Microfluid Nanofluid (2013) 15:859–870
DOI 10.1007/s10404-013-1196-7
RESEARCH PAPER
High field asymmetric waveform for ultra-enhanced
electroosmotic pumping of porous anodic alumina membranes
D. Piwowar • M. E. Tawfik • F. J. Diez
Received: 9 December 2012 / Accepted: 19 April 2013 / Published online: 7 May 2013
Ó Springer-Verlag Berlin Heidelberg 2013
Abstract An electroosmotic EO process is presented for
nanoporous membranes capable of generating EO flow
rates over thirty times higher than previously possible with
the same membrane and solution. In generating high EO
flows, a limiting factor is faradaic reactions which appear
at high electric fields. A process is presented capable of
limiting and even canceling these reactions allowing
electric field between one and two orders of magnitude
higher. This is achieved by applying an asymmetric bipolar
rectangular voltage waveform. The results show the
enhanced EO pumping capabilities of membranes under a
high electric field asymmetric waveform which prevents
gas generation at high voltages. A baseline is established
by measuring the EO pump performance when a constant
voltage is applied to SiO2-coated nanoporous anodic alu-
minum oxide membranes. The analysis compares the effect
of the applied voltage type on the maximum flow rate,
power consumption, and maximum pressure. Results show
that large gas generation prevents membrane operation
when direct current DC voltages above 50 V are applied.
On the other hand, it operates normally under an asym-
metric voltage ?1,800/-900 V applied, with negligible
gas generation. This results in a thirty-time flow rate
increase. Larger flow rates/voltages are possible but were
not considered due to hardware limitations.
Electronic supplementary material The online version of this
article (doi:10.1007/s10404-013-1196-7) contains supplementary
material, which is available to authorized users.
D. Piwowar  M. E. Tawfik  F. J. Diez (&)
Department of Mechanical and Aerospace Engineering,
Rutgers, The State University of New Jersey,
98 Brett Road, Piscataway, NJ 08854, USA
e-mail: diez@rutgers.edu
1 Introduction
An EO pump generates a flow rate or pressure change due
to the ion migration in the electric double layer toward the
electrode which causes viscous drag creating a net flow.
These pumps can be fabricated using a variety of methods
from capillary columns packed with particles (Zeng et al.
2001; Wang et al. 2006; Chen 2005) to porous membranes
(Yao et al. 2003; Chen et al. 2005; Yao et al. 2006; Cao
et al. 2012; Kwon et al. 2012; Kim et al. 2013; Garg et al.
2012). EO pumps have several advantages compared
to typical mechanical pumps. These include having no
moving or mechanical parts, being compatible with lab-on-
a-chip components and being used for accurate and
adjustable flow rate control. EO pumps are used in a
variety of applications including power electronic cooling
(Berrouche et al. 2009), fuel cells (Buie et al. 2006),
actuation (Prakash et al. 2006), drug delivery (Pikal 1992),
chromatography (Chen et al. 2004), and lab-on-a-chip
systems (Kim et al. 2013; Li and Harrison 1997).
High flow rates are of interest in many applications, but they
require high electric fields. A well-known issue at those con-
ditions has been faradaic reactions. These limit the maximum
electric field that can be applied and the flow rate that can be
obtained. Faradaic reactions can result in pH changes, elec-
trode degradation, and electrolysis. Gas evolution due to
electrolysis occurs at the electrodes and can decrease the
efficiency, the flow rate performance and increase the possi-
bility of channel blockage. To reduce pH changes, large res-
ervoirs can be used, but are not practical for on chip devices. To
prevent electrode erosion, many EO pumps utilize platinum
electrodes due to their high inert properties. Nevertheless, they
still produce bubbles from the reaction of hydrogen and oxy-
gen at the cathode and anode, respectively. One method of
reducing this reaction is to stay below the voltage limit of 2 V
123
860 Microfluid Nanofluid (2013) 15:859–870

of the reaction (Ai et al. 2010). This method is impractical for
high flow rate EO pumps where higher voltages are needed.
Several other methods have been proposed including the use of
ion exchange membranes and meshes among others to block
the gas bubbles from entering the pores (Yao et al. 2003;
Berrouche et al. 2009; Burke et al. 2010). In some cases, this
has been effective but leaves the device with an overload of
gases or the need for platinum catalysts for recombination. Lin
et al. (2007) used Nafion tubing to funnel the gases out of the
electroosmotic pump into a recombiner to create water. Some
EO pumps have replaced the inert electrodes with disinte-
grating electrodes such as in Heuck et al. (2011). In that case,
the electrodes erode with the reactions to prevent bubble
production but these pumps have a short life usage before the
electrodes need to be replaced. The use of palladium electrodes
has also been proposed (Brask et al. 2006) due to their
hydrogen absorbing properties, but this is only effective for
small quantities of hydrogen gas such as that typically pro-
duced from de-ionized water EO flow. It has also been sug-
pumping characteristics of membranes are presented. These
show that under an asymmetric applied voltage, flow rates
over an order of magnitude higher than under an applied DC
voltage can be obtained. Also, applied voltages over an
order higher are possible without faradaic reactions.
2 Theoretical model
2.1 Bipolar rectangular voltage
A process for canceling or limiting faradaic reactions in EO
flow will be presented by applying an asymmetric bipolar
rectangular voltage waveform. As will be shown, this method
not only cancels the gas generation during EO flow, but it can
also significantly enhance the flow rate capabilities of
membranes used as EO pumps. The type of applied voltage
proposed is shown schematically in Fig. 1 and given by,
(
Vþ if tþ þt 1
t\ tþtþt
þ

gested to include additives in the liquid solution to reduce the V¼ 1


tþ

ð1Þ
hydrogen and oxygen reaction at the electrodes (Kohlheyer V if tþ þt t [ tþ þt
et al. 2008). But, it is limited or impractical for high molarity
with a high voltage value V? and a low value V-, and time
solutions and high electric fields applied or in cases where the
periods t? and t-, respectively. Also, the average voltage
solution cannot be contaminated.
can be obtained by integrating over one wave period to give,
Recently, the idea of pulsing the voltage or current has 0 1
been proposed for electroosmotic-driven flows. Selvagana- Ztþ þ þt
tZ

pathy et al. (2002) applied a periodic zero net current pulse to ¼ 1 B C tþ Vþ þ t V

