Sie sind auf Seite 1von 12

This article appeared in a journal published by Elsevier.

The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
Author's personal copy

Chemical Engineering Science 66 (2011) 1962–1972

Contents lists available at ScienceDirect

Chemical Engineering Science


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

Investigations of the unsteady diffusion process in microchannels


Diana Broboana a, Catalin Mihai Balan a, Thorsten Wohland b, Corneliu Balan a,n
a
Department of Hydraulics and Fluid Machineries, REOROM Laboratory, Politehnica University of Bucharest, Splaiul Independentei 313, 060042 Bucharest, Romania
b
National University Singapore, Department of Chemistry, 3 Science Drive 3, Singapore 117543, Singapore

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

Article history: This paper is concerned with the investigations and modeling of the unsteady diffusion process along a
Received 12 August 2010 straight micro-channel with a cross section of 380 mm  360 mm. The studied process is characterized
Received in revised form by small Reynolds numbers (Re o10) and high Péclet number (Pe 41000). The 3D computations of the
20 January 2011
coupled momentum and diffusion equations for isochoric motions are performed with the FLUENT code
Accepted 24 January 2011
using the unsteady solvers for both equations. In the limit of stationary solutions, the numerical results
Available online 2 February 2011
are validated by direct flow visualizations and experimental data using confocal microscopy. The
Keywords: computed distributions of concentration provide qualitative and quantitative information on the
Micro-channel hydrodynamics transitory diffusion process and the rate of solute spreading within the investigated geometry. In
Diffusion process
particular, the pattern of the ‘‘butterfly effect’’ is represented and analyzed during the non-stationary
Butterfly effect
dynamical process. The work is relevant for the design of novel microfluidics applications where the
Flow visualizations
Confocal microscopy control of diffusion processes at the walls are important (absorption, extraction, capture of molecules or
Numerical computation nano-particles).
& 2011 Elsevier Ltd. All rights reserved.

1. Introduction At present, the diffusion phenomena are mostly investigated


in bifurcated planar channels of T- or Y-shapes. The cross section
One method to determine the diffusion coefficient of a solute of the channels is rectangular, with aspect ratios between the
dispersion in a liquid solvent is considering the evolution of width (W, x-direction) and the height (H, y-direction) starting
solute concentration along a micro-channel with well established from one (square section, Sullivan et al., 2007), to very large
hydrodynamics. The experimental pattern of concentration is values (shallow channels, W/H4100, Kamholz and Yager, 2001;
then compared, for a range of the diffusion coefficient Di, either Kamholz et al., 2001). Two liquid specimens (one enriched with
with the corresponding analytical solution of homogeneous the dispersed particles in small concentration) are introduced
diffusion (if it is available), or with the numerical solution of through the branches in the main channel of length L (LbW),
the Navier–Stokes equation coupled with the diffusivity equation. where the diffusivity of the dispersed particles is quantified by
Finally, the best fit of the experimental data with the computa- optical measurements along the flow (which defines the z-direc-
tional results, in some established fixed space domains of the tion). Using classical visualizations techniques (micro-PIV, Wu
micro-geometry, decides the appropriate value of D (for details and Nguyen, 2005) or molecular tagging velocimetry (Garbe et al.,
see Adeosun and Lawal, 2009; Kamholz et al., 1999; Pan et al., 2008), coupled with confocal microscopy (Ismagilov et al., 2000;
2007a, 2007b; Tian and Wohland, 2008). Kamholz et al., 1999; Tian and Wohland, 2008), NMR imaging
One of the most common investigated coupled diffusion- (Akpa et al., 2007), Raman imaging (Dambrine et al., 2009),
momentum equations is the Taylor dispersion phenomena, where time-resolved fluorescence imaging (Benninger et al., 2005), or
the dispersion of a Dirac-pulse of a passive scalar (tracer) is micro-interferometry (Garvey et al., 2008), the concentration
studied in a channel with laminar flow (Gan et al. 2007; distribution is measured in the main channel at fixed distances
Hartmann and Sasso, 2007; Kamholz et al., 1999). The results downstream the junction (at different heights within the channel,
show the contribution of the convective transport term to diffu- if proper devices are available).
sion and the influence of the Péclet number on the process, which The fundamental studies of steady diffusion in the T-sensor
determines through its power of two the increase of the effective with rectangular shallow geometry were presented by Ismagilov
diffusion coefficient (Ajdari et al., 2006; Stone et al., 2004). et al. (2000), Kamholz et al. (2001) and Kamholz et al. (1999). The
main phenomena investigated in these studies is ‘‘the butterfly
effect’’, which characterizes the distribution of the scalar concen-
n
Corresponding author. Tel.: +40 214029705; fax: + 40 214029865. tration in normal planes to the main flow direction. This asym-
E-mail address: corneliu.balan@upb.ro (C. Balan). metry in concentration distribution is induced by the convective

0009-2509/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2011.01.048
Author's personal copy

