Journal of the Korean Physical Society, Vol. 48, No. 5, May 2006, pp.

902∼907

Modeling for Fluid Mixing in Passive Micromixers Using the Vortex Index

Joo-Sung Maeng, Kisoo Yoo and Simon Song∗

School of Mechanical Engineering, Hanyang University, Seoul 133-791

Shin Heu
School of Electrical Engineering and Computer Science, Hanyang University, Ansan 425-791

(Received 23 February 2006)

The present paper addresses the effects of molecular diffusion and the vorticity of microchannel
flows on mixing in passive micromixers, which are essential components of a microfluidic chip. A
model that can predict the mixing performance of passive micromixers is developed based on the
physical characteristics of mixing such as the Peclet number and the vortex index that is newly
defined in this paper. In order to investigate the flow physics in the passive mixers, we performed
numerical simulations for a wide range of Peclet numbers and Reynolds numbers. The model is
found to be able to accurately predict the mixing performance of passive mixers without solving
the coupled, complex diffusion and momentum equations.

PACS numbers: 47.15.-X

Keywords: Microfluidics, Passive mixer, Mixing index, Vortex index

I. INTRODUCTION duce the analysis time [6].

Among the various functionalities, rapid mixing is cru-
cial because biological analyses, like enzyme reactions,
Recently microfluidics has been paid attention for the protein folding, and cell activation, require a rapid re-
development of automated miniaturized analytical de- action process that can be controlled by the mixing of
vices in (bio)analytical chemistry. Microfluidics deals reactants. Unfortunately, mixing at a microscale mainly
with microscale, physical phenomena of fluid and particle depends on molecular diffusion, resulting in an extremely
flows in microchannels that connect various functional slow process and long microchannel for complete mixing.
sites on a miniaturized analytical device. Analysis of This is because almost all microchannel flows are lami-
bio-samples in a miniaturized device requires transport nar, and the Reynolds number is so low that turbulent
of various forms of the bio-samples in appropriate flu- mixing can hardly be achieved.
ids from a functional site to another through microchan- Many micromixers have been developed in order to
nels on the device [1]. As a result, the device is often enhance and control mixing in a microchannel [7]. The
called a microfluidic chip. In addition, when the multiple micromixers can be categorized into two types: passive
functionalities required for a complete procedure of the mixers and active mixers. An active mixer uses external
analysis are integrated onto a single microfluidic chip, it energy sources to perturb a laminar flow. The perturba-
is called Lab-on-a-Chip or micro Total Analysis System tion methods include oscillation of a flow-driving source
(µTAS) [2–4]. Typical functionalities contained on a mi- like pressure field [8,9] or electric field [10,11], acoustic
crofluidic chip include sampling, purification, separation, wave generation [12–14], magnetic field variation [15],
mixing, reaction, concentration amplification, detection, and so on. In general, active mixers stir up fluids bet-
etc. Lithography adapted from semiconductor produc- ter than passive mixers, but they require complicated
tion is commonly used to fabricate microfluidic chips [5]. fabrication techniques and sophisticated operation pro-
Analysis using a microfluidic chip has many advantages cedures [7].
against conventional biochemical analysis. It can auto- A passive mixer uses special geometries embedded in
mate the analysis processes, reduce consumption of pre- a microchannel like grooves, rivets or posts, as shown
cious bio-samples such as DNA, RNA, proteins, cells, in Fig. 1, to increase the vorticity and, subsequently, to
bacteria, viruses or blood, reduce contamination of the cause a chaotic advection [16–22]. Chaotic advection im-
bio-samples by minimizing human interaction, and re- proves mixing by enhancing the transversal motion of a
fluid in a microchannel. The other type of passive mix-
∗ E-mail: [email protected]; Fax: +82-2-2298-4634 ers is a lamination mixer, which decreases the diffusion
-902-
Modeling for Fluid Mixing in Passive Micromixers· · · – Joo-Sung Maeng et al. -903-

Fig. 1. Various flow geometries for passive mixers. (a) Central post mixer: L = 2375 µm, w = 300 µm, h = 100 µm, c =
200 µm, and post diameter = 20 µm. (b) Zigzag post mixer: L = 2375 µm, w = 300 µm, h = 100 µm, a = 150 µm, c = 200
µm, and post diameter = 20 µm. (c) Square wave mixer: L = 2800 µm, w = 300 µm, h = 100 µm, a = 800 µm, and c = 400
µm. (d) Serpentine mixer: L = 2600 µm, w = 300 µm, h = 100 µm, a = 800 µm, and c = 40 µm.