V @ Vþ dt þ V dtA ¼ :
generate a flow rate while preventing electrolysis. They tþ þ t tþ þ t
0 tþ
report that this is possible at low frequencies where the
ð2Þ
current–voltage response of the system is nonlinear and can
generate an EO flow. Xu et al. (2011) used positive voltage For the case when the area per pulse are equal to each
pulsing with different duty cycles to digitally control the other,
flow. They found it to be an effective way to stabilize the Areaþ Vþ tþ
¼ ¼ 1; ð3Þ
flow rate (no decay of the flow rate with time and with Area V t
minimal fluctuation). It also helped the EO pump to behave
the net current in the system is zero, and faradaic reactions
linearly with increasing electric field. There has also been
are canceled. McBride (1999) has suggested that for this
work done on using arrays of electrodes where an alternating
equal area case, there is a net fluid flow due to the pressure
current AC electric field is used to generate an EO flow with
developed by ion drag pumping. Although this is important
no electrolysis (Chen et al. 2009). This flow is driven by the
for electro-hydro-dynamics EHD applications such as in
electric double-layer EDL around the electrodes and the
corona discharges (Stuetzer 1959), ion drag pumping as
vortices generated between adjacent electrodes in the array.
The main application is for micromixers, but the low voltages
used, and the need for assembly of electrode arrays in micro-
channels is not suitable for large EO pumping in membranes.
The present work is motivated by the lack of EO pro-
cesses capable of preventing faradaic reactions that will
allow high electric fields and flow rates. The work starts
with a brief analytical description of the proposed asym-
metric voltage waveform and the EO theory for membranes.
A brief description of the experimental setup follows. Next,
baseline measurements for DC voltage EO pumping for
Fig. 1 A sketch of the bipolar rectangular voltage waveform used for
different membranes are obtained. Also, measurements the electroosmotic pump. The positive and negative areas do not need
showing the effect of asymmetric voltage in the EO to be equal to each other

123
Microfluid Nanofluid (2013) 15:859–870 861

defined by (McBride 1999; Stuetzer 1959), is negligible for Za

pa4 efpa2 f
EO flow applications. Therefore, maintaining the ratio in Q ¼ 2p ruðrÞdr ¼ Pz  Ez ð5Þ
Eq. (3) equal to 1 will result in a negligible flow rate for EO 8g g
0
pumps as we will show. But more importantly, the exper-
iments will prove that by relaxing the condition imposed by where
Eq. (3) to values as low as 0.1, high flow rates are possible Za  
wðrÞ 2r
with negligible faradaic reactions. f ¼ 1 dr: ð6Þ
The response of the electrokinetic flow in a typical nano- f a2
0
pore of radius a to an applied bipolar rectangular pulse wave
described by Eq. (1) can be approximated by a simple f tends to unity when the channel radius a is much larger
hydrodynamic model and classical steady EO flow theory. than the electric double-layer EDL thickness j-1. Briefly, an
Quasi-steady EO flow conditions can be expected by con- EDL is a thin region of nonzero net charge density near the
sidering that the motion induced in the double layer by either a wall that forms when a solid surface (such as the walls of a
symmetric or an asymmetric applied electric field is equiva- channel) enter in contact with an aqueous solution resulting
lent to a variation of Stokes’ problem for the flow induced by an interfacial charge that causes a rearrangement of the free
viscous diffusion above a moving wall. A bipolar rectangular ions in the solution. The EDL thickness can be estimated by
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
time-varying applied field produces a nonsymmetric oscil- j1 ¼ ekT=2e2 z2 MNA , where T is the temperature of the
lating motion in the thin double layer directly adjacent to the fluid, k = 1.381 9 1023 l/mol is the Boltzmann constant,
wall which induces a motion in the bulk electrolyte outside the e = 1.602 9 1023 C is the elementary charge, z = 1 is the
double layer. This motion is confined to a ‘Stokes layer’ with charge number, M is the molarity (i.e., 0.1, 1, 10 mM), and
pffiffiffiffi
thickness d  mt, where v is the kinematic viscosity. For the NA = 6.022 9 1023 l/mol is the Avogadro constant. For
present work in AAO membranes with a * 75 nm and instance, for a molarity M of 0.1, 1, and 10 mM, the cal-
applied frequency of 1,000 Hz, it can be shown that culated j-1 are 30, 9, and 3 nm, respectively. Normalizing
d  a. For instance, defining d as the thickness where these values by the radius a = 75 nm, we obtain the elec-
ud = 0.9uw (uw is the wall velocity) results in d = 1.88 lm. trokinetic radius ja with values 2.5, 8.3, 25 for M of 0.1, 1,
For the membranes in the present work, the upper limit fre- and 10 mM, respectively. For large ja values, f tends to
quency is 353 kHz. This ensures quasi-steady flow conditions unity, but for smaller values such as the one in this work
during each voltage pulsed applied so that a total flow rate (i.e., ja = 2.5 and 8.3), Eq. (6) was solved.
Q = Q? - Q- results from such an asymmetric-induced From Eq. (5), it can be shown that the maximum flow rate
motion during each volumetric displacement ‘‘stroke’’ of the occurs when the back pressure is zero, and the maximum
actuator and classic EO flow theory can be applied. pressure occurs when the flow rate equals zero yielding,
efpa2 f
2.2 Electroosmotic theory Qmax ¼  Ez ð7Þ
g

An analytical model is briefly presented that will be com- 8eff

Pz;max ¼  Ez ð8Þ
pared with experimental results and help to better understand a2
the behavior of the EO pump. Using the electroosmotic Thus, both Qmax and Pz,max are linearly related to the
theory for flow in a cylindrical pore (Levine et al. 1975; Rice applied electric field. Also, the current can be obtained by
and Whitehead 1965), the maximum flow rates, maximum integrating the total current density over the channel to give
pressure, and current consumption can be obtained. The
velocity profile in a nano-pore in presence of both an axial Za
pressure field Pz and axial electric field Ez assuming steady, I ¼ 2p ðqðrÞuðrÞ þ KEz Þrdr
low Reynolds number flow and a 1:1 electrolyte is 0

1  2  e Za   Za  2
uðrÞ ¼ a  r 2 Pz  ðf  wðrÞÞEz ð4Þ pa2 efPz w 2r pa2 e2 Ez dw 2r
4g g ¼ 1 dr þ dr
g f a2 g dr a2
0 0
where g is the viscosity, f potential is the effective surface Za  
electric potential, and e the permittivity of the liquid. The 2r zew
þ pa2 KEz cosh dr; ð9Þ
induced potential w(r) in a pore can be described using the a2 kT
0
solution to the Poisson–Boltzmann (PB) equation from
Levine et al. (1975). Then, the volumetric flow rate can be where q(r) is the charge density at a distance r from the
obtained by integrating the velocity over the cross- axis, and K is the electrolyte conductivity. The first term in
sectional area of the channel to give the integration gives the convection current, and it is due to

123
862 Microfluid Nanofluid (2013) 15:859–870

the transport of ions through the bulk fluid by its velocity.

