Sei sulla pagina 1di 10

Home Search Collections Journals About Contact us My IOPscience

Scale invariant hydrodynamic focusing and sorting of inertial particles by size in spiral micro

channels

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2015 J. Micromech. Microeng. 25 084013

(http://iopscience.iop.org/0960-1317/25/8/084013)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 203.135.190.2
This content was downloaded on 21/08/2015 at 08:39

Please note that terms and conditions apply.


Journal of Micromechanics and Microengineering

J. Micromech. Microeng. 25 (2015) 084013 (9pp) doi:10.1088/0960-1317/25/8/084013

Scale invariant hydrodynamic focusing and


sorting of inertial particles by size in spiral
micro channels
Phanindra Tallapragada, Nilesh Hasabnis, Kalyan Katuri,
Senbagaraman Sudarsanam, Ketaki Joshi and Melur Ramasubramanian
Department of Mechanical Engineering, Clemson University, Clemson, SC 29634, USA

E-mail: ptallap@clemson.edu

Received 2 March 2015, revised 29 April 2015


Accepted for publication 14 May 2015
Published 22 July 2015

Abstract
The hydrodynamic separation of spherical particles in flows at low Reynolds numbers is a
very active area of research in microfluidic engineering due to the many important biomedical
applications. In particular, curved channels such as spiral channels are of growing interest because
the lift and drag force exerted on inertial particles can be used to hydrodynamically separate the
particles. In this paper we present a scale invariant classification of the lateral focusing of particles
in highly curved spiral micro channels with a square cross section. We then use this scale invariant
classification to demonstrate the separation of particles in two-particle mixtures across a large
range of sizes. We thus show that our results can be used to systematically design the geometry of
devices and select flow parameters to separate particles by size in a mixture.

Keywords: inertial microfluidics, hydrodynamic separation, particle separation

(Some figures may appear in colour only in the online journal)

1. Introduction size, shape, and deformability. Thus, in theory differently


sized particles could be forced to position in different regions
The ability to sort particles, in particular cells, by size has of a channel. Such particle focusing in straight channels has
many important diagnostic or therapeutic applications. The been known about since at least 1961 [10, 11]. Such hydro-
common cell separation methods rely on immunoselection by dynamic particle focusing, especially differential focusing
exploiting the selectivity of biological markers. These methods and sorting of particles based on size, has now become an
are often limited due to variations in cell surfaces and the non- active area of research in the microfluidics community, see
uniform expression of cell markers. These methods also often for example [12–22]. A number of possible biomedical appli-
damage the cells and have a low throughput. Label-free tech- cations have emerged for the use of passive methods; such
niques to sort particles and cells have gained much attention in as the identification and separation of circulating tumor cells
recent years. Among label-free techniques, active methods for [23–25], the separation of leukocytes [19], the separation of
cell separation rely on external forces such as electric [1–4], plasma from blood [26], and the encapsulation of cells [27].
magnetic [5, 6], and acoustic [7–9] to achieve particle/cell sepa- One means to achieve hydrodynamic focusing and sorting
ration. While these active methods provide excellent separation of particles is through controlling the geometry of the channel,
accuracy, they have low throughput and the application of high such as by using channels with curved walls. Spiral channels
electric or magnetic fields can adversely affect cell viability. and serpentine channels have been used to demonstrate the
Passive methods employ hydrodynamic forces to manipu- separation of particles by size via differential lateral focusing.
late the motion of particles in a channel. Hydrodynamic forces For example, in [12, 13, 28] using a spiral channel with a rec-
acting on particles depend on several factors such as channel tangular cross section, Papautsky and collaborators showed
geometry, channel curvature, carrier fluid viscosity, fluid flow the separation of particles of diameter 1.9 μm and 7.32 μm
rate, and the physical properties of the particles such as their from a two particle mixture as well as the separation of particles

