You are on page 1of 19

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/366123306

Experimental and Numerical Study of the Effect of the Channel Curvature


Angle on Inertial Focusing in Curvilinear Microchannels

Article  in  Journal of Applied Physics · December 2022


DOI: 10.1063/5.0117224

CITATIONS READS

0 40

4 authors:

Deniz Ince Hatice Turhan


Istanbul Technical University 6 PUBLICATIONS   2 CITATIONS   
3 PUBLICATIONS   2 CITATIONS   
SEE PROFILE
SEE PROFILE

Sertac Cadirci Levent Trabzon


Istanbul Technical University Istanbul Technical University
73 PUBLICATIONS   239 CITATIONS    119 PUBLICATIONS   662 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Investigation on the Effects of Mechanical Forces on Endothel Cells Behaviour by Using Microfludics Systems in the domain of Mechanobiology View project

Thermally Conductive Graphene Textiles (Granted by ASELSAN) View project

All content following this page was uploaded by Levent Trabzon on 12 December 2022.

The user has requested enhancement of the downloaded file.


Experimental and numerical study of the
effect of the channel curvature angle on
inertial focusing in curvilinear microchannels
Cite as: J. Appl. Phys. 132, 224703 (2022); https://doi.org/10.1063/5.0117224
Submitted: 31 July 2022 • Accepted: 17 November 2022 • Published Online: 12 December 2022

D. Ince, H. Turhan, S. Cadirci, et al.

J. Appl. Phys. 132, 224703 (2022); https://doi.org/10.1063/5.0117224 132, 224703

© 2022 Author(s).
Journal of ARTICLE scitation.org/journal/jap
Applied Physics

Experimental and numerical study of the effect


of the channel curvature angle on inertial focusing
in curvilinear microchannels
Cite as: J. Appl. Phys. 132, 224703 (2022); doi: 10.1063/5.0117224
Submitted: 31 July 2022 · Accepted: 17 November 2022 · View Online Export Citation CrossMark
Published Online: 12 December 2022

D. Ince,1,2 H. Turhan,1 S. Cadirci,1 and L. Trabzon1,2,3,a)

AFFILIATIONS
1
Department of Mechanical Engineering, Istanbul Technical University, Istanbul 34437, Turkey
2
MEMS Research Center, Istanbul Technical University, Istanbul 34469, Turkey
3
Nanotechnology Research and Application Center—ITUnano, Istanbul Technical University, Istanbul 34469, Turkey

Note: This paper is part of the Special Topic on Multiphysics of Microfluidics and Nanofluidics.
a)
Author to whom correspondence should be addressed: levent.trabzon@itu.edu.tr. Tel.: +90 212 293 13 00.

ABSTRACT
Passive cell separation methods have attracted great attention due to their superiority over the other methods stemming from their easy
fabrication, precise manipulation, cost-effectiveness, sensitivity, and simplicity. The fluid inertia in these methods is the main factor that is
affected by the channel design; thus, the channel design parameters should be chosen accordingly. Even though all channel design parame-
ters are well addressed in inertial microfluidics, the curvature angle of the channel has not yet been extensively studied. In this study, three
different curvilinear microchannels with curvature angles of 180°, 210°, and 270° were designed, keeping all other remaining parameters the
same. The focusing ability of the fluorescent polystyrene microparticles with diameters of 1.1, 3.3, and 9.9 μm was investigated both experi-
mentally and numerically to understand focusing efficiency affected by the curvature angle of the microchannel. The first set of experiments
was to determine the effect of the channel curvature and indicated the favorable design as channel C, which showed focusing qualities of
0.85 and 0.92 for 9.9 μm particles at volumetric concentrations of 2% and 5%, respectively. The remaining set of experiments and CFD sim-
ulations were conducted to observe the interaction of 3.3 and 9.9 μm particles and reveal the distortion of the focusing line and particulate
phase contours for 9.9 μm particles at the flow rates between 0.3 and 0.7 ml/min, which was further confirmed by enriched mixtures con-
taining 1.1, 3.3, and 9.9 μm particles. The study showed that mixtures comprising low diameter particles could not satisfy the focusing crite-
ria, which emphasized the importance of an appropriate particle size and concentration for a single focus line. On the other hand, it was
shown that geometric features of the microchannel such as the hydraulic diameter and the curvature angle together with the particle size
determine the focusing quality both experimentally and numerically. To sum up, the increment of the channel curvature angle is a deter-
mining factor for particle focusing, and a single focusing line was observed on the particles maintaining the focusing criteria even in many
particle conditions. While the focusing quality of the particles was reduced by multi-particle interactions, they were proven to be separable
achieving the appropriate concentration ratio.

Published under an exclusive license by AIP Publishing. https://doi.org/10.1063/5.0117224

I. INTRODUCTION systems is to create a microstructure that can control the overflow


Microfluidics’ easy fabrication, precise manipulation, cost- field, thus separating and sorting cells at high rates.3,12–14 By doing
effectivity, sensitivity, and simplicity make them superior to conven- so, microfluidics manifests its achievement, regarding fractionation
and enrichment of particles from a heterogeneous mixture, in the
tional technologies.1–7 Among LOC (Lab on Chip) and μTAS scope of cancer diagnosis, water purification, and sedimentation.15–20
(micro-total-analysis-system) applications, in addition to the subjects One of the foremost fields of microfluidics is particle–particle
of chemical science, engineering disciplines, and clinics, these separation and focusing that can be achieved by using active and
systems have become more significant.8–11 The purpose of these passive methods. These methods are classified according to the

J. Appl. Phys. 132, 224703 (2022); doi: 10.1063/5.0117224 132, 224703-1


Published under an exclusive license by AIP Publishing
Journal of ARTICLE scitation.org/journal/jap
Applied Physics

requirement of an external force.1,3,10,21 In general, active methods behavior of 9.9 μm fluorescent polystyrene microparticles was
require external forces such as magnetic,22–28 dielectric,29–38 examined solely. The purpose of the first set was to estimate the
acoustic,39–43 optical,44–46 Marangoni flow,47–49 and hydrophore- most favorable channel design in terms of focusing quality. In the
sis.50,51 Moreover, the active particle separation method is known second set, mixtures of 9.9 and 3.3 μm particles were injected with
to provide more accurate results; however, adapting them to a an increment in the particle volumetric concentration to determine
device has certain complexities.52 In other words, the requirement the favorable constant concentration value of 9.9 μm particles for
of expensive equipment and finding optimum parameters for the better focusing quality. In the third set, the effect of volumetric
separation phenomena restricts their ability for rapid separation. concentrations of 3.3 μm particles on the constant volumetric con-
On the other hand, passive particle separation methods utilize centration of 9.9 μm particles’ focusing ability was investigated to
hydrodynamic forces solely that make them more compatible with determine the favorable concentration for 3.3 μm particles for
medical and biological applications. Moreover, passive methods better focusing. In the last set, the particle–particle interaction was
ensure high flow rates as they operate in the continuous mode, and tested by using three particles, i.e., particles with diameters of 9.9,
the less equipment requirements make them more affordable for 3.3, and 1.1 μm. All the experiments were conducted at different
the same purpose compared to the active particle separation flow rates to observe the effect of different Reynolds and Dean
methods.3,53 numbers. In summary, a single focusing line was provided by the
The channel variations and designs are the most noteworthy particles, which met the focusing criteria. A more efficient decom-
parameters in passive cell separation. The channel variations in position was achieved by increasing the degree of the curvature.
passive methods are strictly associated with hydrodynamic forces; The particles, under the focusing criteria, jeopardized the focusing
therefore, many researchers wend their way to the channel type quality of the particles that met the focusing criteria. More impor-
investigation for better understanding of the effects of channel vari- tantly, it was revealed that as long as a sufficient particle composi-
ations on passive cell separation, thus inertial focusing.4,52,54–56 tion was retained, particle separation could be achieved at high
Microchannels used for passive cell separation can be basically clas- rates.
sified into four configurations: straight,57,58 curved,59–63
spiral,55,64–67 and serpentine.68–72 In straight channels, the equilib-
rium positions of the particles are slightly interlinked with hydro- II. THEORY
dynamic forces and the hydraulic diameter of the channel.52 Pure Hydrodynamic forces play a significant role for particle migra-
equilibrium position in straight channels is limited compared to tion in inertial microfluidics. Particle migration can be controlled
curved channels, but some researchers proved that well-designed by two dominant forces, which are literally known as inertial lift
principles lead to more effective designs, in which full separation and Dean drag forces. Their magnitudes are not only affected by
can be observed.73 In contrast, curved channels enhance their effi- the flow rate, viscosity, and particle dimension but also by the
ciency by the presence of an additional force called the Dean shape, size, curvature angle, and radius of the microchannel. Many
force.64,74 This force, induced by two counter flow rotations, i.e., researchers, thus, concentrated on the underlying physics of the
Dean vortices—occurring at the cross-sectional area of the channel, lateral migration for better understanding of these factors.
is responsible for the particles’ migration in the transverse direction The lateral migration of the particles was first mentioned by
and creates an equilibrium position in the channel cross section. Segre and Silberberg and labeled the tubular pinch effect.77 They
The channel design in passive cell separation is another significant found that the particles in the laminar flow are prone to become
parameter. Since the hydrodynamic forces are restricted by channel stable in an equilibrium position. The foundation of the lateral
design parameters such as the channel length, width, height, radius migration made a breakthrough in the understanding of inertial
of curvature, and curvature angle, a better interpretation of the focusing.52 Saffman came up with a hypothesis, in which he men-
sample acquisitions is required in order to design appropriate chan- tioned the variance of the velocity profile surrounding particles,
nels and to correlate flow parameters with the microfluidic systems which was responsible for particles’ movement due to the slipping
correctly. Although these channel parameters are embedded into velocity.78 However, the importance of the particle rotations in the
the hydrodynamic forces, the effect of the channel curvature angle focusing phenomena was not noted. Lately, plenty of experiments
has not yet been addressed in the literature. showed that this mismatched with the theory of particle migration.
In this study, three different channel geometries, literally On the other hand, Saffman’s theory triggered some analytical
named symmetric curvilinear channel geometries, which can research studies related to inertial focusing. The huge contribution
supply high flow rates, were designed. These channels possessed to inertial focusing theory was done by leading researchers.52,56,79
curvature angles of 180°, 210°, and 270°, where all other geometri- They used analytical approaches to observe the effects of the
cal parameters such as the height, width, length of the channel, and Poiseuille flow on a solid particle and put forth the shear gradient
the number of loops remained the same. A CFD model was created force. The shear gradient force was assumed as one of the three
and applied to the cases investigated also experimentally. The simu- fundamental forces for inertial microfluidics; however, it was noted
lation results were compared to benchmark studies in the literature, as the most effective one.52
and the CFD model was proven as a robust model capable of Even though working principles of the forces slightly differ
solving inertial microfluidics59,75 and found to be appropriate for from one another, the wall interaction force and the shear gradient
particle–particle separation and focusing.76 After the channels were lift force are together known as the inertial lift force, which is a
directed to standard manufacturing processes and produced, four combined force maintaining the lateral movement. Due to the par-
experimental sets were carried out. In the first set, the focusing abolic velocity profile of the flow in the channel, fluid velocity

