 mV). How-
a b
ever, if one wants to study the flow behavior of a highly
charged system where only the outer region of the diffuse Q(t) = u(x, y, t) dx dy. (26a)
double layer is important, the linear approximation can give −a −b
a good prediction even when the zeta potential is as high as Because of the symmetry, the flow rate can be further ex-
100 mV [1], especially so when κDh  1. pressed in the form
 h b/D
a/D  h
4. Solving the double integration (Cmn ) Q(t) = 4U Dh2 ū(X, Y, t¯ ) dX dY. (26b)
0 0
Having obtained the electrical potential distribution,
Ψ (X, Y ), and using the Debye–Hückel linear approxima- Case 1: transient electroosmotic flow under dc electric
tion, sinh Ψ ≈ Ψ , we can carry out the double integration, field (ω = 0)
Cmn , in Eq. (10) analytically, and obtain
In the case of dc applied electric field, substituting
 h a/D
b/D  h
 
2m − 1 Dh  Eq. (11) into Eq. (26b), and making use of Eq. (4b) to con-
Cmn = cos πX vert the dimensionless pressure and electric field into the
2 b
 
X =0 Y =0

  dimensional ones, we can obtain the flow rate as
2n − 1 Dh  dP0
× cos πY Q(t) = L11 + L12 E0 , (27)
2 a dz
× Ψ (X , Y  ) dX dY  . (24) where the Onsager transport coefficients are
Substituting Eq. (23) into Eq. (24), we can show that ∞ ∞ D2 
1024ab 1 − exp − 4h Rmn t¯
2ζ̄ (−1)m+n b L11 = − 4 , (28a)
Cmn = π µ (2m − 1)2 (2n − 1)2 Rmn
 2
(2m−1)2 π 2 Dh  m=1 n=1
1+ K2 2n−1  Dh ∞ ∞
(2m − 1)πDh 4K 2 b2
2n−1  Dh + 2 a π 512zv e0 n0 Dh2
2 a π L12 = − (−1)m+n Cmn
π 2µ
2ζ̄ (−1)m+n a m=1 n=1
+  2 . D2 
(2n−1)2 π 2 Dh
1+ K2 2m−1  Dh 1 − exp − 4h Rmn t¯
(2n − 1)πDh 4K 2 a 2
2m−1  Dh + π × . (28b)
2 b π
2 b (2m − 1)(2n − 1)Rmn
(25) In general, the axial component of the electric current
density is expressed as [2]
 
5. Onsager’s principle of reciprocity ∂n+ ∂n−
iz (t) = uzv e0 (n+ − n− ) − zv e0 D+ + D−
∂z ∂z
Eq. (9) shows that the flow in a closed-end rectangu- + λ0 Ez , (29)
lar microchannel results from the applied electric field and
the induced backpressure gradient. Further examination of where λ0 is the electric conductivity of the fluid, and D±
Eq. (9) indicates the phenomenological relationships be- are the mass diffusivity of cations and anions, respectively.
tween the flow field and the driving electric field and the Since no ionic concentration gradient is built up along the
pressure gradient are linear, suggesting a possible constitu- axial direction during the flow, the second term on the right-
tion of the Onsager principle of reciprocity familiar in the hand of Eq. (29) drops off. The electric current across a
theory of nonequilibrium thermodynamics. In the literature, straight rectangular channel can be expressed as
the Onsager reciprocal relations for electrokinetic flows un- a b
der steady-state situation were demonstrated by Burgreen
I (t) = 4 ρe u(x, y, t) dx dy + 4abλ0E0 . (30)
and Nakache [25] for a parallel slit and by Rice and White-
head [26] and Levine et al. [27] for a cylindrical capillary. 0 0
Recently, Bhattacharrya et al. [15] and Yang et al. [16] also With the Debye–Hückel linear approximation, substituting
demonstrated the validity of Onsager reciprocal relations in Eq. (2) into Eq. (30) gives
oscillating laminar electrokinetic flow in circular and rectan-
 h b/D
a/D  h
gular microchannels, respectively. In the following, we show
the validity of the Onsager reciprocal relations for both tran- I (t) = −8zv e0 n0 U Dh2 Ψ (X, Y )ū(X, Y, t¯ ) dX dY
sient electroosmotic flow and ac oscillating electroosmotic 0 0
flow in a closed-end rectangular microchannel. + 4λ0 E0 ab. (31)
Marcos et al. / Journal of Colloid and Interface Science 275 (2004) 679–698 685

