Formation of high electromagnetic gradients through a particle-based microfluidic approach Yuh ‘Adam’ Lin

by user








Formation of high electromagnetic gradients through a particle-based microfluidic approach Yuh ‘Adam’ Lin
J. Micromech. Microeng. 17 (2007) 1299–1306
Formation of high electromagnetic
gradients through a particle-based
microfluidic approach
Yuh ‘Adam’ Lin1, Tak-Sing Wong2, Urvashi Bhardwaj3,
Jia-Ming Chen4, Edward McCabe3,4 and Chih-Ming Ho2,4
Department of Biomedical Engineering, University of California, Irvine, CA, 92697,
Department of Mechanical and Aerospace Engineering, University of California,
Los Angeles, CA, 90095, USA
Department of Pediatrics, David Geffen School of Medicine, University of California,
Los Angeles, CA, 90095, USA
Institute for Cell Mimetic Space Exploration, University of California, Los Angeles,
CA, 90095, USA
E-mail: [email protected], [email protected], [email protected], [email protected],
[email protected], [email protected] and [email protected]
Received 6 March 2007, in final form 14 May 2007
Published 5 June 2007
Online at stacks.iop.org/JMM/17/1299
The ability to generate strong magnetic field gradients is a prerequisite for
efficient magnetic-based cell/bio-particle separation or concentration.
Creating these gradients is difficult under microscale fluidic devices.
Conventional MEMS magnetic-based microfluidic devices involve the use
of non-trivial and expensive multi-layer fabrication processes in order to
produce magnetic field generators/concentrators (e.g. metal
coil/ferromagnetic structures) around the microfluidic channels. A
microfluidic device with simplified fabrication procedures while achieving
the same functional purposes of magnetic separation/concentration of
particles is highly desirable. Here, we propose a simple single-layer,
single-mask fabrication technique for magnetic MEMS fluidic device
construction, where nickel microparticles can be monolithographically
integrated into any configurations. We constructed the microfluidic device
through conventional PDMS replicate molding, with injection of nickel
microparticles into a side channel 25 µm apart from the main separation
channel. The nickel microparticles are responsible for bending and
concentrating the external magnetic field for gradient generation. This
magnetic field gradient induced magnetic forces on the particles present in
the main channel. The force generated by the presence of the nickel
particles is 3.31 times greater than that without the use of a magnetic field
concentrator (i.e. nickel particles). The proposed methodology can be
extended for the development of automated high-throughput microfluidic
cell separation devices. The simplicity of fabrication and enhanced
magnetic separation efficiency shows great promise for future microfluidic
(Some figures in this article are in colour only in the electronic version)
© 2007 IOP Publishing Ltd
Printed in the UK
Y (Adam) Lin et al
1. Introduction
Creating large electromagnetic field gradients is crucial for
numerous cell or bio-particle separation or concentration
applications, such as using dielectrophoresis in achieving
particles separation [1]. Different from macro-scale devices,
high magnetic field gradients in microscale fluidic system
are difficult to generate. Previous developments to generate
large magnetic field gradients were achieved by changing
the shape and position of magnets that surrounded main
fluidic channels. Quadrupole and dipole magnetic systems had
been successful to separate cells in channels with diameters
in the millimeter range [2, 3]. The purity of the separated
sample is high (99%) but the recovery rate, defined as
the percentage of target cells recovered from the original
sample, is unstable (37–86%) [4]. Recent developments
use MEMS technology to generate a magnetic field
gradient through the use of micro-coils and magnetic pillars
[5, 6]. Although these platforms can easily manipulate the
magnetic beads in batches, they do not provide a continuous
The above-mentioned MEMS magnetic devices require
non-trivial and expensive multi-layers fabrication processes in
order to integrate the magnetic materials with the microfluidic
channels to achieve magnetic particles separation.
