DEP Particle Dynamics and the Steady Drag Assumption

DEP Particle Dynamics and the Steady Drag Assumption

S.T. Wereley*, I. Whitacre*, R. Bashir** and H.B. Li**
*
Purdue University, School of Mechanical Engineering,
West Lafayette, IN 47907-2088, USA, [email protected]
**
Purdue University, School of Electrical and Computer Engineering,
West Lafayette, IN 47907-2088, USA, [email protected]

ABSTRACT a 0.3 Pm thick layer of PECVD silicon dioxide, which

insulates the electrodes from the liquid medium,
The interaction of fluid drag, dielectrophoretic forces, suppressing electrolysis. The electrodes are arranged in
and Brownian motion on a nanoparticle's motion is studied interdigitated pairs so that the first and third electrodes from
using a microfluidic chip with interdigitated electrodes. The Figure 1 are always at the same potential. The second and
flow domain is a 11.6 Pm deep by 350 Pm wide channel fourth electrodes are also at the same potential, but can be
with 23 Pm wide electrodes located at the bottom surface of at a different potential than the first and third electrodes. An
the channel and a glass top surface. The electrodes are alternating electric potential is applied to the interdigitated
covered by a thin layer of silicon dioxide to insulate them electrodes to create an electromagnetic field with steep
from the fluid medium, suppressing electrolysis and local spatial gradients. Particle motion through the resulting
Joule heating. Although these phenomena have been electric field gradients causes polarization of the suspended
considered by other researchers, our experiments and components, resulting in a body force that repels particle
modeling reveal it to be a considerably more complicated motion into increasing field gradients. In the experiments,
phenomenon than previously thought. Using an adapted sample solutions were injected into the chamber using a
micro Particle Image Velocimetry technique along with syringe pump (World Precision Instruments Inc., SP200i)
microscopic imaging, particle motion in three-dimensions is and a 250Pl gas-tight luer-lock syringe (ILS250TLL, World
measured and compared to predicted results, showing not Precision Instruments Inc.). The flow rate could be adjusted
only the expected horizontal DEP retarding force but also a and precautions were taken to avoid air bubbles. An HP
vertical force away from the electrodes. Further, because of 33120A arbitrary waveform generator was used as the AC
the spatially varying nature of both the DEP force as well as signal source to produce sinusoidal signal with frequency
the drag force, one of the main assumptions made in many specified at 1MHz.
previous DEP studies must be seriously questioned—
whether steady low Reynolds number particle dynamics are
insufficient to predict the particle behavior.

Keywords: dielectrophoresis, DEP, PIV, particle, velocity

1 INTRODUCTION AND SET UP

1.1 DEP Device

A dielectrophoretic device has been designed to trap,
separate, and concentrate biological components carried in
solution. This is done by creating a dielectrophoretic
interaction between the spheres and the fluid. The device
was designed and manufactured by Haibo Li at Purdue
University, a student in Prof. Bashir’s research group [1].
The device consists of a microchannel with a depth of 11.6
µm, width of 350 µm, and length of 3.3 mm. The channel
was anisotropically etched in silicon to produce a
trapezoidal cross-section. The channel was covered by a
piece of anodically bonded glass. A schematic view and
digital photo of the device are shown in Figure 1. Bright
regions represent platinum electrodes and the dark regions Figure 1. (a) Schematic view of experimental apparatus
represent the electrode gaps. The electrodes are covered by and (b) photo of apparatus.

