Microfluidic flow reversal at low frequency by AC ... - Springer Link

Microfluid Nanofluid (2009) 7:757–765 DOI 10.1007/s10404-009-0433-6

RESEARCH PAPER

Microfluidic flow reversal at low frequency by AC electrothermal effect Meng Lian Æ Jie Wu

Received: 15 December 2008 / Accepted: 11 March 2009 / Published online: 27 March 2009 Ó Springer-Verlag 2009

Abstract This paper describes a flow reversal phenomenon for fluids with moderate conductivity. Fluids with conductivities of 2 9 10-4 S/m, 0.02 S/m and up to 0.1 S/ m were experimented at frequencies ranging from 1 to 110 kHz. Flow reversal was observed only at *1 kHz and 5.3 Vrms for r = 0.02 S/m, and our analysis indicates that AC electrothermal effect could be responsible. Analysis of the system impedance and simulation of power consumption show that the distribution of electric power consumption is dependent on conductivity and AC frequency. At low frequencies, possibly more electric power is consumed at surface/electrolyte interface rather than within the fluid, which consequently changes the location of temperature maximum and the directions of temperature gradients. The direction of AC electrothermal force is reoriented, causing the flow reversal. Numerical simulation is also performed and agrees within the experiments. Keywords Microfluidics Flow reversal AC electrothermal effect

1 Introduction The ability to direct fluid flow at length scales of the order of tens of microns is essential in laboratory-on-a-chip applications, and various mechanisms have been investigated, such as magnetism and optical force. Among them, electrical field is particularly well suited for miniaturization, because high field strength can be easily generated. M. Lian J. Wu (&) Department of Electrical Engineering and Computer Science, The University of Tennessee, Knoxville, TN 37996, USA e-mail: [email protected]

The improvements in microfabrication where devices incorporate both microelectrodes and microchannels have further led to a widespread use of various electrokinetic methods. EK methods can be implemented by applying DC or AC potentials over electrodes immersed in the fluid. Traditional EK, or DC EK, has almost 200 years’ history and has been rather thoroughly investigated. EK by AC potentials, or ACEK, has attracted increasing research interest recently. ACEK has many attractive features. The nonlinear nature of ACEK produces higher transport efficiency than DCEK. Because time varying low voltage signal is used, ACEK also minimizes undesirable byproducts of electrochemical reactions that are unavoidable with DC excitation. Fluid or particle motion generated by ACEK is local, so complex flow patterns can be generated by addressing electrodes individually. ACEK manipulates fluid by two mechanisms, AC electroosmosis (ACEO) and AC electrothermal effect (ACET). ACEO is the fluid motion induced by moving charges in the electric double layer at the interface of electrode/electrolyte. ACET is caused by uneven Joule heating of the fluid, which induces gradients of conductivity and permittivity, and as a result mobiles charges in the fluid bulk that move under the influence of the electric fields to generate flows. ACEO is mainly effective in low ionic strength fluid (e.g., deionized water) at low frequencies when most electric energy is stored at the interface, while ACET is more pronounced at higher ionic strength. Both ACEO and ACET have already been applied to develop micropumps. While ACEK micropumps have advantages of easy implementation and compatibility with microchip fabrication, it has been observed by several groups that ACEK micropumps may reverse the flow direction upon a small

