Journal of Colloid and Interface Science 233, 12–22 (2001)

doi:10.1006/jcis.2000.7262, available online at http://www.idealibrary.com on

Electro-Viscous Effects on Liquid Flow in Microchannels

Liqing Ren, Dongqing Li,1 and Weilin Qu
Department of Mechanical & Industrial Engineering, University of Toronto, Toronto, Ontario, Canada M5S 3G8

Received June 8, 2000; accepted September 28, 2000

pattern as Tuckerman’s results. Urbanek et al. (3) investigated

The presence of the electrical double layer near a solid-liquid in- liquid (1-,2-propanol and 1-,3-pentanol) flow through 5, 12, and
terface results in the electro-viscous effect on pressure-driven liquid 25 µm hydraulic diameter microchannels with liquid tempera-
flow through microchannels. The objective of this paper is to exam-
ture varying from 0 to 85◦ C . The friction coefficient dependence
ine the magnitude of the additional flow resistance caused by the
electrokinetic effect in microchannels. Deionized ultrafiltered wa-
on fluid temperature and channel size was reported. Pfahler (4)
ter, 10−4 and 10−2 M aqueous KCl solutions, 10−4 M AlCl3 solution, measured the friction coefficient in microchannels. Nonconven-
and 10−4 M LiCl solution were used as the testing liquids. Carefully tional flow behavior was observed for both isopropanol and sil-
designed flow measurements were conducted in three silicon mi- icon oil . His results indicate that polar fluids (isopropanol) and
crochannels with a height of 14.1, 28.2, and 40.5 µm, respectively. nonpolar fluids (silicon oil) behave differently, and that the po-
The measured dP/dx for the pure water, the 10−4 M KCl solution, lar nature of the fluid may play a role in microchannel flow.
and the 10−4 M LiCl solution was found to be significantly higher An experimental study by Mala et al. (5) showed that the flow
than the prediction of the conventional laminar flow theory at the behavior of water and electrolyte solutions in glass and silicon
same Reynolds number. Such a high flow resistance and the result- microchannels depends on channel material, channel size, and
ing high apparent viscosity strongly depend on the channel’s height, liquid properties. Mala and Li (6) also showed that the pressure
the ionic valence, and the concentration of the liquids. The zeta
drop of water flow through microtubes is significantly higher
potentials for the liquid-solid systems were calculated by using the
measured streaming potential data. The experimentally determined
than that predicted by the conventional theory. Under the same
dP/dx ∼ Re relationships were compared with the predictions of a flow rate and for the same diameter, the pressure drop in metal
theoretical electro-viscous flow model, and a good agreement was tubes is lower than that in silica glass tubes. Generally, the ex-
found for pure water, 10−4 M KCl solution, and 10−4 M AlCl3 solu- perimental data of friction coefficient and apparent viscosity in
tion systems. The present electrokinetic flow model cannot interpret these previous investigations are contrary to the conventional
the flow characteristics of the LiCl solution. °C 2001 Academic Press theories. The smaller the channels, the larger the differences.
Key Words: electrokinetic flow; microchannel; electro-viscous One possible explanation to these unusual behaviors of the
effect; streaming potential; electrical double layer. microchannel flow is the interfacial electrokinetic or the electro-
viscous effect. It is known that most solid surfaces have electro-
static charges, i.e., an electrical surface potential. If the liquid
INTRODUCTION contains a very small number of ions (for instance, due to impuri-
ties), the electrostatic charges on the nonconducting solid surface
The recently rapid development of MEMS and microfluidics will attract the counterions in the liquid. The rearrangement of
makes it highly desirable to understand the fundamental charac- the charges on the solid surface and the balancing charges in the
teristics of electrokinetic flow in microchannels. An important liquid is called the electrical double layer (EDL) (7). Because of
issue is the deviation of the flow resistance from the prediction the electrostatic interaction, the counterion concentration near
of the conventional fluid mechanics. Tuckerman and Pease (1) the solid surface is higher than that in the bulk liquid far away
employed rectangular microchannels as high-performance heat from the solid surface. Immediately next to the solid surface,
sinks for cooling electronic components. They found that the ex- there is a layer of ions that are strongly attracted to the solid sur-
perimental results of flow friction were slightly higher than those face and are immobile. This layer is called the compact layer,
predicted by classical theories. Several other investigators con- normally less than 1 nm thick. From the compact layer to the
firmed this similar, anomalous behavior for microchannel flow uniform bulk liquid, the counterion concentration gradually re-
in their studies. Rahman and Gui (2) measured water flow in sili- duces to that of bulk liquid. Ions in this region are affected less by
con wafer microchannels with different ratios of height to width. the electrostatic interaction and are mobile. This layer is called
Their results showed that the friction coefficient has the same the diffuse layer of the EDL. The thickness of the diffuse layer
depends on the bulk ionic concentration and electrical proper-
1 To whom correspondence should be addressed. E-mail: dli@mie.utoronto. ties of the liquid, ranging from a few nanometers for high ionic
ca. concentration solutions up to 1 mm for distilled water and pure