Recall that the velocity field, ū(X, Y, t¯ ), and the electri- From Eqs. (35) and (37), it can be noted that J12 = −J21 ,
cal potential distribution, Ψ (X, Y ), are given by Eqs. (11) which fulfills Onsager’s principle of reciprocity for oscil-
and (23), respectively. We then can show that the electric lating electroosmotic flow in a closed-end rectangular mi-
current can be written in the Onsager form by substituting crochannel.
Eqs. (11) and (23) into Eq. (31),
dP0
I (t) = L21 + L22 E0 , (32) 6. Results and discussion
dz
where the Onsager transport coefficients are
In all the calculations, NaCl solution with valence zv of
∞ ∞
512zv e0 n0 Dh2 unity is used as the electrolyte solution. The dielectric con-
L21 = (−1)m+n Cmn stant, εr , density, ρ, and viscosity, µ, of the electrolyte are
π 2µ
m=1 n=1 taken to be the same as those of water, and these are 80,
D2  1000 kg/m3, and 9 × 10−4 N s/m2 , respectively. The tem-
1 − exp − 4h Rmn t¯
× , (33a) perature T of the electrolyte is taken as 293 K. The reference
(2m − 1)(2n − 1)Rmn
D2  velocity U is chosen as 1 mm/s. The strength of the electric
∞ ∞
256zv2 e02 n20 Dh4 2 1 − exp − 4h Rmn t¯ field E0 is fixed at 10,000 V/m. The reference geometric
L22 = Cmn size of the channel is 2a × 2b, which is 40 × 100 µm.
abµ Rmn
m=1 n=1
+ 4λ0 ab. (33b) 6.1. Transient velocity profile under dc electric field
From Eqs. (28) and (33), it can be noted that L12 = −L21 ,
suggesting the validity of the Onsager principle of reci- Fig. 2 shows the time evolution of the EOF in a quarter of
procity for transient electroosmotic flow in a closed-end rec- the closed-end rectangular microchannel cross section under
tangular microchannel. a constant dc electric field. When time t¯ = 0, the fluid veloc-
ity anywhere in the channel is at rest, and hence the velocity
Case 2: oscillating electroosmotic flow under steadily equals zero.
alternating ac electric field According to the results shown in Fig. 2, it can be seen
that on application of the electric field, the flow begins to de-
Similarly for the sinusoidal alternating applied electric velop in the region close to the channel wall and the flow ve-
field, we can obtain the flow rate by substituting Eq. (16) locity increases rapidly from zero at the wall to its peak value
into Eq. (26b), within the EDL region. This scenario reveals the unique fea-
dP (t) ture of the electroosmotic flow; it is driven by the electrical
Q(t) = J11 + J12 E(t), (34) body force resulting from the interaction of the electric field
dz
and the net charge density. Such driving force exists only
where the Onsager transport coefficients are within the EDL region where the nonzero electrical charge
∞ ∞
1024ab 1 density is present. The flow outside the EDL region is due
J11 = − , to viscous shear diffusion. However, different from the elec-
4
π µ (2m − 1)2 (2n − 1)2 Rmn + i 4ω
ν
m=1 n=1 troosmotic flow in an open-end microchannel, there exists a
(35a) reverse flow in the bulk liquid region in a closed-end mi-
512zv e0 n0 Dh2 crochannel. Such a reverse flow is generated by induced
J12 = −
π 2µ backpressure to counterbalance the electroosmotic flow so
∞ ∞ that the net flow rate is zero. Furthermore, Fig. 2 shows that it
(−1)m+n Cmn
× , (35b) takes a unit dimensionless time, t¯ = 1, or dimensional time,
m=1 n=1
(2m − 1)(2n − 1) Rmn + i 4ω
ν tc = Dh2 /ν, for the flow to attain its steady state. Physically,
and the electrical current is given by the time tc denotes the time scale for viscous diffusion to
travel a distance of channel hydraulic diameter, Dh . Finally,
dP (t)
I (t) = J21 + J22 E(t), (36) at the steady state, the velocity profile at the bulk liquid re-
dz gion exhibits a parabolic profile due to the presence of the
where the Onsager transport coefficients are induced backpressure gradient. In addition, an important fea-
512zv e0 n0 Dh2 ture that can also be observed from Fig. 2 is the “corner
J21 = effect,” which is demonstrated during the early stage of the
π 2µ
∞ ∞
evolution of the EOF. The presence of such “corner effect”
(−1)m+n Cmn is due to the overlap of the two EDLs formed at the two ad-
× , (37a)
m=1 n=1
(2m − 1)(2n − 1) Rmn + i 4ω
ν
jacent sides of the rectangular channel wall.
∞ ∞ As mentioned earlier, study of electroosmotic flow in
256zv2 e02 n20 Dh4 2
Cmn a closed-end microchannel is related to the development
J22 =  + 4λ0 ab. (37b)
abµ
m=1 n=1
Rmn + i 4ω
ν
of novel electrokinetic micropumps and microactuators. As
686 Marcos et al. / Journal of Colloid and Interface Science 275 (2004) 679–698

