Chemical Engineering Journal 327 (2017) 268–277

Contents lists available at ScienceDirect

Chemical Engineering Journal

journal homepage: www.elsevier.com/locate/cej

A micromixer with two-layer serpentine crossing channels having

excellent mixing performance at low Reynolds numbers
Shakhawat Hossain a, Insu Lee a,b, Sun Min Kim a,b,⇑, Kwang-Yong Kim a,⇑
a
Department of Mechanical Engineering, Inha University, Incheon 22212, Republic of Korea
b
WCSL of Integrated Human Airway-on-a-Chip, Inha University, Incheon 22212, Republic of Korea

h i g h l i g h t s

Experimental and numerical analyses of a novel micromixer with two-layer crossing channels were performed.

The micromixer showed at least 96% mixing throughout a Reynolds number range (0.2–120).

At low Reynolds numbers (0.2–10), the micromixer showed about 99% mixing at the exit.

The proposed micromixer showed lower pressure drop than TLCCM for Re larger than 10.

a r t i c l e i n f o a b s t r a c t

Article history: A novel design is presented for a chaotic micromixer using two-layer serpentine crossing microchannels.
Received 7 May 2017 The performance of the micromixer was analyzed both numerically and experimentally. The numerical
Received in revised form 13 June 2017 analysis was performed using three-dimensional Navier-Stokes equations with a convection–diffusion
Accepted 20 June 2017
model for the species concentration in a Reynolds number range of 0.2–120. An experimental model of
Available online 21 June 2017
the micromixer was fabricated by soft lithography with polydimethylsiloxane (PDMS). Two working flu-
ids, water and dye-water mixture were used for numerical analysis except for the experimental valida-
Keywords:
tion of numerical results. Both the numerical and experimental analyses confirm that the micromixer
Mixing
Chaotic micromixer
achieves a high level of mixing over a wide range of Reynolds numbers through splitting, enlarging,
Two-layer serpentine crossing recombination, and folding mechanisms. The micromixer showed over 95% mixing throughout the tested
microchannels range of Reynolds number. Especially, about 99% mixing was achieved at Reynolds numbers less than ten.
Experimental analysis Thus, the proposed micromixer can be used in microfluidic systems which require fast mixing at low
Navier-Stokes equations Reynolds numbers.
Ó 2017 Elsevier B.V. All rights reserved.

1. Introduction Various types of micromixers have been developed to achieve

rapid and homogenous mixing. Micromixers are generally catego-
Microfluidics technology enables the miniaturization of labora- rized according to the mixing principles as active or passive mix-
tory instruments into lab-on-a-chip (LOC) devices for various ers. Active micromixers require an external energy source, unlike
chemical and biological applications [1–4]. This technology is gain- passive micromixers [7,8]. Passive micromixers are more favorable
ing attention from researchers due to the faster analysis speeds, than active micromixers due to their simple fabrication, simple
higher efficiency, and small sample volumes. Efficient mixing of integration with microfluidic systems, and greater robustness
liquid samples plays an important role in LOC [5] and micro- and stability [9,10].
scale total analysis systems [6]. Generally, fluid flow at the The performance of the passive micromixers mostly depends on
micro-scale is laminar, so mixing is governed by molecular diffu- the channel geometry. Thus, clear understanding of the flow
sion, which requires a greater channel length to achieve a homoge- phenomena within the microchannel is required to develop
neously mixed solution than turbulent diffusion. micromixers. Computational fluid dynamics (CFD) based on three-
dimensional (3D) Navier-Stokes equations is a promising technique
for developing efficient micromixers [9,10]. Mixing in passive
⇑ Corresponding author at: Department of Mechanical Engineering, Inha Univer- micromixers relies on chaotic advection as well as molecular diffu-
sity, Incheon 22212, Republic of Korea (S.M. Kim and K.Y. Kim). sion [10]. Molecular diffusion can be enhanced by enlarging the
E-mail addresses: [email protected] (S.M. Kim), [email protected] (K.-Y. Kim). contact surface between fluid samples and reducing the diffusion

http://dx.doi.org/10.1016/j.cej.2017.06.106
1385-8947/Ó 2017 Elsevier B.V. All rights reserved.
 S. Hossain et al. / Chemical Engineering Journal 327 (2017) 268–277 269