0021-9797/01 $35.00 12
Copyright ° C 2001 by Academic Press
All rights of reproduction in any form reserved.
 ELECTRO-VISCOUS EFFECTS ON LIQUID FLOW 13

organic liquids. The boundary between the compact layer and In practice, microchannels used in microfluidic devices are
the diffuse layer is usually referred to as the shear plane. The made by modern micromachining technology. The cross section
electrical potential at the solid-liquid surface is difficult to mea- of these microchannels is close to a rectangular shape. In such a
sure directly. The electrical potential at the shear plane, however, situation, the two-dimensional channel cross-section shape and
is called the zeta potential, ζ , is a property of the solid-liquid especially the corner of the channel have important contributions
pair, and can be measured experimentally (7). to the EDL field, subsequently to the flow field. A recent model
When a liquid is forced through a microchannel under an ap- developed by Yang and co-workers (11–13) revealed significant
plied hydrostatic pressure, the counterions in the diffuse layer electro-viscous effects on liquid flow in rectangular microchan-
(mobile part) of the EDL are carried toward the downstream end, nels in terms of the aspect ratio of the channel’s cross section,
resulting in an electrical current in the pressure-driven flow di- the surface potential, and the ionic concentration.
rection. This current is called the streaming current. Correspond- A key question: Is the interfacial electro-viscous effect as
ing to this streaming current, there is an electrokinetic potential large as predicted by these theoretical models? So far there has
called the streaming potential. This flow-induced streaming po- been no direct experimental verification to these models. The
tential is a potential difference that builds up along a microchan- objectives of this paper are to examine if the electro-viscous
nel. This streaming potential acts to drive the counterions in the effect can be measured, and to compare the measured effect with
diffuse layer of the EDL to move in the direction opposite to the theoretical model prediction. To do so, careful experimental
the streaming current, i.e., opposite to the pressure-driven flow studies were conducted. If the nonconventional flow behavior is
direction. The action of the streaming potential will generate an due to the interfacial electrokinetic or electro-viscous effects, it
electrical current called the conduction current. It is obvious that must depend on the ionic concentrations and the ionic valence
when ions move in a liquid, they will pull the liquid molecules to of the testing liquids. For this purpose, pure water, aqueous KCl
move with them. Therefore, the conduction current will produce solutions of two different concentrations, an AlCl3 solution, and
a liquid flow in the opposite direction to the pressure driven flow. a LiCl solution were chosen as the testing liquids. It was also
The overall result is a reduced flow rate in the pressure drop di- noted that micromachined channels on glass or silicon plates
rection. If the reduced flow rate is compared with the flow rate have a trapezoidal cross section and relatively rough surface (14).
predicted by the conventional fluid mechanics theory without To avoid the complication due to cross-section geometry and the
considering the presence of the EDL, it seems that the liquid surface roughness of the flow channels, microchannels formed
would have a higher viscosity. This is usually referred to as the by two parallel, smooth silicon plates (the same material used in
electro-viscous effect (7). making most MEMS and microfluidic devices) were used in our
Generally, for macrochannel flow the EDL effects can be studies. The experimental data show significant electro-viscous
safely neglected, as the thickness of the EDL is very small com- effects and were compared with an electrokinetic flow model.
pared with the characteristic size of channels. However, for di-
lute electrolyte solutions flowing in microchannels, the thickness EXPERIMENTAL
of the EDL may be comparable with the characteristic size of
flow channels. Thus the electro-viscous effect described above Figure 1 shows the experimental system used to study the
should be considered. Because the electro-viscous effect origi- electrokinetic effect on the flow characteristics of a liquid
nates from the EDL field which in turn depends on the material
of the channel wall, the size and shape of the channel, and the
ionic concentration, dielectric constant, and other properties of
the liquid, it is not difficult to understand that microchannel
flow may depend on the liquid and on the material and size of
the channel, as reported in the literature.
In the literature there are some theoretical models considering
the EDL effects on microchannel flow characteristics. Burgreen
and Nakache (8) studied the effect of the surface potential on
liquid transport through fine capillary slits by using an approx-
imation valid only for cases of low surface potentials. Rice and
Whitehead (9) discussed the same problem in narrow cylindri-
cal capillaries. Levine et al. (10) extended Rice and Whitehead’s
model for cylindrical capillaries to a higher surface potential by
developing an approximate solution. Levine’s model predicted
significantly high electro-viscous effects, the ratio of the ap-
parent viscosity to the true viscosity µa /µ can be as high as
1.3, depending on the channel size, the zeta potential, and ionic FIG. 1. Schematic of the experimental system used to measure the flow in
concentration, etc. microchannels.
14 REN, LI, AND QU