length and increases the contact area of fluids by split- II. DEFINITIONS OF THE
ting incoming streams into multiple substreams and then CHARACTERISTIC QUANTITIES FOR
laminating them into one stream again [23–27]. Passive MIXING
mixers are frequently adapted in the development of in-
tegrated microfluidic chips due to the simple concept. Important physical mechanisms involved in microscale
Predicting mixing in a microchannel requires solving mixing are diffusion and chaotic advection. Diffusion can
the coupled diffusion and momentum equations, which be characterized by the Peclet number that depends on
leads to complex numerical calculations. However, it is the diffusivity of a species and the residence time of the
expected that one can effectively predict and evaluate species in the channel. Chaotic advection depends on the
mixing in passive mixers by simply considering the mass flow geometry, and the strength of the chaotic motion can
diffusivity of the mixed fluids and the vorticity of the flow be expressed by the vorticity. If mixing is to be predicted,
without solving the coupled equations. This is because the effects of both diffusion and chaotic advection should
passive mixers, especially chaotic mixers, have been de- be considered. The first step is to quantify mixing and
signed to utilize the vorticity generated in laminar flows. then relate it to the vorticity and the Peclet number of
Maeng et al. developed a model to predict mixing in mi- a microchannel flow.
crochannels by using the so-called vortex index [28]. The
vortex index was defined for 2-D planar mixers and rep-
resents only one-dimensional effects of vorticity on mix- 1. Mixing Index
ing for a fixed inlet flow condition. Thus, modification
of the vortex index is required to utilize it for predict-
ing the mixing performance of general 3-D mixers or for Let us consider a mixing of two identical fluids in-
different inlet flow conditions. serted into a Y-shaped microchannel as shown in Fig. 2
The present paper proposes a simple model to predict (a). However, the upper fluid is assumed to contain a
the degree of mixing in passive mixers operated over a species like protein or DNA, and the dimensionless con-
practical, wide range of Reynolds number from 0.1 to centration is set to unity. The concentration of the other
1000. The model relates effects of both diffusion and is assumed to be zero. To perform a numerical analy-
chaotic advection to the mixing. In order to determine sis on a microchannel flow, we divide the microchannel
the vorticity and the mixing in various passive mixers, into infinitesimal volumes by generating grids as shown
we performed numerical analyses by using a commercial in Fig. 2 (b).
software, CFD-ACE. The mixing index and the vortex A mixing index (M) to quantify mixing in the mi-
index are newly defined to accurately quantify mixing crochannel can be defined as follows:
and vorticity in commonly used passive mixers.
-904- Journal of the Korean Physical Society, Vol. 48, No. 5, May 2006

Fig. 3. Vortex index variations with Reynolds numbers.

with Q being the flowrate in the channel. The integration

interval is from the inlet to the cross section of interest.
The vortex index is defined separately for 2-D mixers
Fig. 2. Y-shaped microchannel and generated grid. (a) Y- (e.g., a post mixer with posts and a square-wave mixer)
shaped microchannel (b) Microchannel divided into infinites- and 3-D mixers (e.g., a serpentine mixer) because the
imal volumes by generating a grid spanwise vorticity (ωy ) does not contribute to mixing at
all for 2-D mixers. A vorticity normalization with Q is
v required to reflect variations of the inlet flow conditions,
u N like the flow velocity or the flowrate, from one case to
u1 X
M= t (Ci − Cin )2 (1) another. Thus, the normalization makes the vortex index
N i=1 depend on the flow geometry only and be independent
of Reynolds number for a given geometry, as shown in
where Ci is the local, dimensionless concentration at a Fig. 3.
node of a given cross section of the channel. Cin is the The effects of diffusion on mixing can be represented
dimensionless, average concentration at the inlet, which by the Peclet number:
is the junction of the Y-shape channel, and N is the num-
UL
ber of nodes at the cross section. The mixing index is Pe = , (5)
equal to 0 when the two fluids are mixed completely and D
0.5 at the inlet. where U is the inlet velocity, L is the hydraulic diameter
of the channel, and D is diffusivity. Physically, the Peclet
number implies the ratio of the longitudinal convection
2. Vortex Index and Peclet Number to the lateral diffusion times. Fig. 4 shows the effects of
Peclet number on mixing in a simple Y-shaped channel
In order to consider the effects of chaotic advection on for different Peclet numbers. A Smaller Peclet number
mixing, one should quantify the proper vorticity of the results in a more rapid and more uniform mixing, as
flow that induces chaotic motions and that subsequently expected.
contributes to mixing. However, it is not the local vortic-
ity of the flow but the vorticity history that determines
the degree of mixing at a cross section of the channel III. RELATIONS OF THE MIXING INDEX,
because the degree of mixing at a position depends on VORTEX INDEX, AND PECLET
the upstream history of the flow. Thus, a vortex index NUMBER
representing the vorticity history of a flow can be defined
as follows:
p Numerical analyses are performed for four represen-
Ω = Ω2x + Ω2z f or 2 − D mixers, (2) tative geometries of passive mixers as shown in Fig. 1,
q to develop a model that relates the mixing index to the
Ω = Ω2x + Ω2y + Ω2z f or 3 − D mixers, (3) vortex index and Peclet number. The inlet velocity for
each channel varies from 1 mm/s to 5 mm/s, but is as-
sumed to be uniform. Incoming fluids are assumed to be
where
R R R a dilute water solution of bio particles with a diffusivity
v
|ωx |dV v
|ωy |dV v
|ωz |dV of 1.7 × 10−9 m2 /s, and the pure water. The numerical
Ωx = , Ωy = , Ωz = (4) results for each mixer are shown in Figs. 5 and 6 for a
Q Q Q
Modeling for Fluid Mixing in Passive Micromixers· · · – Joo-Sung Maeng et al. -905-