123

758

change in pumping parameters, such as AC frequency or voltage. To expand the application scope of ACEK, it is imperative that ACEK microflows can be very well predicted and controlled, which motivates the research on ACEK microflow reversal. Flow reversal happens mostly in ACEO devices (Lastochkin et al. 2004; Studer et al. 2004; Urbanski et al. 2006; Wu et al. 2005b; Yang and Wu 2008). Using an array of asymmetric planar electrodes and KCL solutions at about 1.5e-3 S/m, Studer et al. (2004) reported the reversal of their ACEO pumping direction at relatively high frequencies (50–100 kHz) and high voltages (up to 6 Vrms with 4.2 lm in-pair electrode gap), as ACEO operates below 50 kHz. Wu et al. (2005b) and Lastochkin et al. (2004) reported ACEO flow reversal by Faradaic charging in deionized water. Urbanski et al. (2006) observed flow reversal in their 3D step ACEO micropump at frequencies over 10 kHz and voltages beyond 2 Vrms using deionized water, and they reported that flow reversal threshold voltage increases as the increase of the operating frequency. Yang and Wu (2008) produced back and forth microflows in their T-shape electrodes by changing AC frequency between 1 kHz and 1 MHz. Gregersen et al. (2007) also observed flow reversal at low frequency (*1 kHz) and low AC voltage (*1 Vrms). They attributed this phenomenon to possible Faradaic electrode reactions. Several different mechanisms have been put forward to explain the flow reversal. At low frequency and moderate voltage, induction of co-ions from Faradaic reactions is expected to be responsible, e.g., flow reversal at 2 Vrms, 100 Hz in Wu et al. (2005b). At high voltage and high frequency as in Yang and Wu (2008), ACET dominates over ACEO and reverse the microflows. In the middle frequency range (1–100 kHz), where ACEK devices most likely to operate, Storey et al. (2008) and Kilic and Bazant (2007) suggest steric effect as responsible for flow reversal. Even in very dilute electrolytes, because the external voltage applied over the electrodes usually exceeds the requirement for small perturbation (more than a few tenths of a volt), traditional Poisson–Boltzmann distribution of charges in the double layer breaks down. The charges induced in the diffuse layer of the double layer are crowded and cannot be regarded as closely packed at the electrode surface. Therefore, the double-layer capacitance decreases, which affects the charging time (response time) of the electrodes. A pair of electrodes with unequal width will have different charging times for its two members, and the narrower electrode can respond to a higher frequency than the wider electrode. At relatively high frequency (10– 100 kHz), the wider electrode loses its ability to generate ACEO surface flow, and the net flow will follow the surface flow on the narrow electrode. As a result, the pumping direction reverses. In steric effect, the surface flows on each electrode maintain their directions from the inner edge

123

Microfluid Nanofluid (2009) 7:757–765

to the outer edge, and the flow reversal is caused by the change of dominance from the wide electrode to the narrow one. In our experiments, we have observed strong flow reversal around 1 kHz over a pair of symmetric electrodes. The electrolytes used in our experiments have conductivities of 2 9 10-4 S/m, 0.02 S/m (close to tap water), and up to 0.1 S/m (saline). We have observed the flow reversal only for the case of r = 0.02 S/m at low frequencies (*1 kHz). Since the flow direction over each individual electrode is reversed, i.e., going toward the gap, steric effect is not the responsible mechanism. Faradaic charging alone is unlikely responsible for flow reversal since it did not occur at higher ionic strengths (where Faradaic charging is more prone to occur). Observation of the microflow patterns leads us to believe that the microflows are induced by ACET effect, not ACEO. A signature of ACEO flows, the counter rotating vortices at electrode outer edges, were not observed. The explanation is that for the ionic strength of 0.02 S/m, ACEO and ACET are of comparable strength. As the signal frequency changes, the electric energy distribution in the fluidic cell changes, and the dominant ACEK mechanism could transit from ACEO to ACET. Furthermore, the temperature field distribution changes with electric field energy, which affects the directions and magnitudes of ACET forces. As a result, ACET forces become dominant and they dictate microflows moving in opposite direction. To support the above hypothesis, we measured the impedance spectrum of the electrolytic cell to extract the power consumption at the electrode interface and in the fluid bulk. Then the impedance data were used in numerical simulation (Comsol MultiphysicsTM), which showed changes in temperature gradients and produced ACET flow patterns similar to the experimental observation.