flowing through a microchannel. This system consists of a flow The width and length of the microchannel can be accurately
loop, a test section including a slit microchannel, instruments for measured by using a precision gauge. The height of the mi-
measuring flow and electrokinetic parameters, and a computer crochannels was first directly measured by a microscope (Leica
data acquisition system. MS5 Stereomicroscope)–computer image analysis system with
Deionized ultrafiltered water (DIUF) (Fisher Scientific) and a resolution of 0.8 µm. Then the channel height was calibrated
aqueous KCl solutions of two different concentrations, a AlCl3 by an indirect method that involves the flow of a high ionic con-
solution, and a LiCl solution were used as the testing liquids. The centration solution through the microchannel. For a high con-
concentrations of the KCl solutions are 10−4 M (kmol/m3 ) and centration electrolyte solution, the EDL thickness is very small
10−2 M (kmol/m3 ), respectively. The concentration of both the (about a few nanometers) and the electrokinetic effect on the
AlCl3 and the LiCl solutions is 10−4 M (kmol/m3 ). In the exper- flow is negligible. The liquid flow in such a case is basically
iments, the testing liquid was pumped from a liquid reservoir to a Poiseuille laminar flow. Therefore, the channel height can be
the flow loop by a high precision displacement pump (Ruska determined from the measured pressure drop and flow rate by
Instruments, Model 2248-WII) which has a flow rate range using the Poiseuille flow equation. The channel heights deter-
of 2.5–560 cm3 /h and can generate a pressure up to 4000 psi mined in this way were used in this study. It was found that
(27.6 MPa). A 0.1-µm filter was installed in the flow loop be- the difference between the measured channel heights from these
tween the outlet of the pump and the inlet of the test section. two methods was less than 0.5 µm.
The liquid was forced to flow through this submicrometer filter A microchannel was placed in a two-part symmetrical Plex-
before entering the test section to avoid any particles or bub- iglas assembly to form a test section, as shown in Fig. 1. The
bles from flowing through the test section and blocking the mi- epoxy resin was applied to bond the microchannel and the as-
crochannel. In order to minimize the environmental electrical sembly together to avoid leaking. It was found that the height
interference on the measurement of the electrokinetic parame- of the microchannel may be altered if the pressure of the liquid
ters, such as the streaming potential across the microchannel and in the microchannel were too high. This is because neither the
the bulk conductivity of the liquid, the whole flow loop is made clapper nor the epoxy resin can stand very high pressure and de-
of plastic tubes and plastic valves. formation may happen. This limited our experiments to a small
The silicon plates (30 mm in length, 14 mm in width and Reynolds number range especially for smaller microchannels
1 mm in thickness) used to form the microchannels in this study (maximum height change is less than 0.5 µm).
were supplied by Alberta Microelectronic Center (Edmonton, Two pairs of sumps were machined in the assembly and were
Canada). The surface roughness of these plates is approximately used for the pressure drop and streaming potential measurement.
20 nm. To form a microchannel, first two strips of a thin plastic A diaphragm-type differential pressure transducer (Validyne
shim (Small Parts Inc.) were used as the spacer and put be- Engineering Corp., Model DP15) with ±0.5% FS accuracy was
tween a pair of silicon plates in the length direction along the connected to one pair of sumps to measure the pressure drop
sides of the plates, so that a flow passage of 5 mm width was along the microchannel. The pressure transducer was calibrated
formed. Then, a specially designed clapper was used to fix the by using a standard deadweight pressure source before use in
relative position of the plates and the thin spacers. Finally, epoxy the experiments. The details of the calibration procedures and
resin was applied to bond the clapper and the silicon plates to- the results can be found elsewhere (15). The flow-induced elec-
gether and to seal all the openings except the inlet and outlet trokinetic potential, the streaming potential, was measured by a
of the microchannel. The cross section of such a microchannel pair of Ag/AgCl electrodes (Sensortechnik Meinsberg GmbH)
is illustrated in Fig. 2. Three microchannels were made in this and an electrometer (Keithley Instruments Inc., Model 6517).
way and tested in this study. The channels have identical width The volume flow rate of water flowing through the microchan-
and length, which are 5 and 30 mm, respectively, and different nels was measured by the weighting method as described below.
heights. By choosing different shim thickness, the heights of the The liquid exiting the test section was accumulated in a glass
three microchannels are 14.1, 28.2, and 40.5 µm, respectively. beaker whose weight was measured before. A stopwatch was
employed to measure the time spent for the accumulation. Then,
an electronic balance (Mettler Instrument AG, Model BB240)
with an accuracy of 0.001 g was used to measure the weight
of the accumulated liquid. Usually 1 to 1 g of the liquid was
collected over approximately 20 to 30 min, depending on the
channel size and the flow rate. Evaporation effect was examined
and found to be negligible. The total volume of the liquid was
determined by dividing the weight by the liquid’s density. The
flow rate was then obtained by dividing the total volume of the
accumulated liquid by the time. The accuracy of the flow rate
measurement was estimated to be ±2%.
FIG. 2. Illustration of the microchannel formed between two parallel silicon In an experiment, the pump was set to maintain a constant flow
plates. rate. The readings of the pressure drop along the microchannel
 ELECTRO-VISCOUS EFFECTS ON LIQUID FLOW 15