Fig. 4. Effects of Peclet number on mixing in a simple Y- Fig. 5. Vortex index variations along the channel length.
shaped channel. The gray scale represents the concentration The length is defined along the streamline at the center of
(0 ∼ 1) that is measured at the outlet. each channel for the square-wave and the serpentine mixers.

Fig. 6. Mixing index variations with the vortex index for different Peclet numbers.

wide range of Peclet numbers. The vortex index is the mixer type. As long as the values of the vortex index are
largest for the serpentine mixer at the same longitudinal the same despite different flow geometries, the degrees
length from the inlet as shown in Fig. 5 (a), indicating of mixing remain similar. In addition, Fig. 6 indicates
that the serpentine mixer induces chaotic motions most that the mixing index approaches zero more rapidly for
violently. The vortex index of the serpentine mixer is a given mixer as the Peclet number decreases.
2.4 times higher than that of the central post mixer at Based on the mixing index variations, we can define a
a longitudinal length of 1900 µm. Fig. 6 shows mixing- functional form of the mixing index dependence on the
index variations with the vortex index for different Peclet vortex index and the Peclet number as
numbers (100 ∼ 1500). For a fixed Peclet number, the
mixing index data collapse onto a single line, indicating M = A exp (−BΩ), (6)
that vorticity dominates fluid mixing, regardless of the where the coefficient A is identical to the value (0.5) of
-906- Journal of the Korean Physical Society, Vol. 48, No. 5, May 2006

Fig. 9. Prediction of the mixing performance using the

model for a square-wave mixer.
Fig. 7. Variations of coefficient B with Peclet number.

(8) to the mixing index (Mnum ) calculated from the nu-

merical results obtained by solving the coupled diffusion
and momentum equations. As Fig. 8 shows, most esti-
mates are close to the numerical values, and the maxi-
mum discrepancy among the data is 8 % near Mnum =
0.2.
This model is also tested to check if it can accurately
estimate the mixing performance of a square-wave mixer
with an elongated channel. The channel length is in-
creased by factor of two compared to that of the pre-
vious case. The model’s estimate is in good agreement
with the mixing index calculated from the numerical re-
sults within 5 % error at most. This is evidence that this
model can be used to estimate the mixing performance
of a passive mixer operating over a practical range of
Reynolds number and Peclet numbers.

Fig. 8. Comparison of the mixing indices from estimates

and numerical calculations.
V. CONCLUSIONS
the mixing index at the inlet because the vortex index
is zero at the inlet. The coefficient B is a function of The performance of a passive mixer was predicted by
the Peclet number, and it can be determined by fitting simply considering the Peclet number and the vorticity
all the data of Fig. 6 for different Peclet numbers. The of the flow. To quantify the degree of mixing and the
fitting results are plotted in Fig. 7, and the functional effects of vorticity on the mixing, we defined a mixing in-
relation between B and the Peclet number is dex and a vortex index based on the physical characteris-
2.7 tics of fluid mixing. Numerical analyses were performed
B= + 0.004. (7)
Pe to develop a model that determined the functional re-
Therefore, the mixing index can be modeled as lation between the mixing index, the vortex index and
2.7 the Peclet number. The model could accurately predict
M = 0.5 exp (−( + 0.004)Ω). (8) the performance of passive mixers operated in the Peclet
Pe number range of 100 to 1500 and the Reynolds number
This model can be used to readily estimate the per- range of 0.1 to 1000 at least.
formance of a passive mixer that utilizes diffusion and
chaotic advection without solving the coupled diffusion
and momentum equations.
ACKNOWLEDGMENTS
IV. DISCUSSION
This research was supported by The Center of Innova-
In order to examine the accuracy of the model, we tive Design Optimization Technology (ERC of the Korea
compare the mixing index (Mest ) estimated using Eq. Science and Engineering Foundation).
Modeling for Fluid Mixing in Passive Micromixers· · · – Joo-Sung Maeng et al. -907-

REFERENCES [14] K. Yasuda, Sensor Actuat. B-Chem. 64, 128 (2000).