320 NSTI-Nanotech 2004, www.nsti.org, ISBN 0-9728422-7-6 Vol. 1, 2004

1.2 Micro Particle Image Velocimetry array is 1392 by 1040 pixels, but the camera has the
capability of pixel binning, which can drastically increase
Particle image velocimetry (PIV) is a well accepted the acquisition frame rate by reducing the number of pixels
technique for acquiring full field velocity information from that need to be digitized. A three-by-three pixel binning
experimental fluid flows [2]. Micro particle image scheme was used in this experiment, producing images
velocimetry (µPIV) has expanded conventional PIV to measuring 464 by 346 pixels, which were captured at a
utilize microscopic imaging devices as developed by [3]. speed of 20 frames per second. The average focused
µPIV results have been used in a number of applications particle diameter in the images was approximately 3 pixels.
such as a detailed investigation of pressure driven flow Shallow Channel Considerations
through a microchannel [4], micron resolution temperature When performing µPIV measurements on shallow
measurements [5], and investigations of dielectrophoretic microchannels, the depth of focus of the microscope can be
particle trapping [6,7]. comparable in size to the depth of the flow. A PIV cross-
Figure 2 shows the typical layout of a µPIV system. correlation peak, the location of which is the basis for PIV
The measurement volume can be illuminated by either a velocity measurements, is a combination of the velocity
broad wavelength constant light source or a pulsed laser, distribution in the interrogation region and some function of
such as a frequency doubled Nd:YAG. In these average particle shape. PIV velocity measurements
experiments a standard microscope mercury lamp is used. containing velocity gradients can substantially deviate from
To ensure that all particles have the same properties, 0.7 the ideal case of depthwise uniform flow. Gradients in the
Pm polystyrene latex (PSL) microspheres (Duke Scientific) in-plane direction have been addressed by image correction
are suspended in de-ionized water in dilute concentrations techniques [8], but gradients in the depthwise direction
(less than 0.1% by volume). Tracer particles in the remain problematic. They can cause inaccurate velocity
measurement volume are coated with a red fluorescing dye measurements due to the presence of multiple velocities
(Oemit=542 nm, Oemit=612 nm). Images are captured with within an interrogation region that are independent of mesh
precise time delay separating exposures. The image sets refinement. One problem with depthwise velocity gradients
are then divided into interrogation regions. The is cross-correlation peak deformation which reduces the
corresponding interrogation regions from consecutive signal to noise ratio of a PIV measurement [9]. Cross-
images are cross-correlated to determine the most likely correlation peak deformation can also reduce the
relative displacement of the particles in the interrogation effectiveness of subpixel peak fitting schemes which are
regions in the form of a cross-correlation peak. Reducing based on a particular cross-correlation shape, such as a
the size of an interrogation region reduces the measurement common five point Gaussian fit.
volume and increases the spatial resolution of a
measurement so long as displacements are less than half the 2 EXPERIMENTAL RESULTS
size of an interrogation region, i.e. the Nyquist criterion.
The images were acquired using a Photometrics The experiments presented here designed to quantify the
CoolSNAP HQ interline transfer monochrome camera dielectrophoretic performance of the device. The
(Roper Scientific). This camera is capable of 65% quantum experiments used six sets of 800 images each to analyze the
efficiency around the 610 nm wavelength. The largest effect of dielectrophoresis on particle motion in the test
available image size that can be accommodated by the CCD device. These images are high quality with low readout
noise, as can be seen in the example fluorescent image of
Fluorescent Microspheres
Figure 3. The top image demonstrates the many different
Focal Plane particle intensity distributions which are typically present in
Cover Slip
a µPIV image in which the particles are distributed
randomly within the focal plane. The bottom figure shows
Microscope Objective how, as the result of the DEP force, the particles migrate to
Exciter 532 nm the top of the channel and all have nearly identical images.
The top image also shows that when a significant DEP
Epi-Flourescent
Filter Cube force exists, the particles are trapped at the electrode
Laser
locations by the increase in DEP force there. In general, the
Emitter 610 nm
Beam Expander observed particle image shape is the convolution of the
geometric particle image with the point response function
12 bit CCD Camera of the imaging system. The point response function of a
(1280 x 1024 pixels) microscope is an Airy function when the point being
imaged is located at the focal plane. When the point is
displaced from the focal plane, the Airy function becomes a
Lommel function [10]. Both Airy and Lommel functions
Figure 2. Diagram of typical µPIV system. feature concentric rings, although at different intensities

NSTI-Nanotech 2004, www.nsti.org, ISBN 0-9728422-7-6 Vol. 1, 2004 321

 70

y (microns)
60
50

40
30

20
10

20 40 60 80
x (microns)
70

y (microns)
60

50

40
30
Figure 3. A photo of PSL particles with 0.0 volts (top) and
4.0 volts (bottom). 20

and diameters. For a standard microscope the diffraction 10

limited spot size is given by
20 40 60 80
1.22 ˜ O x (microns)
de (1)
NA
Figure 4. PIV vector plots for electrode voltages of 0.5
A numerical aperture (or NA) of 1.00 and an incident light volts (top) and 4.0 volts (bottom).
wavelength Ȝ of 540nm results in a diffraction limited spot more easily interpreted. In all cases, the electrode voltage
size of 0.66µm, while the particles used are 0.69µm. signal frequency was set to 580 kHz. This was determined
Consequently the particle intensity distributions as recorded by sweeping between frequencies of 100 kHz to 10MHz
by the camera are partly due to the geometric image of the and qualitatively determining the most effective trapping
particle and partly due to diffraction effects. Hence, the frequency.
distance of any particle from the focal plane can be The PIV results are summarized in Figure 5 which
determined by the size and shape of the diffraction rings. shows the average axial velocities for the six electrode
A conventional µPIV analysis was performed to obtain
40
an initial estimate of the average particle velocity field. 0.5V
Because the goal of the conventional µPIV analysis is not to 35
1.0V
2.0V
extract velocity distribution, only the median velocity is 2.5V
Average Axial Velocity (microns/sec)

reported. The spatial resolution can be higher than when 30 3.5V

4.0V
using a velocity distribution analysis method such as the
25
synthetic image method. In section 5.4, the velocity field
will be extracted using the synthetic image approach. The 20
PIV images were analyzed with EDPIV1, an advanced PPIV
interrogation package written in part at Purdue University 15

by Lichuan Gui. An interrogation region of 32 by 32 pixels

10
and a grid spacing of 8 by 8 pixels were used. Figure 4
shows vector plots for the experimental cases of 0.5 volts 5
(top) and 4.0 volts (bottom). A uniform color scale was
applied to these figures for representing velocity 0
0 20 40 60 80 100 120 140
magnitude, such that velocity changes between cases can be Axial Position (microns)

1 Figure 5. Average axial velocity from PIV results for all

Evaluation version available at
http://www.ecn.purdue.edu/microfluidics/Edpiv.zip electrode voltage cases.