Fig. 1. Micromixer geometry with two-layer serpentine crossing microchannels and geometric parameters. (a) Schematic diagram of the top and bottom channel layers with
inverse-N and N-shaped segments, respectively. (b) 3D view of the micromixer, where both the layers are interconnected at the middle of the ‘‘X” shape and the vertical
sections. (c) Locations of cross-sectional planes A1–A9 at the nodes of the crossing structures. (d) Optical image of the micromixer with nine mixing units. (e) Optical image of
a micromixer composed of four mixing units used only for the parametric study (the dotted ellipse in the left figure is enlarged in the right figure).

length, and chaotic advection enlarges the contact surface by efficiency in the laminar flow regime. A chaotic flow structure
manipulating the bulk flow inside the microchannels [11]. can be generated by modifying the channel shape so that the lam-
Several studies reveal that micromixers based on chaotic advec- inar flow continually stretches, folds, splits, and recombines within
tion show efficient mixing in an extensive range of Reynolds num- the channel. This flow phenomenon is reported for various
ber [10,12]. Chaotic advection caused by the periodic perturbation microchannel shapes [13,14], including a staggered herringbone
of the two-dimensional flow can remarkably enhance the mixing structure [15], split and recombine structure [16], Tesla structure
270 S. Hossain et al. / Chemical Engineering Journal 327 (2017) 268–277