J. Appl. Phys. 132, 224703 (2022); doi: 10.1063/5.0117224 132, 224703-2


Published under an exclusive license by AIP Publishing
Journal of ARTICLE scitation.org/journal/jap
Applied Physics

surrounding the particles relatively varies relative to the wall. resulting in an increased force at the center by which fluid travels
Differences in the velocity result in a compensation movement that along the inner wall to the outer wall of the microchannel and
makes particles migrate from the center of the channel toward the causes two symmetrical vortices in the top and bottom halves of
walls. The shear gradient lift force is given in the following equa- the channel, which are perpendicular to the flow direction.64,83,84
tion: The definition of the non-dimensional number related to Dean
flow is given in the following equation:
(CSG ρUmax
2
a3p )
FSG ¼ , (1) rffiffiffiffiffiffi
Dh Dh
De ¼ Re , (6)
2R
where CSG is the shear gradient lift force coefficient, ρ and μ are the
fluid density and viscosity, respectively, Umax is the maximum
where Re is the channel Reynolds number and R is the channel
velocity, and Dh is the hydraulic diameter of the channel. The
curvature radius. The channel Reynolds number is calculated
diameter of the particles is denoted by ap. For a channel with a
based on the maximum velocity and the hydraulic diameter of
rectangular cross section, the hydraulic diameter is defined in the
the microchannel as shown in Eq. (7a). In a similar fashion, the
following equation:
particle Reynolds number can be defined based on the particle
diameter and the λ ¼ ap /Dh ratio as given in the following equa-
2HW
Dh ¼ , (2) tion:
HþW

where H is the channel height and W is the channel width. Since ρUmax Dh
Re ¼ , (7a)
particles drift along with the flow, they undergo dissimilar effects μ
of the flow on either side, where the fluid velocity is relatively
higher than the particle velocity. When the particles approach the
wall of the channel, wall interaction forces push them from ρUmax a2p
Rep ¼ λ2 Re ¼ : (7b)
the channel walls toward to the center, which is associated with μDh
the distance between the wall and the particle. In other words, the
distance between the particle and the wall of the channel creates Equation (6) shows that the Dean number (De) is associated
wall interaction force,80–82 which is given in the following equa- with the Reynolds number in Eq. (7) and the curvature radius of
tion: the channel. Thus, the Dean number is affected by the flow veloc-
ity and geometric features of the flow domain such as the hydrau-
CWI ρUmax
2
a6p lic diameter and the curvature radius of the channel. When the
FWI ¼ , (3) curvature radius increases, the channel can be assumed to be
D4h
plain, where the Dean number is approaching zero confirming
where CWI is the wall-induced force coefficient. that the Dean flow cannot be observed in the plain microchan-
During the lateral migration, in some way, the two intrinsic nels. A correlation for the average and maximum secondary flow
forces working in the opposite directions occupy an equilibrium velocities with respect to the changing Dean number was pro-
position in the channel cross section.52,60 Therefore, the inertial posed in Refs. 59 and 85.
force formulas are further reduced to formulas that depend on par- U Dean is the average Dean flow velocity, which is expressed in
ticles’ lateral migration. When the shear gradient force is dominant the following equation:
in the channel centerline, the net lift force is expressed by the fol-
m
lowing equation:  Dean ¼ 1:8  104 De1:63
U : (8)
s
ρUmax
2
a3p
FL ¼ : (4) The following equation shows the drag force exerted on a par-
Dh
ticle due to Dean flows that can be obtained by assuming the
In addition, it changes by the increase in the wall interaction Stokes drag:
force near to the channel wall, thus becoming the following equa-
tion:  Dean ap :
FD ¼ 3πμU (9)

ρUmax
2
a6p The Dean flow leads to the Dean drag force (FD), which is sig-
FL ¼ : (5)
D4h nificant for obtaining equilibrium positions of the particles. Based
on this force, particles move along the Dean vortices regardless of
The number of focusing positions is interdependent with the their size. Considering the Stokes drag force, the Stokes flow can be
channel geometry. Introducing a curvature to the straight channels modified for the corresponding flow, which implies that the drag
initiates the secondary flow. The secondary flow, first observed by force is directly proportional to the particle diameter and μ, the
Dean, originates from the dominant lift force at the channel center, dynamic viscosity of the fluid. The Dean drag force can be

J. Appl. Phys. 132, 224703 (2022); doi: 10.1063/5.0117224 132, 224703-3


Published under an exclusive license by AIP Publishing
Journal of ARTICLE scitation.org/journal/jap
Applied Physics

expressed in the following equation: models used in the study and their dimensions are given in
Figs. 1(a)–1(c) and in Table I, respectively.
FD ¼ (5:4  104 ) πμDe1:63 ap (N): (10) The geometric design parameters of the microchannels were
selected in a way to meet the inertial focusing criteria,
Particles were found to be stable in four equilibrium positions λ ¼ ap /Dh . 0:07 and Rep ¼ λ2 Re . 1, as well as the required
in the channel cross section.62 The shear gradient and the wall channel length. Based on the inertial microfluidic studies, the
interaction lift are the main driving forces in such channels, and required channel length for the channel aspect ratio varying
these forces push the particles in the opposite directions. Thus, par- between 0.5 and 2 should be provided, where CL is the lift coeffi-
ticles become stable by the aid of these forces. Moreover, it is worth cient ranging from 0.02 to 0.05.54,56,88 For each operating condition
to mention that particles, without effective driving forces, cannot and particle diameter, the required channel length, Lf, can be calcu-
be stabilized in the channel cross section. In curved channels, on lated using the following equation, where the total length of the
the other hand, the Dean drag force restricts these four equilibrium channel is 7.54 mm:
positions to a single position. In these channels, the net lift force
and the Dean drag force work in the opposite directions to limit πμD2h
particles’ focusing positions.56,79,81,84,86,87 Lf ¼ : (11)
CSG ρUmax a2p

Important parameters for inertial focusing including channel Re


A. Channel design and particle Re based on the particle diameter and thresholds for
The channel geometries with three different curvature angles high quality focusing are tabulated in Table II, where some of
were designed, and the other geometrical parameters including the the conditions do not satisfy the optimum inertial focusing
channel height, width, and length were kept the same. The channel criteria.

FIG. 1. Microchannels used in the study: (a) channel A, (b) channel B, and (c) channel C.

J. Appl. Phys. 132, 224703 (2022); doi: 10.1063/5.0117224 132, 224703-4


Published under an exclusive license by AIP Publishing
Journal of ARTICLE scitation.org/journal/jap
Applied Physics

TABLE I. The dimensions and features of the microchannels.

Name of the channels A B C


Curvature of the channels, R (μm) 900 900 900
Length between two loops, L (μm) 900 1280 1650
Channel width, W (μm) 50 50 50
Channel height, H (μm) 50 50 50
The number of the loops 48 48 48
The curvature angle of each loop 180 [360–(θ = 180)] 210 [360–(θ = 150)] 270 [360–(θ = 90)]

III. FABRICATION AND METHODS (Carl Zeiss Jena GmbH, Thüringen, Germany), and then, the
A. Fabrication features were measured to assess the accuracy of the pattern.