D. Broboana et al. / Chemical Engineering Science 66 (2011) 1962–1972 1963

term of the diffusion equation, due to the parabolic velocity rhodamine (TMR), in aqueous buffer solution (phosphate buffered
distribution of the solvent in the micro-channel. As consequence saline (PBS)) (see Tian and Wohland, 2008). The experimental
of the fluid adherence at the walls and non-uniformity velocity results, represented as distributions of the signal intensity (which
distribution in the planes z¼constant, in the vicinity of the upper is proportional to concentration of the solute) across the width of
and lower walls of the channel (y¼ H/2, respectively, y¼ H/2) the channel, are found consistent with the numerical solutions.
the iso-concentration lines are curved, taking the form of a ‘‘C’’ The unsteady diffusion was investigated exclusively by numer-
shape. In this case, the inter-diffusion length d ¼ d(y, z) – defined ical simulations. We represent the iso-concentration distributions
by the distance between the original interface at z ¼0 (assumed to in planes normal to the channel in order to characterize the
be the line x¼ 0) and the current interface – is not only a function of dynamics of the ‘‘butterfly effect’’ and the rate of solute spreading
‘‘convection length’’ (i.e. z-coordinate), but also a function of the across the channel width. The computed spectrum of concentra-
height in the planes with z¼zi ¼ constant, respectively d(H/2, zi)4 tion gives original and detailed information about the transitory
d(0, zi), with d(y, 0) ¼0. diffusion which take place at contact with the channel walls
In the last decade, most of the applications of diffusion channel. These results have potential influence on the design of
processes in microchannels are directed to: (i) diffusion-based novel micro-channel applications based on extraction and absorp-
cell extraction from cell suspension (diffusion with reactive term) tions processes, where the control of the solute concentration
(Mata et al., 2008), (ii) transport with stable concentration gradient is needed (Smith et al., 2010) and the quantification of
gradients (Gorman and Wikswo, 2008), (iii) protein deposition mixing between the contact phases is important (Glasgow et al.,
in microgeometries (Bransky et al., 2008), or (iv) improving the 2004; Sun and Sie, 2010).
mixing of the fluids in order to increase the diffusivity of the
solute in simple or complex microgeometries (Abonnenc et al.,
2009; Beutel, 2003; Lee et al., 2005; Truesdell et al., 2003), see 2. Experimental
also the last review by Aubin et al. (2010). In all these studies
(which belong to the Lab-on-a-chip applications, Stone et al., In this paper, the investigated geometry is a symmetric
2004) flow visualizations are normally corroborated with numer- Y-branching channel with the angle of 751 between the entrances
ical simulations, in order to obtain a better description of the and the length L¼100 mm of the main channel, along the
hydrodynamics and diffusion process within the tested micro- z-direction (see Fig. 1). The geometry of the transverse cross
geometry (Liu et al., 2004; Shih and Chung, 2008; Yamaguchi section of the channel is W¼380 mm (width) and H¼360 mm
et al., 2006). (height), W/H¼ 1.2, so the characteristic space scale is R ¼370 mm
Solutions of the coupled equations of motion and diffusion are (i.e. four times the hydraulics radius).
based on various numerical methods and numerical codes: The Y-microchannels were manufactured from polydimethyl-
commercial FLUENT code (Adeosun and Lawal, 2009), Fourier siloxane prepolymer (PDMS) (Dow Corning, Singapore), which
transform (Estévez-Torres et al., 2007), Lattice Boltzman (Sullivan was poured into the corresponding mold profile. It was then
et al., 2007), or FD specialized programs (Morf et al., 2008). degassed and cured at 601 C overnight in an oven. Following that,
Numerical computations are performed normally in 3D micro- the hardened PDMS was peeled off from the mold. Holes were
channel configurations. These investigations are focused on the then made on the inlets and outlet, both PDMS and glass slide
study of mixing processes (Liu et al., 2004; Shih and Chung, 2008), were washed with de-ionized water and blow dried with nitro-
the detection of the butterfly effect or the demonstration of the gen. Next, they were placed into the air plasma for 1 min to
influence of the local viscosity dependence on concentration of enable irreversible bonding of the PDMS and glass slide before
the solute on the diffusion processes (Sullivan et al., 2007). placing on the hot plate at 105 1C for 30 min. Lastly, needles were
The main goal of the present paper is to model the unsteady introduced into the inlets and outlet and the lines at 5, 15, 20, 25
diffusion process in a symmetric Y-geometry, in particular to and 35 mm from the Y-junction were made on hardened PDMS
determine the evolution of the non-stationary butterfly effect in a substrate, so that measurements can be taken at these marked
straight channel with the cross section of 380 mm  360 mm. The positions.
values of the characteristic Reynolds and Péclet numbers are In experiments the solvent is a PBS buffer solution and the
Re ffi1.85 and Pe ffi3000, respectively. The 3D computations of the fluorophore solute is Tetramethylrhodamine (TMR) (TMR is a red
coupled momentum and diffusion equations for isochoric fluorophore with an excitation wavelength of 550 nm and emit-
motions are performed (in the absence of reaction) with the ting at 580 nm). TMR has a non-planar structure and sterically
FLUENT code, using the unsteady solvers for both equations. hindered substituents that prevents the formation of short p–p
The applied numerical procedure is similar to the computa- interactions, which allows fluorophores to exhibit strong
tions presented in the recent paper by Adeosun and Lawal (2009), fluorescence.
where transient diffusion of a pulse tracer injection was studied The visualizations have been performed using two setups:
in a 3D T-junction geometry, in order to appreciate the residence (i) confocal laser scanning microscope and (ii) inverted micro-
time distribution of particles in a micro-channel mixing process scope and a CCD camera (see for details Pan et al., 2007a, 2007b;
under a steady velocity distribution. Tian and Wohland, 2008). The dye solution, c0 ¼1% volumetric
The transient dispersion regime was also investigated concentration of TMR—1 mM in PBS (i.e. fluid A +c0), was pumped
by Goulpeau et al. (2007), but in relation with the control of from one inlet while PBS buffer solution (i.e. fluid A) was pumped
concentration gradients along microchannels of parabolic cross- from the other inlet at the same flow rate (see Fig. 1). There were
sections. The unsteady diffusion in the neighborhood of the walls used two identical syringe pumps, with the flow rate within the
is also a topic of interest in establishing the correct boundary range of (0.1, 2) ml/h, which determine an average velocity in the
conditions at solid surfaces in the presence of adsorption process, main channel V0A(0.5, 10) mm/s, corresponding to ReA(0.25, 5).
as investigated by Brenner and Ganesan (2000), or the analysis of Diffusion of the TMR in PBS buffer was detected by monitoring
Brownian diffusion process in the very vicinity of the walls (Sadr the change in fluorescence intensity across the channel width.
et al., 2007). The 543 nm laser beam of the confocal microscope (FV300,
The numerical procedure is tested making a comparison of Olympus, Singapore) was scanned in an x–z directions from a
the stationary computed concentration along the channel with depth of 0 mm (cover slide) to a depth of 180 mm (middle of the
the measured distributions of a fluorescent tracer, teramethyl- channel). The fluorescence observed in the channel is reflected as
Author's personal copy

1964 D. Broboana et al. / Chemical Engineering Science 66 (2011) 1962–1972

Fig. 1. Geometry of the channel and location of the local frame {x, y, z}: (a) the Y-geometry and dimensions corresponding to the model used in numerical simulations (the
length of main channel in the real geometry is 100 mm instead of 40 mm), (b) steady path lines distributions at the inlet in the main channel, and (c) steady parabolic
velocity distribution at z¼ 0.