2 AC electrothermal effect and flow reversal ACEK can generate fluid motion through ACEO and ACET. ACEO flow relies on the movement of mobile ions in the electric double layer induced at electrode surfaces. When a low frequency AC signal is applied over an electrode pair, the electrode surfaces become capacitively charged, i.e., forming counter-ion accumulation, which is referred to as ‘‘capacitive charging.’’ The counter-ions will migrate under the influence of an electric field that is tangential to the electrode surface, which in turn produces fluid motion due to fluid viscosity. ACEO velocity depends on the strength of the tangential electric field and the quantities of mobile ions in the diffuse double layer. The latter is indicated by the electric potential drop over the electric double layer. Low operating frequencies will favor ACEO as the interface


759

impedance dominates, and ACEO velocity has a bell shape dependence on frequency with an optimal frequency around (r/e)(kD/L), where kD is the double-layer thickness and L is the characteristic length of the system, often the electrode gap. ACEO flow velocity is also highly dependent on the ionic strength of the fluid. With increasing ionic strength, the thickness of the double layer is greatly compressed, leading to a higher double-layer capacitance and less impedance in the electric path. That could mean less electric energy dissipation at the interface to induce ACEO, and the fluid velocity reduces. For the same applied voltage, more conductive fluids have a lower peak ACEO velocity, and the highest fluid conductivity with which ACEO has been observed is 0.085 S/m. On the other hand, increased ionic strength results in more current through the fluids, and consequently more heating of the fluid. As a result, ACET flows may become enhanced. ACET effect induces fluid motions from the gradients in conductivity and permittivity of the fluid. When an electric field Erms is applied over a fluid body, 2 electric energy is dissipated as hPi ¼ rErms (r: electrolyte conductivity). A non-uniform electric field will lead to a non-uniform temperature rise, i.e., temperature gradient rT, which will produce gradients in conductivity and permittivity, rr and re, as re = (qe/qT)rT, rr = (qr/qT)rT. In turn, rr and re generate mobile charges, qq, in the fluid bulk, by qq ¼ r ð~eEÞ ¼ r~e E þ ~er E and qqq/qt?r(rE) = 0 with q/qt = ix in AC fields. The free charges within solution are driven by electric fields, resulting in bulk fluid motion, which is known as ACET microflow. The electrical force per unit volume for an incompressible fluid can be expressed as: 1 fE ¼ qq E E2 re; 2

ð1Þ

where qq is the volume charge density. By assuming small perturbation of permittivity and conductivity (Ramos et al. 1998), the electric field in Eq. 1 can be written as the sum of applied component E0 and perturbation component E1 : Equation 1 can be rewritten as: 1 fE ¼ ðr~e E0 þ ~er E1 ÞE0 E02 re: 2

ð2Þ

Here ~e is a complex permittivity, and ~e ¼ e jr=x;

gradient rT, determines the magnitude of ACET force. rT can be imposed internally by Joule heating or externally such as by light illumination. Green et al. (2000a, 2001) presented qualitative observations of induced electrothermal flow from strong illumination. In Gonzalez et al. (2006), flow reversal was observed by adjusting the external illumination, where the heating effect by illumination dominated other heat sources (e.g., Joule heating). Flow directions in an electrothermal device can also be reversed by modifying its thermal boundary conditions. Perch-Nielsen et al. (2004) have shown by simulation how thermal conductivities of the substrates affect electrothermal flow directions. Their work studies traveling wave electrothermal systems with silicon and glass substrates. Electrothermal flows on a glass substrate move in opposite directions to those on a silicon substrate. Silicon is a better thermal conductor than glass, and conducts more heat away from the Joule heating. Thus, when no external heating is applied, for glass substrate the temperature maxima is located within the fluid, while the silicon substrate pulls the temperature maxima down onto the solid surface. This subtle difference changes the direction of thermal gradients and therefore, causes electrothermal forces to be opposite in directions. The above two situations for flow reversal do not involve a change in the applied AC signals. As a matter of fact, it is common belief that ACET flows will not be affected by AC frequency. In our experiments, we observed microflow reversal around 1 kHz at a fluid conductivity of 0.02 S/m. At frequencies much higher or lower than 1 kHz, flows resume their normal direction, i.e., from the inner gap outward. Different from Gonzalez et al. (2006) and Perch-Nielsen et al. (2004), we did not introduce external heat sources or modify the device materials to change the thermal gradients. Through numerical simulation of ACET forces, we find that for electrolytes with appropriate ionic strength, the location of temperature maxima is a function of frequency. In the simulation, the electrode surfaces are treated as heat sources since the double-layer polarization possesses resistive impedance. As AC frequency shifts, the distribution of energy dissipation at the electrode/fluid interfaces and in the fluid bulk also changes. The thermal field distribution changes and hence temperature gradients change with AC frequency, which results in the reversal of ACET flows.