The fabrication process of the microfluidic chip is explained 3. Negative photoresist coating
in this section.
Separately, a 100 mm silicon wafer was cleaned first with
acetone, then with isopropyl alcohol, and finally with DI water.
1. Pattern mask lithography
The wafer was then blow-dried using nitrogen gas and left on a
The channel pattern masks were developed based on the CFD hotplate set to 115 °C for 15 min. The silicon wafer was placed in a
analysis. The channel design was prepared using PCB design soft- WS-400-6NPPB single substrate spin processor (Laurell
ware Klayout v0.23.10 (Matthias Köfferlein) and Ledit v13 (Tanner Technologies Corporation, North Wales, USA) and spin coated
EDA, California, USA). The designs saved first as a GDSII format with a SU-8 3050 negative photoresist (Microchemicals GmbH,
were then converted to JOB files using proprietary conversion soft- Ulm, Germany) using a program set to spin at 3000 rpm for 30 s to
ware provided by Heidelberg Instruments Mikrotechnik GmbH. obtain 50 μm channel thickness and 1000 rpm for 30 s to obtain
The pattern masks were written on a 120 × 120 × 8 mm3 chromium 100 μm channel thickness. After the first layer of spin coating, the
plate that fused quartz substrates with a 540 nm thick AZ1500 photoresist was soft baked at 95 °C for 10 min.
series positive photoresist coating (CS Hardmask Blanks, Clean
Surface Technology Co., Kanagawa, Japan) using a 4 mm write 4. Photolithography and development
head on a DWL 66FS series High Resolution Pattern Generator
(Heidelberg Instruments Mikrotechnik GmbH., Heidelberg, The photoresist coated wafer and the pattern mask were
Germany). placed in a Suss MA6 Mask Aligner (SUSS MicroTec, Garching,
Germany) and exposed to an 850 W 365 nm UV light source in
soft contact configuration. Immediately after exposure to UV light,
2. Mask development and etching the wafer was placed on a hotplate set to 95 °C for 5 min for a post
After exposure to the AZ1500 series, the photoresist was exposure bake. Once the post exposure bake was completed, the
developed in an AZ 351B solution (Microchemicals GmbH, Ulm, wafer was submerged in a puddle of SU-8 developer solution
Germany), which was diluted with de-ionized (DI) water to 20% (Microchemicals GmbH, Ulm, Germany) in a Petri dish gently
solvent for 40 s and rinsed with DI water. The development of the shaken for 5 min to allow the photoresist to develop. The wafer was
photoresist was first assessed qualitatively for color changes indicat- then removed from the puddle, rinsed first with a fresh developer
ing underdevelopment under an Axio Imager Upright Microscope solution for 10 s and then with isopropyl alcohol, and blow-dried

TABLE II. Inertial focusing criteria calculations.

Flow rate (ml/min) Dh (μm) Umax (m/s) Re Rep1 Rep2 Rep3 λ1 λ2 λ3 ap1 (μm) ap2 (μm) ap3 (μm)
0.1 50 0.667 33.3 1.307 0.145 0.016 0.198 0.066 0.022 9.9 3.3 1.1
0.2 50 1.333 66.7 2.614 0.290 0.032 0.198 0.066 0.022 9.9 3.3 1.1
0.3 50 2.000 100.0 3.920 0.436 0.048 0.198 0.066 0.022 9.9 3.3 1.1
0.4 50 2.667 133.3 5.227 0.581 0.065 0.198 0.066 0.022 9.9 3.3 1.1
0.5 50 3.333 166.7 6.534 0.726 0.081 0.198 0.066 0.022 9.9 3.3 1.1
0.6 50 4.000 200.0 7.841 0.871 0.097 0.198 0.066 0.022 9.9 3.3 1.1
0.7 50 4.667 233.3 9.148 1.016 0.113 0.198 0.066 0.022 9.9 3.3 1.1
0.8 50 5.333 266.7 10.454 1.162 0.129 0.198 0.066 0.022 9.9 3.3 1.1
0.9 50 6.000 300.0 11.761 1.307 0.145 0.198 0.066 0.022 9.9 3.3 1.1
1.0 50 6.667 333.3 13.068 1.452 0.161 0.198 0.066 0.022 9.9 3.3 1.1

J. Appl. Phys. 132, 224703 (2022); doi: 10.1063/5.0117224 132, 224703-5


Published under an exclusive license by AIP Publishing
Journal of ARTICLE scitation.org/journal/jap
Applied Physics

with nitrogen gas. After development, the accuracy of the features with flow solver software, ANSYS Fluent. Modifications of the
was assessed under an Axio Imager Upright Microscope (Carl geometry were completed with the Siemens-Nx program, and the
Zeiss Jena GmbH, Thüringen, Germany). After the assessment of geometries were transferred to ANSYS Fluent in a Parasolid format
the accuracy of the features was made in the 100 class laminar flow where the geometry was specified as the control volume with
clean room with safe lighting, the microchannel mold was then appropriate boundary conditions. The boundary conditions used in
taken out of the 100 class room into the 1000 class cleanroom with the analyses were as follows: the uniform velocity inlet at the single
fluorescent lights, where the Axio CSM 700 (Carl Zeiss Jena opening of the channel was defined and the pressure-outlet was
GmbH, Thüringen, Germany) was located. To further assess the used at the exit or the triple opening of the channel. The entrance
accuracy and usability of the microchannel mold, its depth and length of the microchannel is sufficient to provide fully developed
surface roughness were measured. flow condition in the flow visualization section. All the walls satis-
fied no-slip boundary conditions. In addition, based on the analy-
5. Silicone microchannel casting sis, volume fractions were defined for the secondary particulate
phases at the inlet of the channel.
A Sylgard® 184 polydimethylsiloxane (PDMS) silicone elasto- As the mesh type, hexahedral elements were preferred since
mer and a curing agent (Dow Corning, Minnesota, USA) were the convergence and accuracy of the analyses were expected to be
thoroughly mixed for 2 min in 10:1—the elastomer-to-curing agent better than those of tetrahedral elements. The following step was
ratio. The mixture was then placed inside a vacuum chamber at mesh-independency tests since the reliability of the CFD analysis
50 mbar to avoid air bubbles caused by mixing. A 2 × 2 × 90 mm3 strictly depends on the number of elements. Mesh-independency
rubber ring was placed on top of the wafer mold, which was pre- analyses were performed for the single-phase flow at various
pared in the earlier step to prevent spills, and then, the silicone number of elements meeting the same convergence criteria, and
mixture was poured on the wafer mold. The wafer was placed the analyses continued until the residuals were below 10−5. As
inside a vacuum chamber with heating capability to remove any air Fig. 2 shows, the velocity profiles in the streamwise direction in the
bubbles that got trapped inside the PDMS mixture while poring, fully developed region of the channel indicated that a mesh consist-
and the polymer was cured for 15 min at 50 mbar pressure and ing of approximately 1.4 × 106 elements was found to be sufficient
115 °C temperature. After the curing process was completed by for further particle–particle analyses.
poking the edge of the PDMS layer with the end of a pair of The details of two-phase CFD analyses are tabulated in
forceps, the PDMS layer on top of the wafer was peeled off and the Table III. The finite-volume-based flow solver was utilized, which
excess PDMS around the microchannel was cut out with a scalpel can predict the flow field through the microchannel with high
and 0.5 mm holes for attaching the syringes were punctured. accuracy. Laminar, steady, incompressible Navier–Stokes equations
with a two-phase flow approach were solved at different flow rates,
6. Glass substrate preparation and after post-processing the simulations, detailed flow patterns
First, 70 × 25 × 3 mm3 microscope slides (Gold Seal Products, including the velocity and Dean vortices at a certain cross section
Hungary) were cleaned in acetone for 10 min at 35 kHz using a of the microchannel were obtained. To simulate the particles that
UW Qi-100 ultrasonic cleaner (Ultrawave Precision Ultrasonic drifted along with the flow, the Eulerian–Eulerian based VOF
Cleaning Equipment, Cardiff, U.K.). The substrates were rinsed (volume of fluid) multiphase approach was found to be the optimum
first with isopropyl alcohol and then with water and blow-dried model that matched to the experimental results since this approach
with nitrogen gas and put on a hot plate at 115 °C for 10 min. predicts the interface shape between immiscible fluid phases. In this
approach, both the fluid and particles were treated as continuous
liquid phases. However, the granular form as a characteristic was uti-
7. Assembly
lized for the secondary or particulate phase. The SIMPLE algorithm
The PDMS microchannel and the glass substrate were placed is utilized together with the first-order upwind scheme for spatial
inside a PDC-32G-2 plasma cleaner (Harrick Plasma, New York, discretization. For the Eulerian multi-fluid VOF model, the implicit
USA) with their surfaces facing up and exposed. The chamber was formulation is used for the volume fraction parameter.
vacuumed down to 200 mTorr pressure, and oxygen gas was In the phase-interaction, the Saffman–Mei lift coefficient was
pumped into the chamber while it was still being vacuumed. The preferred in the VOF model. The reason for adding a lift coefficient
oxygen gas inside the chamber, at 200 mTorr, was subjected to was to represent phase-interaction lift in the laminar flow, and
18 W RF radiation to produce oxygen plasma. Once stable plasma Saffman buoyancy is suitable for small particles at a submicron
was produced, the parts inside the chamber were exposed to the scale. However, the Saffman–Mei model is not fully compatible
plasma for 50 s and then removed. Immediately after removal from with the flow of interest, but the model was chosen since there was
the chamber, the surface of the glass substrate and the microchan- no adequate alternative and it was proven itself by benchmarking
nel were placed on top of the affected surface. studies in the literature. The Wen–Yu drag coefficient was used in
the VOF model since it is applicable to dilute phase flows where
B. Methods the total volume fraction of the secondary phase is significantly
lower than that of the primary phase, water.89 Since the particles’
1. Computational fluid dynamic analyses density was close to water’s density, the interphase surface tension
Since the masks were designed as solid bodies with the was assumed to be zero. The primary phase had a density of
AutoCAD program, flow-domains were extracted to be compatible 1000 kg/m3 and an absolute viscosity of 10−3 kg/ms. The secondary