0960-1317/15/084013+9$33.00 1 © 2015 IOP Publishing Ltd  Printed in the UK


J. Micromech. Microeng. 25 (2015) 084013 P Tallapragada et al

of diameter 5 μm, 10 μm, and 15 μm with a high efficiency. 2.  Theoretical background
Toner et al [29] showed the separation of particles of size 10 μ
m and 7 μm, reported using a spiral channel with a rectangular The following notation will be used throughout the rest of
cross section. Go et al demonstrated differential lateral focusing manuscript; U is the average velocity of the fluid through the
of 20 μm and 40 μm beads in a curved rectangular channel [15]. channel cross section, ν is the viscosity of the fluid, 2h is the
Hu et al [23] demonstrated the separation of 5 μm and 15 μm height and width of the square cross section of a channel, R
polystyrene beads in a double spiral channel. Han’s group dem- is the radius of the curvature of the centerline of the channel,
onstrated the differential focusing and separation of particles a is the diameter of a spherical particle and the particle con-
of diameters 15 μm and 25 μm in spiral channels with a very a
finement ratio, λ = 2h . The two Reynolds numbers of interest
high aspect ratio trapezoidal cross section [19, 21]. Tracey et al are the channel Reynolds number, Re, which characterizes the
recently showed that particles of diameters 2.1 μm and 3.2 μm undisturbed flow of the fluid in the channel and the particle
could be separated using a spiral channel with a square cross Reynolds number, Rep, which characterizes the influence of
section, of 20 μm × 20 μm [30]. Di Carlo et al [14] showed the the fluid flow on a particle. These are defined in the following
possibility of the separation of particles of size 2 μm and 10 μ manner,
m by multiple passes through a serpentine channel. Recently
Nguyen et al [22] demonstrated the separation of particles of 2Uh
Re =
(1)
size 3 μm and 10 μm, and the separation of blood and plasma in ν
a serpentine channel with a high efficiency [26]. and
Despite the many investigations exploiting the role of cur-
vature of the channel to sort particles by size, a scale invariant ⎛ a ⎞2
Rep = Re⎜ ⎟ = Reλ2
⎝ 2h ⎠
(2)
criteria to enable differential focusing and sorting of particles
by size has not been reported. A scale invariant criteria to
Furthermore, if the centerline of the channel is curved, with
separate particles of any prescribed size in curved channels is
a radius of curvature R, the relevant non-dimensional geo-
necessary to systematically design micro channels and select h
the flow parameters. metric parameter is the curvature ratio, δ = R . The interaction
In this paper we report on such criteria for spiral micro- of the fluid flow and the curvature of the channel is described
channels with a square cross section that exploit scale invariant by the Dean number,
differential lateral focusing. We demonstrate that there are
De = Re δ
(3)
two distinct sets of non-dimensional sizes of particles which
focus on different lateral positions in the channel. This dif- The motion of particles in a straight channel, the case
ferential lateral focusing reported here is different from those where δ = 0, is well understood and has been examined both
of earlier findings [12, 13, 28, 29, 31, 32], due to the square experimentally and theoretically in numerous works [10, 11,
cross section  of the channel and the use of spiral channels 34–38]. In a straight channel, due to shear gradient in the
with significantly higher values of δ. We then show that if in direction transverse to the axial flow, particles experience a
a two-particle mixture the two particles’ sizes fall into these lift force, FL, which pushes the particles towards the channel
distinct sets of non-dimensional sizes, then they can be sorted walls [39]. As the particles near the wall, a wall-induced lift
completely in a single pass through the spiral channel. We force, FW, pushes the particles towards the centerline of the
demonstrate this by sorting six two-particle mixtures, ranging channel. The wall-induced lift force is the result of two, pos-
in size from 95 μm to 7 μm using device geometries that are sibly competing, effects. First, the presence of a wall breaks
informed by the criteria for the scale invariant lateral focusing the symmetry of the spatial distribution of the wake vorticity
of particles. While this size range is unusual in microflu- of the particles resulting in a lift force directed away from the
idics, we are motivated by an important practical problem; walls. Secondly, from potential flow theory, it can be shown
the encapsulation of pancreatic whole islets, whose diameter that the faster flow between a sphere and a wall results in a
can vary between 50 μm to 400 μm in rodents and between low pressure on the surface of the sphere adjacent to the wall.
100 μm–200 μm in humans [33]. Separating the islets into Particles accumulate in a circular annulus at a distance of 0.6h
groups of similar sizes is a required step before encapsulation. from the centerline, where the lift forces are balanced. Note
Our findings enable the systematic design of channel geome- that the Stokes drag in the transverse direction is absent at the
tries and the selection of flow parameters to hydrodynamically equilibrium position since the relative transverse velocity of
separate particles across a range of sizes from a mixture. the particle with respect to the fluid is zero.
The outline of the paper is as follows. In section  2 we If the centerline of the channel is curved, i.e. when δ ≠ 0, a
review the forces acting on inertial particles in a channel and secondary flow in the cross sectional plane exists. Due to the
define the non-dimensional parameters governing their motion centrifugal effect, the axial velocity is no longer symmetric
and in section 3 we summarize the fabrication method, experi- about the axis of the channel. Instead, the axial velocity of the
mental set up and image analysis. In section 4.1 we present fluid is the highest between the centerline and the outer wall
experimental evidence of scale invariant differential focusing of the channel. Thus, the curvature of the channel induces an
for particles ranging in diameter from 8 μm to 95 μm. In sec- asymmetry in the axial velocity and a secondary vortical flow.
tion 4.2 we use the scale invariant differential focusing criteria For low channel Reynolds number this secondary flow takes
to demonstrate the sorting of six two-particle mixtures by size. the form of two vortices [40, 41], the so-called Dean vortices.