The second term is the conduction current, and it is due to
the movement of ions relative to the bulk fluid. This is
derived from Ohm’s Law and simplified in terms of the
solution’s conductivity when the surface conductance is
sufficiently small as suggested by Rice and Whitehead
(1965). The ratio of the flow rate to the current can be
written in terms of a nondimensional parameter g (Yao and
Santiago 2003), given by
,8 Z a   Za  
9
< e2 dw 2r2
zew 2r =
g¼f dr þ cosh dr :
:gK dr a 2 kT a2 ;
0 0
ð10Þ
The functions f and g are evaluated using the solution to
the PB equation from Levine et al. (1975). In most cases of
interest, ja is large so that f and g tend to unity. For the
present study with 150-nm-pore size membranes and a
typical 1 mM phosphate buffer solution, the calculated ja Fig. 2 A 3D model of the EO pump housing showing the anodic
is 8, and f and g are 0.766 and 0.212, respectively, aluminum oxide membrane in the middle, with platinum wire mesh
electrodes and reservoirs in each side
showing that these correction factors cannot just be
approximated by 1.
To obtain the membrane performance, the single chan- degradation. These are fabricated from 0.25-mm-diameter
nel theory needs to be modified to include the multiple platinum wire (99.9 % pure) and platinum 52 mesh (99.9 %
channels in a membrane. This is done by comparing the pure) from Sigma Aldrich. A Keithley 6517B electrometer or
open area of the membrane to the single pore area which a Trek model 5/80 amplifier with a Tektronix AFG 3021B
gives the number of pores in the membrane as function generator is used to supply power to the EO pump.
An SEM photograph of the anodic aluminum oxide
HAmem (AAO) membranes obtained from Synkera Technologies,
N¼ ; ð11Þ
pa2 Inc is shown in Fig. 4. These have a 150-nm pore size and a
where H is the porosity, and Amem is the membrane’s area. 50-lm thickness. Their characteristics include symmetric
Flow rate and current calculations for single channel theory cylindrical pore radius throughout the thickness of the
can now be multiplied by N to obtain the membrane values. membrane, porosity of 30–32 %, and tortuosity of 1. These
membranes are heat-treated up to 1,000 °C by the manu-
facturer, so they can be operated in the pH range of 4.7–9.
3 Experimental setup Vajander et al. (2007) coated their AAOs with a 5-nm layer
of SiO2 and reported a performance increase. Thus, we had
To study the performance of the membranes, they are Synkera coat the membranes with a custom coating of 10 nm
mounted in an enclosure machined in acrylic sketched in of SiO2 which is added to the heat-treated AAOs by atomic
Fig. 2. This housing for the EO pump has a reservoir in each layer deposition. This resulted in a measured pore size
side of the membrane. The inner reservoirs are connected to opening of 135 ± 10 nm. Miao et al. (2007) coated their
the large diameter outer reservoirs using 00 tubing for flow AAOs with a layer of Pt on both faces to act as electrodes,
rate measurements. Optical access to the inner reservoirs in whereby maximizing the effective voltage. To evaluate this
each side of the membrane allowed monitoring gas genera- effect, a 100-nm Pt-coating is also added to a batch of SiO2-
tion. A scale (Ohaus Scout Pro ± 0.001 g) monitored the coated AAOs. This resulted in a membrane porosity of only
output flow as seen in Fig. 3. For maximum pressure mea- 13 % and a measured pore size opening of 99 nm. This
surements, the tubing is replaced with a connector to a coating is only added to a small batch due the high cost which
pressure transducer (Omegadyne PX309) as seen in Fig. 3. resulted in limited testing. During EO testing, the average
Silicone rubber gaskets are used to seal the pump. Mem- exposed area for all AAOs was 0.400 cm2.
branes are epoxied to PVC sheets of 0.8 mm thickness to The primary solutions used are deionized ultra-filtered
hold the membranes inside the housing. Platinum wire and water and phosphate buffer made with Sodium Phosphate
mesh is used as electrodes for the electroosmotic pump. Monobasic (C99 % pure) and Sodium Phosphate Dibasic
Using inert platinum electrodes minimizes electrode (C99 % pure) from Sigma Aldrich. The corresponding

123
Microfluid Nanofluid (2013) 15:859–870 863

Fig. 3 Sketch showing the

experimental setup used to
study the EO pump
performance. For the pressure
measurements, a pressure
transducer is used, and for the
flow rate measurements, this is
replaced with a reservoir and
precision scale