J. Appl. Phys. 132, 224703 (2022); doi: 10.1063/5.0117224 132, 224703-6


Published under an exclusive license by AIP Publishing
Journal of ARTICLE scitation.org/journal/jap
Applied Physics

FIG. 2. Mesh-independency tests for channel A.

phase (polystyrene particles) had a density of 1050 kg/m3 and an taken the same as that of the experimental studies, which are
absolute viscosity of 10−3 kg/ms, and the diameters of the particles shown in Table IV. The flow solver included the effect of the parti-
were selected as 9.9, 3.3, and 1.1 μm. In the simulations, the initial cle diameter to the volume fraction calculation. As an example, the
volume fractions (volumetric percent) for particulate phases were particle diameter was set in the phase definition section and the

TABLE III. Summary of CFD studies in detail.

CFD CFD CFD CFD


Parameters Set-1 Set-2 Set-3 Set-4
Dh (μm) 50 50 50 50
Curvature angle 180, 210, and 270° 270° 270° 270°
Channel type A, B, C C C C
ap (μm) 9.9 3.3 and 9.9 3.3 and 9.9 1.1, 3.3, and 9.9
Flow rate (ml/min) 0.5 0.3, 0.5, and 0.7 0.3, 0.5, and 0.7 0.3, 0.5, and 0.7
Multiphase model Euler–Euler Euler–Euler Euler–Euler Euler–Euler
Eulerian parameters Multi-fluid VOF model Multi-fluid VOF model Multi-fluid VOF model Multi-fluid VOF model
Volume fraction parameters Implicit Implicit Implicit Implicit
Interface modeling Dispersed Dispersed Dispersed Dispersed
Phase-interaction (drag) Wen–Yu Wen–Yu Wen–Yu Wen–Yu
Phase-interaction (lift) Saffman–Mei Saffman–Mei Saffman–Mei Saffman–Mei
Phase-1 Water Water Water Water
Phase-2 Polystyrene (9.9 μm) Polystyrene (9.9 and 3.3 Polystyrene (9.9 and 3.3 Polystyrene (9.9; 3.3 and
μm) μm) 1.1 μm)
Phase-1 characteristics Fluid Fluid Fluid Fluid
Phase-2 characteristics Granular (spec. gravity Granular (spec. gravity Granular (spec. gravity Granular (spec. gravity =
= 1.05) = 1.05) = 1.05) 1.05)
Granular viscosity Gidaspow Gidaspow Gidaspow Gidaspow
Phase-2 volumetric fraction, 2 and 5 1.5 1.4 1.5; 1.4; and 1.4
α (%)

J. Appl. Phys. 132, 224703 (2022); doi: 10.1063/5.0117224 132, 224703-7


Published under an exclusive license by AIP Publishing
Journal of ARTICLE scitation.org/journal/jap
Applied Physics

TABLE IV. Summary of experimental studies in detail.

Exp. Exp. Exp. Exp.


Parameters Set-1 Set-2 Set-3 Set-4
Dh (μm) 50 50 50 50
Curvature angle 180°, 210°, 270° 270° 270°
and 270°
Channel type A, B, C C C C
ap (μm) 9.9 3.3 and 9.9 3.3 and 9.9 1.1, 3.3, and 9.9
Number of exp. 60 36 30 18
Flow rate (ml/min) 0.1–1 0.3, 0.5, and 0.7 0.3, 0.5, and 0.7 0.3, 0.5, and 0.7
Phase-2 volumetric 2% and 5% 9.9 μm (1%–2% and 5%) 9.9 μm (1.5%) 9.9 μm (1.5%)
concentration (%) 3.3 μm (1%–2% and 5%) 3.3 μm (0.1, 0.2, and 0.5 and 3.3 μm (1.4%)
1%–2%)
(same volumetric (9.9 μm at fixed concentration 1.1 μm (0.1%, 0.2%, 0.5%, 1%, 1.4%,
concentration and and 3.3 μm with various and 2%)
increment) concentrations)
(9.9 and 3.3 μm at fixed
concentrations and 1.1 μm with
various concentrations)

effect of its size on the volume was integrated into the volume frac- details of the governing equations and the forces denoted on the
tion section, where the numerical meaning of the volume fraction, right-hand side of the equation,
α, is defined in the following equation:
@
(αq ρq !
vq ) þ ∇:(αq ρq !
vq !
vq ) ¼ αq ∇p þ ∇:τ q þ αq ρq~
g
Volume of phase in cell @t
α¼ : (12) X
n
Volume of cell !!
þ (K pq (v p vq ) þ m_ pq !
vq )
The parameters in Table III were found by completing a series p¼1
of benchmarking studies based on Eulerian-model-theory such that ! ! !
þ αq ρq (Fq þ Flift,q þ Fvm,q ):
the current results were confirmed by them. Therefore, it was
assumed that the CFD model had the capacity to solve inertial (14)
focusing problems. The model was not controlled by a user defined
function, UDF, that makes the model distinctive among CFD
2. Experimental setup
studies in the literature.
Although the total computation time and the number of itera- Although the CFD calculations provide a deep insight into the
tions required for the steady-state analysis varied according to the flow structures and hydrodynamics of the problem of interest, they
number of phases of various diameters involved, the convergence are numerical predictions based on certain assumptions and
criteria of 10−5 were met for all analyses. The governing equations include acceptable numerical errors due to truncation, order of sol-
such as the conservation of mass and momentum for the fluid and ution scheme, and robustness of the flow solver. Therefore, the best
particulate phases were solved in an iterative scheme until the con- way to understand the behavior of the particulate phase within the
vergence criteria were achieved. In Eulerian-model-theory assuming flow field should be experimentation, specifically flow visualizations
a VOF-based multiphase flow approach, the governing equations to observe the focusing quality of mixtures. In this section, the
for fluid–fluid and granular multiphase flows, as solved by ANSYS experimental setup and the results are discussed in detail.
Fluent, are presented for the general case of the n-phase flow. The Four types of experimental sets were created to observe the
volume fraction of each phase is calculated from a continuity equa- focusing quality in the symmetric curvilinear microchannels. The
tion given in the following equation: first set was carried out to decide for the most favorable channel
design in terms of inertial focusing. To achieve this, a suspension
@ Xn of the 9.9 μm polystyrene particles was sent into the channels at ten
(αq ρq ) þ ∇:(αq ρq !
vq ) ¼ _ pq ,
m (13) different flow rates ranging from 0.1 to 1 ml/min and two different
@t p¼1
concentration values of 2% and 5%.
where α is the volume fraction and p and q represent the phases. The second set was arranged to determine the most effective
The fluid–fluid momentum equations, i.e., laminar Navier– particle concentration proportion. To do so, the concentration of
Stokes equations including all external force sources, are given in 9.9 and 3.3 μm diameter particles was increased equally, and the
Eq. (14). Readers may refer to Ref. 89 to learn more about the concentration ratio, for which the highest focusing quality was

J. Appl. Phys. 132, 224703 (2022); doi: 10.1063/5.0117224 132, 224703-8


Published under an exclusive license by AIP Publishing
Journal of ARTICLE scitation.org/journal/jap
Applied Physics

achieved, was selected. For all selected concentration values, the image to ImageJ, a line was drawn between the microchannel walls.
second experimental set was carried out at three different flow The program then conducted an intensity analysis along this line
rates: 0.3, 0.5, and 0.7 ml/min. and plotted intensity vs location.
The third set was created to investigate the effect of the con- These results were compared based on the focusing quality
centration change of 3.3 μm particles on the focusing quality of Q, which, in turn, depends on these two factors: the normalized
9.9 μm particles. The concentration ratio giving the best focusing focusing intensity (ni ) and the normalized focusing width (nw ).
value was adjusted as a constant value for 9.9 μm particles. The The focusing width is identified as the span of the focusing
concentration value of 3.3 μm particles varied between 0.1% and line along the channel width. This is the width of the path
2% at flow rates of 0.3, 0.5, and 0.7 ml/min. taken by the particles concentrating under the effect of inertial
The fourth experiment, a mini-experimental set consisting of and Dean drag forces. In the equation, ap is the particle diame-
18 tests, was conducted including particles with a diameter of ter, Wch is the channel width, and WF is the focusing width of
1.1 μm. In this set, 1.1 μm particles at different concentrations were the particle (see Fig. 3). The dispersed particles do not count in
sent through the microchannel along with mixtures of 9.9 and the focus; however, their effect on the focus quality is reflected
 
3.3 μm particles in the combination of 1.5% and 1.4% volumetric in the second part of the focus quality equation with the
Ap
AT
concentrations, respectively. The last experimental set investigated
the particle–particle interaction behavior in microchannels within ratio. The focusing behavior improved as this span’s width
the scope of three particle interaction. The details of the conducted decreased. The ideal situation arose when the width was equal-
experiments are tabulated in Table IV. ized to the diameter of the particle used, resulting in the situa-
tion where the particles were sequentially focused along a line.
The normalized focusing width is defined in the following equa-
3. The post-processing of the experiments tion:
The results were obtained by performing time-lapse fluores-  
cence microscopy in which two images of the last loop of the WF  ap
nw ¼ 1 : (15)
microchannel were taken at every flow rate using an epifluorescence Wch  ap
Olympus EX51 microscope. The analyses of the taken images were
performed using open source software (ImageJ), which was dedi- In the second part of the post-processing of the experiments,
cated to scientific processing of images. After uploading the desired Ap is the gray area within the focus width and AT is the sum of all

FIG. 3. Intensity profile and distance from the inner wall of 9.9 μm particles at the flow rate of 0.7 ml/min with a volumetric concentration of 2% through the curved micro-
channel during steady-state flow regime.