Fig. 2. Dimensionless transient velocity, u(x, y)/U under dc electric field versus dimensionless width, x/Dh and dimensionless height, y/Dh . The reference
velocity is taken as 1 mm/s, the channel size is 40 × 100 µm, concentration n0 = 10−5 M, and zeta potential ζ = 50 mV. Snapshots are presented at six
dimensionless times t¯ = 10−5 , 10−4 , 10−3 , 10−2 , 0.1, 1.
Marcos et al. / Journal of Colloid and Interface Science 275 (2004) 679–698 687

channel size increases. This observation may be understood


as follows: According to the Smoluchowski equation, the
maximum magnitude of fluid velocity is independent of the
channel size; this is an important feature of electroosmotic
flow. However, hydrodynamic drag increases as channel size
decreases. Hence, to counterbalance the electroosmotic flow,
higher backpressure is required for smaller-sized microchan-
nels.

6.2. Oscillating velocity profile under a steadily alternating


ac electric field

(a) For a rectangular channel cross section of 2a × 2b = 40 ×


100 µm, the characteristic time of the electroosmotic flow
can be estimated from Eq. (11), t ∗ = (4ρ/µπ 2 )a 2b2 /(a 2 +
b2 ) = 155.3 µs. The corresponding eigenfrequency of the
system is f ∗ = 1/t ∗ = 6.44 kHz. In this work, three differ-
ent frequencies of the applied electric field, 0.1f ∗ , f ∗ , and
10f ∗ (f = ω/2π ) are considered to study the characteris-
tics of the oscillating electroosmotic flow in a closed-end
channel. The results are summarized in Figs. 4A–4C which
present the steadily oscillating velocity distributions in a
symmetric quarter of the rectangular channel for five differ-
ent characteristic moments: ωt = 0, π/4, π/2, 3π/4, and π .
Further, the results of the electroosmotic flow field under the
(b) slip velocity approximation are presented in Figs. 5A–5C
for comparison with the analytical method developed in this
Fig. 3. (a) Induced pressure gradient under dc electric field versus dimen- study. The detailed mathematical derivation of the slip veloc-
sionless time t¯ = 10−5 , 10−4 , 10−3 , 10−2 , 0.1, 1. (b) Steady-state induced ity approximation is presented in Appendix A. In addition, to
pressure gradient under dc electric field versus dimensionless channel’s hy-
provide quantitative comparison of the characteristics of the
draulic diameter with respect to a reference diameter based on a geometric
size of 40 × 100 µm. electroosmotic flow fields under the ac electric field of three
different frequencies, two-dimensional plots corresponding
to the cases presented in Figs. 4A–4C and 5A–5C are shown
such, the capacity of pressure building up is also analyzed. in Figs. 6A–6C, in which the results obtained from both the
The induced pressure gradient over time is displayed in analytical scheme (represented by the solid lines) and the
Fig. 3a. It can be seen that the induced backpressure gra- slip velocity approximation (denoted by the dotted lines) are
dient drops with time and reaches its minimum at the steady compared.
state. The highest pressure gradient is generated at the ini- As discussed earlier, when an electric field is applied, the
tial state as soon as electric field is applied to the system (at electroosmotic flow begins within the EDL region near the
t = 0+ ). An analogy of static friction and dynamic friction charged channel wall. Because of the generation of the in-
concept in a solid body can be drawn to illustrate this effect, duced backpressure gradient in the close-end channel, fluid
where it requires a stronger force to initiate the movement of at the central region flows in the opposite direction of the
the body as compared with the force required to overcome flow in the EDL region. When the frequency of the elec-
the friction when the body is on the move. The decrease in tric field is low (e.g., 0.1f ∗ as shown in Fig. 4A), the flow
the pressure gradient subsequently is due to the effect of the extends through the entire channel and the changes in the
fluid’s inertia. When the fluid begins to flow, it requires very flow frequency follow closely those of the alternating elec-
high pressure to overcome the inertia such that it can pro- tric field. As the applied frequency increases, the perturbed
vide reverse flow. Once the fluid starts to flow, less pressure flow region becomes smaller, and the flow response in the
is needed to maintain the reverse flow. central region of the channel lags behind that of the fast al-
Fig. 3b presents the effect of rectangular channel size ternating electric field (e.g., f ∗ as shown in Fig. 4B). While
(with fixed aspect ratio) on the induced pressure gradient un- at a high applied frequency (e.g., 10f ∗ as shown in Fig. 4C),
der steady state. The graph is plotted using log–log axis. The the flow in the central channel remains virtually motionless
horizontal axis is a ratio of the actual channel hydraulic di- regardless of the fast oscillating flow occurring within the
ameters to the reference hydraulic diameter of the rectangu- EDL region which is close to the channel wall. This phe-
lar channel of 2a × 2b = 40 × 100 µm. It is shown in Fig. 3b nomenon is logically anticipated because as the frequency
that the pressure gradient monotonically decreases as the goes very high, the electric field changes its direction so fast
688 Marcos et al. / Journal of Colloid and Interface Science 275 (2004) 679–698