microfluidic system that allows a simple fabrication procedure
while achieving the same functional purpose of magneticbased separation is highly desirable. Here, we present a
simple single-layer, single-mask fabrication technique for
magnetic MEMS fluidic devices that is capable to perform
cell separation. Generally speaking, magnetic cell separation
or manipulation requires a carrier (i.e. magnetic bead) to
attach to the target cells. The magnetic beads, also known as
Dynabeads (Invitrogen, CA), are 4.5 µm superparamagnetic
cores with polystyrene shells. The surfaces of the beads are
coated with antibodies targeted toward specific cell membrane
markers for certain cell types. Methods for handling the
magnetic beads have been very crucial for biochemical
and analytical applications [7, 8]. A large interest in
cell separation within automated systems has grown among
the medical field especially for oncology or hematology
Different from the conventional methods of fabricating a
magnetic structure within or close to a microfluidic channel
to achieve magnetic field focusing, our device exploits a side
channel that was located close to the main sample channel
(figure 1), where small metal particles can be injected. Small
metal particles, such as nickel, were utilized as the media
to concentrate magnetic fields. The presence of the nickel
particles in an adjacent side channel increases the magnitude
of the magnetic field density gradient which corresponds to an
increase in the translational force exerted on the magnetic
beads. Apart from the device fabrication simplicity, this
method also has the potential to achieve stable and high
recovery rates due to sophisticated force control within the
microenvironment. In addition, the fabrication cost for
the device can be relatively low, which may lead to mass
production and commercialization for clinical or research
Figure 1. A schematic showing the concept of separation of
cells/particles attached to the magnetic beads using metal (nickel)
particles as a medium to generate a large magnetic field gradient.
2. Theory
The magnetic force, Fb, generated on a magnetic bead is
governed by the following equation [9]:
Fb =
χ · Vb · ∇B 2
where µ0 is the magnetic permeability of free space, χ is
the difference of susceptibility between the magnetic bead and
the surrounding medium, Vb is the volume of the bead and
B is the magnetic field density. It is important to recognize
that a gradient of the magnetic field density is required for a
translational force. A strong uniform magnetic field can only
cause rotational force but not translational force.
The total magnetic force acting on a cell with magnetic
beads attached is
F m = Ac · α · β · Fb
where Ac is the total surface area of the cell, α is the number
of target cell surface markers per membrane surface area, β is
the number of antibody bound per marker and Fm is the force
acting on one magnetic bead.
Countering the magnetic force is the drag force, Fd,
defined by the Stokes drag law:
Fd = 6π · η · r · v
where η is the viscosity of the medium, r is the radius of the cell
and v is the velocity of the cell moving through the medium.
Assuming that gravity and buoyant forces are negligible,
the two forces combine into
Fm + Fd = ma
where m is the mass of the cell and a is the acceleration of the
cell. The inertial term (∼10−11) is several orders of magnitude
smaller than the total magnetic force and the Stokes drag
force (∼10−6) [10]. Thus, we can neglect the inertial term in
equation (4). This assumption allows us to find the relationship
between the lateral velocity caused by the induced magnetic
force and the minimum magnetic field density gradient (∇B2)
Formation of high electromagnetic gradients through a particle-based microfluidic approach
(A )
(B )
(D )
(C )
(E )
Figure 2. (A) Simulation of the magnetic field density with Ni particles, Ni bar and magnet only. The nickel particles and the nickel bar
were placed in between 0 and 50 µm. (B) A graph showing the magnetic field density across the center of each simulation case. (C) A
magnified portion of (B) showing the magnetic field density of the center line from 50 to 100 µm. (D) The discrete one-dimensional
gradient (B2/x) for each simulation case. (E) A magnified portion of (D) showing the discrete one-dimensional gradient (B2/x)
between 50 and 100 µm.