were monitored and recorded. The flow was considered to have the exit is then
reached a steady state when the readings of the pressure drop
did not change any more. At such a steady state, the flow rate, 1Pnet = 1Pmeasured − 1Pin − 1Pout . [5]
the pressure drop, the streaming potential, and the bulk liquid
conductivity were measured. The data reported in this paper are The pressure drop reported in this paper is the net pressure drop.
for steady state flow. For a given channel and a given testing
liquid, the measurements for all the parameters were repeated THEORETICAL ELECTRO-VISCOUS FLOW MODEL
at least twice for the same flow rate and in the same flow di-
rection. The flow direction was then switched by adjusting the Electrical Double Layer in a Rectangular Microchannel
control valves, and the measurements were conducted for the According to the theory of electrostatics, the relationship be-
same flow rate (i.e., the same pump setting) after the steady tween the electrical potential, ψ, and the net charge density per
state was reached. Again, the measurements were repeated at unit volume, ρe , at any point in the liquid is described by the
least twice. After the measurements for both flow directions two-dimensional Poisson equation,
were completed, the pump was set to a different flow rate and
the measurements described above were repeated for the same ∂ 2ψ ∂ 2ψ ρe
microchannel, and so on. + =− , [6]
∂ y2 ∂z 2 ε0 ε
When changing a different testing liquid, the flow loop and
the test section were flushed thoroughly by DIUF water and where ε is the dielectric constant of the solution and ε0 is the per-
then the testing liquid for several hours to remove all the ions mittivity of vacuum. Assuming that the equilibrium Boltzman
and/or other possible contamination left from the previous test. distribution is applicable, which implies uniform dielectric con-
Because the electrical conductivity of the liquid is very sensitive stant and neglect of fluctuation, the number of ion distribution
to an even very small change in the ionic concentration, an online in a symmetric electrolyte solution is of the form
electrical conductivity sensor (Model CR 7300, Metter-Toledo
µ ¶
Process Analytical) was used as a monitor. The flushing process z i eψ
was continued until a steady liquid conductivity reading was n i = n i∞ exp − , [7]
kb T
achieved and was the same as the standard value for that liquid.
Consider a rectangular microchannel of width 2W , height 2H , where n i∞ and z i are the bulk ionic concentration and the valence
and length L, as illustrated in Fig. 2. In the entrance region, the of type i ions respectively, e is the charge of a proton, kb is
liquid flow was not fully developed laminar flow. The entrance Boltzmann’s constant, and T is absolute temperature. The net
region length is given by (16) volume charge density ρe is proportional to the concentration
µ ¶ difference between cations and anions, via
ρu m Dh µ ¶
L in = 0.02(2H )Re Re = , [1] zeψ
µ ρe = ze(n + − n − ) = −2zen ∞ sinh . [8]
kb T
where Dh = 4H W/(H + W ) is the hydraulic diameter of the
rectangular channel, and u m is the mean velocity. In this region, Substituting Eq. [8] into the Poisson equation, the well-known
the pressure drop is calculated by Poisson-Boltzmann equation is obtained:
µ ¶
kin · ρ · L in 2 ∂ 2ψ ∂ 2ψ 2n ∞ ze zeψ
1Pin = um, [2] + = sinh . [9]
2Dh ∂ y2 ∂z 2 εε0 kb T

where kin is the friction coefficient given by (16) By defining the Debye-Hückel parameter as k 2 = 2z 2 e2 n ∞ /
εε0 kb T (1/k is normally referred to as the EDL thickness), the
· ¸ hydraulic diameter of the rectangular microchannel as
96 1 0.774 0.00089
kin = + − . [3]
Re Re [L in /2(2H )Re] [L in /2(2H )Re]2
Dh = 4H W/(H + W )
At the exit of the flow, the cross section is greatly increased as
the liquid leaves the slit microchannel and enters a big channel and introducing the dimensionless variables,
(about 10 mm in diameter). The liquid flow becomes turbulent.
The pressure loss at the exit is estimated by y ∗ = y/Dh , z ∗ = z/Dh , and ψ ∗ = zeψ/kb T,

ρ 2 the above equation can be nondimensionalized as

1Pout = u . [4]
2 m
∂ 2ψ ∗ ∂ 2ψ ∗
+ = k 2 Dh2 sinh(ψ ∗ ). [10]
The net pressure drop without the losses at the entrance and at ∂ y ∗2 ∂z ∗2
16 REN, LI, AND QU

Because of symmetry of the EDL in a rectangular channel, (H − 1/k) and (W − 1/k) instead of 0. That is,
Eq. [10] is subjected to the following boundary conditions in a
quarter of the rectangular cross section: ZH ZW
Is = 4 u(y, z)ρe (y, z) dy dz. [16]
∗ ∂ψ ∗ H zeζ
y = 0, = 0, y ∗ = , ψ∗ = [11a] H −1/k W −1/k
∂ y∗ Dh kb T
∂ψ ∗ W zeζ The conduction current, the transport of the excess charge in the
z ∗ = 0, ∗
= 0, z ∗ = , ψ∗ = . [11b]
∂z Dh kb T EDL region driven by the electrokinetic potential, is given by
· ¸
Where the zeta potential ζ is a measurable electrical potential λs Ps
at the shear plane, i.e., the boundary between the compact layer Ic = λ0 E x Ac + λs E x Ps = λ0 E x Ac 1 + , [17]
λ0 A c
and the diffuse layer of the EDL.
where λ0 is the electrical conductivity of the liquid, Ac is the
Flow Field in a Rectangular Microchannel cross-sectional area of the channel, λs is the surface conduc-
The general equation of motion for laminar conditions in a tivity, and Ps is the wetting parameter of the channel. Once again,
liquid with constant density and viscosity is given by since the net charge density is essentially zero outside of the EDL
region whose characteristic thickness is given by 1/k, the effec-
∂ ū tive area of the rectangular channel’s cross section for the bulk
ρ + ρ ū · ∇ ū = −∇ P + µ∇ 2 ū + F̄. [12] conduction current is approximately
∂t
µ ¶
Assuming the flow is steady, two-dimensional, and fully devel- 1 1 1
A = 2 2W + 2H = 4(W + H ) .
oped, the velocity components satisfy u = u(y, z) and v = w = k k k
0. Then both the time term, ∂ ū/∂t, and the inertia term, ū · ∇ ū,
vanish. Also, the hydraulic pressure P is a function of x only The wetting parameter Ps for the rectangular microchannel is
and the pressure gradient, dP/dx, is constant. If the gravity effect given by
is negligible, the body force, F̄, is only from an induced elec-
trical field, E x (V/m). Under the above conditions, the general Ps = 4(W + H ).
equation of motion is reduced to
So the conduction current in the rectangular microchannel can
∂ 2u ∂ 2u 1 dP 1 be expressed as
+ 2 = − E x ρe . [13]
∂y 2 ∂z µ dx µ
· ¸
1 λs
Ic = 4λ0 E x (W + H ) 1+ k . [18]
In Eq. [13], the electrokinetic potential E x can be obtained k λ0
through the balance between streaming current and electrical
conduction current at the steady state. At a steady state, the net Putting Eq. [16] and Eq. [18] into the steady state condition,
electrical current should be zero, which means Eq. [14] yields