ð3Þ

by substituting Eq. 3 into Eq. 2, we can re-write the electrothermal force as: " # ~ 2 1 rr re eE 1 ~ ~ E re ; F et ¼ þ ~ ð4Þ E 2 r e 1 þ ðxsÞ2 2 where s = e/r is charge relaxation time, x = 2pf is radian frequency. Electric field strength, together with temperature

3 Microfluidic experiments A pair of symmetric coplanar electrodes was used in our experiments. Since the flow motions over electrodes exhibit mirror symmetry, any observed flow reversal must originate from the reversed flows above each electrode. The structure of the microfluidic cell is schematically

123

760


Fig. 1 Schematic of the microfluidic cell for simulation and its boundary conditions

shown in Fig. 1. The electrodes are 160 lm in width with a gap of 10 lm. A polymer microwell of 500 lm in height (SA8R-0.5, Grace Bio-Labs, USA) is used to form the fluidic chamber. To observe the movement of fluids, 1 lm latex particles are seeded in the liquid and their trajectories were tracked by a CCD camera (Photometrics CoolSnap ES, USA). The AC voltage magnitude in experiments was fixed at 5.3 Vrms. The conductivities of fluid were 2 9 10-4 S/m (referred to as solution A) and 0.02 S/m (referred to as solution B). More conductive saline solutions (up to 0.1 S/m) were also used in the experiments, and no flow reversal was observed. The fluidic velocities were measured at six different frequencies ranging from 1 to 110 kHz. Interestingly, flow reversal phenomenon was observed only for solution B around 1 kHz. At the instant when the AC voltage was turned on, the particles would initially move from inner electrode gap to the middle of electrodes, similar to ACEO flows. Within a few seconds, before the flows were fully developed, counter flows started to form, which was indicated by tracer particles moving in from outer edges of both electrodes and going toward the center gap. Eventually, the counter flows become strong enough to reverse the original flow directions and form consistent fluid vortices. When the focal plane of the microscope was elevated from the electrode surface to about 40–60 lm above surface, the particles was observed to move in opposite directions, indicating two counter rotating vortices. Figure 2a, b shows the velocity curve for solutions A and B. The velocity data in Fig. 2 were acquired at the inner edges of the electrodes at the surface level. Five measurements were taken at each frequency. Except for the case of 1 kHz, the two solutions exhibit similar velocity

123

Fig. 2 Experiments results of particle velocities for different solution. a Velocity curve for solution A. b Velocity curve for solution B

variation patterns. For solution A, fluid velocity decreases noticeably from 1 to 30 kHz, which is likely due to the transition from ACEO to ACET flows. ACEO flows exhibit strong frequency dependence and are often faster than ACET flows at the same voltage (Ramos et al. 1998). In this frequency range, double-layer impedance still dominates, so ACET forces are weak. However, the electrode polarization cannot respond fast enough to the switching AC field, as a result, ACEO velocity goes down rapidly as the frequency increases. At 50 kHz and above, the interfacial impedance becomes negligible compared with the fluid bulk, ACET forces dominate and fluid velocity only vary slightly. For solution B, there is only small variation in fluid velocity, which is probably because ACET dominates at this ionic strength and there is not much difference in Joule heating of the fluid except for at 1 kHz.