2
J. Micromech. Microeng. 25 (2015) 084013 P Tallapragada et al

Figure 1.  A representation of a flow in a curved channel. Secondary flows (Dean vortices) are created in the cross sectional plane.

A schematic representation of a flow through a curved channel corresponding experimental observations. Neutrally buoyant
with a square cross section  is shown in figure  1, with Dean polyethelene microspheres (Cospheric, California, USA)
vortices shown in the two left-most panels. The secondary and polysterene microspheres (Thermoscientific) of six dif-
flow exerts a drag force, FD, on a particle that is transverse to ferent diameters; 7 μm, 13 μm, 25 μm, 45 μm, 65 μm and 95
the axis of the channel. As in the case of the straight channel, μm were used in the experiments. Water with 0.08% (w/v)
a particle also experiences a lift force due to the shear gradient Tween-20 surfactant solution was used as a carrier fluid. In
and the wall-induced lift force. The resulting interaction of all the experiments, the particle volume fraction was main-
the shear-induced lift force, the wall-induced lift force, and tained between 0.45%–0.5%, a volume fraction range where
the drag force due to the Dean flow, all of which depend on the inter-particle interactions are negligible.
the size of the particles, can lead to size dependent lateral
focusing of the particles.
3.2.  Experimental set up

3.  Materials and methods The particle suspension was introduced into the microflu-
idic chip using two 3 ml plastic syringes (BD, Breda, the
3.1.  Microfluidic chip design and fabrication and materials Netherlands). A New Era NE-300 series syringe pump was
used to achieve the desired flow rate. A magnetic stirrer bar was
A standard soft-lithography procedure was used to fabricate inserted into the particle suspension containing the syringe to
microfluidic devices. High viscosity photoresists (SU-8 2150 prevent the suspension from settling during the injection. The
and SU-8 2075 from Microchem, Berlin, Germany) were used flow rate was varied from 5 μl min −1 to 2000 μl min −1. The
to generate resist molds with channel thicknesses up to 500 μm. microfluidic chip was mounted onto an XYZ translation and
The chip was made in PDMS (Sylgard 184, Dow Corning, rotational stage. The fluorescence microscope (Nikon, Japan)
Midland, MI, USA). The curing and base agents were mixed had two filter blocks, with excitation filters with wavelength in
at a ratio of 1:10 and poured onto the silicon wafer mold, blue and green. In addition, a computer-controlled high-speed
degassed, and cured at 60 °C for 4 h. After curing, inlet and camera (Andor Zyla 5.5) was mounted onto the microscope
outlet access holes were made using a biopsy punch before for image recording, using the accompanying Nikon software.
plasma bonding the PDMS microfluidic chip onto a glass
microscope slide (2’ × 3’ slide, Corning, USA). To improve 3.3.  Image acquisition and analysis
the bonding strength the device was placed on a hot plate at
120 °C for 10 min. Videos and images of the particle ordering in the spiral micro-
Channels with the following cross sectional dimen- channel were captured using the camera at a frame rate of
sions were fabricated; 70 μm × 70 μm, 125 μm × 125 μm, 18–25 frames per second. The exposure time for imaging
230 μm × 230 μm, 300 μm × 300 μm, 400 μm × 400 μm, fluorescent particles varied between 30 and 100 millisec-
450 μm × 450 μm and 500 μm × 500 μm. Microfluidic chan- onds at low flow rates to 400 and 800 milliseconds at high
nels were designed as Archimedean spirals. The channels with flow rates. The motion of the particles in each of the mix-
a hydraulic diameter of 300 μm and less consisted of 5 loops, tures was imaged separately using blue and green excitation
and channels with a hydraulic diameter greater than 300 μm wavelengths. When the excitation wavelength is green, only
consisted of 4 loops. Here we define an arm or loop of the spiral the red fluorescent particles in the two-particle mixture are
to be the section of the channel, that is traversed in moving visible. When the excitation wavelength is blue, only the
along the spiral by 360°. The figures showing the geometry of green particles in the two-particle mixture are visible. The
the channels are presented in the next section along with the individual movies and pictures collected were superimposed