Fig. 2. Diffusion evolution along the flow direction (CCD pictures at V0 ¼ 2.4 mm/s). The dye solution (fluorescent TMR in PBS buffer solution) is represented with
white color.

images in the computer. This was done at distances of 5, 15, 20, However, as the CCD camera cannot capture images at a parti-
25 and 35 mm from the Y-junction along z-direction of the flow. cular depth, only one image of averaged fluorescence intensity
The experiment was also repeated using a normal fluorescence across the whole channel depth was taken at each position from
microscope, the pictures being recorded with a CCD camera. the Y-junction.
Author's personal copy

D. Broboana et al. / Chemical Engineering Science 66 (2011) 1962–1972 1965

Data from the linescans are then related to the diffusion Table 1
coefficient (D) of the fluorescent compound (TMR), the graphs Variation of diffusion coefficient along z-direction at V0 ¼2.4 mm/s (here tc is the
characteristic time) (see Figs. 2 and 3). The values are automatically computed by
being fitted with the program Igor Pro (Vers. 6.0, Wavemetrics,
the program Igor Pro.
Lake Oswego, OR, USA) according to the classical one-dimensional
solution of the diffusion equation, i.e. z (mm) tc (s) D (m2/s)
 
x 5 2.1 (2.95 7 0.03)  10  10
cðxÞ ¼ 0:5erfc pffiffiffiffiffiffiffiffi ð1Þ 15 6.2 (3.10 7 0.05)  10  10
4Dt
20 8.2 (3.30 7 0.06)  10  10
25 10.3 (3.42 7 0.06)  10  10
where c is the solute concentration and t ¼tc ¼(z+ a)/V0 is con- 35 14.4 (3.22 7 0.08)  10  10
sidered the characteristic time of the diffusion along the channel
length, V0 being the average velocity in the main channel
(see Fig. 1). We have to mention that diffusion time scales are
different on the other directions, e.g. the diffusion across the
channel width has the characteristic time td ¼W2/4D (Garvey
et al., 2008; Hartmann and Sasso, 2007). We also remark that
(1) is obtained with imposed infinite boundary conditions for the
x-direction, but here the formula is used to approximate at
t ¼constant (i.e. z¼constant) the distribution of solute in the
vicinity of the symmetry axis of a finite channel (i.e. at x ¼0).
Direct visualization of the TMR diffusion process in PBS is
shown in Fig. 2. The pictures are taken at V0 ¼2.4 mm/s in the Fig. 4. Experimental fluorescence intensity recorded across the depth of the
main channel, the process being considered steady. The measured channel with the confocal microscope at z ¼20 mm (see Fig. 2), between two
y-coordinates. The presence of the butterfly effect is easily observable in the
distributions of the fluorescence intensity (proportional to the
vicinity of the wall (y¼150 mm).
concentration of the TMR) across the channel width is repre-
sented in Fig. 3. From (1), the average computed diffusion
3. Numerical simulations
coefficient is D ¼(3.270.2)  10  10 m2/s (see Table 1), the value
which fits the best all distributions from Fig. 3 (for details about
Numerical computations of the diffusion process within the
procedure see also Munson et al. (2005)).
channel were performed using the commercial FLUENT 6.3.26
The corresponding average value of the diffusion coefficient
code, interfaced with the Gambit pre-processor to construct the
computed from the confocal measurements data is D ¼(3.717
geometry and the corresponding meshes. At each time step, the
1.1)  10  10 m2/s (Tian and Wohland, 2008). The error is larger in
Navier–Stokes equation (2) is solved in parallel with the diffusion
the last case because the average is made not only between the
equation (3) for the concentration field c,
values recorded at different lengths and flow rates, but also  
between the measurements took at different depths: y¼0 (mid- @v
Re Shi þ grad vv ¼ grad p þ Dv ð2Þ
dle of the channel) and y¼150 mm. In this case the presence of the @t
‘‘butterfly effect’’ is evident (see Fig. 4), but using our set-up we  
could not perform exact measurements of the inter-diffusion @c
Pe Shd þ grad cUv ¼ Dc ð3Þ
length d at the channel walls. @t
At this moment, quantitative results of the inter-diffusion where Re ¼ rRV=Z is the Reynolds number, Pe ¼ RV=D is the
length and the evolution of the‘‘butterfly effect’’ along the flow Péclet number, Shi ¼ R=ðVti Þ and Shi ¼ R=ðVtd Þ being the corre-
can be obtained only by numerical simulations. sponding Strouhal numbers. The expressions (2) and (3) are non-
dimensional, scaled with space dimension R , the average velocity
(V ¼V0), momentum characteristic time ti ¼ rR2 =Z, and the diffu-
sivity characteristic time td ¼ R2 =D (here Z is the dynamic viscos-
ity, r is the mass density and D is the diffusion coefficient, all
these material parameters are being considered constant within
the flow field).
The ratio between the two time scales (td =ti ) defines the
Schmidt number, Sc ¼ Z=ðrDÞ, one of the most important para-
meter which characterizes the diffusion process in a viscous fluid.
The present simulations have been performed for a single fluid
phase with constant viscosity and density (details on the numer-
ical simulations and modeling of mixing immiscible fluids in
microchannels using the VOF code implemented in FLUENT are
given in Balan et al. (2008) and Balan et al. (2010)).
A solute concentration below 1% (i.e. c r 0:01 [-]) does not
change the fluid properties and the diffusion coefficient is con-
stant, so the only coupling between the Eqs. (2) and (3) is the
convective term from the diffusion equation, i.e. grad cUv, where v
is the Navier–Stokes solution for the velocity field (see Fig. A1
from the Annex for comparison of results with constant viscosity
Fig. 3. Experimental normalized measured fluorescence intensity (arbitrary units) and variable viscosity).
across the channel (x-direction), at different z-distances from the junction
(see Fig. 2) (steady process); a slight asymmetry of the measurements is observed.
All the simulations are performed with water (density
From the best fitting of the curves with relation (1) the diffusion coefficients r ¼1000 kg/m3 and viscosity Z ¼1 mPa s) in a Y-geometry with
from Table 1 are obtained. length scale R ¼0.37 mm and diffusion coefficients in the range
Author's personal copy

