Upload
others
View
25
Download
0
Embed Size (px)
Citation preview
1
COMPUTATIONAL FLUID DYNAMICS (CFD) SIMULATIONS APPLIED
TO THE DESIGN OF PASSIVE MIXING MICROFLUIDIC SYSTEMS
by
SERGIO LEONARDO FLOREZ GONZALEZ
A thesis submitted to the Universidad de
los Andes in partial
fulfilment for the degree of
Master in Electronic Engineering
Supervised by Dr. Johann F. Osma and Dr. Juan Carlos Cruz
School of Engineering
Department of Electrical and Electronic Engineering
2020
2
Abstract
Computational Fluid Dynamics CFD simulations of microfluidics and microsystems allow
to rapidly design and test the performance of multiple prototypes by identifying the
variables that impact the most the efficiency of the device for a particular task. This work
aimed at implementing multiphysics simulations for helping to prototype two different
microfluidic systems for passive mixing. First, a low-cost microsystem for the synthesis of
magnetite nanoparticles via chemical co-precipitation. Second, a toroidal microreactor to
conduct the continuous enzyme-based degradation of dyes in wastewater using magnetic
nanoparticles. The simulations show different changes in velocity profiles, shear rate, and
homogeneous mixing in each device tested. These variations in the behavior of the fluid can
improve the formation of nanoparticles by controlling their growth. Likewise, the torus
microsystem, with the implementation of a permanent magnetic field, increases the
retention time of the nanoparticles in the device, increasing the interaction and treatment of
the wastewater. The use of computational tools allows to quickly and economically
determine the impact of changes in the geometrical configuration of the systems on their
overall performance.
3
Contents
1. Introduction .............................................................................................................................. 4
2. Materials and Methods ............................................................................................................ 5
2.1. Tutorial videos of Comsol Multiphysics simulations for microsystems .................. 5
2.2. Micromixer synthesis of magnetite nanoparticles: geometry design and
simulation ...................................................................................................................................... 6
2.3. Torus microreactor: geometry design and simulation ................................................ 8
3. Results and Discussion .......................................................................................................... 10
3.1. Tutorials videos .............................................................................................................. 10
3.2. Micromixer synthesis of magnetite nanoparticles ..................................................... 10
3.3. Torus Microreactor ......................................................................................................... 14
4. Conclusions ............................................................................................................................. 17
5. Acknowledgement ................................................................................................................. 18
6. Annexes ................................................................................................................................... 18
6.1. Comsol Introduction. ..................................................................................................... 18
6.2. Interface. .......................................................................................................................... 18
6.3. Pos-processing. ............................................................................................................... 18
6.4. CFD modeling focusing on microfluidics. .................................................................. 18
6.5. Mass transfer model focusing on microfluidics. ........................................................ 18
6.6. Heat transfer model focusing on microfluidics. ........................................................ 18
6.7. Supplementary information.......................................................................................... 18
7. References ................................................................................................................................ 18
4
1. Introduction
Since the 1980s, with the advances in microelectromechanical systems (MEMS), the field of
microfluidics has had an exponential growth due to the development of easier and cheaper
ways of manufacturing at the microscale [1]. Today, this research field has reached an
important level of maturity and has been implemented by different industries including
food, pharmaceutical and biomedical [2]–[5]. One of the most attractive features of these
devices is that they can carry out chemical processes with low reagent consumption due to
their small sample volumes [1]. There are different device configurations to perform several
types of analysis and functions including sampling, preparation of the sample, analysis and
processing of the collected data. These systems are known as miniaturized Total Analysis
Systems, and an important number of them have been incorporated into in vitro testing
schemes of biological systems (µTAS) [1][6]. Moreover, an emerging trend is to assemble
microsystems capable of mimicking the performance of organs at the physiological level,
which are known as organs on a chip [7]. All these devices facilitate the processing of
samples, thereby reducing costs and processing times, which is very attractive for different
applications such as disease diagnosis [8], manipulation of materials at the micro- and nano-
scales [9][10], and the generation of complex and multifunctional nanocompounds and
encapsulates [11][12], among many others.