[17], twisted channels [18], grooved channels [19], and barriers on in repeating patterns of splitting and recombination of the fluid
the channel walls [20]. streams.
Liu et al. [21] designed a 3D serpentine microchannel to gener- For the reference micromixer geometry, the main channel
ate rapid stretching and folding of the fluid streams, which pro- width (H), diagonal channel width (w), length of the vertical seg-
duced chaotic advection. However, a relatively high Reynolds ment (b), pitch length (P), depth of the single channel (d), and
number (>25) was required to generate the chaotic advection. number of mixing units were fixed at 1.07 mm, 0.30 mm,
Stroock et al. [15] achieved a 3D twisting flow mechanism in a 0.15 mm, 0.64 mm, 0.15 mm, and 9, respectively. Thus, the depth
microchannel with diagonally oriented ridges on the bottom wall. of the micromixer (D) is equal to 2 d. The cross-sectional dimen-
They also developed staggered herringbone structure patterns at sions of the inlets and outlet are 0.15  0.3 mm and
the bottom of the microchannel, in which chaotic mixing is 0.3  0.3 mm, respectively. Figs. 1(d) and (e) show optical images
obtained by alternating velocity fields. Aubin et al. [22] observed of the proposed micromixer with nine mixing units and four mix-
counter-rotating vortices in a mixer with staggered herringbone ing units, respectively. The nine-unit mixer was considered for the
grooves (SHG) using numerical simulation. Johnson et al. [23] used flow analyses, while the four-unit mixer was considered in a para-
slanted grooves at the bottom surface of a T-shaped micromixer metric study to save computational time.
and obtained remarkably enhanced mixing by transverse fluid
transport. Kim et al. [20] proposed a micromixer with barriers on
3. Numerical analysis
slanted grooves, which generated chaotic mixing due to the regular
perturbation of the flow field. Periodically situated barriers were
Numerical analysis of the flow and mixing was accomplished
capable of creating alternating velocity fields within the helical
using the commercial CFD code ANSYS CFX 15.0Ò [27]. The finite
motion, which resulted in remarkable enhancement of chaotic
volume technique was used to solve the governing differential
mixing.
equations. The governing equations for fluid flow are the 3D steady
Hong et al. [17] developed an in-plane passive micromixer
continuity and Navier-Stokes equations:
using two-dimensional Tesla structures to create transverse dis-
Mass conservation:
persion, which produced chaotic advection and significantly
enhanced the mixing. Wang and Yang [24] obtained a chaotic flow r  ðq~
VÞ ¼ 0 ð1Þ
pattern using staggered overlapping crisscross microchannels,
which resulted from two chaotic mixing mechanisms involving Momentum conservation:
stretching and folding of the fluid streams. Kim et al. [25] devel- 1
oped a serpentine lamination micromixer (SLM) based on chaotic ð~ V ¼  rp þ mr2 ~
V  rÞ~ V ð2Þ
q
mixing mechanisms, which included F-shaped mixing segments.
The mixing performance was enhanced by merging two chaotic where V, q, and m represent the velocity, density, and kinematic vis-
mixing mechanisms. Park et al. [26] improved the mixing effi- cosity of the fluid, respectively. Numerical diffusion errors are
ciency of an SLM by geometric modification. induced by the discretization of the convection terms in the govern-
Xia et al. [12] proposed one of the most interesting passive ing equations. These errors can be minimized by introducing a high-
micromixers using a chaotic flow pattern. The micromixer con- resolution numerical scheme of second-order approximation [28].
sisted of two-layer crossing channels (TLCCM), which showed an The SIMPLEC (Semi-Implicit Method for Pressure Linked
excellent overall mixing performance in a wide range of Reynolds Equations-Consistent) algorithm [29] was used for the pressure
numbers. This micromixer created chaotic advection through split- velocity coupling.
ting, enlarging, recombination, and folding, and a high mixing The conservation of mass transport equation [30] for each fluid
index of 0.96 was achieved even at a low Reynolds number of 0.2. component with constant viscosity and density is:
In the present study, a chaotic micromixer with two-layer ser-
pentine crossing channels is proposed for further enhancement ðv  rÞC ¼ ar2 C ð3Þ
of the mixing performance at low Reynolds numbers while retain-
where a and C are the diffusivity coefficient and concentration of
ing high performance in high Reynolds number range. The fluid
the fluid component, respectively. This equation was used to model
flow characteristics and mixing phenomena were investigated
the diffusive mixing of multi-component fluids [31]. The boundary
both numerically and experimentally. The numerical analysis is
conditions were a no-slip condition at the walls, atmospheric pres-
based on 3D Navier-Stokes equations with a convection–diffusion
sure at the outlet, and uniform velocity at the inlets (in a range of
model for the species concentration. The micromixer was fabri-
0.001–0.6 ms1). Pure water and a dye-water mixture enter
cated using a soft lithography technique with PDMS for experi-
through inlet 1 and inlet 2, where the mass fractions of water are
mental analyses. The numerical results for mixing index were
zero and one, respectively. The dynamic viscosity (m) and density
validated using experimental data in a wide Reynolds number
(q) of water were measured at 25 °C as 8.8  104 kg/m-s and
range of 0.2 to 120. Superiority of the proposed micromixer
997 kg/m3, respectively [32]. The diffusion coefficient of the dye-
was proved compared to three previous micromixers including
water mixture was 1  1011 m2/s [12]. To achieve maximum accu-
TLCCM [12].
racy of the numerical solutions, the governing equations were
solved iteratively until normalized root-mean-square residual val-
ues of less than 1  106 were obtained.
2. Micromixer model
The mixing performance was evaluated by a variance-based
method. The variance of the liquid species can be calculated in a
The proposed micromixer comprises two layers of serpentine
cross-sectional plane perpendicular to the flow:
structures, as shown in Fig. 1. The top and bottom channels are
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
arranged in a periodic fashion with segments that have ‘‘inverse- u N
u1 X
N” and ‘‘N” shapes, as shown in Fig. 1(a). The main serpentine r¼t ðci  cm Þ2 ð4Þ
channels have an angle of 90° with respect to the inlets, as shown N i¼0
in Figs. 1(b) and (c). Fluids in the two-layered channels are inter-
connected through the middles of the ‘‘X” shapes and the vertical where N is the number of sampling points in the cross-sectional
sections. The successive arrangement of the mixing units results plane, and cm and ci are the optimal mass fraction and mass fraction
 S. Hossain et al. / Chemical Engineering Journal 327 (2017) 268–277 271

Fig. 2. Experimental setup.

at sampling point i, respectively. Finally, the mixing index of the

micromixer at any cross-sectional plane is defined as:
sffiffiffiffiffiffiffiffiffiffi
r2
M¼1 ð5Þ
r2max
where rmax represents the maximum standard variance over the
data range. A mixing index of zero represents completely unmixed
fluids (r = rmax), while a value of one represents completely mixed
fluids (r = 0). The Reynolds number was defined as follows:
qVD
Re ¼ ð6Þ
l
where D represents the hydraulic diameter of the inlet, and m is the
dynamic viscosity. The fluid properties were taken for water at
25 °C.