3
J. Micromech. Microeng. 25 (2015) 084013 P Tallapragada et al

Figure 2.  Differential lateral focusing of particles in a spiral channel with cross section 300 μm × 300 μm. The picture at the top shows the
geometry of the device in which the experiment was performed. The region of the device bounded by the red rectangle in this picture is the
imaged area. In the table of images the left-most column shows the time lapse images of the positions of particles of diameter 25 μm in the
imaged region. The right-most column shows the time lapse images of the positions of particles of diameter 45 μm. Along each row the
corresponding normalized histograms of particle distribution f (d ) as a function of d are shown. The green histograms show the distribution
of the smaller particles, while the red histogram shows the distribution of the larger particles. The value of δ is fixed along each column
and this corresponds to the average value of the parameter δ along the segment of the spirals shown. Note—all the images in the left-most
column are time lapse images obtained by overlaying the particle position in every frame of a 20 s video.

Table 1.  The table shows the particle confinement ratio for various analyzed with the image processing toolbox of MATLAB and
combinations of particles and different sized channels in the the imaging software Nikon Elements.
experiments.
Particle diameter, a p, in μm
4.  Results and discussion
8 13 25 45 65 95
4.1.  Scale invariant differential lateral focusing of particles
2h 70 0.114 0.186 — — — —
125 — 0.104 0.2 — — — Inertial particles of different sizes are focused to different
230 — — 0.108 0.195 — — lateral positions in the channel. This differential focusing
300 — — 0.083 0.15 0.216 — and whether focusing occurs are a function of the channel
400 — — 0.0625 0.1125 0.1625 0.2375 Reynolds number, the curvature ratio, δ, and the particle con-
450 — — 0.055 0.1 0.144 0.211
finement ratio, λ. An example of differential lateral focusing
500 — — 0.05 0.09 0.13 0.19
of particles is shown in figures  2. The left-most and right-
most columns show the overlayed streaklines of 25 μm and
45 μm particles respectively. To make these experimental
to clearly show the differential focusing of the particles in the
observations of particle focusing in the micro channels
mixtures. The low frame rate, when combined with longer
objective, we define a spatial particle density function, f (d ),
exposure times, helped us obtain better composite images that
as follows,
showed tight bands of focused particles. The high exposure
time ensured that the thickness of the focused bands was not N (d )
underestimated. When capturing bright field images of lateral f (d ) = 1 .
(4)
focusing of the particles the frame rates varied between 18 ∫ N (y)dy 0
and 25. Many (about 600) image frames of moving particles
were overlayed to generate composite images of focused Here the normalized distance from the inner walls of the
x
bands of particles. Observing particle focusing consistently in channel, d = 2h , as shown in figure 1, varies from 0 to 1. The
such composite images gave us confidence in the reproduc- number of particles passing through the channel at a distance
ibility of the results. All the videos and images captured were d from the inner wall of the channel is denoted by N (d ).

4
J. Micromech. Microeng. 25 (2015) 084013 P Tallapragada et al

0.25
Large particles 30
0.2 focus at the
25
center
Small particles
0.15 20 focus at the inner
λ wall

Re
15
0.1 Small particles 10
Large particles
focus at the 5
0.05 inner wall
focus at the
center
0
0 0.025 0.05 0.075
0 δ
0 0.25 0.5 0.75 1
d

Figure 3.  (a) The mean position of the spatial density function when the particles are focused. The green squares show the mean positions
of focused bands for the smaller particles, while the red triangles show the mean positions of focused bands for the the larger particles.
There are clearly two regimes in particle focusing, particles with λ ⩽ 0.12 focus towards the inner wall, while particles with λ > 0.12 focus
towards the center of the channel. (b) The lowest channel Reynolds numbers and the curvature ratio δ when focusing of the particles is first
observed.