Table 1 Measured Phosphate buffer conductivity for the different prevented gas generation. A baseline is established first by
molarities used measuring the EO pump performance for a constant
Molarity mM pH Conductivity applied voltage. This analysis will allow us to compare and
(lS/cm) show the effect of the type of voltage applied on the
maximum flow rate, current consumption, and maximum
0.1 5.7 6.2
pressure of the membranes.
0.5 6.1 45
1 6.2 90
4.1 Constant voltage EO pump performance
2 6.3 168
10 6.3 861
The EO pumping capabilities of membranes are evaluated
using the setup described in Fig. 3. Flow rate measurements
are conducted in two different types of membranes. The first
is a SiO2-coated AAO membrane where the electrode is a
thin Pt film coated on both sides of the membrane surface.
The second is also a SiO2-coated AAO membrane, but the
electrode is formed by a Pt wire, and mesh not in contact
with the membrane but rather positioned 1–2 mm from the
membrane. The flow rate measurements are normalized by
the applied voltage and open area of the membrane in order
to compare these results to published work. This normalized
flow rate can now be plotted against the applied electric
field in Fig. 5, and it allows for the comparison of mem-
branes with different open areas, thickness, and applied
voltage. For instance, the SiO2-coated AAOs with the Pt
wire mesh achieved for at 30 V and a 1 mM buffer solution,
a maximum flow rate of 0.561 ml/min which when nor-
malized gives 0.102 ml/min/V/cm2. This is comparable to
the flow rate measured for SiO2-coated AAO membranes in
the literature (Vajandar et al. 2007) and also for porous
silicon membranes (Yao et al. 2006) as shown in Fig. 5. The
best performing EO pumps were the SiO2-coated AAO with
Fig. 4 SEM image of SiO2-coated AAO with 150-nm pore size Pt film-coated electrodes. The Pt coating minimizes any
voltage loss from the electrodes to the membrane and thus,
phosphate buffer (pH of 6.2) conductivities are listed in allows the applied voltage to be approximately equal to the
Table 1 and the measured deionized water conductivity is effective voltage. The flow rate from these AAOs improved
1.51 lS/cm. Conductivity and pH measurements are con- by an order of magnitude compared with the non-Pt-coated
ducted using an Oakton 510 series meter. AAOs and outperformed the SiO2- and Pt-coated AAOs
from Miao et al. (2007). They generated a flow rate of
2.10 ml/min at 10 V which equates to a normalized value of
4 Results 4.69 ml/min/V/cm2 shown in Fig. 5. To our knowledge,
this normalized flow rate is the highest obtained to date with
The objective is to evaluate the enhanced EO pumping a membrane.
capabilities of membranes when a bipolar rectangular In order to compare the experimentally measured flow
voltage waveform is applied which limited and even rate results to those predicted by the theory, the effective

123
864
10
Q/V/Am (mL/min/V/cm2)
1
0.1
0.01
0.01 0.1 1
V/L (V/µm)
Fig. 5 Flow rate measurements of SiO2-coated AAOs normalized by
the applied voltage and open area of the membrane. Results are
compared to published work for various types of membranes. The
obtained value of 4.69 ml/min/V/cm2 for the platinum and SiO2-
coated AAOs represents the highest normalized flow rate measured to
date for an EO pump membrane
voltage Veff across the membrane needs to be calculated.
This is the actual voltage that generates the EO flow. This
Veff is lower than the voltage V applied by the power supply
as a result of voltage losses between the electrodes which
can be substantial even for millimeter-spaced electrodes.
Miao et al. (2007) reduced these losses by sputtering the
platinum electrodes on both surfaces of the AAO. To solve
for Veff, a simple potential model by Yao et al. (2003) is
used
Veff ¼ V  Vdec  2IRel ð12Þ
where Vdec is the electrode decomposition voltage and
combines the voltage needed for the reactions at the
cathode and the anode and the over-potential voltage
needed to start the process. It tends to be small (i.e., 4 V for
1 mM borate buffer Yao et al. 2003), and it varies with the
molarity, the reactants, the electrode material, and current
supplied. In the other hand, the voltage drop occurring
between the electrodes and the membrane can be much
larger and is given by IRel where the resistance of the fluid
between the electrode and membrane is Rel = D/AelK
where D is the distance between the electrodes, and Ael is
the electrode area. It can also be approximated experi-
mentally by removing the membrane, calculating the I–V
curve and measuring the slope (Cao et al. 2012).
The theoretical flow rate Eq. (5) can now be calculated
by using the Veff for each applied V. As predicted by the
theory, the measured flow rates of the SiO2-coated AAOs
increased linearly with voltage as shown in Fig. 6a. It
shows that the flow rate is higher at higher molarities.
123
Microfluid Nanofluid (2013) 15:859–870
The 10 mM case could not be run at 30 V due to the
presence of electrolysis while the 0.1 mM case could be
run at 50 V (not shown in the figure) with a maximum flow
rate of 0.099 ml/min before gas generation is observed. A
more detailed effect of the molarity on the flow rate is
shown in Fig. 6b. It compares the flow rates generated by
varying the molarity of the phosphate buffer using the
SiO2-coated AAOs. It also shows that molarities greater
than 2 mM do not contribute to a significant increase in
flow rate. This is comparable to the result obtained by
Vajandar et al. (2007) where a 2.5 mM solution generated
their greatest normalized flow rate. As discussed by
Vajandar et al. (2007), in the regime of 10 C ja C 1, the
flow rate decreases as the EDL overlaps. Present results
have the same tendency with 1 mM phosphate buffer
(ja = 8) having a greater flow rate than a 0.1 mM phos-
phate buffer (ja = 2). On the other hand, the increase in
molarity has the undesirable effect of increasing the current
as shown in Fig. 6c and by extension the power con-
sumption. Similar to the flow rate results, the increase in
current consumption is linear with the applied voltage as
predicted by the theory. The deviation between the theory
and the experiments are due to the errors in estimating Veff
needed for the theoretical calculations. There, the largest
error comes from measuring the electrodes spacing (i.e.,
1.0 ± 0.3 mm) which varies between experiments and
showed the largest deviation in the theoretical predictions
of the flow rate in Fig. 6a.
Another method for evaluating the performance of EO
pumps is measuring the maximum pressure achieved. The
measurements are conducted using the setup described in
Fig. 3 with the pressure transducer. A 1 mM phosphate
buffer is used due to the higher flow rates obtained previ-
ously. Results are compared to published data in Fig. 7. It
shows that the SiO2-coated AAOs generated a slightly
greater maximum pressure than the APS (NH2–(CH2)3–
Si(OCH3)3)-coated AAOs from Chen et al. (2010). This is
most likely due to the pore size difference of 150–200 nm.
As noted by Chen et al. (2007) and shown by Eq. (8), smaller
pore sizes tend to produce a greater maximum pressure.
The f potential is an important parameter in character-
izing membranes’ EOP capabilities. It can be calculated
from Pz;max ; Qmax ; and Imax measurements using the fol-
lowing relations (Yao et al. 2003; Reichmuth et al. 2003)
Pz;max 8ef
¼ 2 ð13Þ
Veff a
Qmax ef
¼ g ð14Þ
Imax gK
Theoretically, the f potentials for the 10-nm SiO2-coated
AAOs and for borosilicate glass capillaries should be
comparable due to similar chemical composition. Our
Microfluid Nanofluid (2013) 15:859–870 865

a
0.6
30
0.5

0.4
20
Q (mL/min)