To design and test microsystems before their final manufacture, there are several in silico
tools including rapid and economic prototyping [13][14], hydraulic circuit analyses [15][16]
and Computational Fluid Dynamics (CFD) [17]–[22]. CFD simulations of the microsystems
allow to rapidly design and test the performance of multiple prototypes by identifying the
variables that impact the most the efficiency and overall performance of the device for a
particular task [17][19][20]. CFD simulations require to discretize the computational
domain, which can be done through several methods including Finite Difference Method
(FDM), Finite Volume Method (FVM) and Finite Element Method (FEM). In the case of
FDM, discretization relies on an approximation by Taylor series expansion[23]. FVM uses
an integral approach to solve the equations and discretizes the domains into volumes where
the equations are simultaneously solved [23] Finally, the Finite Element Method (FEM),
which also uses an integral scheme but in this case the domain is discretized into elements
that form intersections or nodes where the equations are solved [23]. All these methods
allow to solve partial differential equation systems through algebraic equations [24].
Commercial software packages have facilitated the implementation of FEM methods. A
popular one is Comsol Multiphysics®, which incorporates several physics that can be
coupled to investigate rather complex situations. Some authors have conducted
multiphysics simulations to optimize the performance of the systems. These improvements
vary depending on the type of microsystem but can range from the design of the
microsystem [25][26] to the mode of operation [27] and allow the integration of governing
equations for different phenomena such as fluid mechanics, heat transfer, mass transport,
reaction kinetics, electrochemistry, etc [28]–[30]. This means that the governing equations of
different physics can be solved simultaneously, allowing the multiparametric analysis of a
myriad of variables and their relationships [31][32].
5
Microsystems geometries vary depending on their function. For instance, they can be
intended for the separation of particulate matter, for drop generation or for mixing
purposes. In the case of separators, they can be tailored to handle inorganic or polymeric
microparticles, cells, or in general any micro-object by exploiting differences in size or
density [33]–[35]. The drop generators can be operated to produce homogeneous size drops
for different applications ranging from encapsulation of pharmacological compounds to
aromas or fragrances with application in cosmetics or food industries [36]–[38]. Likewise,
the micromixers allow to carry out mixtures of different components or reagents involved
in chemical reactions, blend preparation or sample preparation [39][40]. A major challenge
when working with microfluidic systems is the difficulty to generate efficient mixing
patterns mainly due to the low Reynolds numbers (Re ≈ 1). To tackle this issue, it is necessary
to generate turbulence through special geometries such as Zigzag channels, 3-D serpentine
structures, and twisted channels [41]. Micromixing can be achieved either through active
mixing or passive mixing [42][43]. In the case of active mixing, integration of components is
accomplished via an external energy source such as ultrasound, acoustic vibrations, small
impellers, or electrokinetic instability. In contrast, passive micromixing requires a device to
intimately put in contact the components through disturbances in the mixing patterns.
These devices include chaotic flow configuration, flow recirculation configuration, colliding
jet, split and recombine flow configurations [42][43].
This work aimed atimplementing multiphysics simulations for helping to prototype two
different microfluidic systems for passive mixing. This approach was explored for two
different microsystems. First, a low-cost microsystem for the synthesis of magnetite
nanoparticles via chemical co-precipitation. The main objective was to evaluate whether this
synthesis was attainable in a continuous operation mode and that the obtained material
exhibited the appropriate size distribution and nanoscale morphology. This was
accomplished by coupling the CFD, chemistry and transport of diluted species modules.
Second, toroidal microreactor to conduct the continuous enzyme-based degradation of dyes
using magnetite nanoparticles. The objective was to evaluate the interaction of the magnetite
nanoparticles with the fluid and the residence time and retention within the system. In both
cases, the performance of the microsystems was determined in terms of mixing efficiency.
This was accomplished by coupling the CFD and particle tracing modules of Comsol
Multiphysics®.