Plugging in equations (1), (2) and (3) into equation (4),
the relationship between the magnetic field gradient and the
lateral velocity of the cell moving in media is obtained:
∇B 2 =
12π · µ0 · η · r
Ac · α · β · χ · Vb
By attempting to calculate the relationship between ∇B2 and v,
the following assumptions were made. First, the number of the
magnetic beads bound to each surface marker (β) is assumed
to be a constant, which, in this case, equals 1 bead/marker.
Second, we assume that the number of markers per area of
cell surface (α) is also a constant. If one bead is bound to
each cell, α equals 8.84 × 109 makers m−2 [11]. Third, the
susceptibility of the media (∼10−6) is negligible compared
to the susceptibility of the magnetic beads (i.e. χ = 1.52)
[12]. Fourth, the diameter of the cell is between 3 µm and
10 µm. We assume that the diameter of the cell is 6 µm.
Other constants are the permeability of free space, µ0 = 4π ×
10−7 N A−2, and the viscosity of media, η = ∼10−3 N s m−2.
The relationships established in this section will be used to
estimate the total magnetic force acting on the cell/bead
complexes for determining the cell separation efficiency of
the device.
3. Numerical simulations
To predict the performance of the resulting magnetic
separation scheme in the presence of the nickel particles
as a magnetic field concentrator, simulations were carried
out using a simplified one-dimensional magnetostatic model
by commercial software (COMSOL Multiphysics). Three
different scenarios were simulated. In the first scenario
(case I), a 100 µm long square magnet with a magnetic
strength of 1 T was positioned behind the origin, in the
absence of the nickel particles. From the simulation results,
the magnetic field decreased dramatically within 100 µm from
the magnet and reached a steady state after 100 µm from the
magnet (figure 2(B)). This showed that the maximum force
can only be obtained near the magnet (i.e. within 100 µm
from the magnet). To implement this physically, magnets
need to be fabricated in extremely close proximity to the
sample channel in order for this scheme to be effective for
cell separation.
In the second scenario (case II), the nickel particles were
put in between the magnet and the fluid to extend the effective
range of the magnetic field, and the resulting effects were
simulated. The concentration of the nickel particles (of 20 µm
in diameter) used in the simulation was approximately 6 ×
107 particles ml−1. As shown in figure 2(A), the presence
Y (Adam) Lin et al
of the nickel particles concentrates the magnetic field by
bending the field lines. This results in a substantially localized
magnetic field gradient, translating in enhanced magnetic
force on the magnetic beads. According to equation (1), the
force is directly proportional to the gradient of the squared
magnetic field density (∇B 2 ). Therefore, in order to estimate
the amount of force generated on the magnetic beads, the
simulated magnetic field data were converted into discrete,
one-dimensional gradient values (B2/x) (figure 2(D)).
The ratio between the gradient values with and without the
nickel particles indicated that the presence of the nickel
particles induces a maximum force (at x = 50 µm) that is
approximately nine times larger than that in the absence of the
particles. At 200 µm away from the edge of the magnet, this
ratio converges to around 3.
Instead of using nickel particles as a magnetic field
concentrator, a nickel bar could be used as an alternative
to serve the same functional purposes. Therefore, in the
last scenario (case III), a nickel bar, instead of the nickel
particles, was placed in between the magnet and the fluid,
and the corresponding effects were simulated. The simulation
results can be described into two different regimes (i.e. first
regime: 0–100 µm and second regime: >100 µm from
the origin). In the first regime, the nickel particles are
capable of generating a stronger magnetic field gradient
compared to the nickel bar, which translates into higher force
generation on the magnetic particles that are being manipulated
(figure 2(E)). In the second regime, there are no significant
differences between the effects contributed by the nickel
particles or the nickel bar (figure 2(D)). In addition, one
disadvantage of using a nickel bar is that it can only induce
a noticeable magnetic field gradient close to its edges, which
implies that the effective particle manipulation area is highly
restricted. For example, consider that the adjacent nickel
channel is replaced with the nickel bar (figure 1), the magnetic
fields would not be bent or concentrated effectively except at
the ends of the bar. Therefore, no translational forces can
be generated on the magnetic particles passing by the middle
portion of the bar. Comparatively, the nickel particles were
able to concentrate magnetic fields more efficiently all over the
surfaces. Based on the simulation results, we can conclude that
using nickel particles as a magnetic field concentrator can be
more beneficial in generating enhanced magnetic forces over
the nickel bar.
