HIGH-PURITY SEPARATION OF RARE SPECIES IN DROPLET
MICROFLUIDICS USING DROPLET-CONDUIT STRUCTURES
Gaurav J. Shah and Chang-Jin “CJ” Kim
Mechanical and Aerospace Engr. Dept., University of California, Los Angeles (UCLA), CA, U.S.A.
ABSTRACT
We present a technique to achieve high-purity
isolation of the target species in electrowetting-on-
dielectric (EWOD)-based droplet microfluidics by
forming a long and slender liquid path between droplets
that acts as a “conduit” for actively transported target
species, while minimizing non-specific transport due to
diffusion and fluidic movement. By stabilizing the conduit
chemically (e.g., using surfactants) and controlling the
fluid by EWOD, we demonstrate, as an example, very
high magnetic bead (MB) collection efficiency (> 99%)
while eliminating ~97% non-magnetic beads (nonMBs) in
just one separation step.
INTRODUCTION
Concentration/separation is critical in many
biochemical assays, particularly where purity of isolated
target species (“TS”) (e.g. specific cells, proteins, DNA,
either by themselves or bound to beads) is vital to their
effectiveness. Unlike continuous microfluidics where TS
are immobilized while wash-buffer is flowed through the
channels to remove impurities, purification in droplet
microfluidics (e.g. by EWOD) typically involves serial
(i.e., repeated steps of) dilution of the non target species
(“nonTS”) [1],[2]. In each wash step, a buffer droplet is
added to the sample. The TS is actively collected in one
region of this combined (“parent”) droplet using one or
more differentiating properties, such as magnetic, electric,
optical or dimensional (e.g., [3],[4]). The droplet is then
split so that most of the TS are collected in one of the
daughter droplets.
The distribution of the nonTS between the two
daughter droplets, on the other hand, is governed by their
original distribution in the parent droplet and non-specific
phenomena such as diffusion and microfluidic movement
that occur during the purification step. If the nonTS were
uniformly distributed in the parent droplet, they would be
distributed between the daughter droplets roughly in
proportion to their volumes. Thus, even though many
nonTS from the parent droplet will be removed in the
form of the “depleted” droplet (from which TS have been
depleted), this would still leave a significant proportion of
nonTS in the “collected” droplet (wherein the TS has been
collected). In order to improve the TS purity, therefore,
TS is collected across the incoming buffer droplet [5],
which is initially free from the nonTS (Fig. 1(1a-1b)). The
droplet is then stretched (Fig. 1(1c)) and split (Fig. 1(1d))
into the collected (left) and depleted (right) droplets.
Even though the original distribution of nonTS in the
parent droplet is favorable for high purity (Fig. 1(1b)),
their non-specific transport and redistribution of the
nonTS during the subsequent collection and droplet
splitting (Fig. 1(1b-1c)) may lower purity of the TS in the
collected droplet (Fig. 1(1d)), entailing multiple wash
cycles (Fig. 1(2a-3a)). Two main mechanisms for the non-
specific transport are: (a) diffusion and (b) microfluidic
movement.
Fick’s law relates the diffusion flux (Jdif) of a species
978-1-4244-2978-3/09/$25.00 ©2009 IEEE
across a fluidic section to its concentration gradient:
∂φ
J dif = − D
∂x
where D is the diffusion coefficient determined by the
particle radius, temperature and viscosity of the medium.
For a given particle size, concentration gradient and
viscosity of the medium, the rate of diffusion-driven
transport across a fluidic section varies directly with its
cross-sectional area and inversely with its length.
The second mechanism for contamination is fluidics-
driven transport, i.e. species transported due to the viscous
forces during fluidic movement. Since the nonTS are
usually not actively manipulated, they are therefore highly
susceptible to travel with the flow. Although appropriate
choice of droplet actuation sequence can reduce the flow
into the collected droplet, some flow is inevitable during
neck creation and pinch-off required for droplet splitting.
As the droplet is stretched, the ensuing flow drags the
nonTS along with it. This is particularly pronounced
along the droplet meniscus, where flow velocity is higher
[6].
NonTS contamination into the collected droplet due
to both the mechanisms described above could be reduced
if a slender neck was created in the buffer droplet prior to
sample introduction. While this neck could be a “conduit”
for active TS transport, its slender (long and narrow)
structure would make it an excellent diffusion barrier.