(A)

Fig. 4. Steadily oscillating velocity distributions along the dimensionless channel width and height for three different aspect frequencies of the ac electric
field, 0.1f ∗ , f ∗ , and 10f ∗ , with channel size 40 × 100 µm, concentration n0 = 10−5 M, zeta potential ζ = 50 mV, and eigenfrequency f ∗ = 6.44 kHz.
Snapshots are presented at five different characteristic moments: ωt = 0, π/4, π/2, 3π/4, π . Reference velocity is taken as 1 mm/s. Results are obtained from
the analytical scheme developed in this study. (A) Low frequency of the external field, f = 0.1f ∗ . (B) Eigenfrequency of the external field, f = f ∗ . (C) High
frequency of the external field, f = 10f ∗ .
Marcos et al. / Journal of Colloid and Interface Science 275 (2004) 679–698 689

(B)

Fig. 4. Continued.
690 Marcos et al. / Journal of Colloid and Interface Science 275 (2004) 679–698

(C)

Fig. 4. Continued.
Marcos et al. / Journal of Colloid and Interface Science 275 (2004) 679–698 691

(A)

Fig. 5. Steadily oscillating velocity distributions along the dimensionless channel width and height for three different aspect frequencies of the external field,
0.1f ∗ , f ∗ , and 10f ∗ , with channel size 40 × 100 µm, concentration n0 = 10−5 M, zeta potential ζ = 50 mV, and eigenfrequency f ∗ = 6.44 kHz. Snapshots
are presented at five different characteristic moments: ωt = 0, π/4, π/2, 3π/4, π . Reference velocity is taken as 1 mm/s. Results are obtained from the slip
velocity approximation based on the Stokes second problem. (A) Low frequency of the external field, f = 0.1f ∗ . (B) Eigenfrequency of the external field,
f = f ∗ . (C) High frequency of the external field, f = 10f ∗ .
692 Marcos et al. / Journal of Colloid and Interface Science 275 (2004) 679–698

(B)

Fig. 5. Continued.
Marcos et al. / Journal of Colloid and Interface Science 275 (2004) 679–698 693

(C)

Fig. 5. Continued.
694 Marcos et al. / Journal of Colloid and Interface Science 275 (2004) 679–698

(A) (B)

(C)