4 Impedance analysis In order to explain the fluidic behavior with respect to AC frequency, an equivalence circuit model as previously


761

Impedance data of microfluidic systems

(a)

|z| (Ohm)

100000

10000

1000

Fig. 3 Equivalence circuit model for the microfluidic cell Measured impedance for Solution A Simulated impedance for Solution A Measured impedance for Solution B Simulated impedance for Solution B

100

100

1k

10k

100k

Frequency (Hz)

Impedance phase of microfluidic systems

(b)

Phase Angle (degree)

described in Lian et al. (2007), also shown here in Fig. 3, is adopted to extract the power dissipation at different parts of the fluidic cell. We assume that the fluid bulk obeys Ohm’s law and can be regarded as a resistor. The double-layer polarization impedance is approximated by a constant phase element (CPE), whose impedance has the form of Z = A/ (jx)b. When b is 1, the CPE model describes the behavior of an ideal capacitor. In our case, the values of b are between 0.65 and 0.85. The power dissipation at the electrode surface is then calculated as the real part of the product of the electric current and the voltage for the double-layer polarization. At very high frequency, AC electric field can also pass through the fluid by dielectric coupling between the electrodes, i.e., the fluid is equivalent to a dielectric medium and it is represented by Cdielectric in Fig. 3. The impedance from dielectric coupling (1/jxCdielectric) is usually several magnitudes larger than the double-layer impedance at low frequencies. Therefore, its influence will not manifest until at high frequencies. Figure 4 shows the impedance spectra for solution A and B (bottom). Both spectra show a decrease at low frequency that is attributed to the double-layer capacitance. At middle frequency range, the fluid resistance causes most of the voltage drop and the plots exhibit a more flat slope. At higher frequencies, the dielectric capacitance dominates and the impedance magnitude decreases at -20 dB/decade, as in any simple RC filters. The values of the components in the equivalent circuit of Fig. 3 were extracted by curvefitting of Fig. 4 and impedance calculation with Matlab. Good agreement can be seen between the measurements and the extracted equivalent circuit. Table 1 lists the extracted component values. With the impedance values, the heat generation at each impedance component is calculated as a function of AC frequency. Figure 5 compares the heat dissipation at the electrode/fluid interfaces and in the fluid. For solution A, the large resistance in the fluid bulk dictates that, at all frequencies, much more voltage is dropped over the fluid than over the fluid/electrode interface, leading to consistent higher power consumption in solution, as shown in Fig. 5a. For solution B, the heat dissipation has a very different pattern from that of solution A at low frequencies, as shown

20 Measured phase for Solution A Simulated phase for Solution A Measured phase for Solution B Simulated phase for Solution B

0

-20

-40

-60

-80 100

1k

10k

100k

Frequency (Hz)

Fig. 4 Impedance spectra of the microfluidic cell. Lines are the fitted values. a Impedance magnitude, and b phase angle

in Fig. 5b. At low frequencies (*1 kHz), more voltage drops at the electrode interfaces than over the fluid, 3.5 Vrms as opposed to 1.8 Vrms, hence more heat dissipates at the electrode surfaces. With increasing frequency, the impedance of the double layer goes down, and the differences of heat dissipation between two parts are reduced. According to Fig. 5b, the heat dissipations at the electrode and in the fluid equal each other at approximately 3 kHz, beyond which the fluid heating again dominates. As a result, the temperature maximum in solution B will move as a function of frequency, leading to the flow reversal as observed in experiments.