4. Material and methods
4.1. Channel fabrication
Different channel geometries were designed in conventional
computer-aided design software and printed out onto a
negative transparency mask (Photoplot, CO). The channels
were fabricated using a replicate molding technique. The mold
was fabricated using SU-8 negative photoresist (MicroChem,
MA) on a silicon wafer. The thickness of the mold was
∼50 µm. Then, a polydimethylsiloxane mixture (PDMS),
in a ratio of curing agent to PDMS at 1 to 10 by weight, was
poured onto the mold and subsequently cured at 60 ◦ C. After
the curing process, the PDMS replicate was peeled off and
punched with inlets and outlets at designated locations. To
(A )
(B )
Figure 3. (A) A mask layout for the microfluidic device. B was the
inlet for the sample. A, C and D were the inlets for media. E was
the outlet of the waste sample and F was the outlet for the separated
sample. G was the inlet for the nickel particles. H was the outlet for
the nickel particles. The G–H channel was the adjacent nickel
channels for enhanced magnetic field gradient generation. (B) A
schematic illustration showing the corresponding channel
dimensions, unit in µm.
complete the fabrication procedures, both the PDMS channel
surface and a glass substrate were activated by oxygen plasma
in order to bond the two surfaces together.
All inlets and outlets were 100 µm in width with the
exception of outlet E, which was 150 µm. The main channel
was 200 µm in width while the adjacent channel was 100 µm
in width. The two channels were 25 µm apart (figure 3). In
addition, a 500 µl syringe was used at inlet C while 250 µl
syringes were applied for the rest of the inlet locations
(A, B and D) as indicated in figure 4. Cell and magnetic
beads mixture sample entered the device from inlet B. Cell
growth media were inserted from inlets A, C and D. Inlet A
was designed to serve the purpose of pushing stagnated cells
and beads that were stuck in inlet B into the main channel.
Media from inlets C and D constituted two streams of sheath
flows that focused the sample flow into a fine central stream
through hydrodynamic focusing. This microfluidic focusing
technique allowed us to adjust the position and the width of
the sample stream in the same channel design.
4.2. System setup
Following the Dynabead protocol from Invitrogen, 25 µl of
magnetic beads were added to 1 ml of B-lymphocyte sample
(Coriell institute, NJ), at a cell density of approximately
106 cells ml−1 and mixed for 30 min in a 1.5 ml microcentrifuge
tube. The magnetic beads that are commonly found for
analytical purposes were 4.5 µm in diameter and made
from polystyrene superparamagnetic material [13]. The
B-lymphocytes were cultured in RPMI 1640 (Mediatech, VA)
with 10% FBS and antibiotics 1XPSN (Sigma-Aldrich, MO).
The cells were stained by an addition of 0.5 µl of Mitotracker
red dye (Invitrogen, CA). Roughly 20% volume ratio of
glycerol was added to the sample tube to prevent the
precipitation of cell/beads complexes in the syringe during
the experiment [14]. 100 µl of the prepared mix sample was
put in a 250 µl gas-tight glass syringe (Hamilton, NV) and
connected to inlet B. Then growth media were filled into
two 250 µl syringes (connected to inlets A and D) and a
500 µl syringe (connected to inlet C) (figure 4). Once the
setup was completed, the syringes were connected to the
microfluidic chip with soft tubing. The chip was placed on
Formation of high electromagnetic gradients through a particle-based microfluidic approach
Figure 4. A schematic illustration showing the system connections. The syringes for inlets A and B were placed on one syringe pump
(sample pump) and the other two (C, D) syringes were placed on another syringe pump (media pump). The top small magnet was used in
holding the bottom magnet in place.