Fig. 3. Validation of the mixing indices at the exit of the micromixer with nine
4. Device fabrication and experimental procedure
mixing units (x/w = 25) using experimental data in the Reynolds number range of
0.2–120.
Micromixers were fabricated using PDMS elastomer based
micro-molding technique known as soft lithography. Briefly, a pho- wafer by photo-lithography, and a mixture of PDMS pre-polymer
toresist (SU8-2050, MicroChem) mold was fabricated on a silicon and cross-linker (with a weight ratio of 10:1) was poured onto
272 S. Hossain et al. / Chemical Engineering Journal 327 (2017) 268–277

Table 1
Tested Reynolds numbers and the corresponding flow rates.

Re 0.2 1 10 15 20 30 40 60 80 100 120

Flow rate (mL/min) 2.4 12 120 180 240 360 480 720 960 1200 1440

Fig. 4. Comparison of mixing indices at micromixer exits among the four different
chaotic micromixers in the Reynolds number range of 0.2 to 120. The axial length of
the micromixers was kept constant (x = 7.5 mm).

the mold for polymer replication. For polymerization, the polymer

was baked in a convection oven at 65 °C for 4 h. The complete mix-
ing device was composed of two separate microchannel layers,
which were bonded together by a plasma bonding method. The
bonding surfaces of the PDMS device were exposed to oxygen
plasma (CUTE-100, FemtoScience). Both plasma-treated
microchannel layers were then bonded to each other and heated
at 65 °C for 5 min to enhance the bonding.
For experimental analyses, deionized water (Direct-Q3, Milli-
pore) and 10 wt% black ink (Quink, Parker) were prepared as work-
ing fluids. The fluids were loaded into a syringe and transferred to a
micromixer through TygonÒ tubing using a syringe pump (LEGATO
270, KDScience). A flow rate range of 2.4–1440 lL/min was set for
the syringe pump to achieve a Reynolds number range of 0.2–120.
The whole mixing process in the microchannel was monitored
using a microscope system (Ti-U Eclipse, Nikon). The mixing index
was calculated using image processing software (ImageJ, NIH). The
images were converted to 8-bit format, and their pixel intensities
were measured in the region of interest (at the exit of the
micromixer). The mixing index was then calculated using Eqs. (3)
and (4).
Fig. 2 shows the experimental setup used in this work. A syringe
pump was connected to the micromixer to transfer both deionized
water and 10 wt% black ink contained deionized water to the
inlets. The device was monitored through inverted microscope
and captured images were processed using ImageJ.

5. Results and discussion

For the numerical analysis, a preliminary grid sensitivity test

was executed at a Reynolds number of 40 with four different grid Fig. 5. Developments of mixing index along the proposed micromixer for various
Reynolds numbers (Re = 0.2, 1, 15, 40, and 80). (a) Dye mass fraction distributions at
systems. The number of nodes varied from 0.48  106 to
the first five cross-sectional planes (A1–A5). (b) Mixing index distributions along the
1.56  106. The optimal grid selected for further calculations from micromixer. The mixing indices were calculated at the cross-sectional planes
the test results had 1.24 million nodes, which correspond to a located at the nodes of the crossing structures.
mesh element size of 30 lm.
Fig. 3 shows a quantitative comparison of the numerical and ing unit (x/w = 25) in the Reynolds number range from 0.2 to 120
experimental results for mixing index at the exit of the ninth mix- (Table 1). Considering the large uncertainties involved in taking
 S. Hossain et al. / Chemical Engineering Journal 327 (2017) 268–277 273

challenging, which was one of the main reasons for the geometrical
variation. Thus, the micromixer geometry used in the numerical
calculation was modified to fit the real geometry of the experimen-
tal model for the validation. However, the walls were assumed to
be smooth in the numerical calculations, which might contribute
to the differences between the numerical results and experimental
data.
Moreover, to measure experimental uncertainties in mixing
index, the mixing indices were compared between the numerical
and experimental results using a micromixer of 4 units. Mixing
index error was calculated as,
 