1966 D. Broboana et al. / Chemical Engineering Science 66 (2011) 1962–1972

10  10 oD o10  9 (m2/s). The boundary conditions are specified the upper branch and c¼0 at the entrance of the lower one, with
as follows: constant velocity at the entrances in upper/lower zero diffusive fluxes at the wall.
branches and zero relative pressure at the exit of the main branch, The analyzed nominal case corresponds to D0 ¼6  10  10 m2/s
and constant trace concentration c¼ c0 ¼0.01 at the entrance of and V0 ¼4.8 mm/s, therefore the flow is characterized by the
following values of the non-dimensional parameters: Reffi 1.85,
Peffi 3000, and Sc ffi 1600 (with Sc ¼ Pe=Re). The similitude criteria
Table 2 with experiments is the Péclet number (the same geometry is
Numbers of nodes and cells size for the used meshes.
used, but the values of the diffusion coefficient and the velocity is
Mesh No. of nodes x-axis y-axis (high) z-axis doubled in order to speed up the computations). The motion
(width) (mm) (mm) (length) (mm) within the channel is purely laminar and the velocity profile is
stabilized much faster than the concentration distribution
M1 164,400 40 40 40 (Sc b1). For that reason in similar studies the unsteady part of
M2 857,571 20 30 20
M3 1,843,532 15 20 16
the momentum equation is neglected in comparison to the
transient diffusion (Adeoson and Lawal, 2009).

Mesh M1 Mesh M2
0.005
t = 5s
0.005 t = 5s
0.010 0.010 0.010 0.010
0.015 0.015
0.020 0.020
0.025 0.025
0.008 0.008 0.030
Concentration [-]

0.030 Concentration [-]


0.035 0.035
0.006 0.006

0.004 0.004

0.002 0.002

0.000
0.000
-0.0002 -0.0001 0.0000 0.0001 0.0002
-0.0002 -0.0001 0.0000 0.0001 0.0002
Channel width [m]
Channel width [m]

t = 10s t = 10s
x = 0.005
0.010 x = 0.010 0.010 0.005
x = 0.015 0.010
x = 0.020 0.015
x = 0.025 0.020
0.008 0.008 0.025
Concentration [-]

x = 0.030
Concentration [-]

x = 0.035 0.030
0.035
0.006 0.006

0.004 0.004

0.002 0.002

0.000 0.000

-0.0002 -0.0001 0.0000 0.0001 0.0002 -0.0002 -0.0001 0.0000 0.0001 0.0002
Channel width [m] Channel width [m]

t = 30s
0.005
t = 30s
0.005
0.010 0.010 0.010 0.010
0.015 0.015
0.020 0.020
0.008 0.025 0.025
0.008
Concentration [-]

0.030 0.030
Concentration [-]

0.035 0.035
0.006 0.006

0.004 0.004

0.002 0.002

0.000 0.000

-0.0002 -0.000 10.0000 0.0001 0.0002 -0.0002 -0.0001 0.0000 0.0001 0.0002
Channel width [m]
Channel width [m]

Fig. 5. Time variation of concentration, c¼ c(x), in the middle of the channel (y¼0), at z ¼zi ¼ constant (influence of the mesh size; comparison between M1 and M2
meshes).
Author's personal copy

D. Broboana et al. / Chemical Engineering Science 66 (2011) 1962–1972 1967

Since the flow is perfectly symmetric and the input branches


are long enough, the flow reaches a stable velocity parabola at the
entrance in the main channel at 10 ms from the onset of the
motion (at Re o10). The tracer starts to be present in the main
channel, at z ¼0 mm, not before 1 s, so the process is basically
time dependent only in the diffusion equation (3). Only after 30 s
the solute concentration has a steady distribution along the
channel.
The coupling between (2) and (3) becomes relevant either at
small Schmidt numbers or for non-symmetric entrance conditions
in the main channel and unsteady velocity distributions. These
flows are not explored in this paper, but we mention that the
latest case is at the moment under study in our laboratory.
The numerical simulations were performed with the following
options: laminar solver (Green Gauss cell based gradient option),
unsteady formulation (second order implicit scheme), pressure–
velocity coupling (PISO) and QUICK scheme as the interpolation
method for convective terms. The computations used a time step
of 10  4 s and a 10  10 convergence criteria, Fluent code being
installed on a 64-bit server (Dual 2.66 GHz with 16 GB RAM
memory).
The influence of the grid size was investigated using three
meshes (see Table 2).
Fig. 8. Comparison between the analytic solutions (continuous line in the detail),
experimental data (Fig. 3) and the corresponding 3D solutions (scatter) at
z¼ 20 mm. The qualitative confirmation of the computations are given by the
experimental picture from Fig. 4.

Fig. 6. Time variation of concentration, c ¼c(x), in the middle of the channel


(y¼ 0), at z¼ zi ¼ constant; comparison between M2 and M3 meshes, at time 5 s,
respectively 10 s.