Moreover, since little fluidic movement would be required
from this pre-necked stage to split droplet formation, such
a slender conduit would also help minimize the fluidics-
driven nonTS transport during the splitting.
Making a narrow, physical channel or other such
structures is one way to achieve high purity [7], but has
disadvantages like complexity in fabrication and lack of
re-configurability and control. Instead, we propose a
purely fluidic conduit in this paper to overcome these
issues. The challenge, however, is that for pure DI water
or buffer media, such a slender liquid structure tends to be
hydrodynamically unstable. The addition of surfactants,
for example, can make the thin liquid column more stable
and may therefore stabilize the “droplet-conduit
structure”, forming the basis of the proposed idea.
PROPOSED IDEA: DROPLET-CONDUIT
The use of surfactants on EWOD has recently
received much attention [8],[9]. However, addition of
surfactants to the solution tends to impede droplet
splitting by stabilizing the neck. Although long necks are
routinely found during the cutting steps, their dimensions
and locations are not so controllable. Here, we report
achieving the controllability using a combination of
surfactants and electrode modification. Specifically, a
surfactant was added in a moderate concentration to
increase tendency for stable conduit formation, while a
stabilizing electrode (“SE”) was used to toggle between
stability and breakage.
471

Fig. 1: Known technique of TS purification by serial
dilution: (1a) Wash buffer is added to sample. (1b) TS
(dark) is transported and collected in one region of the
droplet. (1c) The droplet is then stretched and (1d) cut to
form “collected” (wherein TS are collected) and depleted
(depleted of TS) droplets. Many nonTS (bright) are
removed with the depleted droplet, but some nonTS enter
collected droplet due to diffusion, and/or fluidics-driven
transport. (2a, 3a) More wash-buffer is added and the
above steps are repeated to improve TS purity by serial
dilution of the nonTS.
Fig. 2 illustrates the proposed idea. The typical
EWOD electrode layout is modified to incorporate a
slender line electrode (i.e. SE) through the center of the
square EWOD electrodes. The conduit-forming droplet
(left) is stretched towards the mixed sample containing TS
and nonTS. Keeping the SE on, a slender droplet-conduit
is created from the conduit-forming droplet (Fig. 2(a,b)),
whose width is defined by the SE.
On merging with the sample, the TS are transported
across the conduit using an active transport mechanism
(Fig. 2(c)). However, very few nonTS can cross the
diffusion barrier presented by the droplet-conduit. After
TS transport, the SE is turned off, and the droplet is
stretched further (Fig. 2(d)), breaking the conduit and
completing the droplet split (Fig. 2(e)). It should be noted
that much lesser fluidic movement is involved in this
cutting operation as compared to Fig. 1, since the neck
was already formed. As such, the fludically driven nonTS
transport into the collected region is much reduced. The
collected droplet (left) contains the TS and very few
nonTS, most of which are left in the depleted droplet
(right).
472
Fig. 2: Proposed technique for high purity rare TS
separation using droplet conduit. (a) TS are transported
to left edge of sample, while nonTS are randomly
distributed in the droplet. With SE on, the conduit-forming
droplet is stretched, (b) forming a slender “conduit”. On
merging with sample, (c) the conduit allows active TS
transport. But the slender conduit restricts diffusion-
driven nonTS transport. (d,e) When droplet is stretched
with SE off, the droplet splits with minimal fluidically
driven nonTS transport into the high purity “collected”
droplet and the “depleted” droplet.
MATERIALS AND METHODS
The technique was demonstrated for high purity
collection of magnetic beads (MBs) from a mixed
population of MBs and non-magnetic beads (nonMBs) on
an EWOD device. MBs of 4.5 µm diameter (Invitrogen)
were used as the TS, and fluorescent nonMBs of 5.2 µm
diameter (Molecular Probes) were used as the nonTS.
Pluronic F68 surfactant of 0.15% w/v (Sigma-Aldrich) in
PBS was used for both the sample and the buffer droplets.