J. Appl. Phys. 132, 224703 (2022); doi: 10.1063/5.0117224 132, 224703-9


Published under an exclusive license by AIP Publishing
Journal of ARTICLE scitation.org/journal/jap
Applied Physics

gray areas within the channel width as shown in Fig. 3. The equal to the particle diameter and form a particle-chain. On the
assumption here is that the width of the microchannel and the par- contrary, while the width of the particle focusing line increases rela-
ticle diameter is constant. The ratio of the area under the peak’s tive to the diameter of the focused particle, the normalized focus
span to the total area under the intensity plot is the normalized quality decreases. When focusing is achieved on a narrow line for
focusing intensity, which is a measure for the particles extending all the desired particles, the particles are easily separated from the
along the focusing location. The equation of this measurement mixture without the aid of any additional force and can be custom-
based on the normalized focusing intensity is expressed in the ized with respect to the focusing width and the location of the
following equation: focusing line. Focusing quality, therefore, represents the ability of
  the particle separation and enrichment process.
Ap Forward integration was utilized in computing these areas.
ni ¼ : (16)
AT The multiplication of nw by ni gives the focusing quality Q as
indicated in Eq. (17).55 Figure 3 was created to demonstrate a
The ideal focusing quality is 1, and this can be achieved when typical intensity profile, where the channel width and the height
all the particles used are lined up in a single focal line of width were 50 μm, the particle diameter was 9.9 μm, and the flow rate

FIG. 4. The behavior of 2% volumetric concentration of 9.9 μm particles at 0.5 ml/min flow rate: (a) in microchannel A, (b) in microchannel B, (c) in microchannel C [for
subfigures of (a), (b), and (c), the top left shows the experimental focusing line, the bottom left shows the intensity profile of the particles, and the top right and the bottom
right show the CFD contours for the particulate phase from the longitudinal and transverse midplanes of the microchannel, respectively], and (d) 2% volumetric concentra-
tion of 9.9 μm particles at various flow rates ranging from 0.1 to 1 ml/min for all microchannels.

J. Appl. Phys. 132, 224703 (2022); doi: 10.1063/5.0117224 132, 224703-10


Published under an exclusive license by AIP Publishing
Journal of ARTICLE scitation.org/journal/jap
Applied Physics

was 0.7 ml/min with a volumetric particle concentration ratio channel. Figure 4(d) shows the relationship between normalized
of 2%, quality and the flow rate with respect to all channels. The trend
   showed that the focusing quality was around 0.85 for the channel
WF  ap Ap C. Similar to the results shown in Fig. 4, the experiments were
Q ¼ ni nw ¼ 1 : (17)
Wch  ap AT repeated for 9.9 μm particles with 5% volumetric concentration
keeping remaining parameters the same, and the results are dis-
played in Figs. 5(a)–5(c). In this case, the trend showed that the
IV. RESULTS focusing quality was increased to approximately 0.92 for the
In the first set, the focusing was accomplished in all channels channel C as shown in Fig. 5(d).
with calculated quality differences as demonstrated in Figs. 4(d) Since the differences between the focusing quality values were
and 5(d). In Figs. 4(a)–4(c), the relationship between normalized trivial for channel C, it was thought to be the most effective
quality and focusing location is shown along with the comparisons channel design among all. This idea also is consistent with the liter-
of the CFD results with experiments obtained at a flow rate of ature, and it empowers that as the curvature angle increases, the
0.5 ml/min for 9.9 μm particles with 2% volumetric concentration. focusing quality of the particles also increases.90 Although the
Concentration maps are shown on the transverse and longitudinal angle of curvature was not stated in any formula, it could be attrib-
planes of the channel and reveal particle distributions inside the uted to the change in the flow characteristics and the occurrence of

FIG. 5. The behavior of 5% volumetric concentration of 9.9 μm particles at 0.5 ml/min flow rate: (a) in microchannel A, (b) in microchannel B, (c) in microchannel C [for
subfigures of (a), (b), and (c), the top left shows the experimental focusing line, the bottom left shows the intensity profile of the particles, and the top right and the bottom
right show the CFD contours for particulate phase from the longitudinal and transverse midplanes of the microchannel, respectively], and (d) 5% volumetric concentration
of 9.9 μm particles at various flow rates ranging from 0.1 to 1 ml/min for all microchannels.

J. Appl. Phys. 132, 224703 (2022); doi: 10.1063/5.0117224 132, 224703-11


Published under an exclusive license by AIP Publishing
Journal of ARTICLE scitation.org/journal/jap
Applied Physics

vortices affecting the focusing criteria. While the velocity profile in summary, the channel focusing performance diminished from
a plain channel is symmetrical and parabolic, adding a curvature to channels C to A as the curvature angle of the channel decreased.
the channel is responsible for the change of the parabolic shape. In the second set, mixtures consisting of 9.9 and 3.3 μm parti-
Likewise, when the angle of curvature increases, the channel shape cles were sent through channel C at the same volumetric concen-
becomes more dominant and changes the focusing intensity by the trations or the same mixture ratios, as shown in Figs. 6(a)–6(c). As
Dean vortices, causing shifts in the focusing position of the parti- in previous investigations, Figs. 6(a)–6(c) show normalized focus-
cles. This shift, thus, increases the focusing intensity of the parti- ing locations and comparisons of CFD simulations with corre-
cles. Figure 5(d) shows that channel C provided the best focusing sponding experiments indicating particulate phase contours of
quality at flow rates in the range of 0.1– 1 ml/min. 9.9 μm sized particles. For both particle sizes with a volumetric
In addition, channel C minimized the effect of the flow rate concentration of 1.5%, the focusing quality of 9.9 μm sized particles
on the particle focusing showing a negligible change in the quality was found around 0.58 at the flow rates of 0.3, 0.5, and 0.7 ml/min.
values. The CFD analyses confirmed the experimental findings On the other hand, this value was around 0.85 for channel C at the
such that the highest particulate phase distributions altered from same flow rates with the volumetric concentration of 2% for 9.9 μm
channels A to C as shown in Figs. 4(a)–4(c) and 5(a)–5(c). In particles when they were sent through the channel separately. The

FIG. 6. The behavior of 1.5% volumetric concentration of 9.9 and 3.3 μm particles in microchannel C: (a) at the flow rate of 0.3 ml/min, (b) at the flow rate of 0.5 ml/min,
(c) at the flow rate of 0.7 ml/min [for subfigures of (a), (b), and (c), the top left shows the experimental focusing line, the bottom left shows the intensity profile of the parti-
cles, and the top right and the bottom right show the CFD contours for the particulate phase from the longitudinal and transvers midplanes of the microchannel, respec-
tively], and (d) the relation between the particle concentration and the focusing quality at the flow rates of 0.3, 0.5, and 0.7 ml/min for microchannel C.

J. Appl. Phys. 132, 224703 (2022); doi: 10.1063/5.0117224 132, 224703-12


Published under an exclusive license by AIP Publishing
Journal of ARTICLE scitation.org/journal/jap
Applied Physics

decrement of the focusing quality could be expressed by the in Fig. 6(d), focusing quality values for both particle sizes were
increasing number of smaller particles preventing 9.9 μm particles observed to accumulate between 0.55 and 0.65 at all flow rates.
from being focused in the channel cross section. Since the small In the third set, volumetric concentrations of 3.3 μm particles
particles with the size of 3.3 μm did not meet the focusing criteria were changed between 0.1% and 2% at the flow rates of 0.3, 0.5,
of Rep and λ, they dispersed along the channel and reduced the and 0.7 ml/min, and the volumetric concentration of 9.9 μm parti-
focusing intensity of 9.9 μm particles. The focusing ability of the cles was kept constant at 1.5%. The first finding was that increasing
particles sent at a volumetric concentration of 1.5% was higher the concentration ratio to a certain limit did not disturb the focus-
than the other mixing ratios. This is associated with an optimum ing quality of 9.9 μm particles as shown in Fig. 7(d). As the concen-
interaction rate of the particles at a given specific volumetric con- tration of 3.3 μm particles approached that of 9.9 μm particles,
centration of both particles. Moreover, having a similar volumetric there was a decrease in the focusing potential of 9.9 μm particles
ratio of interacting particles revealed an identical focusing quality due to the increasing interaction between different sized particles.
at certain flow rates as shown in Figs. 6(a)–6(c). The experimental After this decrement, the focusing quality of the 9.9 μm particles
findings in the second set were supported by the CFD simulations increased again as the interaction settled down. Although the
that the highest particulate phase distribution areas have enlarged highest focusing quality was found at 0.2% volumetric concentra-
compared to the first experimental set in Figs. 4(a)–4(c). As shown tion of 3.3 μm particles, the optimum focusing was reached at the

FIG. 7. The behavior of 1.5% volumetric concentration of 9.9 μm particles and 1.4% volumetric concentration of 3.3 μm particles in microchannel C: (a) at the flow rate of
0.3 ml/min, (b) at the flow rate of 0.5 ml/min, (c) at the flow rate of 0.7 ml/min [for subfigures of (a), (b), and (c), the top left shows the experimental focusing line, bottom
left shows the intensity profile of the particles, and the top right and the bottom right show the CFD contours for the particulate phase from the longitudinal and transverse
midplanes of the microchannel, respectively], and (d) the relation between the 3.3 μm particle concentration change and the 9.9 μm focusing quality at the flow rates of
0.3, 0.5, and 0.7 ml/min.