I = Is + Ic = 0. [14] RH RW
H −1/k W −1/ku(y, z)ρe (y, z) dy dz
Ex = − £ ¤ . [19]
λ0 (W + H ) k1 1 + λλ0s k
Because of symmetry, the electrical streaming current, transport
of the net charge in the EDL region with the liquid flow, is given
by Substituting Eq. [19] for E x and Eq. [8] for ρe into Eq. [13] and
employing the following non dimensional parameters
ZH ZW y z x
Is = 4 u(y, z)ρe (y, z) dy dz. [15] y∗ = , z∗ = , x∗ =
Dh Dh ρ Dh U/µ
2
0 0
u P − P0 4z 2 e2 n 2∞ Dh4
u∗ = , P∗ = , M = £ λs
¤,
Realizing that the net charge density is non-zero essentially only U ρU 2 µλ0 (W + H ) k1 1 + λ0
k
in the EDL region whose characteristic thickness is given by
1/k (k is the Debye-Hückel parameter as defined previously), where U is a reference velocity, and ρ0 is a reference pressure,
the lower boundary of the above integration may be changed to then the following nondimensional equation of motion can be
 ELECTRO-VISCOUS EFFECTS ON LIQUID FLOW 17

obtained:

∂ 2u∗ ∂ 2u∗ d P∗
+ = + M sinh(ψ ∗ )
∂ y ∗2 ∂z ∗2 dx∗
Z h
H/D Z h
W/D

× u ∗ sinh(ψ ∗ ) dy ∗ dz ∗ . [20]
H/Dh −1/k Dh W/Dh −1/k Dh

The above equation of motion is subjected to the following

boundary conditions:

∂u ∗ H
y ∗ = 0, ∗
= 0, y ∗ = , u∗ = 0 [21a]
∂y Dh
∂u ∗ W
z ∗ = 0, = 0, z ∗ = , u ∗ = 0. [21b]
∂z ∗ Dh

Numerically solving the Poisson-Boltzmann equation,

Eq. [10], and the equation of motion, Eq. [20], with the boundary FIG. 3. Experimentally determined pressure gradient and Reynolds number
conditions specified in Eqs. [11] and [21], both the EDL field for a microchannel of a height 14.1 µm.
and the velocity field in the rectangular microchannel can be
determined. The EDL or electrokinetic effects on the flow char- for the high concentration solution where there is no EDL ef-
acteristics such as the dP/dx vs Re relationship can be predicted. fect on flow. However, the difference in dP/dx between the pure
water and the 10−2 M solution diminishes as the channel height
RESULTS AND DISCUSSION increases. Furthermore, for all the three microchannels, there
are clear differences between the dP/dx ∼ Re relationships for
Using the measured volume flow rate, Q, and the measured the 10−4 M solution and that for the pure water. That is, at the
pressure drop, the pressure gradient and the Reynolds number same Reynolds number, the dP/dx for the pure water is always
can be calculated as follows, higher than that of the 10−4 M solution. These experimental
results show a strong dependence of the flow characteristics in
dP 1Pnet microchannels on the channel size and on the ionic concentration
= [22]
dx L of the liquid.
ρ Q Dh
Re = , [23]
µAc

where 1Pnet is determined from Eq. [5], L is the length of the

channel (30 mm), Dh is the hydraulic diameter of the channel,
and Ac is the cross-sectional area of the channel. For the mi-
crochannels of three different heights, the experimentally deter-
mined pressure gradient and the Reynolds numbers were plotted
in Figs. 3, 4, and 5. Generally, for a given liquid, there are six data
points for one and approximately the same Reynolds number.
All data reported in this paper are at 21◦ C.
For KCl solutions at the high concentration 10−2 M (i.e.,
kmol/m3 ), the EDL thickness is approximately several nanome-
ters. The EDL effect on flow can be neglected as the EDL
thickness is so small in comparison with the channel height.
The flow in the microchannels is considered as the conventional
Poiseuille laminar flow. However, the EDL thickness is approx-
imately 100 nm for the 10−4 M solution and 1 µm for the DIUF
water, respectively. As seen clearly from all these figures, for the
lower concentration (10−4 M) solution and the pure water, the
measured pressure gradients at the same Reynolds number (i.e., FIG. 4. Experimentally determined pressure gradient and Reynolds number
the same flow rate) are significantly (up to 20%) higher than that for a microchannel of a height 28.2 µm.
18 REN, LI, AND QU