Fig. 6. Two-dimensional plots of steadily oscillating velocity distributions along the dimensionless channel height (at x/Dh = 0) for three different aspect
frequencies of the external field, 0.1f ∗ , f ∗ , and 10f ∗ , with channel size 40 × 100 µm, concentration n0 = 10−5 M, zeta potential ζ = 50 mV, and eigenfre-
quency f ∗ = 6.44 kHz. Snapshots are presented at five different characteristic moments: ωt = 0, π/4, π/2, 3π/4, π . Reference velocity is taken as 1 mm/s.
Comparison with the results obtained from the slip velocity approximation represented by dashed lines. (A) Low frequency of the external field, f = 0.1f ∗ .
(B) Eigenfrequency of the external field, f = f ∗ . (C) High frequency of the external field, f = 10f ∗ .

that due to the inertia of the fluid, the fluid cannot response induced by a sinusoidally oscillating infinite flat plate [28].
fast enough to develop the flow across the entire channel. It is demonstrated in Figs. 6A–6C that the velocity distri-
To provide an insight into the observed scenario, we refer to butions at the central region of the channel, given by both
the frequency-dependent Stokes penetration depth, δs , rep- the analytical scheme (developed in the present study) and
resenting a typical length scale of the oscillatory laminar the slip velocity approximation method, are the same. How-
viscous flows in response to a harmonic external excitation ever, in the region adjacent to the channel wall (i.e., the
(e.g., ac electric field or oscillating boundary). According to EDL region), the two methods predict different velocity dis-
Minor et al. [6], δs is defined as tributions. The discrepancy is due to the assumption of the
  moving boundary condition applied in the slip velocity ap-
µ µ
δs = = . (38) proximation, in which, the boundary velocity is assumed to
ρω 2πρf follow the Smoluchowski equation. In the analytical scheme
This equation clearly shows that the penetration depth however, the boundary velocity is set as zero (i.e., no-slip
(i.e., perturbed flow region) is inversely proportional to the condition). The slip velocity approximation neglects the de-
square root of the frequency, f , of the alternating electric tailed flow characteristics in the EDL region. However, as far
field. as the flow field in the central region of the channel is con-
The slip velocity approximation is rooted in an analogy to cerned, the slip velocity approximation method gives a good
the Stokes second problem, i.e., the oscillating viscous flow estimation when κDh  1.
Marcos et al. / Journal of Colloid and Interface Science 275 (2004) 679–698 695

Fig. 7. Induced pressure gradient under ac electric field versus characteristic moments from 0 to 4π .

when the frequency of applied electric field is large, the liq-


uid is excited only within a small region near the channel
wall while the liquid at a large portion of the channel cen-
tral region remains almost stationary. The compression of
the perturbed flow region with increasing frequency behaves
as the flow in a channel of reduced size. Hence, a high in-
duced backpressure gradient is needed to counterbalance the
electroosmotic flow, fulfilling the condition of zero flow rate
in closed-end microchannels.

7. Concluding remarks

The theory of ac electroosmotic flow in a closed-end


Fig. 8. Amplitude of induced pressure gradient under ac electric field versus rectangular microchannel is presented. Onsager’s principle
the ratio of applied frequency to characteristic frequency (f/f ∗ ). of reciprocity for transient and ac electroosmotic flows is
demonstrated. The analyses show that the flow in a closed-
The induced steadily oscillating pressure gradient in a end channel is the superposition of the electroosmosis in
closed-end rectangular microchannel is shown in Fig. 7. It the EDL region and the reverse pressure-driven flow out-
can be noted that the induced pressure gradient response side the EDL region (i.e., the bulk fluid region). The induced
is always behind the applied sinusoidal electric field. The pressure gradient monotonically decreases as time elapses
phase lag becomes larger as the frequency increases due to (on the application of a dc electric field) or the channel in-
the effect of fluid inertia. Moreover, in contrast to the de- creases. Moreover, it is shown that the ac electroosmotic
pendence of fluid velocity on frequency, high backpressure flow is strongly dependent on the frequency of the applied
is generated as the frequency goes up. This scenario is also sinusoidal electric field. It is identified that it is the fre-
clearly demonstrated in Fig. 8 which presents the induced quency of electric field that determines the thickness of the
backpressure gradient versus the frequency of applied elec- Stokes penetration layer, and thus governs the extent of the
tric field. It is interesting to observe that the amplitude of oscillating flow and the velocity distributions. There is al-
the steadily oscillating pressure gradient remains almost a ways a phase lag for the oscillating backpressure gradient
constant value when the applied frequency is less than the built up in a closed-end channel due to fluid inertia. Fur-
characteristic frequency, f ∗ . In this situation, the flow can ther, it is found that the induced backpressure is insensitive
extend to the entire cross section of the channel, and the to the frequency of electric field when the frequency is less
induced backpressure gradient is insensitive to the applied than the characteristic frequency of the fluid system, and
frequency. However, as the frequency becomes larger than increases significantly with increasing the frequency of elec-
f ∗ , the amplitude of the oscillating backpressure gradient tric field when the frequency is larger than the characteristic
increases significantly. The fast increase of the pressure gra- frequency of the fluid system. In addition, it is shown that
dient with frequency is attributed to the compression of the there is a similarity between the slip velocity approximation
perturbed flow region in the channel. As discussed earlier, and the present analytical scheme in the bulk fluid region,
696 Marcos et al. / Journal of Colloid and Interface Science 275 (2004) 679–698