J. Appl. Phys. 132, 224703 (2022); doi: 10.1063/5.0117224 132, 224703-13


Published under an exclusive license by AIP Publishing
Journal of ARTICLE scitation.org/journal/jap
Applied Physics

composition of concentrations of 1.5% and 1.4% for 9.9 and 3.3 μm number of the particles involved, for instance, by adding 1.1 μm
particles, respectively. Particles persist the stable focusing quality triggered the possibility of the interaction between particles. These
independence from the flow rate, and the particles achieved the particles were responsible for increased collisions between the parti-
optimum focusing when the volumetric concentration of 3.3 μm cles and, thus, made the particles instable in the channel flow. In
particles was 1.4%. CFD analyses at the flow rates of 0.3, 0.5, and the third experimental set, the focusing quality of 9.9 μm particles
0.7 ml/min indicating particulate phase distributions of 9.9 μm par- was around 0.65, where a mixture consisting of 9.9 and 3.3 μm par-
ticles are shown in Figs. 7(a)–7(c). It was concluded that there was ticles was sent at the flow rates of 0.3, 0.5, and 0.7 ml/min with vol-
a certain critical concentration ratio for each channel design and umetric concentrations of 1.5% and 1.4%, respectively. This value,
the particle size. Once the particles reached this threshold composi- on the other hand, decreased to almost 0.6 by the addition of
tion, they tended to recover again. 1.1 μm particles at a volumetric concentration of 1.4% as shown in
In the fourth and last experimental set, it was observed that Fig. 8(d). The degradation can be explained by the focusing criteria
the quality of focusing was decreased with increasing number of that the 3.3 and 1.1 μm particles did not satisfy the critical thresh-
involved particles. This was due to the push–pull principle formed old values for Rep and λ. These particles dispersed along the
by the interaction between particles. In other words, increasing the channel since they did not meet the requirements for focusing.

FIG. 8. The behavior of 1.5% volumetric concentration of 9.9 μm particles and 1.4% volumetric concentration of 3.3 μm and 1.1 μm particles in microchannel C: (a) at the
flow rate of 0.3 ml/min, (b) at the flow rate of 0.5 ml/min, (c) at the flow rate of 0.7 ml/min [for subfigures (a), (b), and (c), the top left shows the experimental focusing line,
the bottom left shows the intensity profile of the particles, the top right and the bottom right show the CFD contours for the particulate phase from the longitudinal and
transverse midplanes of the microchannel, respectively], and (d) the effects of the addition of 1.1 μm particles at various concentrations to mixture of 9.9 and 3.3 μm parti-
cles at different flow rates.

J. Appl. Phys. 132, 224703 (2022); doi: 10.1063/5.0117224 132, 224703-14


Published under an exclusive license by AIP Publishing
Journal of ARTICLE scitation.org/journal/jap
Applied Physics

Therefore, equilibrium force acting on 9.9 μm particles did not • The degree of curvature increased the focus quality, and more
remain the same, resulting in non-homogenous focusing in the efficient decomposition was achieved. This revealed the necessity
channel flow, and the particles chose to focus on different lines of a formula to integrate the curvature angle to the channel
rather than lining up like a string as confirmed by particulate phase design phase. If achieved, microchannels would be fully under
contours for 9.9 μm particles shown in Figs. 8(a)–8(c). control in terms of focusing ability.
• The CFD studies also suggest that an appropriate multiphase
V. DISCUSSION CFD model could be harnessed to predict a flow field and a
channel’s focusing potential.
The purpose of this study was to examine the effects of the
particle–particle interaction in detail and to show the importance
of the channel curvature angle in the focusing quality of inertial
ACKNOWLEDGMENTS
microfluidics with CFD modeling. The first finding of the study
was that the gradual increase in the channel curvature angle We acknowledge the financial support of the Scientific and
improved the focusing quality of the particles. Although the curva- Technological Research Council of Turkey (TUBITAK) through the
ture radii of the channels were the same, in other words, the Dean project “Investigation on the Effects of Mechanical Forces on
drag forces acting on the particles were the same, the focusing posi- Endothelial Cells Behavior by Using Microfluidics Systems in the
tions of the particles within the channel cross section changed at domain of Mechanobiology” under Grant No. 114R037 and partial
different flow rates. The change could be addressed by both posi- financial support by the Scientific Research Projects Department
tional shift of the force acting on the particles, i.e., the Dean drag (BAP) of Istanbul Technical University through a research project
and net inertial forces, and the velocity profile altering with the under Project No. 41825.
increment in the channel curvature angle. The flow rate effect was
minimized as the curvature angle was increased, so channel C was AUTHOR DECLARATIONS
chosen because it provided independence from the flow rate. In the
case of FFLD  1, as the velocity increases, CL decreases below 0.5, Conflict of Interest
and the focusing line of the particles approached the channel The authors have no conflicts to disclose.
center. This finding was also consistent with the studies in the liter-
ature—280 degree symmetrical channels (similar to channel C).90 Author Contributions
The last finding of the first experimental set was that the effect of
the particle concentration on particle behavior was so low, which D. Ince: Investigation (equal); Validation (equal); Writing – origi-
was considered benign. In the second experimental set, it was nal draft (equal). H. Turhan: Visualization (equal). S. Cadirci:
observed that the quality of focusing was remarkably reduced by Validation (equal); Visualization (equal); Writing – original draft
the addition of 3.3 μm particles to the 9.9 μm particles as a (lead); Writing – review & editing (equal). L. Trabzon:
mixture. λ and Re p values were responsible for the decrement of Conceptualization (equal); Funding acquisition (lead); Investigation
the focusing quality since these criteria limit the focusing ability of (equal); Methodology (equal); Project administration (lead);
the particles. Even though the focusing ability of the particles was Supervision (lead); Validation (equal); Visualization (equal);
deteriorated, the particle separation would be still possible as long Writing – original draft (lead); Writing – review & editing (lead).
as proper concentration values were set.
In the third experimental set, 3.3 μm particles with different DATA AVAILABILITY
concentration values were added to 9.9 μm particles, where the con- The data that support the findings of this study are available
centration of 9.9 μm particles was kept constant. The result showed from the corresponding author upon reasonable request.
that the change in the focusing positions of the particles was
resembling a parabola. Since 3.3 μm particles created a virtual wall,
by natural migration, they prevented 9.9 μm particles from being REFERENCES
focused. Therefore, the concentration values of 3.3 μm particles, in 1
D. Mark, S. Haeberle, G. Roth, F. von Stetten, and R. Zengerle, “Microfluidic
other words small particles, revealed their matter on the inertial lab-on-a-chip platforms: Requirements, characteristics and applications,” Chem.
focusing performance. In the fourth experimental set, the aim of Soc. Rev. 39(3), 1153–1182 (2010).
the set was to strengthen inertial focusing theory. The addition of
2
J. H. Myung and S. Hong, “Microfluidic devices to enrich and isolate circulating
tumor cells,” Lab Chip 15, 4500–4511 (2015).
1.1 μm particles to the 3.3 and 9.9 μm mixture heavily degraded the 3
P. Sajeesh and A. K. Sen, “Particle separation and sorting in microfluidic
focusing quality of the 9.9 μm particles. This tendency confirmed
devices: A review,” Microfluid. Nanofluid. 17(1), 1–52 (2014).
inertial focusing theory. 4
J. Hansson, J. M. Karlsson, T. Haraldsson, H. Brismar, W. Van Der Wijngaart,
To summarize the results, and A. Russom, “Inertial microfluidics in parallel channels for high-throughput
applications,” Lab Chip 12, 4644–4650 (2012).
• A single focus line was observed on the particles meeting the 5
G. M. Whitesides, “The origins and the future of microfluidics,” Nature 442,
focusing criteria. 368–373 (2006).
• Particles below the focusing criteria impaired the focusing 6
R. Gorkin et al., “Centrifugal microfluidics for biomedical applications,” Lab
quality of the particles above the focusing criteria. On the other Chip 10(14), 1758–1773 (2010).
hand, the separation of the particles in a high yield would be still 7
X. J. Li and Y. Zhou, Microfluidic Devices for Biomedical Applications (Elsevier,
possible if the appropriate concentration rate was achieved. 2013).