Fig. 9. Comparison between the experiments (Fig. 3) and the corresponding 3D


Fig. 7. Flow geometry at z¼ constant. The ‘‘butterfly effect’’ is represented by the solutions in the center of the channel (a) and at the wall (b) (t ¼30 s, Pe ¼ 3000;
curvature of the iso-concentration line c¼ c(x, y) ¼ 0.1c0. experiment: Re¼ 0.925, numerics: Re¼ 1.85).
Author's personal copy

1968 D. Broboana et al. / Chemical Engineering Science 66 (2011) 1962–1972

The comparisons between the results computed with meshes


M1, M2, and M3 are shown in Figs. 5 and 6. One can observe that
the grid size influence is remarkable only in the transitory regime 5s
(see Fig. 5). In the limit of the steady state (tZ30 s) the results are
almost identical. The transient numerical results computed with
M1 and M2 meshes differ significantly (see Fig. 5) (t o30 s). The
precision of the numerical calculus is directly influenced by the
magnitude of the local Péclet number; it is doubled for the M1
grid, so the errors generated during the computations of the
unsteady and convective terms from (3) are larger by a factor of
two. Since the meshes M2 and M3 (Fig. 6) produce minor
differences also during the transitory regime, the mesh M2 was
considered the optimum for the next computations (the compu-
tation time is significant larger for M3 and 1 s of t-time flow 10 s
simulation for mesh M2 consumes up to 12 h of computing time).
However, it is well known that calculus of diffusion fluxes (i.e.
grad cUv) through the cell faces involves (especially at high global
Péclet numbers) a truncation error, referred to as numerical or false
diffusion (see Ferziger and Peric, 1999; Chung, 2003). The present
simulations are characterized by a CFL factor less than one and grid z
lines aligned with the flow, but the local Péclet number computed
with the grid dimension is much larger then one (100oPe-
localo200), so the accuracy of the numerical prediction is limited
(for details see Barz et al., 2008; Soleymani et al., 2008). Since it is
not feasible to refine the grid in order to decrease significantly the 15 s
local Péclet number and we do not have access to other numerical
codes, the only possibility to appreciate the degree of computational
predictions is the direct comparison with measured data (compar-
ison between the performances of different commercial CFD codes

30 s

Fig. 12. Time dependence of the concentration distributions in the main channel
along the flow direction, in sections z¼ 5/10/15/20/25/30/35 mm; c 40.5c0 (red),
co 0.1c0 (blue); D ¼ 6  10  10 m2/s, V0 ¼ 4.8 mm/s (simulations performed at
Fig. 10. Inter-diffusion lengths, dw (wall) and d0 (central), at different diffusion Pe¼ 3000, respectively, Re ¼1.85). (For interpretation of the references to colour
coefficients, along the flow direction (z-coordinate). in this figure legend, the reader is referred to the web version of this article.)

D D D

c > 0.1c0

c = 0.1 c0 c = 0.1c0

c < 0.1c0

Fig. 11. Iso-concentration line c ¼0.1c0 at t¼ 35 s, for different diffusion coefficients, in section z ¼5 mm (average velocity in the channel: V0 ¼ 4.8 mm/s).
Author's personal copy

D. Broboana et al. / Chemical Engineering Science 66 (2011) 1962–1972 1969

and experiments performed in similar microfludic applications are solutions, relation (1), and with the experimental data
discussed by Glatzel et al., 2008). from Fig. 3, respectively. The phenomena is well represented by
the numerical solutions, and both the analytical solution and the
experimental data from Fig. 3 being between the values com-
4. Comparison between experiments and simulations puted in the middle and at the wall of the channel (see Fig. 8).
One observes in Fig. 9 that 3D numerical solutions at y¼0 fit
Since the numerical simulations are performed at the same better the experimental data at small and medium z-coordinates,
Pe-number as the experiments from Section 2, the steady state since at large distances from the junction (z 420 mm) the
numerical patterns of the concentration spectrum have to repro- measured concentration is fair fitted by the numerical solution
duce the experiments, with the remark that the corresponding at the wall (y¼H/2). This fact might be explained by the presence
iso-concentration lines are reached faster in simulations (since of the ‘‘butterfly effect’’, which increases the width of the separa-
the Reynolds number is twice larger in the simulation than in tion band observed by direct visualizations between the marked
experiments, see Fig. A2 from the Annex). and un-marked regions with fluorophore solute. However, some
One phenomena with a major contribution to the complexity discrepancies between experiments and computation might be
of computation is the convection of concentration, and as a generated also by two other factors: (i) propagation of the
consequence the presence of the ‘‘butterfly effect’’ mentioned in numerical errors along the flow direction (numerical diffusion),
Section 1 (see also Kamholz and Yager, 2001). Beyond the length or (ii) lack of isotropy of the real diffusion coefficient within the
of stabilization of the parabolic velocity distribution vz ¼ vz ðx,yÞ, channel and possible dependence of diffusion on the residence
along the main channel at z¼constant, the solute concentration c time.
is not only a function of the x-direction (channel width), but also In Fig. 10 are represented the computed inter-diffusion lengths
dependents on the channel height (y-direction), i.e. c¼c(x, y). for different values of diffusion coefficients, as function of the
In steady state, diffusion in x-direction is higher in the vicinity z-coordinate, at t¼35 s (considered the steady state for all cases).
of the walls then in the center (y¼0), because convection is lower For the geometry under investigation, the dependence of inter-
where velocity is lower, hence dw 4 d0 (see Fig. 7) (see the diffusion lengths discloses an increase with constant slope (n) at
experimental qualitative confirmation from Fig. 4). Here dw is small values of D (Do5  10  10 m2/s), nw on0 (where dw pznw ,
the inter-diffusion length at the wall (y¼ 7H/2) and d0 (y¼0) is respectively, d0 pzn0 Þ: We have to remark that the exponents nw and
the inter-diffusion length at the middle of the channel, both n0 are in this case smaller than the values reported by Ismagilov
lengths correspond to an iso-concentration line (in this case, et al. (2000) and Kamholz and Yager (2001): nw ¼0.24 (instead of
c(x, y) ¼0.1c0) within the plane z¼constant. 0.33) and n0 ¼0.27 (instead of 0.5), but in the mentioned papers the
In Figs. 8 and 9 the numerical 3D results of the steady exponents were obtained for different parameters, i.e. D¼10-9 m2/s,
concentration field (t Z30 s) are compared with the analytic W/H¼2.5 and Pebz/Hb1 (see also the review of Stone et al., 2004).