(B )
Figure 5. (A) Locus of the sample stream under the influence of the external magnetic field. The white particles at the bottom of the
channel were cells that were pulled out of the stream. This only happened with the presence of the nickel particles. The white dotted lines
represent the edges of the main channel. (B) Locus plot showing the locus of the upper, center and lower bounds of the sample stream. In
every ten pixels, the upper and lower bounds of the white stream were taken and averaged. The average of the two created a center line
which was line fitted to obtain the first-order coefficient.
an inverted microscope (Nikon TE2000U) that was connected
to a CCD camera (AG Heinze, CA). All the fluid media were
pumped through digitally controlled syringe pumps (Harvard
Apparatus, MA). The fluid pumping speed for the sample
syringe (inlet B), along with one of the 250 µl media syringes
(inlet A), was set at 0.2 µl min−1, while the other 250 µl media
syringe (inlet D) and the 500 µl media syringe (inlet C) were
set at 1 µl min−1.
In order to prove the validity of the increased magnetic
field gradient in the presence of nickel particles, three different
conditions were tested: (1) in the absence of magnet and
nickel particles (termed as cell trial), (2) in the presence of
a magnet but without nickel particles (termed as magnet trial)
and (3) with the presence of both magnet and nickel particles
(termed as Ni trial). The cell trial was the control experiment
that served as a reference to compare with the later results.
Comparison of the magnet trial and the Ni trial determined
the contribution of the nickel particles to the magnetic field
gradient generation. The magnet in the experiments used was
a NdFeB cube magnet with a side length of 4.76 mm (3/16 )
(Amazing Magnets, CA). In order to hold the magnet in place
on one side of the chip, another small plate magnet, with the
dimensions of 3.18 mm × 3.18 mm × 1.59 mm (1/8 ×
1/8 × 1/16 ), was placed on the other side of the chip
(figure 3). For the Ni trial, the nickel particles, with less
than 20 µm in diameter (Atlantic Equipment Engineers, NJ),
were immersed in silicone oil that carried the particles into the
adjacent side channel from inlet G. The concentration of the
particles was approximately 5.31 g ml−1, which corresponds
to 1.42 × 108 particles ml−1. Fluorescent images were taken
at four different locations of the main channel to quantitatively
measure the locus of the cell/bead complexes that were
subjected to the external magnetic field. At each location,
15 pictures were taken with a 10 s exposure time.
Since the images were taken in four different locations
of the main channel, in order to reconstitute the locus of the
sample stream, the images were combined using pre-defined
alignment points. The images from the first position did not
have any usable alignment points; therefore, the images from
the other three positions were further analyzed. The pictures
from each of the three positions were randomly chosen and
linked together to become partial channel images. The images
were further processed to enhance the signal-to-noise level
for later data analysis purposes (figure 5). The fluorescent
images taken were first leveled with an input black and white
point of 2/1.00/5. The pictures were then joined together and
leveled once more to obtain the movement of the cells. Level 2
was set with an input black and white point of 130/9.99/132.
Y (Adam) Lin et al
Figure 6. A representative set of center locus data from all three trials: Ni trial, magnet trial, cell trial. The starting points of each locus
were offset for easier visual comparison.
The locus of the sample stream was traced and drawn from
the images. The processed images were used for further data
analysis that will be explained in the next section.