2. Materials and Methods
2.1. Tutorial videos of Comsol Multiphysics simulations for microsystems
As part of the training for future students with interest in developing research on the field
of microsystems, we prepared a series of video tutorials for implementing Multiphysics
simulations in Comsol Multiphysics®. These videos explain in detail the steps to setup a
simulation. This includes the design of the geometry using CAD tools as well as the
processing and postprocessing of collected data. The pre-processing includes the definition
of physics, implementation of boundary conditions, initial conditions, properties, meshing
6
and convergence criteria and solvers. In the case of post-processing, we discussed data
visualization, types of graphics, data operations, use of boundary conditions and exporting
data. All videos focused on evaluating the performance of the microsystems via coupling
the governing equations for transport of momentum, transport of species, chemical
reactions, and transport of thermal energy.
The first module is a brief introduction to multiphysics simulations, the finite element
method, governing equations, initial conditions, boundary conditions, symmetry and mesh
convergence. The second module studies the CAD tool of Comsol, materials, physics, mesh
and solvers. The third module studies post-processing, handling of large data sets, types of
graphics and data processing. The fourth module is a worked example of a microsystem
with a serpentine mixing topology. In this case, the governing equations of momentum
transport are introduced. Additionally, a parametric study was carried out in the
simulation. In the fifth module, mass transport was studied where the mass transport
governing equations were introduced, and the concepts of diffusion and convection were
explained. Finally, the sixth module is a microsystem heating system based on the Joule
effect. In this final study case, the energy transport equations were introduced.
2.2. Micromixer synthesis of magnetite nanoparticles: geometry design and simulation
Three micromixing geometries were proposed and analyzed via Comsol Multiphysics 5.3:
a serpentine-based topology (SB), a triangular-based topology (TB), and a 3D-based
topology (3DB) that involves sudden changes in flow direction by 90° elbows, allowing
turbulent mixing patterns (See Figure 1). The depth of systems was 1 mm. All microfluidic
systems were design to handle the same volume i.e., 300 mm3.
These 3 different geometries were implemented in Comsol by coupling the computational
fluid dynamics (CFD), laminar flow, chemical reaction engineering and chemical species
transport modules. To study both fluid flow and extent of reaction during the synthesis of
the nanoparticles, we coupled the Computational Fluid Dynamics (CFD), chemistry and
transport of diluted species modules. Eq.1 describes the chemical reaction to form magnetite
by the coprecipitation method.
7
Figure 1. Microfluidic mixers geometry. A) Serpentine-based mixer, B) Triangular-based mixer, C) 3D-based mixer. (1-3.
Inflow for FeCl2 and FeCl3, 2. Inflow for NaOH and 4. Outflow)
𝐹𝑒𝐶𝑙2 + 2𝐹𝑒𝐶𝑙3 + 8𝑁𝑎𝑂𝐻 → 𝐹𝑒3𝑂4 + 8𝑁𝑎𝐶𝑙 + 4𝐻2𝑂 Eq.1
Also, the mixing phenomena were studied with the aid of the laminar flow and transport
of diluted species modules. In the former, the transport of momentum equation (Eq.2) is
solved to calculate the velocity profile. . In the latter, the conservation of species
was used (Eq.3). The velocity profile calculated by solving Eq.2 is input in the convective
term of Eq.3.
𝜌[(𝑉. ∇)𝑉] = −∇𝑃 + 𝜌𝑔 + 𝜇∇2𝑉 Eq.2
Where V is the vector velocity of the fluid, P is the pressure, ρ is the density of the
fluid, g is the gravity and μ is the viscosity of the fluid
∇(−𝐷𝑖∇𝑐𝑖) + 𝑉∇𝑐𝑖 = 𝑅𝑖
Eq.3
Where Di is the diffusion constant of each species, ci is the concentration of each species and
Ri is the reaction rate for each species.
8
All the simulation parameters are summarized in the Table 1. The 3 geometries were
subjected to a mesh convergence analysis to determine the minimum number of elements
necessary to obtain a stable solution . For this study, 5 random measurement points were
selected along the computational domain and the change in magnitude of speed was
evaluated as the number of mesh elements increased. As a convergence criterion, it was
determined that the velocity magnitude change obtained for a meshing level and the next
one was below 3%. An unstructured mesh with tetrahedral elements was
generated. A stationary study was run with the direct solver PARDISO that allows to
parallelize processes solving large symmetric or structurally symmetric dispersed linear
systems of equations in shared memory multiprocessors [44].