5 ACET numerical modeling We have used Comsol Multiphysics to simulate the effects of heat distribution on ACET microflow field. The

123

762


Table 1 Extracted impedance values corresponding to the equivalent circuit in Fig. 3 Impedance (X) Solution A

Solution B

40

120 K

37 K

100 1K

83 K 46 K

18 K 4.6 K

10 K

12.2 K

2.7 K

50 K

2.7 K

1.7 K

200 K

670

620

A

3.30E ? 06

3.60E ? 06

b

0.6788

0.8424

R_fluids (X)

43 K

1.5 K

C_dielectric (F)

1.20E - 09

1.20E - 09

Surface impedance

Z = A/(jx)b

Frequency (Hz)

Extracted parameters

electrodes are assumed to be infinitely long in the z-direction (perpendicular to the paper). So we can treat the problem as 2D and only simulate the cross sectional area. Because the experiments with solution B present several scenarios of power distribution in the fluid, ACET simulation was done only for solution B (r = 0.02 S/m) as it is of interest to the paper and can be readily validated by the experiments. Three modules are used in our simulation to find the electric field, temperature distribution and fluidic velocity. First, the electric field distribution in the fluidic chamber is derived. The resulting electric field distribution is used to calculate the temperature field according to the energy equation. Then the fluid volume force in the chamber, Fet, is calculated using the temperature gradient and the electric field distribution from the first two steps. Lastly, the fluid flow field is obtained by Navier–Stokes equation. The electric module solves the field distribution according to qq ¼ r ð~e~ EÞ;

ð5Þ

where ~e is a complex permittivity as defined in Eq. 3. Here, the electric field E is within the fluid bulk. Therefore, the voltage drop applied over the electrodes should be the voltage over Rfluid in Fig. 3. For example, in solution B, the potential drops over the fluid between the electrodes are 1.8 Vrms at 1 kHz, and 4.8 Vrms at 10 kHz, respectively. Other solid boundaries are treated as electrical insulator. In a microfluidic system, when the solution has nonnegligible conductivity, the electric field drives substantial conduction current through the fluid, causing localized

123

Fig. 5 Distribution of power consumption as a function of AC frequency for different solutions. a Solution A and b solution B

Joule heating. In order to find the temperature rise in the microfluidic cell (Ramos et al. 1998), the energy balance equation needs to be solved, qm cp v rT þ qm cp

oT ¼ kr2 T þ rE2 : ot

ð6Þ

Order of magnitude estimate in Castellanos et al. (2003) shows that temperature equilibrium is reached quickly after AC voltage is applied and the effect of fluid motion on the temperature distribution is minimal, Eq. 6 can then be reduced to kr2T ? rE2 = 0. It is common practice to consider the electrodes at ambient temperature. However, in this work, we take into account the heat generation at the electrode surface. The electrodes are defined as a heating source in the thermal module. The heating power density is derived from energy dissipation and actual electrode size, and is calculated as 3,125 W/m2 for 1 kHz and 1,172 W/m2 for 10 kHz. Other


763

boundary conditions in thermal module include setting the fluid inlet and outlet as thermal insulation, since no heat exchange happens on these boundaries. The cover of our fluidic chamber is made of plastic. Therefore, it is treated as a contact layer to ambient temperature with a reasonable guess of heat transfer coefficient (h = 600 W/m/K) in heat transfer module. The heating process produces gradients in local conductivity and permittivity by re = (qe/qT)rT and rr = (qr/qT)rT. For aqueous media, we have (Lide 2000) (1/e)(qe/qT) = -0.004 ) (re/e) = (1/e)(qe/qT)rT = -0.004rT and (1/r)(qr/qT) = 0.02 ) (rr/r) = (1/r) (qr/qT)rT = 0.02rT. These relationships are substituted into the calculation of thermal force and incorporated into our thermal module. Then, the fluid velocity, u, is solved for using Navier– Stokes equation, rp þ gr2 u þ ~ F et ¼ 0