t = 35 s
t = 35 s
c < 0.1c0

c < 0.1 c0

10 s

t ≅ 6.3 s
t≅6s

10 s

t = 35 s 15 s
t = 35 s
c < 0.1c0
c < 0.1c0

Fig. 13. Time evolution (tr 35 s) of the iso-concentration line c¼ 0.1c0, at different z-coordinates: (a) z¼ 5 mm, (b) z¼ 10 mm, (c) z ¼20 mm, and (d) z ¼ 35 mm.
The domains with concentration co 0.1c0 at t ¼35 s (steady state) are marked (simulations performed at Pe ¼3000, respectively, Re¼ 1.85).
Author's personal copy

1970 D. Broboana et al. / Chemical Engineering Science 66 (2011) 1962–1972

From the present numerical results, one observes that at


D 45  10  10 m2/s the exponents are increasing, but their values
are not maintained constants along the flow direction of the
tested channel. As was expected, the increase of the diffusion
coefficient (i.e. smaller Péclet number, if velocity is kept constant)
determines the increase of the inter-diffusion lengths and, con-
sequently, the spreading of solute is faster in the channel (see also
Fig. 11) where the ‘‘butterfly effect’’ is represented.
One concludes that the 3D simulations produced fairly con-
sistent results with the measured data and offer a real description
of the investigated diffusion process, at least from the qualitative
point of view. The numerical distributions of concentration can
also provide valuable information about the transitory diffusion
process within the channel, in particular to characterize the
dynamics of the ‘‘butterfly effect’’ and the rate of solute spreading
at the wall.

5. Transient butterfly effect

One goal of the present study is to investigate the transitory


regime (t o35 s), in particular to determine the dynamics of the
iso-concentration patterns. In Figs. 12 and 13 are shown the
computed concentration distributions for the transitory diffusion
process (nominal case; at time t¼ 0 the solute enters the upper
branch of the Y-geometry).
The results shown in Fig. 13 disclose a different pattern of the
inter-diffusion length distribution (corresponding to c¼0.1c0), in
comparison to the steady results from Fig. 10. At each constant
z-coordinate, we remark the existence of ‘‘time thresholds’’ which
separate qualitatively the unsteady distributions of d0 and dw , e.g.
at totcr1 the concentration in the very vicinity of the walls
y¼ 7H/2 is co0.1c0, so the solute actually does not reach the
wall. The most important aspect to be mentioned is the existence
of the second critical time value, tcr2, which defines the time
corresponds to dw d0 (e.g. tcr2 ffi15 s in Fig. 13d). For tcr1 ot otcr2
there coexists two values for dw , both smaller than d0 ,at least one
being negative (see the topology of iso-concentration line at
t ¼10 s in Fig. 13d). Therefore, always a close domain with
co0.1c0 is present in vicinity of the upper corners of the flow
Fig. 14. Concentration rate variation along the width channel at y¼ 0, with
section, beside the region with c o0.1c0 located near the wall z-coordinate parameter at t ¼5 s and with time parameter at z¼ 0.01 m. The
x¼  W/2. For t 4tcr2 the pattern of iso-concentration lines original data are presented in Fig. 5.
became qualitatively similar to the steady case and the region
of co0.1c0 is present only in the vicinity of x¼  W/2.
The unsteady spectrum of concentration from Fig. 13, corro-
borated with the corresponding velocity and wall shear stress
distributions, offer the possibility to obtain detailed information
about transitory diffusion processes which take place at the
contact with the channel walls, relevant for applications as:
(i) surface adsorption (Das and Chakraborty, 2010), nano-particles
(molecules) capture (Munir et al., 2009), liquid–liquid micro-
extraction (Kikutani et al., 2009).

6. Final remarks and conclusions

The modeling of diffusion processes in micro-channel configura-


tions is important not only to determine the diffusion coefficient of a
specific constituent in a solution or to obtain the desired mixing
between different solutes, but also to control the diffusion in the
case of active interfaces laid on the channel wall. In the latest case,
the time variation of the concentration distribution during the
transitory flow regime is a phenomenon of major interest.
The present paper brings an original contribution in this Fig. 15. Numerical computed rate along the width channel at y¼ 0, with
direction, especially in the modeling of the transitory butterfly z-coordinate parameter (steady state).
Author's personal copy

D. Broboana et al. / Chemical Engineering Science 66 (2011) 1962–1972 1971

effect and the analyses of time evolution of the inter-diffusion The further investigations are focused on two directions: (i) to
length at the micro-channel walls. establish the influence on transitory diffusion of the non-symmetric
After the validation of the numerical procedure by the steady and unsteady velocities distributions at the entrance in the channel
experimental results, the study investigated numerically the and (ii) to describe the transitory hydrodynamics and diffusion in
evolution of concentration in the analyzed Y-branching geometry, the vicinity of porous surface, at the micro-channel walls.
from the onset of the flow until the stabilization of the stationary
spreading of the solute. The rates of concentration on different
direction within the channel and the corresponding fluxes of the Acknowledgements
dispersed solute can be also computed. In Fig. 14 are shown the
variations in space and time of the rate dc=dx: Of course, as This work was financially supported through CEEX-NANOINT
computations are approaching the steady state, the graphs are project (Romanian National Authority for Scientific Research—
approaching the normal symmetric distribution, more and more ANCS) and CNCSIS grant BD 73. Corneliu Balan acknowledges also
flat as the z-coordinate is increasing (see Fig. 15). the financial support of the National University of Singapore
A special aim of the paper was to obtain the description of the (EERSS program). Thorsten Wohland acknowledges funding by
transitory ‘‘butterfly effect’’ along the main channel. The time the Singaporean Ministry of Education (R-143-000-358-112). The
evolution of the iso-concentration patterns in the flow domain is authors are very thankful to Dr. Tiberiu Barbat for his advices and
a qualitative characteristic pattern of any diffusion process. In the support in obtaining the numerical results.
view of the authors, the simulation and calculus of transitory
diffusion in micro-channels has a direct application in developing
new techniques and applications, as the modeling, design and Annex
control of cell extraction/absorption processes through porous
walls of micro-channels (see also Zhu, 2009). See Figs. A1 and A2 for more details.