FIG. 6. Apparent viscosity ratio vs nondimensional electrokinetic channel

height for DIUF water.
FIG. 5. Experimentally determined pressure gradient and Reynolds number
for a microchannel of a height 40.5 µm.
netic channel height k H . Since k H = H/(1/k), where 1/k is
the characteristic thickness of the EDL, k H can be understood
KCl solutions have symmetrical 1 : 1 ions (the valence is 1). as the channel height relative to the EDL thickness. Within the
However, the ions in the AlCl3 solution are not symmetrical and limited experimental data range, the electro-viscous effect can
the valence of Al+ ions is 3. Comparing the 10−4 M KCl solution produce an apparent viscosity from several percent up to 18%
with the 10−4 M AlCl3 solution, we see from Figs. 3, 4, and 5 that higher than the true viscosity of the liquid, depending on the
the measured pressure gradient for the 10−4 M AlCl3 solution electrokinetic channel height. These results agree with Levine’s
is always lower than that for the 10−4 M KCl solution under model prediction (10) qualitatively.
the same Re number. In fact, the dP/dx ∼ Re relationship of the The flow-induced streaming potential was measured in this
10−4 M AlCl3 solution is essentially the same as that of 10−2 M study. Figures 9, 10, 11, and 12 show the measured streaming
KCl solution. This clearly shows that the flow characteristics in potential vs the Reynolds number for the 10−4 M KCl, AlCl3 , and
microchannels depend not only on the ionic concentration but LiCl solutions and the DIUF water. For the 10−2 M KCl solution,
also on the valence and possibly the symmetry of the ions. the streaming potential is close to zero and not plotted here.
As seen from Figs. 3, 4, and 5, the 10−4 M LiCl solution
has a very unusual behavior. The measured pressure gradient
of the 10−4 M LiCl solution is always significantly higher than
that of pure water under the same Re number, independent of
the channel size. Considering the LiCl solution has symmetri-
cal 1 : 1 ions, one would expect that its flow characteristics are
very similar to those of the KCl solution. It is difficult for us to
understand the observed behavior.
As discussed in the introduction, because of the presence of
the EDL, the pressure-driven flow induces an electrokinetic po-
tential, the streaming potential. The streaming potential in turn
will generate a conduction electrical current and hence a liq-
uid flow opposite to the pressure-driven flow. The net result is
a reduced flow rate in the pressure-driven flow direction un-
der a given pressure drop or a higher pressure drop for a given
flow rate, in comparison with the conventional Poiseuille flow
model. This is the electro-viscous effect. Using the conventional
Poiseuille laminar flow equation, the measured pressure gradi-
ent and the flow rate, we can calculate the apparent viscosity
µa . Figures 6, 7, and 8 show the ratio of the apparent viscosity FIG. 7. Apparent viscosity ratio vs nondimensional electrokinetic channel
to the true viscosity µa /µ as a function of the half electroki- height for 10−4 M aqueous KCl solution and 10−4 M LiCl aqueous solution.
 ELECTRO-VISCOUS EFFECTS ON LIQUID FLOW 19

FIG. 8. Apparent viscosity ratio vs nondimensional electrokinetic channel FIG. 10. Measured streaming potential vs Reynolds number for 10−4 M
height for 10−4 M AlCl3 aqueous solution. aqueous KCl solution in the three microchannels.

In all these figures, each point represents an average value of viscosity of the liquid, and H is the half of the channel height.
six independent measurements, with an error of approximately Using the measured streaming potential and the measured pres-
±5%. As seen from these figures, for a given liquid, the absolute sure drop for different channel heights, the zeta potential and the
value of streaming potential increases as the Re increases. This is surface conductance can be determined by using this equation.
simply because the flow-induced streaming potential increases The detailed procedures can be found elsewhere (17). However,
with the flow. In the derivation of Eq. [24], the electro-viscous effect was not
In the theory of electrokinetics (17), the streaming potential considered. This approach to determine the zeta potential and
E x and the pressure drop 1P across a slit capillary channel are the surface conductance is not suitable for the electrokinetic
related to the zeta potential ζ through the following equation, flow in small microchannels in the present work. In this study,
Eqs. [10], [19], and [20] were used to predict the E x ∼ Re re-
Ex εζ lationship. By minimizing the difference between the predicted
= , [24]
1P µ(λ0 + λs /H ) E x ∼ Re relationship and the measured E x ∼ Re data as shown
in Figs. 9–12, the zeta potential and the surface conductance are
where λ0 is the electrical conductivity of the liquid, λs is the determined. The results are summarized in Table 1. As seen
surface conductance, ε and µ are the dielectric constant and from Table 1, for a given solution, the surface conductivity

FIG. 9. Measured streaming potential vs Reynolds number for DIUF water FIG. 11. Measured streaming potential vs Reynolds number for 10−4 M
in the three microchannels. aqueous AlCl3 solution in the three microchannels.
20 REN, LI, AND QU