Pz (kPa)
0.3

0.2 10

0.1

0
0.0 0 10 20 30
0 10 20 30
V (V) V (V)
b
0.6 Fig. 7 Pressure measurements for SiO2-coated AAOs using 1 mM
phosphate buffer with theoretical predictions (solid line) along with
0.5 data from Arulanandam and Li (2000) that used an APS-coated AAO
with 200-nm pores and deionized water

0.4
Q (mL/min)

measurements for the f potential in borosilicate capillaries

0.3 using current monitoring (Arulanandam and Li 2000) gives
-97 mV for 0.1 mM phosphate buffer (-90 mV in Kirby
0.2
and Hasselbrink 2004) and -120 mV for 1 mM phosphate
buffer (-100 mV in Kirby and Hasselbrink 2004). These f
potential values are much larger than those obtained for
0.1
SiO2-coated AAO membranes using our measurements in
combination with Eq. (14) and listed in Table 2. Briefly,
0.0
0.01 0.1 1 10 100 when comparing the measured f potential values to those in
Molarity (mM) Xu et al. (2011) and Miao et al. (2007) in Table 2, their
c values are higher. It is speculated that this is due to the
30
fabrication process and membrane treatment applied which
varied significantly from the vendor used (Synkera Inc) to
25
those in Xu et al. (2011) and Miao et al. (2007) which must
affect the properties of the SiO2 coating. Following the
20 discussion by Cao et al. (2012) where it has been reported
significant differences in f potential depending on the
I (mA)

15 method used, this is also calculated using Eq. (13) to give

-11 mV. This similarity among values calculated from
10 two different measurements supports that these membranes
have a f potential between -11 and -13 mV.
5 The EO efficiency of the AAO can be calculated from
the ratio of the hydraulic output power to the electrical
input power. It can also be written in terms of the measured
0
0 10 20 30 variables, Qmax and Pmax (Chen et al. 2007), to yield
V (V)
QP Qmax Pmax
v¼ ¼ : ð15Þ
Fig. 6 Effect of a applied voltage and b solution molarity on flow IV 4IV
rates and effect of c molarity on current for EO pumps. For a and c,
measurements are compared to theoretical predictions with similar
Using Eq. (15), the calculated experimental efficiencies
trends and linear behavior. In b measurements show that molarities for the SiO2-coated AAOs with and without the Pt-coated
greater than 2 mM do not significantly contribute to flow rate increase surfaces were 0.161 and 0.028, respectively. These values

123
866 Microfluid Nanofluid (2013) 15:859–870

Table 2 Measured f potential for SiO2-coated AAOs and published f potential values for different types of membranes
Source Membrane Molarity (mM) Solution pH Zeta potential (mV)

Present AAO w/SiO2 0.5, 1, 2, 10 Phosphate 6.5 -8, -13, -13, -21
Miao et al. (2007) AAO N/A DI water 7.0 -19.3
Miao et al. (2007) AAO w/H2SO4 N/A DI water 7.0 -30
Xu et al. (2011), Miao et al. (2007) AAO w/SiO2 N/A DI water 5.8, 7 -50, -42
Prakash et al. (2006) Alumina 0.2 NaCl Phosphate 10.3 -9, -30
Prakash et al. (2006) Silica 0.2 NaCl Phosphate 10.3 -63,-75, -6.3
Cao et al. (2012) MCP N/A DI water N/A -15
Wang et al. (2012) PC N/A DI water N/A -23

contrast with the higher values in the literature due to their 100
ff = =
1 (1
t+ /+(t−t+) + t− )
higher zeta potentials. For instance, an efficiency of 0.900
is obtained for SiO2- and Pt-coated AAOs by Miao et al.
(2007) although conductivity and pump resistance were 10
similar. Also, an v = 0.100 is obtained for APS-coated
AAOs by Chen et al. (2010) or an v = 0.430 is obtained

Q (mL/min)
for bare AAOs by Chen et al. (2007). These higher values
can be attributed to the lower current consumption. 1

Although we would prefer membranes with a higher

value for improved performance and efficiency values, this
lower values do not affect the scope of the method
proposed for enhancing the flow rate by applying an
asymmetric bipolar rectangular voltage waveform for EO
pumping.
4.2 Bipolar rectangular voltage waveform EO pump
performance
Having established a baseline for EO pump performance
when a constant voltage is applied, we now consider the
EO pump capabilities when an asymmetric bipolar rect-
angular voltage is applied. The same setup used for the DC
voltage tests described in Fig. 3 is used for these flow
measurements. Figure 8 shows the measured flow rates
when a bipolar rectangular pulse wave is applied to SiO2-
coated AAO membranes. Results are plotted as a function
of the ratio V?t?/V-t- where V?t? and V-t- correspond to
the area underneath the positive and negative part of the
waveform in Fig. 1. By varying these positive and negative
areas, it can be shown that the flow rate is driven by the
greater area and that for equal areas (V?t?/V-t- = 1) the
flow rate is negligible. For instance, let’s consider an
example where V?/V- = ? 80/-20 V shown by the solid
circles in Fig. 8. For this case, when the areas are the same
(i.e., t?/t- = 1/4 which corresponds to t? = 0.2 ms, t- =
0.8 ms for a frequency of 1,000 Hz), the flow rate is about
0.04 ml/min. In the other hand, when the areas are not the
same (i.e., t?/t- [ 1/4 or t?/t- \ 1/4) the flow rate
increases significantly until reaching a plateau of *0.8 and
*0.6 ml/min for V?t?/V-t- of 0.2 and 1.7, respectively.
0.1
0.01
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
V+t+/V-t-
Fig. 8 Measured flow rate for an applied asymmetric bipolar
rectangular voltage to SiO2-coated AAO with 1 mM (closed symbols)
and 0.1 mM (open symbols) phosphate buffer. Results are evaluated
in terms of the area ratio V?t?/V-t-. An asymmetric voltage as high
as ?1,800/-900 V is applied generating a flow rate of 2.95 ml/min
with no gas generation
A similar behavior is observed for all the other V?/V-
cases considered in Fig. 8. Also, 1 mM phosphate buffer
(solid symbols) generated a greater flow rate than 0.1 mM
phosphate buffer (open symbols) similar to the case for
constant voltage.
What is remarkable about these results is not so much
that a flow rate can be generated by an asymmetric bipolar
rectangular pulse wave, but rather the unexpected high
voltages that can be applied with negligible gas generation.
For instance, when a constant voltage is applied, gas gen-
eration is observed between 30 and 50 V for the buffers
and concentrations used. In the other hand, when an
asymmetric pulse is applied, there was no gas generation
for voltage as high as 1,800 V. Furthermore, higher volt-
ages might be possible but could not be tested due to
testing hardware limitations. The effect of the frequency,
1/(t? ? t-), on gas generation is also considered. The