while in the EDL region, the discrepancy is observed. This with boundary conditions
leads to the applicability of the second Stokes solution to the  
oscillating electroosmotic flow by use of the instantaneous ∂v  ∂v 
= 0, = 0,
Smoluchowski velocity as appropriate slip boundary condi- ∂X X=0 ∂Y Y =0
tion when κDh  1.  
iωDh2
¯ 
v(b/Dh , Y, t ) = U0 exp ¯
t ,
ν
 
iωDh2
Acknowledgments 0 exp
v(X, a/Dh , t¯ ) = U t¯ .
ν
Marcos acknowledges gratefully the Postgraduate Stu-


dent Scholarship from Nanyang Technological University. ∂ 2 w ∂ 2 w d P0 ωDh2 t¯ ∂w
Fruitful discussions with Mr. Y.J. Kang are greatly appreci- (II) + − exp i + φ = ,
∂X 2 ∂Y 2 dZ ν ∂ t¯
ated.
(A.4)
with boundary conditions
 
Appendix A. Slip velocity approximation ∂w  ∂w 
= 0, = 0,
∂X X=0 ∂Y Y =0
Conceptually, the steadily oscillating velocity profile in-
duced by the application of a sinusoidal alternating electric w(b/Dh , Y, t¯ ) = 0, w(X, a/Dh , t¯ ) = 0.
field is similar to the fluid flow induced by a mechanically Eq. (A.3) can be solved by using the separation of vari-
oscillating motion of the channel wall under the same fre- ables method, v = F1 (X)F2 (Y )H (t¯ ). Assuming H (t¯ ) =
quency. In the literature, the fluid flow induced by moving exp((iωDh2 /ν)t¯ ), we can show that
the boundary of the wall is referred to as the Stokes sec-
ond problem [28]. The idea of introducing the slip velocity v(X, Y, t¯ )
approximation is to neglect the thin EDL region such that  ∞ iωD 2 
the entire flow is driven by the hydrodynamic shear stresses 2Dh U0 (−1)m+1 exp ν h t¯ cos(αm X)
= REAL 2 
originating from the liquid viscosity. Here, the thin EDL b 2 + iωDh 1/2 a
m=1 αm cosh αm ν Dh
region can be regarded as an oscillating moving channel
 
wall. The outer slip boundary velocity is determined by the iωDh2 1/2
× cosh αm +2
Y
Smoluchowski equation, Us (t) = −ε0 εr ζ E0 eiωt /µ. Then ν
the oscillating velocity profile can be obtained by solving ∞ iωD 2 
the following transient Stokes equation, in a dimensionless 2Dh U0 (−1)n+1 exp ν h t¯ cos(βn Y )
+
form, a iωDh2 1/2 b 
n=1 βn cosh βn2 + ν Dh
 