FIG. 12. Measured streaming potential vs Reynolds number for 10−4 M

aqueous LiCl solution in the three microchannels.
FIG. 13. Comparison of the experimentally determined dP/dx ∼ Re rela-
tionships with the predictions of the electro-viscous flow model for a microchan-
nel of a height 14.1 µm.
increases slightly with the channel height. This phenomenon
was reported by others as well (7, 18, 19).
In this study, the electrokinetic flow model, Eq. [10] and by the model is essentially negligible. When the surface con-
Eq. [20], were solved numerically. The model predictions were ductance is not considered in the model (Eqs. [17] and [18]), the
compared with the experimental data, as shown in Figs. 13, 14, predicted dP/dx by the model for the pure water and the 10−4 M
and 15. It was found that the dP/dx ∼ Re relationship of the KCl solution agrees well with the experimental data points. The
10−2 M KCl solution can be predicted very well by the model predictions presented in Figs. 13, 14, and 15 (the solid
conventional Poiseuille laminar flow model, indicating a lines) did not consider the surface conductance effect. These
negligible EDL effect for the high ionic concentration solution. results may lead us to reconsider in the future the surface con-
Furthermore, the dP/dx ∼ Re relationship of the 10−2 M KCl ductance and the way it has been introduced in the electrokinetic
solution is very close to that of the 10−4 M AlCl3 solution. flow model.
Therefore, the experimental data and the model prediction for
the 10−2 M KCl solution were not plotted in Figs. 13, 14, and
15 for graphic reasons.
It was found that for the same Re number, the predicted dP/dx
by the model for the pure water and the 10−4 M KCl solution
is too small in comparison with the measured dP/dx data, if the
surface conductance is considered. In other words, if the surface
conductance is considered, the electro-viscous effect predicted

TABLE 1
The Zeta Potential and the Surface Conductance Determined
by Using Eqs. [10], [19], and [20] and the Measured Streaming
Potential Data
Channel size

Solution 2H1 = 14.1 µm 2H2 = 28.2 µm 2H3 = 40.5 µm

DIUF water ζ = 245 mV ζ = 245 mV ζ = 245 mV

λs = 2.43 × 10−8 S λs = 7.48 × 10−8 S λs = 9.25 × 10−8 S
10−4 M KCl ζ = 107 mV ζ = 107 mV ζ = 107 mV
λs = 1.76 × 10−7 S λs = 4.15 × 10−7 S λs = 6.45 × 10−7 S
10−4 M AlCl3 ζ = 60 mV ζ = 60 mV ζ = 60 mV FIG. 14. Comparison of the experimentally determined dP/dx ∼ Re rela-
λs = 0.215 × 10−7 S λs = 4.8 × 10−7 S λs = 6.2 × 10−7 S tionships with the predictions of the electro-viscous flow model for a microchan-
nel of a height 28.2 µm.
 ELECTRO-VISCOUS EFFECTS ON LIQUID FLOW 21

net charge density in the EDL region of the DIUF water is signif-
icantly higher than that of the 10−4 M KCl solution, as shown in
Fig. 16. Furthermore, the bulk conductivity of the 10−4 M KCl
solution is about one order of magnitude higher than that of wa-
ter (measured values: λ0DIUFwater@21◦ C = 1.053 × 10−4 S/m and
λ010−4 MKCl@21◦ C = 15.025 × 10−4 S/m). The more conductive
the liquid, the smaller the streaming potential that can be gen-
erated. Figures 9 and 10 also confirm this fact. Therefore, as
discussed above, for the 10−4 M KCl solution system, a smaller
streaming potential and a smaller net charge density will result
in a smaller electrokinetic body force (with an opposite sign to
the pressure gradient) in the flow equation, Eq. [13]. This is the
reason that the electro-viscous effect in the 10−4 M KCl solution
system is smaller that in the DIUF water system.
The model was also used to predict the dP/dx ∼ Re relation-
ship for the 10−4 M AlCl3 solution. Because of the nonsymmet-
rical ions, Eq. [8] and Eq. [10] for the 10−4 M AlCl3 solution
are changed into
· µ ¶ µ ¶¸
−3eψ eψ
FIG. 15. Comparison of the experimentally determined dP/dx ∼ Re rela- ρe = 3en ∞ exp − exp
tionships with the predictions of the electro-viscous flow model for a microchan-
kb T kb T
nel of a height 40.5 µm. ∂ 2ψ ∗ ∂ 2ψ ∗ 1
∗2
+ = k 2 Dh2 [exp(ψ ∗ ) − exp(−3ψ ∗ )],
∂y ∂z ∗2 4
As seen in Figs. 13, 14, and 15, for the DIUF water and the
10−4 M KCl solution, the predicted dP/dx ∼ Re lines agree where ψ ∗ and the Debye-Hückel parameter are given by
reasonably well with the experimental data points. Comparing
· ¸1/2
the DIUF water and 10−4 M KCl solution, the water-silicon 12n ∞ e2 eψ
surface system has a much higher zeta potential (ζDIUF water = k= and ψ∗ = .
εε0 kb T kT
245 mV and ζ10−4 MKCl = 107 mV) and a much thicker
EDL field ( k1 |DIUFwater@21◦ C = 305 nm and k1 |10−4 MKCI@21◦ C = The 10−4 M AlCl3 solution system has a smaller zeta potential
30.5 nm). The EDL field is stronger in the DIUF water system (60 mV), a smaller EDL thickness ( k1 |10−4 MAIC3 @21◦ C =
than in the 10−4 M KCl solution system. Correspondingly, the 12.45 nm) and a much higher bulk conductivity
(λ0 10−4 MAlCl3 @21◦ C = 47.028 × 10−4 S/m), in comparison with
the DIUF water system and the 10−4 M KCl solution system.
That is why the electro-viscous effect on the 10−4 M AlCl3
solution system is very small and the dP/dx ∼ Re relationship
of the 10−4 M AlCl3 solution is closer to that of the 10−2 M KCl
solution.
For the LiCl solution system, the ions are 1 : 1 symmetrical.
The measured bulk conductivity is λ0 10−4 MLiCl@21◦ C =
11.95 × 10−4 S/m. The characteristic EDL thickness is
1
| −4
k 10 MLiCi@21 C
◦ = 30.5 nm. From these properties, one would

expect that the 10−4 M LiCl solution would behave similarly

to the 10−4 M KCl solution. The electrokinetic flow model
presented in this paper cannot explain the measured extra-high
flow resistance. However, it was noted that the radius ratio of
Li+ to K+ is 0.68/1.33 ≈ 0.511. We suspected that this unusual
behavior might have some thing to do with the size and the
adsorption of Li+ ions in the compact layer of the EDL.