123
Microfluid Nanofluid (2013) 15:859–870 867

results show that changing the frequency from 100 to a

0.7
1,000 Hz made no difference on the flow rate generated.
Also, negligible faradaic reactions are present except on
0.6
the limits when V?t?/V-t- *0.2 or *1.7 where a small
amount is observed but not to the extent seen for constant
0.5
voltages. Higher frequencies could not be tested due also to

Q (mL/min)
amplifier limitations. In the other hand, frequencies below 0.4
50 Hz resulted in gas generation which shows that this is a
lower bound frequency value where the proposed technique 0.3
cannot be applied. This is also an indicator that the time
scale for gas generation H2(g) and/or O2(g) is probably of 0.2
this order at least for the cases studied here. The highest
flow rate measured is 2.95 ml/min at ?1,800/-900 V for 0.1
0.1 mM phosphate buffer and 1.69 ml/min at ?400/
-200 V for 1 mM phosphate buffer as shown in Fig. 8. 0.0
As with all EO pumps, power consumption is an 0.001 0.01 0.1 1 10
Pin (W)
important characteristic that needs to be considered. The
b
goal is to achieve higher flow rates Q with the lower power 2.5
consumption Pin possible. Figure 9 shows the EO mea-
sured flow rates and corresponding power consumption
2.0
when a bipolar rectangular pulse wave is applied to SiO2-
coated AAO membranes in a 1 mM phosphate buffer
Q/Pin (mL/min/W)

solution. The plot in Fig. 9a shows the effect that the
variation of the area ratio V?t?/V-t- has on Q and Pin
when the bipolar rectangular waveform is applied. It shows
that the asymmetric voltage pulsing follows the same trend
as the constant voltage but it is not as efficient. For these
measurements, the most efficient area ratio is 0.44 and the
least efficient ratio or the worst performer in terms of
power to flow rate is the equal area pulsing which has
already been shown in Fig. 8. Also, the effect that the
variation of the voltage ratio has on Q/Pin when the bipolar
rectangular waveform is applied is considered in Fig. 9b. A
voltage ratio of 2:1 provided the best flow rate per supplied
power, and as the ratio is increased to 4:1, 7:1 and lastly
9:1 higher flow rates are obtained but the power con-
sumption increases at a faster rate resulting in a less
effective system.
Similar to the constant voltage analysis, the performance
of the EO pump can be evaluated by measuring the max-
imum pressure achieved by the SiO2-coated AAOs in
1 mM phosphate buffer. The measurements are conducted
using the setup described in Fig. 3 with the pressure
transducer. The obtained pressures are plotted as a function
of the ratio V?t?/V-t- in Fig. 10. As was the case for flow
rate measurements, pressure was maximized for the greater
negative area to positive area ratio. For a fixed area ratio
V?t?/V-t- and voltage ratio V?/V- such as 0.45 and 4,
respectively, the results show an increase in pressure with
voltage and power with a maximum pressure of 36 kPa for
?200/-50 V. Higher pressures are possible although it
would require modifying the setup to properly support the
membrane which kept breaking at the higher pressures.
1.5
1.0
0.5
0.0
0.0 0.5 1.0 1.5 2.0 2.5
V+ t+ / V− t−
Fig. 9 Measured flow rates and power consumption when a bipolar
rectangular voltage waveform is applied to AAO membranes in a
1 mM phosphate buffer. Results in a show the effect that the variation
of area ratio V?t?/V-t- has on Q and Pin. Results in b show the effect
that the variation of voltage ratio has on Q/Pin
These results are comparable to the linear increase in
pressure with voltage observed for the constant voltage
analysis in Fig. 7. There is nevertheless a loss in efficiency
estimated to be 0.005 using Eq. (15) which is about
*20 % the efficiency obtained in the constant voltage
analysis. The results also show that for an equal area ratio,
the system was able to generate a small pressure of about
2.1 kPa for ?200/-50 V which is responsible for the small
flow rate observed.
Last, an overall comparison of flow rates obtained in
terms of the average voltage applied V  from Eq. (2) is
shown in Fig. 11. By using the average voltage applied, a
more direct way of comparing flow rate measurements

123
868 Microfluid Nanofluid (2013) 15:859–870

40

30
Pz (kPa)

20

10

0
0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
V+ t+ / V− t−

Fig. 10 Pressure measurements for SiO2-coated AAOs using 1 mM

phosphate buffer for an applied asymmetric bipolar rectangular
voltage waveform

Fig. 12 Close-up images show a partial frontal view of the platinum

mesh electrode and membrane inside a transparent acrylic housing
during pump operation. a For an applied ?600/-300 V bipolar
rectangular voltage, only a few small bubbles are formed without
affecting the pump operation. b For an applied 50 V DC voltage,
1 many small but also large bubbles form in front and behind the
electrode such as the smaller ones throughout that seem to defocus the
image or the large one in the lower part of the image which prevented
Q (mL/min)

pump operation

by nearly a factor of 30 with a peak voltage increase of

0.1 about 36 times.