5. Experimental results and data analysis
The bending of sample stream locus was caused by the force
pulling on the magnetic beads attached to the cells. From the
center line data of all 15 pictures for the three different trials,
the bending of the line from the Ni trial was significantly larger
than that of the magnet trial and the cell trial (figure 6). In
order to analyze the force exerted on the cell/bead complexes,
the velocity values were extracted from the image data to
quantify the difference between the three trials. The horizontal
velocity of the cell/bead complexes (Vx ) is constant for each
experiment since Vx depends on the flow rate of the sample
and the shear media. On the other hand, the vertical velocity
(Vy ) depends on the force exerted on the complexes. From
equation (5), the total magnetic force is directly proportional
to the lateral velocity of the complexes. Since the vertical
Y range is comparably small, the magnetic force within this
range can be assumed to be constant. Therefore, according
to equation (5), the lateral velocity of the complexes should
be constant. By curve-fitting the locus with equation (6), the
ratio of the velocity components can be obtained:
· x + y0
where Vx and Vy are the velocity components of the complexes,
and y0 is the starting position of the sample stream. The
first-order coefficient in equation (6) can be used to quantify
and compare the vertical velocity (Vy ) of the cell complexes
in different trials, which can then be used to calculate the
magnetic force exerted on the complexes based on equations
(1) and (5).
After running the data through a line fitting function
(MATLAB), the average first-order coefficient over the 15
sets of data for the Ni trial was 8.08 × 10−3 with a standard
error of 1.01 × 10−4 while the average for the magnet trial
was 2.44 × 10−3 with a standard error of 2.66 × 10−4. The
cell trial (i.e. the control experiment) had an average of 1.03 ×
10−3 with a standard error of 2.57 × 10−4. The percentage
errors were 1.2% for the Ni trial, 10.9% for the magnet trial
Figure 7. The average values for the first-order coefficients for all
the three trials.
and 25.0% for the cell trial (figure 7). The ratio of the average
first-order coefficients between the Ni trial and magnet trial
was 3.31. This directly shows that the presence of the nickel
particles has a significant enhancement on the force generated
on the cell/bead complexes during the device operation.
In order to conclude the significance of the experimental
data, a t-test was performed to verify the statistical difference
of our data. The t-value between the Ni trial and magnet trial
was 19.79. The t-value between the magnet trial and cell trial
was 3.81. The t-value between the Ni trial and magnet trial was
25.55. The p-value for the Ni/magnet trial and the Ni/cell trial
was significantly lower than 0.001 for a two-tailed test. For the
magnet/cell trial, the p-value was close to 0.01 since a t-value
of 2.76 corresponded to a p-value of 0.01 for a two-tailed test.
Overall, the three trials were considered statistically different.
In addition, the effective operational flow rates for the
device with the current geometries have been determined by
considering the velocity components (i.e. Vx and Vy ) of the
resulting cell complexes in the sample channel. As mentioned
earlier, Vx was predominately determined by the shearing flow
rate controlled by the media pump whereas Vy was determined
by the magnetic force induced in a non-homogeneous magnetic
field. From the simulation data, the maximum magnetic force
acting on the complexes was estimated which was translated
into Vy based on equation (5). From the current channel
geometries, the ratio between Vy and Vx was estimated to
be at least 5 × 10−3 in order to have significant lateral
(y-direction) displacement of the cell complexes for effective
separation. Therefore, Vx and hence the upper bound of the
flow rate can be determined indirectly using the simulation
data. Based on the calculations, the maximum flow rate for
Formation of high electromagnetic gradients through a particle-based microfluidic approach
the Ni trial is 1.11 µl min−1 and 0.421 µl min−1 for the
magnet trial. In the experimental implementation, we have
tested the sample flow rate from 0.1 µl min−1 to 1 µl min−1
with 0.1 µl min−1 increments and found that 0.2 µl min−1
was a suitable operating flow rate for the current experimental
settings. Optimizing the device geometries would increase the
maximum flow rate for enhancement of the particle separation
throughput while maintaining high separation efficiency.