 1/2 
∂ ū ∂ 2 ū ∂ 2 ū d P0 ωDh2 t¯ iωDh 2
= + − exp i + φ , (A.1) × cosh βn2 +
∂ t¯ ∂X2 ∂Y 2 dZ ν ν
X , (A.5)

which is subjected to the boundary conditions


where αm = ((2m − 1)/2)(Dh /b)π , βn = ((2n − 1)/2) ×
 
∂ ū  ∂ ū  (Dh /a)π . i is the unit of imaginary number, and “REAL”
= 0, = 0, (A.2a)
∂X X=0 ∂Y Y =0 represents the real part of the solution.
    Using the Green’s function formulation, we can show that
b iωDh2
ū , Y, t¯ = U0 exp t¯ , (A.2b) the solution of Eq. (A.4) is
Dh ν 
    64 d P(t¯ )
a iωDh2 w(X, Y, t¯ ) = REAL −
ū X, ¯ 
, t = U0 exp ¯
t , (A.2c)
Dh ν π Dh2 dZ
2



where (−1)m+n cos(αm X) cos(βn Y )
×
(2m − 1)(2n − 1)
0 = − ε0 εr ζ E0 .
U m=1 n=1

µU 1
× (2m−1)2 2  , (A.6)
Letting ū(X, Y, t¯ ) = v(X, Y, t¯ ) + w(X, Y, t¯ ), the solution π2 + (2n−1) + i 4ω
b2 a2 ν
of Eq. (A.1) can be obtained by solving the two new partial
differential equations of v and w as shown: where


∂v ∂ 2v ∂ 2v d P(t¯ ) d P0 ωDh2 t¯
(I) = + , (A.3) = exp i +φ .
∂ t¯ ∂X 2 ∂Y 2 dZ dZ ν
Marcos et al. / Journal of Colloid and Interface Science 275 (2004) 679–698 697

Hence, the total solution of Eq. (A.1) is f Oscillation frequency of applied electric field (Hz)
f∗ Characteristic frequency (Hz)
ū(X, Y, t¯ ) Dh Hydraulic diameter of microchannel (m)
 iωD 2 
0
2Dh U

(−1)m+1 exp ν h t¯ cos(αm X) ρ Density of the solution (kg m−3 )
= REAL 2  µ Dynamic viscosity of the solution (kg m−1 s−1 )
b 2 + iωDh 1/2 a
m=1 αm cosh αm ν Dh ν Kinematic viscosity of the solution (m2 s−1 )

  Re0 Reference Reynolds number
iωDh2 1/2
× cosh αm 2
+ Y E0 Amplitude of applied electric field strength (V m−1 )
ν kb Boltzmann’s constant (J mol−1 K−1 )
∞ iωD 2 
2Dh U0 (−1)n+1 exp ν h t¯ cos(βn Y ) T Absolute temperature (K)
+ zv
iωDh2 1/2 b 
Valence of electrolyte solution
a β cosh β 2 +
n=1 n n ν Dh n0 Bulk ionic concentration (m−3 )

  e0 Elementary charge (C)
iωDh2 1/2
× cosh βn2 + X ζ Zeta potential of the channel wall (V)
ν κ Debye–Hückel parameter (m−1 )
∞ ∞
64 d P(t¯ ) (−1)m+n cos(αm X) cos(βn Y ) Q Volumetric flow rate (m3 s−1 )

π 2 Dh2 dZ m=1 n=1 (2m − 1)(2n − 1) I Electric current (A)
 λ0 Electric conductivity of the fluid (−1 m−1 )
1 ψ Local EDL potential (V)
× 2  . (A.7)
π 2 (2m−1)
2
+ (2n−1) + i 4ω ρe Electric net charge density (C m−3 )
b2 a2 ν
εr Dielectric constant of the solution
For a closed-channel system, ū(X, Y, t¯ ) should fulfill the ε0 Permittivity of the vacuum (C V−1 m−1 )
condition
 h
b/D  h
a/D
References
ū(X, Y, t¯ ) dX dY = 0. (A.8)
−b/Dh −a/Dh [1] R.J. Hunter, Zeta Potential in Colloid Science: Principles and Appli-
cations, Academic Press, New York, 1981.
The pressure gradient expression can be obtained as
[2] J.H. Masliyah, Electrokinetic Transport Phenomena, Alberta Oil Sands
 
d P(t¯ )
Technology and Research Authority, Edmonton, 1994.
iωDh2

= 2Dh U0 exp t¯ [3] D.J. Harrison, K. Fluri, K. Seiler, Z. Fan, C.S. Effenhauser, A. Manz,
dZ ν Science 261 (1993) 895–897.
∞ 2 iωD 2 1/2 a  [4] A. van den Berg, T.S.J. Lammerink, Top. Curr. Chem. 194 (1998) 21.
tanh αm + νh Dh [5] A. Manz, C.S. Effenhauser, N. Burggraf, D.J. Harrison, K. Seiler, K.
× 2 Fluri, J. Micromech. Microeng. 4 (1994) 257.
2 α2 + iωD h 1/2
m=1 bαm m ν [6] M. Minor, A.J. van der Linde, H.P. van Leeuwen, J. Lyklema, J. Col-