A standard two-plate EWOD device (e.g. [10]) was
used for the experiments (Fig. 3). For the bottom plate,
EWOD electrodes were patterned into the 140 nm ITO
layer on glass substrate. A 60 µm wide line going through
the middle of the 1 mm x 1 mm EWOD electrodes defines
the SE. A gap (1-1.5 mm) was left at the left end of the SE
to ensure the breakage of the conduit during droplet
splitting. Silicon nitride (1.1 μm thick) was deposited by
plasma-enhanced chemical vapor deposition (PECVD) to
form the dielectric layer, and Cytop® (~1 μm thick) was
spun-coated to form the top hydrophobic layer. For the
top plate, an unpatterned layer of ITO was coated with
100 nm silicon nitride and 50 nm Cytop®. Double-sided
tape (~100 μm thick) was used as the spacer between the
plates.
Droplet actuation was achieved by sequential
application of voltage (70-80 Vac @1 kHz) to the EWOD
electrodes. Electronic control for the actuation sequence
was controlled using LabVIEW (National Instruments)
via a digital I/O device (DAQPad 6507, National
Instruments). For magnetic collection, an NdFeB
permanent magnet (½” dia x ½” thick) was placed on top
of the device to the left. Droplet actuation movies and
low-magnification (for a larger field-of-view) images
were captured by a video camera (Panasonic KR-222)
mounted onto the microscope (Nikon TE 2000U), while
better quality fluorescence images were taken using a
cooled-CCD camera (Photometrics Coolsnap EZ)
attachment.
Fig. 3: Schematic cross-section of EWOD device used.
RESULTS AND DISCUSSION
Experiments were performed to show high purity
separation of TS (i.e., MBs) from fluorescent nonTS (i.e.,
nonMB) using the slender droplet conduit. In order to
demonstrate the utility of conduit for purification of rare
species, a low (around 1:20) TS:nonTS ratio was chosen.
Fig. 4 shows the image sequence for the experiment
performed. The sample droplet containing MBs and
nonMBs, along with 0.15 %w/v pluronic surfactant F68 is
placed on the right, while the conduit-forming (i.e.,
buffer) droplet, also containing the surfactant, is
introduced from the left (Fig. 4(a)). The magnet is
positioned at the left, so that the MBs collect at the left
meniscus of the sample droplet (see Fig. 4(a) inset). A
stable, slender conduit is formed by stretching the
conduit-forming droplet while keeping the 60 μm wide SE
on (Fig. 4(b)). On merging with the sample (Fig. 4(c)), the
MBs from the sample are actively transported across the
conduit towards the magnet, to the leftmost edge of the
combined droplet, while most of the nonMBs remain
behind at the right (Fig. 4(d)). After all the MBs are
transported, the SE is turned off and the droplet is
stretched further (Fig. 4(e)) so as to split it into the
collected (left) and depleted (right) droplets (Fig. 4(f)).
Since the neck was already formed prior to the merging,
the splitting operation involves very little fluidic
movement.
To evaluate the purity of the separation, MBs (dark)
and nonMBs (fluorescent) are counted in the collected
(Fig. 5(a,b)) and depleted (Fig. 5(c,d)) droplets. Images
were taken using both cameras under fluorescent
excitation, but with some bright field illumination so as to
visualize non-fluorescent features as well. The bright
nonMBs can be easily distinguished from the background.
To distinguish the MBs from the other dark features in the
images, a magnet was introduced on one side of the
droplet, and the magnetically responsive collected at the
edge were counted. The original sample contained ~283
nonMBs and 16 MBs. Comparing Fig. 5 (a) and (b) it is
clear that much fewer nonMBs (bright) can be seen in the
collected droplet compared to the depleted droplet. The
fluorescent nonMB count in the collected droplet is less
than 10 while that in the depleted droplet is ~267. Thus,
~97% of nonMBs were kept out of the collected droplet.
Around 16 MBs were counted in the collected droplet
(Fig. 5(a)), while no (0) MBs were seen in the depleted
droplet (Fig. 5(b)), demonstrating that high purity
magnetic separation was achieved while maintaining high
collection efficiency (> 99%).

Fig. 4: Image sequence for high purity magnetic separation using droplet conduit structures: All droplets contain 0.15%
pluronic F68 in PBS. (a) Magnet is positioned to the left of the sample and conduit-forming droplet, so that MBs (dark)
are attracted to the left edge of sample (see inset). (b) Conduit is formed by stretching the droplet while the SE is on. (c)
Sample is merged with the conduit-forming droplet, (d) allowing MBs to pass through. After transport, very little fluidic
movement is involved as (e) the droplet further stretched with SE turned off, (f) cutting it into “collected” (MBs collected,
very few nonMBs) and depleted droplets (depleted of MBs). Satellite droplets can be cleaned up by depleted droplet.