ð7Þ

along with the mass conservation equation for incompressible fluid r u = 0, where p is pressure, g is fluid

viscosity. The water properties are given as q = 1,000 kg/ m3 and g = 0.001 kg/m/s at 20°C. No slip boundary condition, u = 0, is applied to all the solid boundaries. Figure 6 gives the simulated ACET flow profile at r = 0.02 S/m (solution B), and three situations are considered. Figure 6a, b are both for ACET at 1 kHz. Joule heating in fluid is limited to a low level due to large impedance at low frequencies. Electrodes in Fig. 6a are treated as a thermal conductor with no surface heating. The area with considerable temperature rise is confined within the electrode gap where electric fields concentrate. The thermal force is directed from the inner gap toward the outer electrode edges, pushing fluids outward and pulling down the fluid in the middle because of mass conservation. When surface heating exceeds the bulk heating (Fig. 6b), the temperature gradient points from the surface where the highest temperature is and fans out into the fluid bulk. The resulting thermal force becomes pointing upward in the middle, and form vortices in opposite directions to those in Fig. 6a. The simulation indicates flow reversal as observed in the experiments. Simulated magnitude of reversal flow is

Fig. 6 Numerically simulated ACET microflows in solution B at three conditions. a When no surface heating is included at f = 1 kHz (Vrms = 1.8 V, Q = 0). b Surface heating is included at f = 1 kHz (arrow length not to scale; Vrms = 1.8 V, Q = 3125 W/ m2). c At high frequencies (*10 kHz; Vrms = 4.84 V, Q = 1,172 W/m2)

123

764

however one order smaller than the experiment (tens of lm/s compared to hundreds of lm/s). Figure 6c shows the velocity and temperature profile at 10 kHz. The surface heating becomes less important in magnitude compared with Joule heating in the fluid bulk. Similar to Fig. 6a, the high temperature area is limited to a small region around the electrode gap, so the regular ACET flow direction is again assumed. The simulated velocity in Fig. 6c is on the order of hundreds of lm/s, which is directly comparable to the experiment data.

6 Low frequency electrode processes and conclusions At low frequencies (well below charge relaxation frequency x = r/e), ACEO typically will generate a higher fluid velocity than ACET. ACEO certainly contributes appreciably to the velocity data in Fig. 2 for frequencies below 10 kHz, especially for solution A. The influence of ACEO decreases with frequency and ionic strength of the solution, while ACET velocity is supposed to be independent of frequency. The frequency dependency of ACEO fluid flow was documented by Green et al. (2000b). Using their results as a reference, solution A should shows a peak in velocity well below 1 kHz and the solution B should have a velocity peak between 2 and 10 kHz. Our velocity data in Fig. 2a agree with the ACEO characteristics, while those in Fig. 2b do not, indicating that mechanism other than ACEO should be responsible. While the numerical simulation shows that ACET forces produce similar fluidic patterns to that observed, the magnitude of ACET microflow is much smaller as indicated by the simulation. This leads to speculation about other possible contributing factors. It was recently indicated by our experiments (Lian and Wu 2009) that reactions could dramatically enhance ACET flows by inducing significant conductivity gradient rr/r, as suggested by Eq. 4. In the experiments, the flow reversal occurred at an applied voltage of 5.3 Vrms. At this level, Faradaic reactions probably took place to generate ions and high rr/r, leading to reaction-enhanced ACET. Instead of Faradaic charging ACEO, reaction-enhanced ACET is suggested here as the responsible mechanism, because the flow patterns support such a conclusion. For the coplanar electrodes used in this work, ACEO and ACET produce very similar flow patterns. However, for ACEO flows, there exist two minor counter rotating vortices near the outer edges of electrodes (Lian et al. 2006; Wu et al. 2005a). Such vortices result in the stagnant lines on the electrodes where tracer particles are found to deposit. ACET effect does not produce counter vortices, and hence, no particle assembly lines on the electrodes. Such a difference can be used to distinguish ACEO and

123


ACET effects. In our experiments, we have observed particle line aggregation at 1 kHz for solution A, however, not for higher ionic strengths or higher frequencies, indicating that ACET becomes stronger than ACEO in those situations. To sum up, an interesting flow reversal at low frequency was observed in our experiments. Since conventional ACEO and ACET theories cannot account for this phenomenon, an ACET flow reversal mechanism is presented and discussed in this paper. Using an equivalence circuit for the microfluidic system, we are able to illustrate the frequency dependence of ACET microflows. Increased dissipation of electric energy at the electrode surface leads to the change in temperature gradients and ACET flow pattern, while reactions could enhance rr/r resulting in high fluid velocity. The study of flow reversal helps understand ACET effect and realize reliable control of microfluidic devices. Acknowledgment The project has been supported by the US National Science Foundation under grant number ECS-0448896.