ðMSim  M Exp Þ

Mixing index error ¼   ð7Þ
M Sim 

MSim indicates calculated mixing index value from the simulation,

and MExp is average mixing index value obtained from repeated
experiments (N = 3). The experimental mixing index values showed
Fig. 6. Effect of number of mixing units on mixing performance in Reynolds maximum standard deviation, rmax = 0.120. And, the mixing index
number range, 0.2–120. error range was calculated as 2–13%.
The computational mixing performance of the proposed refer-
and analyzing the experimental images, the numerical mixing ence micromixer was compared quantitatively with those of three
indices show reasonable agreement with the experimental data different micromixers: TLCCM [12], a SHG micromixer [15], and a
with slight over-predictions over the whole range. The experimen- 3D serpentine micromixer [34], as shown in Fig. 3. The mixing
tal images were captured by microscope camera (DS-1QMc, 1.5 indices at the exits (x/w = 25.0) were compared using the same
Megapixel) and those raw images were analyzed after conversion axial lengths in the Reynolds number range of 0.2–120. The pro-
of 8 bit images for mixing index calculation. And, other experimen- posed micromixer and TLCCM show excellent mixing performance
tal uncertainties were also induced from operating condition and (M  0.96) compared to the other two micromixers throughout the
device fabrication process. The working solutions were transferred Reynolds number range, although the 3D serpentine micromixer
using syringe pump (±0.5% of accuracy). Deionized water (Direct- shows similar mixing indices at high Reynolds numbers
Q3, Millipore) and 10 wt% black ink (diffusion coefficient [33], (Re = 100–120). Both the proposed micromixer and TLCCM show
3.23  1010 m2 s1) were used in both the experiment and nearly complete mixing (M > 0.99) at Reynolds numbers larger
numerical calculations for validation. However, water and a dye- than 60 (Re = 60–120). However, the proposed micromixer shows
water mixture (diffusion coefficient, 1.0  1011 m2 s1) were used higher mixing indices than the TLCCM at the Reynolds numbers
for the calculations shown in Figs. 4–9. less than 10 (Re = 0.2–10), while the TLCCM shows better mixing
The dimensions of the fabricated channel differed slightly performance in the middle range (Re = 20–40). At low Reynolds
(±10 mm) from the reference dimensions, and some wall roughness numbers (Re  10), the proposed micromixer achieves nearly 99%
was introduced in the fabrication process. In the fabrication pro- mixing at the exit, whereas the SHG and 3D serpentine micromixer
cess, alignment of the top and bottom channel layers was very show around 58% mixing, and the TLCCM, which has been known

Fig. 7. Optical images of the fluid mixing in the first and last mixing units at Re = 15 and 60.
274 S. Hossain et al. / Chemical Engineering Journal 327 (2017) 268–277

Fig. 8. Velocity vectors on y-z planes at the nodes of the crossing structures at Re = 0.2.

Fig. 9. Velocity vectors at different cross-sectional planes to explain the formation of the saddle- shaped flow pattern at the node of the crossing structure at Re = 0.2.
 S. Hossain et al. / Chemical Engineering Journal 327 (2017) 268–277 275