 = 0  = 0 (1 + 2.5 c)

7.5 s 7.5 s

6.3 s 6.5 s

10 s
10 s

15 s 15 s

20 s 20 s

30 s 30 s

Fig. A1. Time evolution of the iso-concentration line c ¼ 0.1c0 at z ¼35 mm (D ¼ 6  10  10 m2/s, V0 ¼4.8 mm/s), for constant viscosity coefficient (Z0) and variable viscosity
(dependent on the local concentration c, the Einstein formula).

D = 6·10-10 m2/s, V = 2.4 mm/s D = 12·10-10 m2/s, V = 4.8 mm/s

10 s 5s

15 s

10 s
30 s
15 s
30 s

Fig. A2. Time evolution of the iso-concentration line c¼ 0.1c0 at z ¼20 mm. For the same Pe-number, the iso-concentration lines have the same topology at time t¼ (V0/V)t0
(compare the corresponding iso-concentration lines from the two figures at the bold marked times).
Author's personal copy

1972 D. Broboana et al. / Chemical Engineering Science 66 (2011) 1962–1972

References Kamholz, A.E., Weigl, B.H., Finlayson, B.A., Yager, P., 1999. Quantitative analysis of
molecular interaction in a microfluidic channel: T-sensor. Anal. Chem. 71,
5340–5347.
Adeosun, J.T., Lawal, A., 2009. Numerical and experimental studies of mixing Kamholz, A.E., Yager, P., 2001. Theoretical analysis of molecular diffusion in
characteristics in a T-junction microchannel using residence-time distribution. pressure-driven laminar flow in microfluidic channels. Biophys. J. 80, 155–160.
Chem. Eng. Sci. 64, 2422–2432. Kamholz, A.E., Schilling, E.A., Yager, P., 2001. Optical measurement of transverse
Abonnenc, M., Josserand, Jacques, Girault, Hubert H., 2009. Sandwich mixer– molecular diffusion in microchannel. Biophys. J. 80, 1967–1972.
reactor: influence of the diffusion coefficient and flow rate ratios. Lab Chip 9, Kikutani, Y., Mawatari, K., Hibara, A., Kitamori, T., 2009. Circulation microchannel
440–448. for liquid–liquid microextraction. Microchim. Acta 164, 241–247.
Ajdari, A., Bontoux, N., Stone, H.A., 2006. Hydrodynamic dispersion in shallow Lee, N.Y., Yamada, M., Seki, M., 2005. Development of a passive micromixer based
microchannels: the effect of cross-sectional shape. Anal. Chem. 78, 387–392. on repeated fluid twisting and flattening, and its application to DNA purifica-
Akpa, B.S., Matthews, S.M., Sederman, A.J., Yunus, K., Fisher, A.C., Johns, M.L., tion. Anal. Bioanal. Chem. 383, 776–782.
Gladden, L.F., 2007. Study of miscible and immiscible flows in a microchannel Liu, Y.Z., Kim, B.J., Sung, H.J., 2004. Two-fluid mixing in a microchannel. Int. J. Heat
using magnetic resonance imaging. Anal. Chem. 79, 6128–6234.
Fluid Flow 25, 986–995.
Aubin, J., Ferrando, M., Jiricny, V., 2010. Current methods for characterising mixing
Mata, C., Longmire, E.K., McKenna, D.H., Glass, K.K., Hubel, A., 2008. Experimental
and flow in microchannels. Chem. Eng. Sci. 65, 2065–2093.
study of diffusion based extraction from a cell suspension. Microfluid Nano-
Balan, C., Marculescu, C., Calin, A., 2008. Modelling of interfaces between
fluid 5, 529–540.
inmiscible fluids with the VOF method. In: Susan-Resiga, R., Bernad, S.,
Morf, W.E., van der Wal, P., de Rooij, N.F., 2008. Computer simulation and theory of
Muntean, S. (Eds.), Vortex Hydrodynamics and Applications. Eurostampa
the diffusion- and flow-induced concentration dispersion in microfluidic
Publ., Timisoara, pp. 195–203.
devices and HPLC systems based on rectangular microchannels. Anal. Chim.
Balan, C.M., Broboana, D., Balan, C., 2010. Mixing process of immiscible fluids in
Acta 622, 175–181.
micro-channels. Int. J. Heat Fluid Flow. doi:10.1016/j.ijheatfluidflow.2010.06.008.
Munir, A., Wang, J., Li, Z., Zhou, H.S., 2009. Numerical analysis of a magnetic
Barz, D.P.J., Zadeh, H.F., Ehrhard, P., 2008. Laminar flow and mass transport in a
nanoparticle-enhanced microfluidic surface-based bioassay. Microfluid Nano-
twice–folded microchannel. AIChE J. 54, 381–393.
fluid. doi:10.1007/s10404-009-0497-3.
Benninger, R.K.P., Hofmann, O., McGinty, J., Requejo-Isidro, J., Munro, I., Neil, M.A.A.,
Munson, M.S., Hawkins, Kenneth R., Hasenbank, Melissa S., Yager, Paul, 2005.
deMello, A.J., French, P.M.W., 2005. Time-resolved fluorescence imaging of
Diffusion based analysis in a sheath flow microchannel: the sheath flow
solvent interactions in microfluidic devices. Opt. Express 13 (16), 6275–6285.
T-sensor. Lab Chip 5, 856–862.
Beutel, D., 2003. Mixing in microchannels. Ph.D. Thesis, Harvey Mudd College, USA.
Pan, X.T., Yu, H., Shi, X.K., Korzh, V., Wohland, T., 2007a. Characterization of flow
Bransky, A., Korin, N., Levenberg, S., 2008. Experimental and theoretical study of
direction in microchannels and zebrafish blood vessels by scanning fluores-
selective protein deposition using focused micro laminar flows. Biomed.