473
Fig. 5: High-purity separation is obtained using droplet-conduit structures, as estimated by counting the MBs (dark) and
nonMBs (bright) in the collected and depleted droplets. Images were taken under fluorescent excitation but with some
bright-field illumination so as to visualize non-fluorescent features as well. To count MBs (and not dirt etc.), a magnet
was introduced right next to a droplet edge, and the magnetically responsive MBs attracted to it were counted. Original
sample droplet had ~16 MBs and ~283 nonMBs. (a) All the MBs (~16) and very few nonMBs (<10) were counted in the
“collected droplet”. (b) No MBs (0) and a majority of nonMBs (~267) were seen in the “depleted droplet”. The results
correspond to target (MB) collection efficiency over 99% in the collected droplet, with nonMB concentration dropping to
~3% (or by more than 28 times) in just one step.

CONCLUSION organ transplant rejection," in Proc. Solid-State

Purity of the TS in droplet based washing steps is
adversely affected by the non-specific transport, mainly as
a result of diffusion and fluidic movement. (For the
present device geometry and particle size, it turns out that
the latter is particularly dominant). Contamination due to
both the factors is reduced using droplet-conduit
structures, without sacrificing the TS collection
efficiency.
The chemically stabilized slender droplet conduit
structure controlled by EWOD not only creates a diffusion
barrier, but also provides a pre-formed neck to minimize
fluidic movement required for droplet splitting after TS
collection, thus ensuring high purity as well as high TS
collection efficiency.
ACKNOWLEDGEMENT
This work was supported by NASA through Institute
for Cell Mimetic for Space Exploration (CMISE), NIH
through Pacific Southwest RCE (grant AI065359), and
Intramural Seed Grant of UCLA Department of Urology.
REFERENCES
[1] R. B. Fair, "Digital microfluidics: is a true lab-on-a-
chip possible?," Microfluidics and Nanofluidics, vol.
3, pp. 245-281, 2007.
[2] J. Gong and C.-J. Kim, "All-electronic droplet
generation on-chip with real-time feedback control
for EWOD digital microfluidics," Lab on a Chip, vol.
8, pp. 898-906, 2008.
[3] S. K. Cho, Y. J. Zhao, and C.-J. Kim, "Concentration
and binary separation of microparticles for droplet-
based digital microfluidics," Lab on a Chip, vol. 7,
pp. 490-498, 2007.
[4] G. J. Shah, J. L. Veale, Y. Korin, E. F. Reed, H. A.
Gritsch, and C.-J. Kim, "Concentration of CD8+
lymphocytes on EWOD Platform for monitoring
Sensors, Actuators and Microsystems Workshop,
Hilton Head Island, SC, USA, Jun. 2008, pp. 28-31.
[5] Y. Z. Wang, Y. Zhao, and S. K. Cho, "Efficient in-
droplet separation of magnetic particles for digital
microfluidics," Journal of Micromechanics and
Microengineering, vol. 17, pp. 2148-2156, 2007.
[6] J. Fowler, H. J. Moon, and C.-J. Kim, "Enhancement
of mixing by droplet-based microfluidics," in Proc.
IEEE Int. Conf. MEMS, Las Vegas, NV, Jan. 2002,
pp. 97-100.
[7] S. M. Kim, S. H. Lee, and K. Y. Suh, "Cell research
with physically modified microfluidic channels: A
review," Lab on a Chip, vol. 8, pp. 1015-1023, 2008.
[8] O. Raccurt, J. Berthier, P. Clementz, M. Borella, and
M. Plissonnier, "On the influence of surfactants in
electrowetting systems," Journal of Micromechanics
and Microengineering, vol. 17, pp. 2217-2223, 2007.
[9] V. N. Luk, G. C. H. Mo, and A. R. Wheeler,
"Pluronic additives: A solution to sticky problems in
digital microfluidics," Langmuir, vol. 24, pp. 6382-
6389, 2008.
[10] S. K. Cho, H. J. Moon, and C.-J. Kim, "Creating,
transporting, cutting, and merging liquid droplets by
electrowetting-based actuation for digital microfluidic
circuits," Journal of Microelectromechanical
Systems, vol. 12, pp. 70-80, 2003.

474