∞ tanh β 2 + iωDh 1/2 b
2  loid Interface Sci. 189 (1997) 370–375.
n ν Dh [7] A. Ramos, H. Morgan, N.G. Green, A. Castellanos, J. Phys. D 31
+ iωD 2 1/2 (1998) 2338–2353.
n=1 aβn2 βn2 + ν h [8] A. Ramos, H. Morgan, N.G. Green, A. Castellanos, J. Colloid Interface
 ∞ ∞ Sci. 217 (1999) 420–422.
256 1 [9] V.M. Barragán, C.R. Bauzá, J. Colloid Interface Sci. 230 (2000) 359–
π 4 Dh4 m=1 n=1 (2m − 1)2 (2n − 1)2 366.
 [10] M.H. Oddy, J.G. Santiago, J.C. Mikkelsen, Anal. Chem. 73 (2001)
1 5822–5832.
×
(2n−1)2   .
2
(A.9) [11] P. Dutta, A. Beskok, Anal. Chem. 73 (2001) 5097–5102.
π 2 (2m−1)
b 2 + a 2 + i 4ω
ν [12] J. Lyklema, Fundamentals of Interface and Colloid Science, vol. 1:
Fundamentals, Academic Press, London, 1991.
[13] P.M. Reppert, F.D. Morgan, J. Colloid Interface Sci. 254 (2002) 372–
383.
Appendix B. Nomenclature
[14] Y.J. Kang, C. Yang, X.Y. Huang, Int. J. Eng. Sci. 40 (2002) 2203–2221.
[15] A. Bhattacharrya, J.H. Masliyah, J. Yang, J. Colloid Interface Sci. 261
a Half channel height (m) (2003) 12–20.
b Half channel width (m) [16] J. Yang, A. Bhattacharrya, J.H. Masliyah, D.Y. Kwok, J. Colloid Inter-
u Fluid velocity (m s−1 ) face Sci. 261 (2003) 21–31.
U Reference velocity (m s−1 ) [17] D. Erickson, D. Li, Langmuir 19 (2003) 5421–5430.
P0 Induced backpressure (N m−2 ) [18] Marcos, C. Yang, T.N. Wong, K.T. Ooi, Int. J. Eng. Sci., in press.
[19] P.H. Paul, Electrokinetic Pump Applications in Micro-Total Analysis
t Time (s) Systems, Kluwer Academic, Boston, 2000.
ω Oscillation angle frequency of applied electric field [20] J.V. Beck, K.D. Cole, A. Haji-Sheikh, B. Litkouhi, Heat Conduction
(rad s−1 ) Using Green’s Functions, Taylor & Francis, London, 1992.
698 Marcos et al. / Journal of Colloid and Interface Science 275 (2004) 679–698

[21] C. Yang, in: F. Tay (Ed.), Microfluidics and BioMEMS, Kluwer Aca- [25] D. Burgreen, F.R. Nakache, J. Phys. Chem. 68 (1964) 1084–
demic, Dordrecht, 2002. 1091.
[22] C. Yang, Attachment of Fine Bubbles onto a Solid Surface and Elec- [26] C.L. Rice, R.J. Whitehead, Phys. Chem. 69 (1965) 4017–4023.
trokinetics of Gas Bubbles, Ph.D. dissertation, University of Alberta, [27] S. Levine, J.R. Marriott, G. Neale, N. Epstein, J. Colloid Interface
1999. Sci. 52 (1975) 136–149.
[23] C. Yang, D. Li, Colloids Surf. A 143 (1998) 339–353. [28] D.P. Telionis, Unsteady Viscous Flow, Springer-Verlag, New York,
[24] C. Yang, D. Li, J. Colloid Interface Sci. 194 (1997) 95–107. 1981.

You might also like