maintain identical flow patterns because the inertia force of the

fluids is negligibly small at low Reynolds number (Re < 1; Stokes
flow). Thus, Fig. 5(b) shows exactly the same developments of
the mixing index throughout the channel length for both of these
Reynolds numbers. Fig. 5(b) also confirms the minimum mixing
index at Re = 15 due to the shorter residential time and insufficient
inertia force of the fluids. Beyond this Reynolds number, the inertia
force increases, which promotes stretching and folding of the fluid
layers through chaotic advection, as shown in Fig. 5(a). The devel-
oping rate of the mixing index generally increases with the Rey-
nolds number.
Mixing performances were compared between the two
micromixers comprised of four mixing units and nine mixing units,
in the Reynolds number range from 0.2 to 120, as shown in Fig. 6.
Mixing indices at the exits (x/w = 11.5 and 25.0) were considered
for the comparison. The micromixer with nine mixing units shows
far better mixing performance than that with four unit for Rey-
nolds numbers less than 60, and the largest difference in the mix-
ing index (about 24%) is found at Re = 15. Beyond Reynolds number
60, effect of number of mixing units on mixing performance
becomes nearly constant. Fig. 7 shows optical microscope images
of the initial mixing units including the inlets and the last mixing
units including the outlet, at Re = 15 and 60. As the flow proceeds,
the fluid layers are continually subdivided into thinner layers
through splitting and recombination at both Reynolds numbers.
The process occurs due to the periodic perturbations induced by
the channel geometry.
The velocity vectors are plotted on the y–z planes at the nodes
of the crossing channels at Re = 0.2 in Fig. 8. This figure shows a
saddle-shaped flow pattern [12], which consists of two parabolic
(top and bottom) and two helical (left and right) flow structures
due to the two-layer crossing channels. In all the cross-sectional
planes (A1–A9), the fluid interface is stretched along the diagonal
line a–d, while that along the diagonal line b–c is shrunken. This
stretching and splitting phenomenon is one of the key mechanisms
for enlarging the interfacial area of the fluids, which enhances the
mixing performance [12]. Therefore, the strong saddle-shaped flow
structures at the nodes of the crossing channels resulted in excel-
lent mixing performance of the proposed microchannel. The chao-
tic mixing mechanism of the proposed micromixer is thus
determined to have both hyperbolic and parabolic centers. Fig. 8
also shows that the distribution of the induced velocity vectors is
stable and invariant in each cross-sectional plane.
To describe the formation of the saddle-shaped flow pattern,
velocity vectors on different y–z planes are plotted in Fig. 9. At sec-
tion AA, two vertical channels overlap each other, and the velocity
vectors form a small saddle-shaped flow structure at the center of
Fig. 10. Effects of design parameters on mixing index at the exit (x/w = 11.5) of the the cross-sectional plane. The two diagonal channels do not cross
micromixer with four mixing units at Re = 15: (a) H/P, (b) w/P, and (c) D/P.
each other at sections BB and DD due to their orientations. At sec-
tion BB, the velocity vectors in the bottom and the top channels
move downward and upward, respectively, and the velocity vec-
as one of the best chaotic-flow type passive micromixers, shows tors are parallel throughout the cross-sectional plane. At section
less than 97% mixing. CC, where the two diagonal channels cross each other, a saddle-
The effects of the Reynolds number on the chaotic mixing shaped flow structure is formed as discussed for Fig. 8. At section
mechanism were analyzed qualitatively and quantitatively by DD, the velocity vectors are similar to those shown at section BB
numerical calculations, as shown in Fig. 5. The dye mass fraction but with opposite directions. Through the sections DD to EE, the
distributions on cross-sectional planes A1–A5 in Fig. 1(b), are plot- flow is reoriented, and a new saddle-shaped flow pattern is formed
ted in Fig. 5(a) to show the chaotic advection phenomena. Fig. 5(b) at the next crossing node.
represents developments of mixing index along the axis of the A parametric study was performed numerically on mixing per-
micromixer for different Reynolds numbers. The mixing indices formance of the proposed micromixer with four mixing units at a
were estimated at the cross-sectional planes located at the nodes Reynolds number of 15, as shown in Fig. 10, using three dimen-
of the crossing structures. Fig. 5(a) demonstrates that the fluid sionless design parameters: H/P, w/P, and D/P. The axial length of
streams are gradually subdivided into multiple thinner layers as the micromixer was kept constant for the study. As shown in
the flow proceeds. Thus, the interfacial area between the fluids Figs. 10(c), the mixing performance is not significantly varied with
enlarges, and the mixing performance improves significantly. At D/P over the tested range. The results indicate that the mixing
Reynolds numbers of 0.2 and 1, the dye mass fraction distributions index at the exit is more sensitive to H/P (65% variation for
276 S. Hossain et al. / Chemical Engineering Journal 327 (2017) 268–277

H/P = 1.65 with the larger residential time discussed above, is

better than that at H/P = 1.87.
Variations of pressure drop with Reynolds number (0.2–120)
are compared between the proposed micromixer and TLCCM as
shown in Fig. 12. The pressure drop is directly related to the pump-
ing power to drive the working fluids. The pressure drop was cal-
culated as a difference between area-weighted averages of total
pressure on the cross-sectional planes located at the first vertical
segment (x/w = 0) and exit of the last mixing unit (x/w = 25.0). At
low Reynolds numbers (Re  10) the proposed micromixer shows
slightly higher pressure drop, while beyond Re = 10 TLCCM shows
higher pressure drop throughout the working range of Reynolds
number.

6. Conclusion

This study has presented a novel chaotic micromixer with two-

layer serpentine crossing microchannels. Both experimental and
numerical analyses of the micromixer were performed, and the
experimental model was fabricated by soft lithography with PDMS.
The numerical results agreed reasonably well with the experimen-
tal data both qualitatively and quantitatively. The proposed micro-
mixer with nine mixing units showed at least 96% mixing
throughout the whole Reynolds number range (Re = 0.2–120).
Most importantly, at low Reynolds numbers (Re = 0.2–10), the
micromixer showed about 99% mixing at the exit. Saddle-shaped
flow structures occurred at the nodes of the crossing channels,
Fig. 11. Dye mass fraction distributions on y-z planes at Re = 15 for different H/P and improved the mixing. Effect of number of mixing units was
values.
remarkable only for low Reynolds numbers less than 60, and the
largest difference in the mixing index (about 24%) was found at
Re = 15. The proposed micromixer showed lower pressure drop
than TLCCM for Reynolds numbers larger than 10. Therefore, the
proposed chaotic micromixer could be integrated with microfluidic
systems, where fast mixing is required at low Reynolds numbers.