To objectively identify particle focusing, we set a quantita- in rodents and between 100 μm–200 μm in humans [33].
tive criteria to define particle focusing in terms of the standard Table  1 lists the confinement ratio, λ, of the particles used
deviation of f (d ) and the area under the spatial density func- with the various devices. The channel Reynolds number, Re,
tion. Let the mean and standard deviation of f (d ) be μ and σ . varied from 0.03 to 50 and the Dean number, De, varied from
We defined the particles to be focused, when 0.005 to 15. The Dean number range is considerably higher
than in the results reported in the earlier literature. Table 1 lists
μ+b
f (y )dy ⩾ 0.9
(5) ∫μ−b 21 particle-channel combinations. For each of these combi-
nations the focused positions of the particles in the channel
were determined using the same criteria as in 5. In the case of
where b = min[σ , 0.15]. Thus, when the position of at least smaller particles (8 μm and 13 μm) fluorescence images were
90% of the particles are restricted to a narrow band that is analyzed.
either 2σ wide or 0.3h which ever is lower, the particles are The focused positions of particles as a function of the non-
defined to be focused. dimensional distance, d, from the inner wall are shown in
The functions N (d ) and f (d ) were calculated in a dis- figure 3(a). The red triangles represent the focused positions
crete fashion by splitting the range of d, the unit interval, into of particles that focus close to the center of the channel, and
50 equal bins of size 0.02. The number of particles passing the green squares represent the focused positions of particles
through these bins during the experiment was counted. A par- that focus close to the inner wall of the channel. Clearly larger
ticle was defined to lay in a bin if its geometric center lies in particles (λ > 0.12) focus towards the center and the smaller
it. Segments were selected across at the end of each arm of particles (λ < 0.12) focus towards the inner wall. Figure 3(b)
the spiral for these measurements. These segments included shows the smallest channel Reynolds number at which the
the channel close to the inlet and the channel at the end of the focusing occurs as a function of the curvature ratio, δ. Larger
spiral. particles (represented by red triangles) focus at smaller
Figure 2 demonstrates that the smaller 25 μm particles channel Reynolds numbers, while the smaller particles (rep-
focus towards the inner wall of the spiral channel and the resented by green squares) focus at larger Reynolds numbers.
larger 45 μm particles focus towards the outer wall of the It is also apparent from figure 3(a) that there is a very rapid,
spiral channel. The larger particles begin to focus for a smaller bifurcation-like transition in the equilibrium positions of par-
Reynolds number (in this case Re = 5.5) than the smaller par- ticles as the confinement ratio increases.
ticles which focus at a higher Reynolds number (in this case The results shown in figure 3 clearly demonstrate a scale
Re = 16.67). invariant law that governs the differential lateral focusing of
To obtain an empirical scale invariant rule to classify the particles in a spiral channel or more generally in a curved
differential lateral focusing regimes, we performed experi- channel. It should also be noted that the results in figure 2 and
ments across a range of values of the channel Reynolds 3 show novel particle focusing behavior. In earlier works [12,
number and different sized particles in flows through spiral 13, 28, 29, 31, 32], inertial particles were found to be focused
channels. All the channels had square cross sections, with the between the inner wall and the centerline of the channel, with
sides varying in length from 70 μm to 500 μm. The diam- the larger particles focusing closer to the inner wall. Smaller
eters of the particles ranged from 8 μm to 95 μm. While this particles are usually reported to be either focused closer to the
size range is unusual in microfluidics, we are motivated by an center line or unfocused if λ < 0.07, [12–14]. Our results differ
important practical problem; the encapsulation of pancreatic because of the higher values of the curvature ratio, δ, (and cor-
whole islets, whose diameter can vary between 50 μm–400 μm respondingly higher values of the Dean number De = Re δ )

5
J. Micromech. Microeng. 25 (2015) 084013 P Tallapragada et al

Figure 4.  Separation of particles of diameter 25 μm and 65 μm from a mixture at Re = 44.4. The images in (a), (b), and (d) are obtained
from the section of device enclosed by the rectangle in (c). The two-particle mixture is pumped at a constant flow rate through the inlet at
the center of the spiral.

that we considered. This unexplored parametric regime pro- mixture in a single pass is— the hydraulic diameter of a spiral
duced a consistent scale invariant differential lateral focusing channel with a square cross section  should be chosen such
of particles in spiral channels. A crucial observation about the that in a two-particle mixture larger particles have a confine-
scale invariant focusing we observed is the robustness of the ment ratio, λ ⩾ 0.15, and smaller particles have a confinement
focused positions to changes in the channel Reynolds number. ratio, λ ⩽ 0.1. The curvature ratio can be selected based on the
Once focusing of the particles is achieved in the range data in figure 3(b).
5 < Re < 30, the focused positions do not change for Reynolds We test this design rule by demonstrating the sorting of par-
numbers as high as 80. ticles by size in six two-particle mixtures. The particle sizes in
these mixtures and the channel dimensions were chosen such
that the larger of the particles had a particle confinement ratio
4.2.  Application to hydrodynamic sorting of particles by size
greater than or equal to 0.15, and that of the smaller of the
The size dependent differential lateral focusing demonstrated particles was lesser than or equal to 0.1. In every mixture the
in our experiments has a clear transition terms of particle con- volume fractions of the smaller and larger particles are 0.25%
finement ratio, see figure 3(a). This data suggests a potential each. In each of the sorting experiments with two-particle
rule to sort particles by size in a single pass; particles whose mixtures, the fluorescent particles in the flow are visualized
non-dimensional sizes fall into the two distinct categories using a green excitation for half the duration of the experiment
shown in figure 3(a) can be sorted. However, sorting a mixture and using blue excitation during the other half. Therefore, red
of large and small particles whose confinement ratios are both fluorescent particles are visible during one half of the experi-
close to 0.12 in one pass is practically difficult. This is because ment and green fluorescent particles are visible during the
the focused bands of particles could be too close to each other, other half. For example, in the case of the mixture consisting
where the particles in one band influence the motion of the of particles of diameter 25 μm and 65 μm, figure 4(a) shows
particles in the other band. Therefore, a good practical design the focused band (red) of particles of diameter 65 μm close to
rule suggested by our data to separate particles by size from a the center of the channel. Figure 4(b) shows the focused band