(B )
(C )
(D )
6. Discussion
The experimental results, in conjunction with the simulation
results, have strongly supported our hypothesis that the
presence of small nickel particles, close to the sample channel,
is able to induce a large magnetic field gradient which
translates into an enhanced magnetic force on the cell/bead
complexes. The average values for the ratio of the first-order
coefficients in the Ni and magnet trials showed that the induced
magnetic force in the presence of nickel particles was more
than three times stronger than that in the absence of the nickel
particles. The results were shown to be statistically different
from the t-test. Besides, the percentage errors of the firstorder coefficients of the magnet (11%) and cell trials (25%)
showed that the variations among the sample were greater
than that of the Ni trial (1%). For the case of the cell trial,
the relatively large percentage error was believed to originate
from the random diffusion of the complexes or instability of the
system such as disturbance from the tubing. In comparison,
the small percentage error in the case of Ni trial has shown
that a more stable and controllable microenvironment can be
maintained in the presence of the nickel particles as magnetic
field concentrators.
Despite the successful demonstration of the device for
cell separation, one should note that the current device
performance is not completely optimized. A number of
improvements can be done to further enhance the device
performance. Parameters such as the geometries of the main
channel as well as the flow rates for the media and sample are
critical in dictating the resulting cell separation performance.
In the experiments, there were two major considerations for
determining the current geometrical parameters of the device.
The first consideration was the fabrication issue. In the current
design, the gap distance between the sample channel and the
adjacent nickel channel is about 25 µm in width and 50 µm
in depth, which give rise to an aspect ratio of 2:1. The
fabrication of high aspect ratio SU-8 mold requires the use
of highly optimized lithography procedures. Therefore, in
order to keep the fabrication process simple, we set some
design constraints in our devices without compromising too
much for the device functionalities. The second reason was
about the length scale matching of the cell/bead complexes
with the channel geometries. The average diameter of the
complex is on the order of 10 µm. Therefore, in order to
prevent blockage of the channel by the cell aggregates, we
purposely designed the channel dimensions large enough for
the passage of the cell/bead complexes in the medium and
at the same time achieving high sample throughput. Although
the current geometries may not be the optimized design, it
nevertheless was able to demonstrate the working principle
of the device for magnetic cell separation successfully. In
Figure 8. The cell/bead complexes stayed attached to the bottom of
the channel and were trapped. The arrows (orange) indicate the flow
direction. (A) The cell/bead complexes were moving along the
bottom of the main channel. (B) The shadow was a group of
complexes moving along the bottom of the main channel which
completely flowed into the separation outlet channel. (C) The
bottom-left circle was a cell/bead complex moving away from the
main stream due to the induced force from the magnetic field
gradient generated by the nickel particles. (D) The cell bead
complexes could get stuck at the bottom of the main channel due to
the strong magnetic force near the nickel particles.
addition to the channel geometries, the device performance
can be adjusted by changing the intensity of the magnetic field
gradient through manipulating the nickel particles density on
the adjacent channel. Occasionally, some cell/bead complexes
may be attracted toward the sidewall of the channel, close to
the corner of the adjacent nickel channel (figure 8). This can
be prevented by cleverly designing the geometries (e.g. using
rounded corners instead of sharp corners) as well as adjusting
the location of the adjacent nickel channel.
Practically, our prototyping device has a number
of advantages over the conventional micro-magnetic cell
separation devices. One of the most significant advantages
is the fabrication simplicity which leads to low cost of device
production. Recent magnetic bead manipulation platforms
require intensive MEMS fabrication technology which are
economically expensive and time consuming [5, 8, 15–17].
The fabrication of our device is a simple replicate molding
technique which only requires a single-mask layer for the
manufacturing process. In addition, the mold can be reused
for multiple times to fabricate new channels for testing and
optimizing the system. Optimization is the next step to
manufacturing a fully automated magnetic-bead-based cell
separation system.
Upon the development of optimized channel designs
for maximized cell separation recovery rate and purity, our
proposed technique can be extended in producing highthroughput microfluidic cell separation array. For example,
an array of the device can be fabricated in a plastic or acrylic
cube. These cell separator arrays can be disposable due to
their low manufacturing cost. On top of that, the choices of
the metal particles are relatively flexible, provided that their
permeabilities are large enough for the device to be effective.