Acknowledgements

This work was supported by National Research Foundation of

Korea (NRF) grants funded by the Korean government (NRF-
2016R1A2B4006987). This work was also partially supported by
Inha University WCSL Research Grant, Korea.

References

[1] D.J. Beebe, G.A. Mensing, G.M. Walker, Physics and applications of
microfluidics in biology, Annu. Rev. Biomed. Eng. 4 (2002) 261–286.
Fig. 12. Variations of pressure drop with Reynolds number.
[2] H.A. Stone, A.D. Stroock Ajdari, Engineering flows in small devices:
microfluidics toward a lab-on-a-chip, Annu. Rev. Fluid. Mech. 36 (2004)
1.25 < H/P < 1.87) than w/P (60% variation for 0.24 < w/P < 0.56) 381–411.
[3] R.L. Hartman, K.F. Jensen, Microchemical systems for continuous-flow
and D/P (22% variation for 0.32 < D/P < 0.62). The variation of the synthesis, Lab Chip 9 (2009) 2495–2507.
mixing index at the micromixer exit (x/w = 11.5) with respect to [4] Y. Gambin, C. Simonnet, V. VanDelinder, A. Deniz, A. Groisman, Ultrafast
H/P is shown in Fig. 10(a). For the fixed axial length, the diagonal microfluidic mixer with three-dimensional flow focusing for studies of
biochemical kinetics, Lab Chip 10 (2010) 598–6097.
channel width (w) increases with the main channel width (H), [5] G. Whitesides, The lab finally comes to the chip, Lab Chip 14 (2014) 3125–
and the velocity in the diagonal channel decreases as H/P increases. 3126.
The lower velocity corresponds to a longer residential time of the [6] N. Rajabi, J. Bahnemann, T.-N. Tzeng, O. Platas Barradas, A.-P. Zeng, J. Müller,
Lab-on-a-chip for cell perturbation, lysis, and efficient separation of sub-
fluids in the micromixer and weaker fluid inertia. A maximum mix- cellular components in continuous flow mode, Sens. Actuators A 215 (2014)
ing index of 0.91 occurred at H/P = 1.65 (the reference value), 136–143.
where the optimum combination of the residential time and the [7] C.K. Harnett, J. Templeton, K.A. Dunphy-Guzman, Y.M. Senousy, M.P. Kanouff,
Model based design of a microfluidic mixer driven by induced charge
interfacial area is achieved. Further increase in H/P causes a rapid electroosmosis, Lab Chip 8 (2008) 565–572.
decrease in the mixing index. [8] D. Ahmed, X. Mao, J. Shi, B.K. Juluri, T.J. Huang, A millisecond micromixer via
To explain the effect of H/P on mixing performance, dye mass- single-bubble-based acoustic streaming, Lab Chip 9 (2009) 2738–2741.
[9] V. Hessel, H. Lowe, F. Schönfeld, Micromixers – a review on passive and active
fraction distributions are plotted on y–z planes at the nodes of the
mixing principles, Chem. Eng. Sci. 60 (2005) 2479–2501.
crossing structures at Re = 15, as shown in Fig. 11. With the [10] N.T. Nguyen, Z. Wu, Micromixers – a review, J. Micromech. Microeng. 15
increase in H/P, height of the cross-sectional plane increases. Thus, (2005) R1–R16.
the interfacial area between the fluids enlarges as H/P increases [11] H. Aref, Stirring by chaotic advection, J. Fluid Mech. 143 (1984) 1–2.
[12] H.M. Xia, S.Y.M. Wan, C. Shu, Y.T. Chew, Chaotic micromixers using two-layer
from 1.25 to 1.65, however, it does not clearly increase as H/P crossing channels to exhibit fast mixing at low Reynolds numbers, Lab Chip 5
increases from 1.65 to 1.87. Therefore, mixing performance at (2005) 748–755.
 S. Hossain et al. / Chemical Engineering Journal 327 (2017) 268–277 277