[15] H. H. Bau, J. H. Zhong ,and M. Q. Yi, Sensor. Actuat.
B-Chem. 79, 207 (2001).
[1] J. S. Yahng, S. C. Jeoung, D. S. Choi, D. Cho, J. H.
[16] S. W. Johnes and H. Aref, Phys. Fluids 31, 469 (1998).
Kim, H. M. Choi and J. S. Park, J. Korean Phy. Soc.
[17] T. J. Johnson, D. Ross and L. E. Locascio, Anal. Chem.
47, 977 (2005).
74, 45 (2002).
[2] P-A. Auroux, D. Isossifidis, D. R. Reyes and A. Manz,
[18] R. H. Liu, M. A. Stremler, K. V. Sharp, M. G. Olsen, J.
Anal. Chem. 74, 2637 (2002).
G. Santiago, R. J. Adrian, H. Aref and D. J. Beebe, J.
[3] D. R. Reyes, D. Isossifidis, P-A. Auroux and A. Manz,
Microelectromech. S. 9, 190 (2000).
Anal. Chem. 74, 2623 (2002).
[19] V. Mengeaud, J. Josserand and H. H. Girault, Anal.
[4] T. Vilkner, D. Janasek and A. Manz, Anal. Chem. 76,
Chem. 74, 4279 (2002).
3373 (2004).
[20] A. D. Stroock, S. K. W. Dertinger, A. Ajdari, I. Mezic,
[5] N-G. Cha, C-H. Park, H-W. Lim and J-G. Park, J. Ko-
H. A. Stone and G. M. Whitesides, Science 295, 647
rean Phy. Soc. 47, S530 (2005).
(2002).
[6] S. C. Jakeway, A. J. D. Mello and E. L. Russell, Fresse-
[21] H. Z. Wang, P. Iovenitti, E. Harvey and S. Masood,
nius J. Anal. Chem. 366, 525 (2000).
Smart Mate. Struct. 11, 662 (2002).
[7] N-T. Nguyen and Z. Wu, J. Micromech. Microeng. 15,
[22] H. Z. Wang, P. Iovenitti, E. Harvey and S. Masood, J.
R1 (2005).
Micromech. Microeng. 13, 801 (2003).
[8] A. A. Deshmukh, D. Liepmann and A. P. Pisano, Tech-
[23] P. Hinsmann, J. Frank, P. Svasek, M. Harasek and B.
nical Digest of the IEEE Solid State Sensor and Actuator
Lendl, Lab Chip. 1, 16 (2001).
Workshop (Hilton Head Island, South Carolina, 2000).
[24] R. F. Ismagilov, A. D. Stroock, P. J. A. Kenis, G. White-
[9] T. Fujii, Y. Sando, K. Higashino and Y. Fujii, Lab Chip
sides and H. A. Stone, Appl. Phys. Lett. 76, 2376 (2000).
3, 193 (2003).
[25] A. E. Kamholz and P. Yager, Sensor. Actuat. B-Chem.
[10] S. C. Jacobson, T. E. Mcknight and J. M. Ramsey, Anal.
82, 117 (2002).
Chem. 71, 4455 (1999).
[26] D. S. W. Lim, J. P. Shelby, J. S. Kuo and D. T. Chiu,
[11] M. H. Oddy, J. G. Santiago and J. C. Mikkelsen, Anal.
Appl. Phys. Lett. 83, 1145 (2003).
Chem. 73, 5822 (2001).
[27] Z. G. Wu, N. T. Nguyen and X. Y. Huang, J. Micromech.
[12] J. C. Rife, M. I. Bell, J. S. Horwitz, M. N. Kabler, R. C.
Microeng. 14, 604 (2004).
Y. Auyeung and W. J. Kim, Sensor Actuat. A-Phys. 86,
[28] J-S. Maeng, I-D. Cho and B-J. Kim, Trans. KSME B 29,
135 (2000).
232 (2005).
[13] Z. Yang, H. Goto, M. Matsumoto and R. Maeda, Elec-
trophoresis 21, 116 (2000).