4.3 Additional comments on gas generation reduction

When a potential is applied to an electrochemical cell, both

0.01
1 10 100
(V+t++V-t-)/(t++t- )
Fig. 11 Flow rates measured in terms of the average voltage applied
for 1 mM (closed symbols) and 0.1 mM (open symbol) phosphate
buffer. This allowed comparing flow rates obtained by applied DC
voltage with those obtained by applied bipolar rectangular voltage
waveforms
between the DC voltage testing and asymmetric bipolar
rectangular voltage testing is possible. Results show that
when a DC voltage is applied, faradaic reactions are
important above 30 and 50 V for 1 and 0.1 mM borate
buffer solutions, respectively, with maximum delivered
flow rates of 0.56 ml/min for 1 mM and 0.099 ml/min for
0.1 mM. In the other hand, the proposed asymmetric
voltage pulsing is capable of delivering a flow rate of
2.95 ml/min for a 0.1 mM solution at V   400 V corre-
sponding to ?1,800/-900 V. This demonstrates that for
the current case, the proposed method raised the flow rate
hydrogen and oxygen ions, and their respective radicals are
1000
formed. The H2 and O2 gas formed dissolves in the elec-
trolyte until it reaches supersaturation. Beyond this
threshold and in the presence of a nucleation site, crack or
cavity in the electrode surface, bubbles are initiated on the
electrode surface (Donose et al. 2012). Part of the dissolved
gas generated is transformed into the gaseous phase of the
bubbles adhering to the electrode, while the other part
transferred to the liquid bulk (Vogt 2011). For a bipolar
voltage, during the positive pulse, H2 is generated at the
cathode, and O2 is generated at the anode. During the
negative pulse, the electrode polarity is reversed with H2
forming at the electrode where O2 previously formed and
vice versa. For a high frequency asymmetric bipolar rect-
angular voltage, the hydrogen ions and hydroxide ions
generated at the electrode compensate each other, and no
gas is generated. However, in the case of voltage pulsing
with different duty cycle, the excess amount of ions gen-
erated at each electrode react and form gas molecules.