FIG. 16. Comparison of the net charge density distribution near the channel
SUMMARY
wall between DIUF water and 10−4 M aqueous KCl solution in a microchannel
of a height 14.1 µm at 21◦ C. “0” in the horizontal axis indicates the position of The correlation among the pressure drop, the streaming poten-
the wall. tial, and the volume flow rate of DIUF water, 10−4 and 10−2 M
22 REN, LI, AND QU

aqueous KCl solutions, 10−4 M AlCl3 aqueous solution, and major cause of the significantly higher pressure drop for pure
10−4 M LiCl aqueous solution in silicon microchannels was ex- water and dilute aqueous ionic solutions flowing through small
perimentally studied. The flow resistance (i.e., dP/dx) of the microchannels.
10−2 M KCl solution and the 10−4 M AlCl3 solution is essen-
ACKNOWLEDGMENT
tially the same as that predicted by the Poiseuille laminar flow
equation. However, up to 20% higher flow resistance was found The authors acknowledge the support of a Research Grant of the Natural
for pure water and the 10−4 M KCl solution. A 20% to 40% Science and Engineering Research Council of Canada.
higher flow resistance for the LiCl solution was found in compar-
ison with the prediction of the Poiseuille laminar flow equation. REFERENCES
These results show a strong dependence of dP/dx ∼ Re rela-
1. Tuckerman, D. B., and Pease, R. F. W., IEEE Electron Device Lett. 2, 126
tionship on the channel size, the ionic concentration, the ionic
(1981).
valence, and the bulk conductivity of the liquids. The apparent 2. Rahman, M. M., and Gui, F., Adv. Electron. Packaging 199, 685 (1993).
viscosity corresponding to these measured dP/dx ∼ Re relation- 3. Urbanek, W., Zeemel, J. N., and Hau, H. H., J. Micromech. Microeng. 3,
ships was found to be up to 18% higher than the true viscosity, 206 (1993).
depending on the liquids and the ratio of the channel height to the 4. Pfahler, J. N., Liquid transport in micron and submicron size channels.
Ph.D. thesis, Univ. of Pennsylvania, 1992.
EDL thickness. The flow-induced streaming potential was also
5. Mala, G. M., Li, D., Werner, C., Jacobasch, H. J., and Ning, Y. B., Int. J.
measured in this study. The streaming potential data were used to Heat Fluid Flow 18, 489 (1997).
evaluate the zeta potential and the surface conductance of these 6. Mala, G. M., and Li, D., Int. J. Heat Fluid Flow 20, 142 (1999).
liquid-solid systems. The experimental results show an unusual 7. Hunter, R. J., “Zeta Potential in Colloid Science: Principles and Applica-
flow resistance and streaming potential behavior of LiCl solu- tions.” Academic Press, New York, 1981.
8. Burgreen, D., and Nakache, F. R., J. Phys. Chem. 68, 1084 (1964).
tion that cannot be understood at the moment. Except the LiCl
9. Rice, C. L., and Whitehead, R., J. Phys. Chem. 69, 4017 (1965).
solution, at the same flow rate, the pure water has the highest 10. Levine, S., Marriott, J. R., Neale, G., and Epstein, N., J. Colloid Interface
streaming potential, and the high concentration (10−2 M) KCl Sci. 52, 136 (1975).
solution has the lowest (essentially zero) streaming potential. 11. Yang, C., and Li, D., J. Colloid Interface Sci. 194, 95 (1997).
According to the electrkinetic theory, the higher the streaming 12. Yang, C., Li, D., and Masliyah, J. H., Int. J. Heat and Mass Transfer 41,
4229 (1998).
potential, the higher the electro-viscous effect on flow. These
13. Yang, C., and Li, D., Colloids Surf. A 143, 339 (1998).
correspond well with the measured dP/dx ∼ Re relationships. 14. Qu, W., Mala, M., and Li, D., Int. J. Heat Mass Transfer 43, 353
The experimental data were compared with an electrokinetic (2000).
flow model developed in this paper. It was found that with the 15. Mala, G. M., Heat transfer and fluid flow in microchannels. Ph.D. thesis,
consideration of the surface conductance, the model predicts University of Alberta, 1999.
16. Werner, C., Korber, H., Zimmermann, R., Dukhin, S., and Jacobasch, H. J.,
negligible electro-viscous effects. However, without consider-
J. Colloid Interface Sci. 208, 329 (1998).
ing the surface conductance, the model’s predictions agree well 17. Lykelema, J., and Minor, M., Colloids Surf. A 140, 33 (1997).
with the experimental data. The comparison confirms that the 18. Morrison, F. A., and Osterle, J. F., J. Chem. Phys. 43, 2111 (1965).
electrical double layer effect or the electro-viscous effect is the 19. Arulanandam, S., and Li, D., J. Colloid Interface Sci., in press.