J. Appl. Phys. 132, 224703 (2022); doi: 10.1063/5.0117224 132, 224703-15


Published under an exclusive license by AIP Publishing
Journal of ARTICLE scitation.org/journal/jap
Applied Physics

8
H. N. Chan, M. J. A. Tan, and H. Wu, “Point-of-care testing: Applications of 32
N. A. Rahman, F. Ibrahim, and B. Yafouz, “Dielectrophoresis for biomedical
3D printing,” Lab Chip 17(16), 2713–2739 (2017). sciences applications: A review,” Sensors 17(3), 449 (2017).
9
S. Smith et al., “CD-based microfluidics for primary care in extreme 33
C. J. Ramirez-Murillo, J. M. de los Santos-Ramirez, and V. H. Perez-Gonzalez,
point-of-care settings,” Micromachines 7(2), 22 (2016). “Toward low-voltage dielectrophoresis-based microfluidic systems: A review,”
10
M. Tang, G. Wang, S.-K. Kong, and H.-P. Ho, “A review of biomedical centrif- Electrophoresis 42(5), 565–587 (2021).
ugal microfluidic platforms,” Micromachines 7(2), 26 (2016). 34
R. A. Rica and M. Z. Bazant, “Electrodiffusiophoresis: Particle motion in elec-
11
T.-F. Wu, S. H. Cho, Y.-J. Chiu, and Y.-H. Lo, “Lab-on-a-chip device and trolytes under direct current,” Phys. Fluids 22, 112109 (2010).
system for point-of-care applications,” in Handbook of Photonics for Biomedical 35
B. Sarno, D. Heineck, M. J. Heller, and S. D. Ibsen, “Dielectrophoresis develop-
Engineering, edited by A. H.-P. Ho, D. Kim, and M. G. Somekh (Springer ments and applications from 2010 to 2020,” Electrophoresis 42, 539–564 (2021).
36
Netherlands, Dordrecht, 2021), pp. 1–30. Y. Wang, D. I. Adeoye, E. O. Ogunkunle, I. A. Wei, R. T. Filla, and
12
L. Mazutis, J. Gilbert, W. L. Ung, D. A. Weitz, A. D. Griffiths, and M. G. Roper, “Affinity capillary electrophoresis: A critical review of the literature
J. A. Heyman, “Single-cell analysis and sorting using droplet-based microflui- from 2018 to 2020,” Anal. Chem. 93, 295–310 (2021).
dics,” Nat. Protoc. 8, 870–891 (2013). 37
A. Wuethrich and J. P. Quirino, “A decade of microchip electrophoresis for
13
C. Wyatt Shields Iv, C. D. Reyes, and G. P. López, “Microfluidic cell sorting: A clinical diagnostics—A review of 2008–2017,” Anal. Chim. Acta 1045, 42–66
review of the advances in the separation of cells from debulking to rare cell isola- (2019).
38
tion,” Lab Chip 15, 1230–1249 (2015). B. Zhu, W. Li, M. Zhu, P. L. Hsu, L. Sun, and H. Yang,
14
D. R. Gossett et al., “Label-free cell separation and sorting in microfluidic “Dielectrophoresis-based method for measuring the multiangle mechanical prop-
systems,” Anal. Bioanal. Chem. 397(8), 3249–3267 (2010). erties of biological cells,” Biomed. Res. Int. 2020, 1–9 (2020).
15
M. E. Warkiani et al., “Slanted spiral microfluidics for the ultra-fast, label-free 39
P. Augustsson, C. Magnusson, M. Nordin, H. Lilja, and T. Laurell,
isolation of circulating tumor cells,” Lab Chip 14, 128–137 (2014). “Microfluidic, label-free enrichment of prostate cancer cells in blood based on
16
S. Nagrath et al., “Isolation of rare circulating tumour cells in cancer patients acoustophoresis,” Anal. Chem. 84, 7954–7962 (2012).
by microchip technology,” Nature 450(7173), 1235–1239 (2007). 40
J. Shi, H. Huang, Z. Stratton, Y. Huang, and T. J. Huang, “Continuous particle
17
C. M. Court, J. S. Ankeny, S. Hou, H. R. Tseng, and J. S. Tomlinson, “Improving separation in a microfluidic channel via standing surface acoustic waves
pancreatic cancer diagnosis using circulating tumor cells: Prospects for staging and (SSAW),” Lab Chip 9(23), 3354–3359 (2009).
single-cell analysis,” Expert Rev. Mol. Diagn. 15(11), 1491–1504 (2015). 41
A. Lenshof and T. Laurell, “Emerging clinical applications of microchip-based
18
A. A. S. Bhagat, H. W. Hou, L. D. Li, C. T. Lim, and J. Han, “Pinched flow acoustophoresis,” J. Lab. Autom. 16(6), 443–449 (2011).
coupled shear-modulated inertial microfluidics for high-throughput rare blood 42
D. Mark, S. Haeberle, G. Roth, F. Von Stetten, and R. Zengerle, “Microfluidic
cell separation,” Lab Chip 11, 1870–1878 (2011). lab-on-a-chip platforms: Requirements, characteristics and applications,” NATO
19
P. K. Chaudhuri, M. Ebrahimi Warkiani, T. Jing, Kenry, and C. T. Lim, Sci. Peace Security Ser. A Chem. Biol. 39, 1153–1182 (2010).
“Microfluidics for research and applications in oncology,” Analyst 141, 504–524 (2016). 43
Z. Ma, D. J. Collins, and Y. Ai, “Detachable acoustofluidic system for particle sepa-
20
M. Zuvin, N. Mansur, S. Birol, L. Trabzon, and A. Sayi-Yazgan, “Human ration via a traveling surface acoustic wave,” Anal. Chem. 88(10), 5316–5323 (2016).
breast cancer cell enrichment by Dean flow driven microfluidic channels,” 44
K. Dholakia, M. P. MacDonald, P. Zemánek, and T. Čižmár, “Cellular and col-
Microsyst. Technol. 22, 645–652 (2015). loidal separation using optical forces,” Methods Cell Biol. 82, 467–495 (2007).
21
E. Altinagac, H. Kizil, and L. Trabzon, “Biological particle manipulation: An 45
A. Sakaue-Sawano et al., “Genetically encoded tools for optical dissection of
example of Jurkat enrichment,” Micro Nano Lett. 10(10), 25–27 (2015). the mammalian cell cycle,” Mol. Cell 68(3), 626–640 (2017).
22
A. H. C. Ng, K. Choi, R. P. Luoma, J. M. Robinson, and A. R. Wheeler, 46
A. Jonáš and P. Zemánek, “Light at work: The use of optical forces for particle
“Digital microfluidic magnetic separation for particle-based immunoassays,” manipulation, sorting, and analysis,” Electrophoresis 29, 4813–4851 (2008).
Anal. Chem. 84(20), 8805–8812 (2012). 47
A. S. Basu and Y. B. Gianchandani, “Shaping high-speed Marangoni flow in
23
B. D. Plouffe, L. H. Lewis, and S. K. Murthy, “Computational design optimiza- liquid films by microscale perturbations in surface temperature,” Appl. Phys.
tion for microfluidic magnetophoresis,” Biomicrofluidics 5, 013413 (2011). Lett. 90, 034102 (2007).
24
S. Yan, J. Zhang, D. Yuan, and W. Li, “Hybrid microfluidics combined with 48
H. Jeong, J. van Tiem, Y. B. Gianchandani, and J. Park, “Nano-particle separa-
active and passive approaches for continuous cell separation,” Electrophoresis tion using Marangoni flow in evaporating droplets,” in Solid-State Sensors,
38, 238–249 (2017). Actuators, and Microsystems Workshop, Hilton Head Island, SC, 8–12 June
25
J. H. Kang, S. Krause, H. Tobin, A. Mammoto, M. Kanapathipillai, and 2014, pp. 223–226.
D. E. Ingber, “A combined micromagnetic-microfluidic device for rapid capture 49
D. R. Tree, T. Iwama, K. T. Delaney, J. Lee, and G. H. Fredrickson,
and culture of rare circulating tumor cells,” Lab Chip 12, 2175–2181 (2012). “Marangoni flows during nonsolvent induced phase separation,” ACS Macro
26
H. Watarai, M. Suwa, and Y. Iiguni, “Magnetophoresis and electromagnetophore- Lett. 7(5), 582–586 (2018).
sis of microparticles in liquids,” Anal. Bioanal. Chem. 387, 1693–1699 (2004). 50
B. Kim, J. K. Lee, and S. Choi, “Continuous sorting and washing of cancer
27
J. Zeng, Y. Deng, P. Vedantam, T. R. Tzeng, and X. Xuan, “Magnetic separa- cells from blood cells by hydrophoresis,” Biochip J. 10, 81–87 (2016).
tion of particles and cells in ferrofluid flow through a straight microchannel 51
S. Yan et al., “Development of a novel magnetophoresis-assisted hydrophoresis
using two offset magnets,” J. Magn. Magn. Mater. 346, 118–123 (2013). microdevice for rapid particle ordering,” Biomed. Microdevices 18(4), 54
28
R. Zhou and C. Wang, “Microfluidic separation of magnetic particles with soft (2016).
magnetic microstructures,” Microfluid. Nanofluid. 20, 1–11 (2016). 52
J. M. Martel and M. Toner, “Inertial focusing in microfluidics,” Ann. Rev.
29
A. Aghilinejad, M. Aghaamoo, and X. Chen, “Numerical study of joule Biomed. Eng. 16, 371–396 (2014).
heating effect on dielectrophoresis-based circulating tumor cell separation,” in 53
J. Zhang et al., “Fundamentals and applications of inertial microfluidics: A
ASME International Mechanical Engineering Congress and Exposition, review,” Lab Chip 16, 10–34 (2016).
Proceedings (IMECE), Tampa, FL, 3-9 November 2017 (ASME, 2017). 54
D. Di Carlo, J. F. Edd, D. Irimia, R. G. Tompkins, and M. Toner, “Equilibrium
30
A. Aghilinejad, M. Aghaamoo, X. Chen, and J. Xu, “Effects of electrothermal separation and filtration of particles using differential inertial focusing,” Anal.
vortices on insulator-based dielectrophoresis for circulating tumor cell separa- Chem. 80(6), 2204–2211 (2008).
tion,” Electrophoresis 39, 869–877 (2018). 55
U. Sonmez, S. Jaber, and L. Trabzon, “Super-enhanced particle focusing in a
31
H. Hadady, D. Redelman, S. R. Hiibel, and E. J. Geiger, “Continuous-flow novel microchannel geometry using inertial microfluidics,” J. Micromech.
sorting of microalgae cells based on lipid content by high frequency dielectro- Microeng. 27, 065003 (2017).
phoresis,” AIMS Biophys. 3(3), 398–414 (2016). 56
D. Di Carlo, “Inertial microfluidics,” Lab Chip 9, 3038–3046 (2009).