cence correlation spectroscopy. J. Biomed. Opt. 12 (1), 014034.
Microdevices 10, 421–428.
Pan, X.T., Foo, W., Lim, W., Fok, M.H.Y., Liu, P., Yu, H., Maruyama, I., Wohland, T.,
Brenner, H., Ganesan, V., 2000. Molecular wall effects: are conditions at a
2007b. Multifunctional fluorescence correlation microscope for intracellular
boundary ‘‘boundary conditions’’? Phys. Rev. E 61, 6879–6897.
and microfluidic measurements. Rev. Sci. Instr. 78, 053711.
Chung, T.J., 2003. Computational Fluid Dynamics. Cambridge University Press.
Sadr, R., Hohenegger, Ch., Li, H., Mucha, P.J., Yoda, M., 2007. Diffusion-induced bias
Dambrine, J., Géraud, B., Salmon, J.-B., 2009. Interdiffusion of liquids of different
in near-wall velocimetry. J. Fluid Mech. 577 (443–456).
viscosities in a microchannel. New J. Phys. 11, 075015.
Shih, T.R., Chung, C.K., 2008. A high-efficiency planar micromixer with convection
Das, S., Chakraborty, S., 2010. Augmented surface adsorption characteristics by
and diffusion mixing over a wide Reynolds number range. Microfluid Nano-
employing patterned microfluidic substrates in conjunction with transverse
fluid 5, 175–183.
electric fields. Microfluid Nanofluid 8, 313–327.
Soleymani, A., Kolehmainen, E., Turunen, I., I., 2008. Numerical and experimental
Estévez-Torres, A., Gosse, C., Le Saux, T., Allemand, J.-F., Croquette, V., Berthou-
mieux, H., Lemarchand, A., Jullien, L., 2007. Fourier analysis to measure investigations of liquid mixing in T-type micromixers. Chem. Eng. J. 135S,
diffusion coefficients and resolve mixtures on a continuous electrophoresis S219–S228.
chip. Anal. Chem. 79, 8222–8231. Smith, R.L., Demers, R.L., Collins, S.D., C.J., 2010. Microfluidic device for the
Ferziger, J.H., Peric, M., 1999. Commputational Methods for Fluid Dynamics. combinatorial application and maintenance of dynamically imposed diffu-
Springer, Berlin. sional gradients. Microfluid Nanofluid 9, 613–622.
Gan, H.Y., Lam, Y.C., Nguyen, N.T., Tam, K.C., Yang, C., 2007. Efficient mixing of Stone, H.A., Stroock, A.D., Ajdari, A., 2004. Engineering flows in small devices:
viscoelastic fluids in a microchannel at low Reynolds number. Microfluid microfluidics toward a Lab-on-a-chip. Annu. Rev. Fluid Mech. 36, 381–411.
Nanofluid 3, 101–108. Sullivan, S.P., Akpa, B.S., Matthews, S.M., Fisher, A.C., Gladden, L.F., Johns, M.L.,
Garbe, C.S., Roetmann, K., Beushausen, V., Jähne, B., 2008. An optical flow MTV 2007. Simulation of miscible diffusive mixing in microchannels. Sensors
based technique for measuring microfluidic flow in presence of diffusion and Actuators B 123, 1142–1152.
Taylor dispersion. Exp. Fluids 44, 439–450. Sun, C.-L., Sie, J.-Y., 2010. Active mixing in diverging microchannels. Microfluid
Garvey, J., Newport, D., Lakestani, F., Whelan, M., Joseph, S., 2008. Full field Nanofluid 8, 485–495.
measurement at the micro-scale using licro-interferometry. Microfluid Nano- Tian, L.Q., Wohland, T., 2008. Measuring reactions in microchannels. NUS Internal
fluid 5, 77–87. Research Report, P07265.
Glasgow, I., Lieber, S., Aubry, N., 2004. Parameters influencing pulsed flow mixing Truesdell, R.A., Vorobieff, P.V., Sklar, L.A., Mammoli, A.A., 2003. Mixing a contin-
in microchannels. Anal. Chem. 76 (16), 4825–4832. uous flow of two fluids due to unsteady flow. Phys. Rev. E 67, 066304.
Gorman, B.R., Wikswo, J.P., 2008. Characterization of transport in microfluidic Wu, Z., Nguyen, N.-T., 2005. Hydrodynamics focusing in microchannels under
gradient generators. Microfluid Nanofluid 4, 273–285. consideration of diffusive dispersion: theories and experiments. Sensors
Goulpeau, J., Lonetti, Barbara, Trouchet, Daniel, Ajdari, Armand, Tabeling, Patrick, Actuators B 107, 965–974.
2007. Building up longitudinal concentration gradients in shallow microchan- Yamaguchi, Y., Ogura, D., Yamashit, K., Miyazaki, M., Nakamura, H., Maeda, H.,
nels. Lab Chip 7, 1154–1161. 2006. A method for DNA detection in a microchannel: fluid dynamics
Glatzel, T., Litterst, C., Cupelli, C., Lindemann, T., Moosmann, C., Niekrawietz, R., phenomena and optimization of microchannel structure. Talanta 68, 700–707.
Streule, W., Zengerle, R., Koltay, P., 2008. Comput. Fluids 37, 218–235. Zhu, X., 2009. Micro/nanoporous membrane based gas–water separation in
Hartmann, C., Sasso, L., 2007. Convection–diffusion in microchannels, Internal microchannel. Microsyst. Technol. 15, 1459–1465.
report, DTU, Lyngby.
Ismagilov, R.F., Stroock, A.D., Kenis, P.J.A., Whitesides, G., Stone, H.A., 2000.
Experimental and theoretical scaling laws for transverse diffusive broadening
in two-phase laminar flows in microchannels. Appl. Phys. Lett. 76, 2376–2378.

Das könnte Ihnen auch gefallen