6
J. Micromech. Microeng. 25 (2015) 084013 P Tallapragada et al

Figure 5.  Sorting by particles by size from two-particle mixtures—the table shows the particle sizes that are sorted and the hydraulic
diameter of the corresponding channels. The last column in the table gives the Reynolds number at which complete separation was
obtained for the particle pairs. In each of the figures (a)–(e) the device layout is shown on the left. The inlet for the mixture is at the center
of the spiral. The device diagrams have a blue rectangle which shows the area that is imaged. The corresponding image of the fluorescent
particles, one red and the other green, is shown to the right. This image has a white rectangle. A zoomed-in picture of the particle streaks in
this rectangular region showing two separate bands of particles is also shown.

(green) of particles of diameter 25 μm close to the inner wall The last column in the table  gives the Reynolds number at
of the channel. The differential focusing of the particles in the which complete separation was obtained for the particle pairs.
mixture is clarified in figure 4(d) which is obtained by super- Each of the images in 5(a-e) show the geometry of the spiral
imposing figure 4(a) and 4(b). Figure 4(c) shows the layout channel in which the sorting is performed, and an image of
of the device used to sort the mixture of particles. The two- green and red fluorescent streaks in a sub-region of the device,
particle mixture is pumped at a constant flow rate through the along with a zoomed-in image of the well-separated streak
inlet at the center of the spiral. Clearly the particles of each lines in the channel close to the outlet. The ratio of the sizes
a a
size are focused in two distinct bands in the outer arm of the of the particles being sorted range from a1 = 1.8 to a1 = 3.46.
2 2
spiral and the separation distance between the two bands is The channel Reynolds number at which complete differen-
about 28 μm. The Reynolds number at which the separation tial lateral focusing occurs in the two outermost spiral arms in
occurs is Re = 44.4. the case of a mixture is in the range of 40–50. This is signifi-
Figure 5 shows the sorting of particles in five other cantly higher than for the case when only particles of one size
two-particle mixtures. The particle sizes and the hydraulic are present. While a rigorous explanation for this phenomena
diameters of the channels used to the sort the particles are is lacking and is left for future work, we can advance a hypoth-
listed in the table in the left-hand upper corner of the figure. esis to explain this. Particles of the two size categories, small