322 NSTI-Nanotech 2004, www.nsti.org, ISBN 0-9728422-7-6 Vol. 1, 2004

voltage cases. The three lowest voltages share a trend of coefficient captures the mean velocity of the particles
decreasing particle velocity in the downstream direction. within the device and the C coefficient captures the
This phenomenon is evident in the average velocity curves sinusoidal behavior of particles as they encounter each
in Figure 5. One explanation for this behavior may be that electrode. These results are shown graphically in Figure 6.
with each electrode a particle encounters, it lags the fluid The parameters for the cases of 0.5 to volts 2.5 volts are
velocity a little more. The cumulative effect results in a very similar, with the exception of a gradual increase in B.
gradual slowing of the particle. In the two highest voltage cases the linear slope A is
Another interesting result apparent from Figure 5 is that reduced to nearly zero indicating that the particles do not
initially the average particle velocity increases as the speed up or slow down as they pass through the devices
voltage increases, 0.5 volts to 2.0 volts. This phenomenon while the magnitude of the sinusoidal component C is
is explained by particles being displaced from the channel greatly increased. From these results it is apparent that the
bottom into faster areas of the fluid flow. This biases the particle velocities are reduced as particles travel
velocity distribution toward higher velocities, altering the downstream for low electrode voltages, and that the
shape of the cross-correlation peak to favor higher particles are dominated by a periodic trajectory for higher
velocities even though the fluid flow is constant. For higher electrode voltages.
voltages the effect of particles being hindered by axial field 3 CONCLUSIONS
gradients is compounded by particles being forced beyond The dynamics of particles traveling through the device
the high speed central portion of the flow profile by the described in this paper are very complicated, exhibiting
DEP force. It can be qualitatively confirmed that particles migration normal to the electrodes as well as trapping
migrate to the top of the channel by observing particle behavior in the plane of the electrode. Further work is
shapes in the images from the higher voltage cases, i.e. needed to assess the accuracy of the steady drag
comparing the many particle shapes found in Figure 3 (top) assumption.
which is acquired at 0.5 volts with the single particle shape
found in Figure 3 (bottom) which is acquired at 4.0 volts. It REFERENCES
is confirmed that the particles are indeed at the top of the [1] HB Li, Y Zheng, D Akin, R Bashir,
channel by traversing the focal plane throughout the depth “Characterization and Modeling of a Micro-Fluidic
of the device. Dielectrophoresis Filter for Biological Species,”
The experimental results in Figure 5 were fit to the submitted to J. Microelectromechanical Sys. (2004).
simple phenomenological model [2] R.J. Adrian, “Particle-imaging techniques for
experimental fluid mechanics,” Ann. Rev. Fluid
Vaxial A ˜ x  B  C ˜ sin  2S D  E  (2) Mech., Vol. 23, 261-304, 1991.
[3] J.G. Santiago, S.T. Wereley, C.D. Meinhart, D.J.
by minimizing the root mean square difference between the Beebe, R.J. Adrian, “A particle image velocimetry
fit function and the experimental data. The resulting system for microfluidics,” Experiments in Fluids,
constants capture important dynamics of the particle Vol. 25, 316-319, 1998.
velocity. The A coefficient captures particles speeding up [4] C.D. Meinhart, S.T. Wereley, J.G. Santiago, “PIV
or slowing down as they pass through the device. The B measurements of a microchannel flow,”
Experiments in Fluids, Vol. 27, 414-419, 1999.
1.2
A
[5] V. Hohreiter, S.T. Wereley, M. Olsen, and J.Chung,
B “Cross-correlation analysis for temperature
1 C
Normalized Coefficient Value

measurement,” Meas. Sci. Tech., Vol. 13, 1072-

0.8
1078, 2002.
[6] E.B. Cummings, “A comparison of theoretical and
0.6
experimental electrokinetic and dielectrokinetic
flow fields,” AIAA 2002-3193, 2002.
0.4 [7] C.D. Meinhart, D. Wang, K. Turner, “Measurement
of Ac Electrokinetic Flows,” Biomedical
0.2 Microdevices, Vol. 5, No. 2, 139-145, 2002.
[8] S.T. Wereley, L. Gui, C.D. Meinhart, “Advanced
0 Algorithms for Microscale Velocimetry,” AIAA J.,
Vol. 40, No. 6, 1047-1055, 2002.
-0.2
0.5 1 1.5 2 2.5 3 3.5 4
[9] E.B. Cummings, “An image processing and optimal
Voltage (V) nonlinear filtering technique for PIV of
microflows”, Experiments in Fluids, Vol. 29,
Figure 5. Curve fit parameters from Eq. 2 versus electrode [Suppl.]:S42-50, 2001.
voltage. The dashed lines represent linear, piecewise [10] M. Born, E. Wolf, Principles of optics, Oxford
linear, and parabolic curve fits for parameters A, B, and C. Press, Pergamon, 1997.

NSTI-Nanotech 2004, www.nsti.org, ISBN 0-9728422-7-6 Vol. 1, 2004 323