The proposed automated system separation device can be
further coupled with a microfluidic cell and magnetic bead
Y (Adam) Lin et al
mixer [18]. Practitioners using this device are only required
to provide the appropriate magnetic bead and place the sample
in the specified container. The separation can then be done
automatically. This device will be most useful for researchers
who want to study certain cell types or bio-particles from a
tissue or blood sample.
7. Conclusion
In this paper, a simple single-layer, single-mask fabrication
technique for the construction of magnetic-based MEMS
fluidic devices is presented. Compared to the conventional
multi-layer MEMS magnetic devices, this new fabrication
method offers the flexibility to incorporate small metal
particles, such as nickel, monolithically with the fabricated
microfluidic device. The particles can bend and/or concentrate
magnetic fields to achieve large magnetic field gradients,
which in turn produces enhanced magnetic force on magnetic
particles. This method can provide a great and economical
platform for not only cell separation devices but also many
other microfluidic systems.
We would like to thank Dr Branden Brough, Dr Martin Ma and
Nancy Li of the University of California, Los Angeles (UCLA)
for the technical assistance. In addition, we would also like to
thank Leslie S Liu of the University of California, Irvine (UCI)
for the editing of the manuscript. The project was supported
by the NIH-funded research project (1R33DK070328) on
‘automated chip-based metabolomic analysis’, the University
of California Leadership through Excellence Advance Degrees
Scholarship Program (UCLEADS) and UCLA’s Jonsson
Comprehensive Cancer Center gift fund from Wendy and
Ken Ruby.
[1] Krupke R, Hennrich F, von Lohneysen H and Kappes M M
2003 Science 301 344–7
[2] Sun L P, Zborowiski M, Moore L R and Chalmers J J 1998
Cytometry 33 469–75
[3] Hoyos M, Moore L R, McCloskey K E, Margel S, Zuberi M,
Chlamers J J and Zborowski M 2000 J. Chromatogr.
903 99–116
[4] Chalmers J J, Zborowski M, Sun L P and Moore L 1998
Biotechnol. Prog. 14 141–8
[5] Ramadan Q, Samper V, Poenar D P and Yu C 2006 Biosensors
Bioelectron. 21 1693–702
[6] Ramadan Q, Samper V, Poenar D P and Yu C 2006 Biomed.
Microdevices 8 151–8
[7] Gijs M A M 2004 Microfluidics Nanofluidics 1 22–40
[8] Choi J W et al 2001 Biomed. Microdevices 3 191–200
[9] Zborowski M, Fuh C B, Green R, Sun L P and Chalmers J J
1995 Anal. Chem. 67 3702–12
[10] Reddy S, Moore L R, Sun L, Zborowski M and Chalmers J J
1996 Chem. Eng. Sci. 51 947–56
[11] Chalmers J J, Zborowski M, Moore L, Mandal S, Fang B B
and Sun L 1998 Biotechnol. Bioeng. 59 10–20
[12] Derks R J S, Dietzel A, Wimberger-Friedl R and Rrins M W J
2007 Microfluidics Nanofluidics 3 141–9
[13] Dudley M E 2003 J. Immunother. 26 187–9
[14] Hu X, Bessette P H, Qian J, Meinhart C D, Daugherty P S and
Soh H T 2005 Proc. Natl Acad. Sci. USA 102 15757–61
[15] Han K H and Frazier A B 2006 Lab Chip 6 265–73
[16] Inglis D W, Riehn R, Austin R H and Sturm J C 2004 Appl.
Phys. Lett. 85 5093–5
[17] Miwa J, Tan W H, Suzukui Y, Kasagi N, Shikazono N,
Furukawa K and Ushida T 2005 1st Int. Conf. on
Bio-Nano-Information Fusion (Marina del Ray, CA)
[18] Suzuki H, Ho C M and Kasagi N 2004 J. Microelectromech.
Syst. 13 779–90
Fly UP