123
Microfluid Nanofluid (2013) 15:859–870
These dissolve in the liquid until the gas concentration
reaches supersaturation. Considering that the fluid is being
convected by this EO pump in an open circuit, the
incoming fluid is not supersaturated, and the threshold
needs to be reached in each cycle before bubbles can be
generated. It is speculated that this reduces bubble forma-
tion. More importantly, the dynamics of bubble formation
and growth is a complex nonlinear process which is
affected by this voltage pulsing. This process is not well
understood, but it is speculated that the bubbles generated
between cycles do not have sufficient time to grow much
larger than the nanochannel’s diameter. This is supported
by some observations during high field asymmetric bipolar
EO pumping of only small bubbles being carried down-
stream with the flow and minimal small bubbles found in
the electrodes as shown in Fig. 12a (see also attached
supplemental video material). On the other hand, for
[30–50 V DC EO pumping, the bubbles grow quickly and
in just seconds bubbles with millimeter size are present
which continue expanding as shown in Fig. 12b and
eventually prevent pump operation. The presence of bub-
bles for these applied DC voltages ([50 V) are reported by
Vajandar et al. (2007) limiting their tests to values below it.
Furthermore, they reported the current being unstable due
to the bubbles. Thus, the high frequency asymmetric
bipolar rectangular voltage permits overcoming this DC
voltage limit, and it results in much higher flow rates than
previously possible.
5 Conclusion
The experimental results have shown that the use of an
asymmetric bipolar rectangular waveform for EO pumps
allows applying higher voltages than those possible in DC
EO pumping while generating much higher flow rates and
preventing gas generation. Results are compared to a
baseline measurement for DC voltage EO pumping in
SiO2-coated AAO membranes. From the DC EO pumping
results, faradaic reactions are observed at about 30–50 V
depending on the buffer concentration. For instance, for 1
and 0.1 mM phosphate buffers, gas generation is observed
at 30 and 50 V, respectively, while the flow rates measured
are 0.561 and 0.099 ml/min, respectively. These values and
their normalized results are comparable to theory and to
flow rate measurements for AAO membranes in the liter-
ature (Vajandar et al. 2007).
Results from the bipolar rectangular-wave EO pumping
measurements are evaluated in terms of the area ratio
V?t?/V-t-. For instance, for V?/V- = ? 80/-20 V
when the area ratio is 1, then flow rate is negligible
(0.04 ml/min), but it increased significantly until reaching
a plateau of *0.8 and *0.6 ml/min for area ratios of 0.2
869
and 1.7, respectively. These results are independent of the
frequency of the applied waveform as long as it is greater
than *50 Hz to avoid faradaic reactions. Furthermore, the
voltage ratio V?/V- of 2:1 provided the best flow rate per
supplied power. The highest flow rate observed were
2.95 ml/min at ?1,800/-900 V for 0.1 mM phosphate
buffer and 1.69 ml/min at ?400/-200 V for 1 mM phos-
phate buffer.
It is remarkable that when an asymmetric pulse is
applied, there was negligible gas generation for applied
voltages as high as 1,800 V, but for DC, voltage gas gen-
eration is observed between 30 and 50 V for the solutions
used. Furthermore, the results showed that the proposed
method could increase the flow rate output compared to DC
EO pumping by one to two orders of magnitude with
similar increased in applied voltages.
Acknowledgments The authors gratefully acknowledge the finan-
cial support provided for this study by the Office of Naval Research
(ONR), Grant No. N00014-11-1-0019, with Dr. Thomas F. Swean of
the Ocean Engineering and Marine Systems Program serving as
Program Manager. The authors are thankful to Mr. Thomas E. Hansen
for all the support obtaining the EO pump video and images.
References
Ai Y, Yalcin SE, Gu D, Baysal O, Baumgart H, Qian S, Beskok A
(2010) A low-voltage nano-porous electroosmotic pump. J Col-
loid Interface Sci 350(2):465–470
Arulanandam S, Li D (2000) Determining f potential and surface
conductance by monitoring the current in electro-osmotic flow.
J Colloid Interface Sci 225(2):421–428
Berrouche Y, Avenas Y, Schaeffer C, Hsueh-Chia Chang, Ping Wang
(2009) Design of a porous electroosmotic pump used in power
electronic cooling. IEEE Trans Ind Appl 45(6):2073–2079
Brask A, Snakenborg D, Kutter JP, Bruus H (2006) AC electroos-
motic pump with bubble-free palladium electrodes and rectifying
polymer membrane valves. Lab Chip 6(2):280
Buie CR, Posner JD, Fabian T, Cha S-W, Kim D, Prinz FB, Eaton JK,
Santiago JG (2006) Water management in proton exchange
membrane fuel cells using integrated electroosmotic pumping.
J Power Sources 161(1):191–202
Burke JM, Smith CD, Ivory CF (2010) Development of a membrane-
less dynamic field gradient focusing device for the separation of
low-molecular-weight molecules. Electrophoresis 31(5):902–909
Cao Z, Yuan L, Liu Y-F, Yao S, Yobas L (2012) Microchannel plate
electro-osmotic pump. Microfluid Nanofluid 13(2):279–288
Chen L (2005) Fabrication and characterization of a multi-stage
electroosmotic pump for liquid delivery. Sensors Actuators B
Chem 104(1):117–123
Chen L, Ma J, Guan Y (2004) Study of an electroosmotic pump for
liquid delivery and its application in capillary column liquid
chromatography. J Chromatogr A 1028(2):219–226
Chen W, Yuan JH, Xia XH (2005) Characterization and manipulation
of the electroosmotic flow in porous anodic alumina membranes.
Anal Chem 77(24):8102–8108
Chen Y-F, Li M-C, Hu Y-H, Chang W-J, Wang C-C (2007) Low-
voltage electroosmotic pumping using porous anodic alumina
membranes. Microfluid Nanofluid 5(2):235–244
123
870
Chen J-K, Weng C-N, Yang R-J (2009) Assessment of three AC
electroosmotic flow protocols for mixing in microfluidic chan-
nel. Lab Chip 9(9):1267
Chen Y-F, Hu Y-H, Chou Y-I, Lai S-M, Wang C-C (2010) Surface
modification of nano-porous anodic alumina membranes and its
use in electroosmotic flow. Sensors Actuators B Chem
145(1):575–582
Donose BC, Harnisch F, Taran E (2012) Electrochemically produced
hydrogen bubble probes for gas evolution kinetics and force
spectroscopy. Electrochem Commun 24:21–24
Garg R, Kumar V, Kumar D, Chakarvarti SK (2012) Electrical
transport through micro porous track etch membranes of same
porosity. Mod Phys Lett B 26(31):1250209
Heuck F, Van der Ploeg P, Staufer U (2011) Deposition and
structuring of Ag/AgCl electrodes inside a closed polymeric
microfluidic system for electroosmotic pumping. Microelectron
Eng 88(8):1887–1890
Kim Y, Cha M, Choi Y, Joo H, Lee J (2013) Electrokinetic separation
of biomolecules through multiple nano-pores on membrane.
Chem Phys Lett 561–562:63–67
Kirby BJ, Hasselbrink EF (2004) Zeta potential of microfluidic
substrates: 1. Theory, experimental techniques, and effects on
separations. Electrophoresis 25(2):187–202
Kohlheyer D, Eijkel JCT, Schlautmann S, Van den Berg A,
Schasfoort RBM (2008) Bubble-Free Operation of a Microflu-
idic Free-Flow Electrophoresis Chip with Integrated Pt Elec-
trodes. Anal Chem 80(11):4111–4118
Kwon K, Park C-W, Kim D (2012) High-flowrate, compact electro-
osmotic pumps with porous polymer track-etch membranes. Sens
Actuators, A 175:108–115
Levine S, Marriott JR, Robinson K (1975) Theory of electrokinetic
flow in a narrow parallel-plate channel. J Chem Soc, Faraday
Trans 2(71):1–11
Li PCH, Harrison DJ (1997) Transport, manipulation, and reaction of
biological cells on-chip using electrokinetic effects. Anal Chem
69(8):1564–1568
Lin C-W, Yao S, Posner JD, Myers AM, Santiago JG (2007) Toward
orientation-independent design for gas recombination in closed-
loop electroosmotic pumps. Sensors Actuators B Chem 128(1):
334–339
McBride SE (1999) Balanced asymmetric electronic pulse patterns
for operating electrode-based pumps. US Patent 5,964,977, filed
21 Mar 1997 and issued 12 Oct 1999
Miao J-Y, Xu Z-L, Zhang X-Y, Wang N, Yang Z-Y, Sheng P (2007)
Micropumps based on the enhanced electroosmotic effect of
aluminum oxide membranes. Adv Mater 19(23):4234–4237
123
Microfluid Nanofluid (2013) 15:859–870
Pikal MJ (1992) The role of electroosmotic flow in transdermal
iontophoresis. Adv Drug Deliv Rev 9(2):201–237
Prakash P, Grissom MD, Rahn CD, Zydney AL (2006) Development
of an electroosmotic pump for high performance actuation.
J Membr Sci 286(1–2):153–160
Reichmuth DS, Chirica GS, Kirby BJ (2003) Increasing the perfor-
mance of high-pressure, high-efficiency electrokinetic micro-
pumps using zwitterionic solute additives. Sensors Actuators B
Chem 92(1–2):37–43
Rice CL, Whitehead R (1965) Electrokinetic flow in a narrow
cylindrical capillary. J Phys Chem 69(11):4017–4024
Selvaganapathy P, Yit-shun Leung Ki, Renaud P, Mastrangelo CH
(2002) Bubble-free electrokinetic pumping. J Microelectromech
Syst 11(5):448–453
Stuetzer OM (1959) Ion drag pressure generation. J Appl Phys
30(7):984–994
Vajandar SK, Xu D, Markov DA, Wikswo JP, Hofmeister W, Li D
(2007) SiO2-coated porous anodic alumina membranes for high
flow rate electroosmotic pumping. Nanotechnology 18(27):
275705
Vogt H (2011) On the gas-evolution efficiency of electrodes
I-Theoretical. Electrochim Acta 56(3):1409–1416
Wang P, Chen Z, Chang H-C (2006) A new electro-osmotic pump
based on silica monoliths. Sensors Actuators B Chem 113(1):
500–509
Wang C, Wang L, Zhu X, Wang Y, Xue J (2012) Low-voltage
electroosmotic pumps fabricated from track-etched polymer
membranes. Lab Chip 12(9):1710
Xu Z, Miao J, Wang N, Wen W, Sheng P (2011) Digital flow control
of electroosmotic pump: Onsager coefficients and interfacial
parameters determination. Solid State Commun 151(6):440–445
Yao S, Santiago JG (2003) Porous glass electroosmotic pumps:
theory. J Colloid Interface Sci 268(1):133–142
Yao S, Hertzog DE, Zeng S, Mikkelsen JC, Santiago JG (2003)
Porous glass electroosmotic pumps: design and experiments.
J Colloid Interface Sci 268(1):143–153
Yao S, Myers AM, Posner JD, Rose KA, Santiago JG (2006)
Electroosmotic pumps fabricated from porous silicon mem-
branes. J Microelectromech Syst 15(3):717–728
Zeng S, Chen CH, Mikkelsen JC, Santiago JG et al (2001) Fabrication
and characterization of electroosmotic micropumps. Sensors
Actuators B Chem 79(2–3):107–114