Table 1. Parameters of reaction simulations
Parameters Value Units
Density of the fluid 1000 Kg/m3
Viscosity of the fluid 1 mPa.s
Rate constant 1*10-24 m30/(s*mol10)
FeCl2 molar mass 0.199 Kg/mol
FeCl3 molar mass 0.270 Kg/mol
NaOH molar mass 0.040 Kg/mol
Fe3O4 molar mass 0.232 Kg/mol
NaCl molar mass 0.058 Kg/mol
H2O molar mass 0.018 Kg/mol
FeCl2 inflow concentration 100 mM
FeCl3 inflow concentration 200 mM
NaOH inflow concentration 800 mM
Central inlet normal inflow rate 1 ml/min
Lateral inlets normal inflow rate Outlet
0.5
1
ml/min
Atm
2.3. Torus microreactor: geometry design and simulation
Three different geometries were studied in silico via Comsol Multiphysics 5.3® for the torus
microreactor. Figure 2 shows such geometries, namely, one-loop, two-horizontal-loop, and
two-vertical-loop microreactors. Each microreactor was equipped with a cylindrical
neodymium magnet with a diameter of 5 mm and a height of 6 mm. The magnet was directly
inserted in the hole created by the loop (Figure 2). The channel width of each microreactor
is 0.5 mm while its depth is 1 mm in the sections previous to the loop. Then, in the section
where the magnet is inserted the depth increases to 2 mm. Finally, it is reduced again to a
height of 1 mm as the fluid exits the loop. This device was conceptualized by considering
its prototyping through the assembly of layers of PMMA. For each case, the Computational
Fluid Dynamics (CFD) (Eq. 4) and Magnetic Field, no current (MF) (Eq. 5-6) modules of
Comsol were coupled. To evaluate the impact of magnetic fields on the transport of the
9
magnetite nanoparticles, the Particle Tracing (PT) module was implemented solely for the
one-loop microreactor.
Figure 2. Simulation geometries for the studied microreactors. A) One-loop, B) Two-horizontal-loop and C) Two-vertical-
loop. The numbers 1, 2 and 3 in the figure represent the inputs, outputs, and the neodymium magnet, respectively.
The laminar flow module (Eq. 4) was used here to describe the fluid flow due to the low
Reynolds number (Re) calculated for the microreactors (Re = 4.4). The inflow to the
microreactor was 12 mL/h while the output pressure was set to 1 atmosphere. Density and
viscosity of water were assumed as the properties of the flowing fluid. The non-slip
boundary condition was imposed at the walls of the microchannels. The boundary
conditions are summarized in Figure 3.
0 = −𝛻𝑃 + 𝜇𝛻2𝑉 Eq. 4
𝐵 = 𝜇0𝜇𝑟𝐻 Eq. 5
𝐵 = 𝜇0𝜇𝑟𝐻 + 𝐵𝑅 Eq. 6
Where ∇𝑃, is the pressure gradient in the fluid, µ is the viscosity of the fluid and 𝑉 is the
velocity of the fluid. For the equations of MF, 𝐵 is a magnetic flux density, 𝜇0 is the magnetic
permeability of the vacuum, 𝜇𝑟 is the magnetic permeability of the fluid, 𝐻 is the magnetic
field and 𝐵𝑅 is the remanent flux density of the neodymium magnet. The PT module was
used to study the dynamics of the magnetic particles within the microreactor. The particle
diameter was assumed as 1.2 µm with a density of 5180 kg/m3. The particle size was decided
according to previous experimentation and small agglomerations of NPs forming clusters.