[13] C.Y. Lee, W.T. Wang, C.C. Liu, L.M. Fu, Passive mixers in microfluidic systems: a [24] L. Wang, J.T. Yang, An overlapping crisscross micromixer using chaotic mixing
review, Chem. Eng. J. 288 (2016) 146–160. principles, J. Micromech. Microeng. 16 (2006) 2684–2691.
[14] L. Capretto, W. Cheng, M. Hill, X. Zhang, Micromixing within microfluidic [25] D.S. Kim, S.H. Lee, T.H. Kwon, C.H. Ahn, A serpentine laminating micromixer
devices, Top. Curr. Chem. 304 (2011) 27–68. combining splitting/recombination and advection, Lab. Chip 5 (2005) 739–
[15] A.D. Stroock, S.K. Dertinger, A. Ajdari, I. Mezic, H.A. Stone, G.M. Whitesides, 747.
Chaotic mixer for microchannels, Science 295 (2002) 647–651. [26] J.M. Park, D.S. Kim, T.G. Kang, T.H. Kwon, Improved serpentine laminating
[16] F. Schönfeld, V. Hessel, C. Hofmann, An optimized split-and-recombine micromixer with enhanced local advection, Microfluid. Nanofluid. 4 (2007)
micromixer with uniform ‘chaotic’ mixing, Lab Chip 4 (2004) 65–69. 513–523.
[17] C.C. Hong, J.W. Choi, C.H. Ahn, A novel in-plane passive microfluidic mixer [27] CFX-15.0, Solver Theory, ANSYS Inc., Canonsburg, PA, USA. (2013).
with modified Tesla structure, Lab Chip 4 (2004) 109–113. [28] S. Hardt, F. Schönfeld, Laminar mixing in different interdigital micromixers. II.
[18] C.P. Jen, C.Y. Wu, Y.C. Lin, C.Y. Wu, Design and simulation of the micromixer Numerical simulations, AIChE J. 49 (3) (2003) 578–584.
with chaotic advection in twisted microchannels, Lab Chip 3 (2003) 77–81. [29] J.P. Van Doormaal, G.D. Raithby, Enhancement of the SIMPLE method for
[19] P.B. Howell, D.R. Mott, S. Fertig, C.R. Kaplan, J.P. Golden, E.S. Oran, F.S. Ligler, A predicting incompressible fluid flows, Numer. Heat Transfer 7 (1985) 147–163.
microfluidic mixer with grooves placed on the top and bottom of the channel, [30] R.B. Bird, W.E. Stewart, E.N. Lightfoot, Transport Phenomenon, Wiley, New
Lap Chip 5 (2005) 524–530. York, USA, 1960.
[20] D.S. Kim, S.W. Lee, T.H. Kwon, S.S. Lee, A barrier embedded chaotic micromixer, [31] C.K. Chung, T.R. Shih, T.C. Chen, B.H. Wu, Mixing behavior of the rhombic
J. Micromech. Microeng. 14 (2004) 798–805. micromixers over a wide Reynolds number range using Taguchi method and
[21] R. Liu, M. Stremler, K. Sharp, M. Olsen, J. Santiago, R. Adrian, H. Aref, D. Beebe, 3D numerical simulations, Biomed. Microdevices 10 (5) (2008) 739–748.
Passive mixing in a three-dimensional serpentine microchannel, J. [32] B.J. Kirby, Micro- and Nanoscale Fluid Mechanics: Transport in Microfluidic
Microelectromech. Syst. 9 (2000) 190–197. Devices, Cambridge University Press, New York, 2010.
[22] J. Aubin, D.F. Fletcher, J. Bertrand, C. Xuereb, Characterization of the mixing [33] C.K. Chung, T.R. Shih, B.H. Wu, C.K. Chang, Design and mixing efficiency of
quality in micromixers, Chem. Eng. Technol. 26 (12) (2003) 1262–1270. rhombic micromixer with flat angles, Microsyst. Technol. 16 (2010) 1595–
[23] T.J. Johnson, L.E. Locascio, Characterization and optimization of slanted well 1600.
designs for microfluidic mixing under electroosmotic flow, Lab Chip 2 (2002) [34] M.A. Ansari, K.Y. Kim, Parametric study on mixing of two fluids in a three-
135–140. dimensional serpentine microchannel, Chem. Eng. J. 146 (2009) 439–448.