J. Appl. Phys. 132, 224703 (2022); doi: 10.1063/5.0117224 132, 224703-16


Published under an exclusive license by AIP Publishing
Journal of ARTICLE scitation.org/journal/jap
Applied Physics

57
I. Kobayashi, T. Takano, R. Maeda, Y. Wada, K. Uemura, and M. Nakajima, 73
A. A. S. Bhagat, S. S. Kuntaegowdanahalli, and I. Papautsky, “Enhanced parti-
“Straight-through microchannel devices for generating monodisperse emulsion cle filtration in straight microchannels using shear-modulated inertial migra-
droplets several microns in size,” Microfluid. Nanofluid. 4, 167–177 (2008). tion,” Phys. Fluids 20, 101702 (2008).
58
C. Liu, C. Xue, X. Chen, L. Shan, Y. Tian, and G. Hu, “Size-based separation 74
A. A. S. Bhagat, S. S. Kuntaegowdanahalli, and I. Papautsky, “Continuous par-
of particles and cells utilizing viscoelastic effects in straight microchannels,” ticle separation in spiral microchannels using Dean flows and differential migra-
Anal. Chem. 87(12), 6041–6048 (2015). tion,” Lab Chip 8, 1906–1914 (2008).
59
S. Ookawara, D. Street, and K. Ogawa, “Numerical study on development of 75
T. T. Chandratilleke, N. Nadim, and R. Narayanaswamy, “Analysis of second-
particle concentration profiles in a curved microchannel,” Chem. Eng. Sci. 61, ary flow instability and forced convection in fluid flow through rectangular and
3714–3724 (2006). elliptical curved ducts,” Heat Transf. Eng. 34, 1237–1248 (2013).
60
J. M. Martel and M. Toner, “Particle focusing in curved microfluidic chan- 76
S. Cadirci, D. Ince, I. Ghanem, S. Z. Birol, L. Trabzon, and H. Turhan,
nels,” Sci. Rep. 3, 3340 (2013). “Experimental and numerical investigation on particle–particle interaction of
61
N. Xiang, K. Chen, Q. Dai, D. Jiang, D. Sun, and Z. Ni, “Inertia-induced multi-particle separation in an alternating and repetitive microchannel,”
focusing dynamics of microparticles throughout a curved microfluidic channel,” Microsyst. Technol. 25, 307–318 (2018).
Microfluid. Nanofluid. 18, 29–39 (2014). 77
G. Segré and A. Silberberg, “Radial particle displacements in Poiseuille flow of
62
A. Russom, A. K. Gupta, S. Nagrath, D. Di Carlo, J. F. Edd, and M. Toner, suspensions,” Nature 189, 209–210 (1961).
“Differential inertial focusing of particles in curved low-aspect-ratio microchan- 78
P. G. Saffman, “The lift on a small sphere in a slow shear flow,” J. Fluid Mech.
nels,” New J. Phys. 11, 075025 (2009). 22(2), 385–400 (1965).
63
J. Sun et al., “Size-based hydrodynamic rare tumor cell separation in curved 79
B. P. Ho and L. G. Leal, “Inertial migration of rigid spheres in two-
microfluidic channels,” Biomicrofluidics 7(1), 011802 (2013). dimensional unidirectional flows,” J. Fluid Mech. 65(2), 365–400
64
N. Nivedita, P. M. Ligrani, and I. Papautsky, “Evolution of secondary Dean (1974).
vortices in spiral microchannels for cell separations,” in 17th International 80
P. Sajeesh and A. K. Sen, “Particle separation and sorting in microfluidic
Conference on Miniaturized Systems for Chemistry and Life Sciences, Freiburg, devices: A review,” Microfluid. Nanofluid. 17, 1–52 (2014).
Germany, 27–31 October 2013, pp. 639–641. 81
H. Fallahi, J. Zhang, H.-P. Phan, and N.-T. Nguyen, “Flexible microfluidics:
65
N. Nivedita, J. Zhou, and I. Papautsky, “Spiral inertial microfluidic devices for Fundamentals, recent developments, and applications,” Micromachines 10(12),
cell separations,” in Encyclopedia of Microfluidics and Nanofluidics, edited by 830 (2019).
D. Li (Springer, New York, 2015), pp. 3058–3067. 82
J. M. Martel and M. Toner, “Inertial focusing in microfluidics,” Ann. Rev.
66
J. M. Martel and M. Toner, “Inertial focusing dynamics in spiral microchan- Biomed. Eng. 16, 371–396 (2014).
nels,” Phys. Fluids 24(3), 032001 (2012). 83
W. R. Dean, “Fluid motion in a curved channel,” Proc. R. Soc. A Math. Phys.
67
N. Nivedita and I. Papautsky, “Continuous separation of blood cells in spiral Eng. Sci. 121(787), 402–420 (1928).
microfluidic devices,” Biomicrofluidics 7(5), 054101 (2013). 84
Q. Zhao, D. Yuan, J. Zhang, and W. Li, “A review of secondary flow in inertial
68
J. Zhu, R. C. Canter, G. Keten, P. Vedantam, T. R. J. Tzeng, and X. Xuan, microfluidics,” Micromachines 11, 461 (2020).
“Continuous-flow particle and cell separations in a serpentine microchannel via 85
S. Ookawara, R. Higashi, D. Street, and K. Ogawa, “Feasibility study on con-
curvature-induced dielectrophoresis,” Microfluid. Nanofluidics 11, 743–752 (2011). centration of slurry and classification of contained particles by microchannel,”
69
X. Chen, T. Li, H. Zeng, Z. Hu, and B. Fu, “Numerical and experimental inves- Chem. Eng. J. 101, 171–178 (2004).
86
tigation on micromixers with serpentine microchannels,” Int. J. Heat Mass S. Razavi Bazaz, A. Mashhadian, A. Ehsani, S. C. Saha, T. Krüger, and M. Ebrahimi
Transf. 98, 131–140 (2016). Warkiani, “Computational inertial microfluidics: A review,” Lab Chip 20, 1023–1048
70
J. Zhang, W. Li, M. Li, G. Alici, and N. T. Nguyen, “Particle inertial focusing (2020).
and its mechanism in a serpentine microchannel,” Microfluid. Nanofluid. 17, 87
J. De Jong, R. G. H. Lammertink, and M. Wessling, “Membranes and micro-
305–316 (2014). fluidics: A review,” Lab Chip 6(9), 1125–1139 (2006).
71
J. Zhang, S. Yan, R. Sluyter, W. Li, G. Alici, and N. T. Nguyen, “Inertial parti- 88
H. Amini, W. Lee, and D. Di Carlo, “Inertial microfluidic physics,” Lab Chip
cle separation by differential equilibrium positions in a symmetrical serpentine 14(15), 2739–2761 (2014).
micro-channel,” Sci. Rep. 4, 4527 (2014). 89
Inc.: ANSYS, “ANSYS FLUENT Theory Guide,” Release 15, 2013.
72
J. Wang, J. Wang, L. Feng, and T. Lin, “Fluid mixing in droplet-based micro- 90
A. Ozbey et al., “Inertial focusing of cancer cell lines in curvilinear microchan-
fluidics with a serpentine microchannel,” RSC Adv. 5, 104138–104144 (2015). nels,” Micro Nano Eng. 2, 53–63 (2019).

J. Appl. Phys. 132, 224703 (2022); doi: 10.1063/5.0117224 132, 224703-17


Published under an exclusive license by AIP Publishing

View publication stats

You might also like