References Castellanos A, Ramos A, Gonza´lez A, Green NG, Morgan H (2003) Electrohydrodynamics and dielectrophoresis in microsystems: scaling laws. J Phys D Appl Phys 36:2584–2597 Gonzalez A, Ramos A, Morgan H, Green NG, Castellanos A (2006) Electrothermal flows generated by alternating and rotating electric fields in microsystems. J Fluid Mech 564:415–433 Green NG, Ramos A, Gonzalez A, Castellanos A, Morgan H (2000a) Electric field induced fluid flow on microelectrodes: the effect of illumination. J Phys D Appl Phys 33:L13–L17 Green NG, Ramos A, Gonzalez A, Morgan H, Castellanos A (2000b) Fluid flow induced by nonuniform ac electric fields in electrolytes on microelectrodes. I. Experimental measurements. Phys Rev E 61:4011–4018 Green NG, Ramos A, Gonzalez A, Castellanos A, Morgan H (2001) Electrothermally induced fluid flow on microelectrodes. J Electrostat 53:71–87 Gregersen M, Olesen L, Brask A, Hansen M, Bruus H (2007) Flow reversal at low voltage and low frequency in a microfabricated ac electrokinetic pump. Phys Rev E 76:056305 Kilic MS, Bazant MZ (2007) Steric effects in the dynamics of electrolytes at large applied voltages. I. Double-layer charging. Phys Rev E 75:021502 Lastochkin D, Zhou R, Wang P, Ben Y, Chang H-C (2004) Electrokinetic micropump and micromixer design based on ac faradaic polarization. J Appl Phys 96(3):1730–1733 Lian M, Wu J (2009) Ultra fast micropumping by biased alternating current electrokinetics. Appl Phys Lett 94:064101 Lian M, Islam N, Wu J (2006) Particle line assembly/patterning by microfluidic AC electroosmosis. J Phys Conf Ser 34:589–594 Lian M, Islam N, Wu J (2007) AC electrothermal manipulation of conductive fluids and particles for lab-chip applications. IET Nanobiotechnol 1(3):36–43 Lide DR (2000) CRC handbook of chemistry and physics, 81st edn edn. CRC, New York Perch-Nielsen IR, Green NG, Wolff A (2004) Numerical simulation of traveling wave induced electrothermal fluid flow. J Phys D Appl Phys 37:2323–2330

Microfluid Nanofluid (2009) 7:757–765 Ramos A, Morgan H, Green NG, Castellanos A (1998) AC electrokinetics: a review of forces in microelectrode structures. J Phys D Appl Phys 31:2338–2353 Storey BD, Edwards LR, Kilic MS, Bazant MZ (2008) Steric effects on ac electro-osmosis in dilute electrolytes. Phys Rev E 77:036317 Studer V, Pepin A, Chen Y, Ajdari A (2004) An integrated AC electrokinetic pump in a microfluidic loop for fast and tunable flow control. Analyst 129:944–949 Urbanski JP, Thorsen T, Levitan JA, Bazant MZ (2006) Fast AC electro-osmotic pumps with non-planar electrodes. Appl Phys Lett 89:143508

765 Wu J, Ben Y, Battigelli D, Chang H (2005a) Long-range AC electrokinetic trapping and detection of bioparticles. Ind Eng Chem Res 44(8):2815–2822 Wu J, Ben Y, Chang H-C (2005b) Particle detection by microelectrical impedance spectroscopy with asymmetric-polarization AC electroosmotic trapping. Microfluid Nanofluidics 1(2):161– 167 Yang K, Wu J (2008) Investigation of microflow reversal by AC electrokinetics in orthogonal electrodes for micropump design. Biomicrofluidics 2:024101

123