7
J. Micromech. Microeng. 25 (2015) 084013 P Tallapragada et al

(λ < 0.12) and large (λ > 0.12), focus in two laterally distinct particles could be larger. The advantages of different channel
bands in the channel. The small particles focus closer to the aspect ratios and curvature ratios for robust scale invariant dif-
inner walls, while the larger particles focus towards the center. ferential focusing of particles could be explored further.
At low volume fractions the longitudinal distance between
two particles in the same band is large, on the order of ten
times the particle diameter. At this distance the influence of References
particles on each others’ motion is small. However, the lat-
eral distance between particles in two adjacent focused bands [1] Jaramillo M C, Torrents E, Martinez-Duarte R, Madou M
is small. The resulting interaction between the wake of the and Juarez A 2010 On-line separation of bacterial cells by
carbon-electrode dielectrophoresis Electrophoresis
particles could explain the increase in the Reynolds number 31 2921–8
at which differential focusing occurs. However, the classifica- [2] Salomon S, Leichl T and Nicu L 2011 A dielectrophoretic
tion chart for differential lateral focusing 3(a) is also valid for continuous flow sorter using integrated microelectrodes
two-particle mixtures. coupled to a channel constriction Electrophoresis
32 1508–14
[3] Khoshmanesh K, Baratchi S, Tovar-Lopez F J, Nahavandi S,
5. Conclusion Wlodkowic D, Mitchell A and Kalantar-zadeh K 2012
On-chip separation of lactobacillus bacteria from yeasts
using dielectrophoresis Microfluidics Nanofluidics
With the increasing interest in the applications of inertial
12 597–606
hydrodynamics in curved channels and the potentially impor- [4] Elitas M, Martinez-Duarte R, Dhar N, McKinney J and
tant biomedical applications, it is important to identify the Renaud P 2014 Dielectrophoresis-based purification of
scale invariant behavior of inertial particles and their differ- antibiotic-treated bacterial subpopulations Lab Chip
ential lateral focusing. In this paper we have demonstrated the 14 1850–7
[5] Kolostova K, Pinterova D, Hoffman R M and Bobek V 2011
existence of a new parametric regime in which inertial par-
Circulating human prostate cancer cells from an orthotopic
ticles focus in a hitherto unknown manner in spiral channels. mouse model rapidly captured by immunomagnetic beads
We also demonstrated a scale invariant principle for hydro- and imaged by gfp expression Anticancer Res. 31 1535–9
dynamic focusing and the sorting of particles by size across [6] Zborowski M and Chalmers J J 2011 Rare cell separation and
a very large size range. The data in figure 3, demonstrates a analysis by magnetic sorting Anal. Chem. 83 8050–6
[7] Svennebring J, Manneberg O, Skafte-Pedersen P, Bruus H
parametric regime for differential lateral focusing of particles
and Wiklund M 2009 Selective bioparticle retention and
by size. This scale invariant differential lateral focusing of characterization in a chip-integrated confocal ultrasonic
particles is a useful rule to systematically engineer devices cavity Biotechnol. Bioeng. 103 323–8
to sort particles by size. Our experiments have demonstrated [8] Muller P B, Rossi M, Marin A G, Barnkob R, Augustsson P,
a consistent principle for sorting particles by size where the Laurell T, Kahler C J and Bruus H 2013 Ultrasound-
a induced acoustophoretic motion of microparticles in three
ratio of the sizes of the particles, a1 , is as low as 1.8. The data dimensions Phys. Rev. 88 023006
2
in figure  3 indicates that the non-dimensional resolution for [9] Antfolk M, Muller P B, Augustsson P, Bruus H and Laurell T
sorting particles by size could be as low as 1 : 1.5. 2014 Focusing of sub-micrometer particles and bacteria
enabled by two-dimensional acoustophoresis Lab Chip
Our experiments on sorting two-particle mixtures indi- 14 2791–9
cate a possible mechanism to sort multi-particle mixtures [10] Segre G and Silberberg A 1961 Radial particle displacements
sequentially. In each sequence particles can be separated into in poiseuille fow of suspensions Nature 189 209–10
two size categories with confinement ratios, λ < 0.12 and [11] Segre G and Silberberg A 1962 Behavior of macroscopic rigid
λ > 0.12, respectively. Devices with appropriate curvature spheres in poiseuille flow J. Fluid Mech. 14 136–57
[12] Bhagat A A S, Kuntaegowdanahalli S S and Papautsky I 2008
can be designed based on the data in figure  3. The channel Continuous particle separation in spiral microchannels
Reynolds number at which this sorting is possible is in the using dean flows and differential migration Lab Chip
range of 40–50, which allows high throughput. At the same 8 1906–14
time, these Reynolds numbers are not high enough to cause [13] Bhagat A A S, Kuntaegowdanahalli S S and Papautsky I 2009
high pressure on the channel walls that could deform the walls Inertial microfluidics for continuous particle filtration and
extraction Microfluid. Nanofluid. 7 217–26
of devices made of PDMS. [14] Di Carlo D, Irimia D, Tompkins R G and Toner M 2007
The scale invariant size-based differential focusing of Continuous inertial focusing, ordering, and separation of
particles could possible change with the changes in the particles in microchannels Proc. Natl Acad. Sci.
two geometric parameters, aspect ratio and curvature ratio, 104 18892–7
δ. Decreasing the curvature ratio to zero (a straight channel) [15] Yoon D H, Ha J B, Bahk Y K, Arakawa T, Shoji S and Go J S
2009 Size-selective separation of micro beads by utilizing
obviously changes the focusing behavior. It is therefore pos- secondary flow in a curved rectangular microchannel Lab
sible that when the curvature ratio is significantly lower than Chip 9 87–90
the values used in our findings, the focusing behavior of par- [16] Gossett D R, Weaver W M, Mach A J, Hur S C, Tse H T K,
ticles changes significantly. At high aspect ratios, the balance Lee W, Amini H and DiCarlo D 2010 Label-free cell
between Dean drag and lift forces can occur at different lateral separation and sorting in microfluidic systems Anal.
Bioanal. Chem. 397 3249–67
positions in the channel; see for example [42] for a detailed [17] Hur S C, Mach A J and DiCarlo D 2011 High-throughput
numerical and experimental exploration of different cases. size-based rare cell enrichment using microscale vortices
The resulting gap between differentially focused bands of Biomicrofluidics 5 022206

8
J. Micromech. Microeng. 25 (2015) 084013 P Tallapragada et al

[18] Lee M G, Choia S and Park J 2011 Inertial separation in a [29] Nagrath S, Gupta A K and Toner M 2009 Differential
contraction?expansion array microchannel J. Chromatogr. inertial focusing of particles in curved low-aspect-ratio
1218 4138–43 microchannels New J. Phys. 11 075025
[19] Wu L, Guan G, Hou H W, Bhagat A A S and Han J 2012 [30] Johnston I D, McDonnell M B, Tan C K L, McCluskey D K,
Separation of leukocytes from blood using spiral channel Davies M J and Tracey M C 2014 Dean flow focusing and
with trapezoid cross-section Anal. Chem. 84 9324–31 separation of small microspheres within a narrow size range
[20] Nivedita N and Papautsky I 2013 Continuous separation of Microfluid. Nanofluid. 17 509–18
blood cells in spiral microfluidic devices Biomicrofluidics [31] Martel J M and Toner M 2012 Inertial focusing dynamics in
7 054101 spiral microchannels Phys. Fluids 24 032001
[21] Guan G, Wu L, Bhagat A A S, Li Z, Chen P C Y, Chao S, [32] Martel J M and Toner M 2013 Particle focusing in curved
Ong C J and Han J 2013 Spiral microchannel with microfluidic channels Sci. Rep. 3 1–8
rectangular and trapezoidal cross-sections for size based [33] Harvat T, Yong W and Eddington D T 2010 Size-based
particle separation Nat. Sci. Rep. 3 1475 separation and collection of mouse pancreatic
[22] Zhang J, Yan S, Sluyter R, Li W, Gursel A and Nguyen N T islets for functional analysis Biomed. Microdevices
2014 Inertial particle separation by differential equilibrium 12 865–74
positions in a symmetrical serpentine micro-channel Nat. [34] Schonberg J A and Hinch E J 1989 Inertial migration of a
Sci. Rep. 4 4527 sphere in poiseuille flow J. Fluid Mech. 203 517–24
[23] Sun J, Liu C, Li M, Wang J, Xianyu Y, Hu G and Jiang X 2013 [35] Vasseur P and Cox R G 1976 The lateral migration of a
Size-based hydrodynamic rare tumor cell separation in spherical particle in two-dimensional shear flows J. Fluid
curved microfluidic channels Biomicrofluidics 7 011802 Mech. 78 386–413
[24] Sun J, Li M, Liu C, Zhang Y, Liu D, Liu W, Hu G and Jiang X [36] Asmolov E S 1999 The inertial lift on a spherical particle in
2012 Double spiral microchannel for label-free tumor cell a plane poiseuille flow at large channel reynolds number J.
separation and enrichment Lab Chip 12 3952–60 Fluid Mech. 381 63–87
[25] Hou H W, Warkiani M E, Khoo B L, Li Z R, Soo R A, [37] Matas J P, Morris J F and Guazzelli E 2004 Inertial migration
Tan D S W, Lim W, Han J, Bhagat A A S and Lim C T of rigid spherical particles in poiseuille flow J. Fluid Mech.
2013 Isolation and retrieval of circulating tumor cells using 515 171–95
centrifugal forces Nat. Sci. Rep. 3 1259 [38] Takemura F and Magnaudet J 2009 Lateral migration of a
[26] Zhang J, Yan S, Li W, Alicia G and Nguyen N T 2014 High small spherical buoyant particle in a wall-bounded linear
throughput extraction of plasma using a secondary flow- shear flow Phys. Fluids 21 083303
aided inertial microfluidic device RSC Adv. 4 33149 [39] Saffman P G 1965 The lift on a small sphere in a slow shear
[27] Kemna E W M, Schoeman R M, Wolbers F, Vermes I, flow J. Fluid Mech. 22 385
Weitz D A, van den Berg A and van den Berg A 2012 [40] Dean W R 1928 Fluid motion in a curved channel Proc. R.
High-yield cell ordering and deterministic cell-in-droplet Soc. 121 402–20
encapsulation using dean flow in a curved microchannel [41] Berger S A, Talbot L and Yao L S 1983 Flow in curved pipes
Lab Chip 12 2881 Ann. Rev. Fluid Mech. 15 461–512
[28] Kuntaegowdanahalli G K S S, Bhagat A A S and Papautsky I [42] Liu C, Hu G and JiaX X 2015 Inertial focusing of spherical
2009 Inertial microfluidics for continuous particle particles in rectangular microchannels over a wide range of
separation in spiral microchannels Lab Chip 9 2973–80 reynolds numbers Lab Chip 4 1168–77