Additionally, the magnetic permeability of the particle was set to 5000 H/m. The particles
were delivered at the main inlet of the device. The PT was solved in a time-dependent study
for 10 seconds using 0.1s steps. In this case, a projected conjugate gradient iterative solver
was chosen due to the high demand for computational resources. Also, a parametric
analysis was performed to evaluate the impact of the remanent flux density of the
neodymium magnet on the particles transport. Equations Eq.7 to Eq.10 describe the forces
experienced by a particle. This study considered the interactions between the particles and
the fluid (drag force) and the magnetic attraction between particles. Particle-particle
interactions, lifting force and Brownian motion were disregarded due to the complexity of
the model and the need for extra computational resources:
10
𝐹𝐷 =1
𝜏𝑃𝑚𝑃(𝑈 − 𝑉) Eq. 7
𝜏𝑃 =𝜌𝑃𝑑𝑃
2
18𝜇 Eq. 8
𝐹𝑚𝑝 = 2𝜋𝑟𝑝3𝜇0𝜇𝑟𝑘∇𝐻2 Eq. 9
𝑘 =𝜇𝑟𝑝 − 𝜇𝑟
𝜇𝑟𝑝 + 2𝜇𝑟 Eq. 10
Where 𝐹𝐷 is the drag force, 𝑚𝑃 is the mass of the particle, 𝑈 is the velocity of the particle, V
is the fluid velocity, 𝜌𝑃 is the density of the particle and 𝑑𝑝 is the particle diameter. 𝐹𝑚𝑝 is
the magnetophoretic force, 𝑟𝑃 is the particle radius and 𝜇𝑟𝑝 is the magnetic permeability of
the particle. As described before, the meshing was subjected to a convergence analysis to
determine the minimum number of elements necessary to arrive to a meaningful solution.
Figure 3. Boundary conditions and mesh for simulation.
3. Results and Discussion
3.1. Tutorials videos
The annexes section includes 6 introductory videos to multiphysics simulations in Comsol
Multiphysics. Each presentation includes a video tutorial, the video presentation and the
example model implemented in Comsol Multiphysics 5.3®.
3.2. Micromixer synthesis of magnetite nanoparticles
Figures S1, S2 and S3 in supplementary information show the mesh convergence analysis
for SB, TB and 3DB, respectively. In the case of SB, converge was achieved with 300,000
mesh elements. For TB, this was achieved with 200,000 mesh elements while for the 3DB
400,000 were required. All simulations were performed on a computer with an AMD Ryzen
5 2100 Mhz processor, 4-core and 16 GB of RAM.
11
Although all the input flows to the systems are the same, they develop different velocity
and shear rate profiles due to the differences in geometries. Figure 4 show the velocity
profile for each micromixer. The SB microsystem shows no noticeable zones of dead
volume. However, the TB and 3DB microsystems have zones of dead volumes along sharp
corners of the geometries. This could be due to low pressure areas of the microsystem where
reagents or air bubbles will begin to accumulate in the corners [45][46]. The SB and TB
microsystems achieved higher maximum velocities compared to the 3DB. Also, the TB
developed a velocity profile that resembles that of the SB system. This is due to the
sinusoidal flow pattern that we can see in Figure 4.B. Due to the generation of the dead
zones described above, the fluid tends to move to areas with less hydrodynamic resistance,
generating such flow pattern.
Figure 4. Velocity profiles for the three studied micromixers. A) Serpentine-based mixer, B) Triangular-based mixer, C)
3D-based.
Figure 5 shows a shear rate in each microsystem. The SB and TB developed shear rate values
that approached similar maximum values of about 15 s-1. In contrast, the 3DB system has
some areas with high shear rate (around 80 s-1). This is due to the smaller width of the
channels and the abrupt changes in flow direction along the path. All these phenomena
affect the formation of NPs and their growth rate [47][48]. Although there is a lack of studies
relating the formation of nanoparticles and the shear rate, some studies suggest that by
increasing the shear rate in the system the energy barrier to form crystalline structures
increases as well [48]. This suggests that the nucleation rates strongly depend on the shear
rate. At low shear rates the nucleation rate increases due to the increase in advective
12
transport while at high shear rates the opposite occurs [48]. The competition between these
two processes results in an intermediate regime where a maximum nucleation rate might
be expected [48].
Figure 5. Shear rate values for the studied micromixers. A) Serpentine-based mixer, B) Triangular-based mixer, C) 3D-
based.
Figure 6 shows the concentration of NPs along the microchannels. NPs tend to accumulate
in the corners of the TB and 3D geometries. This zones match those of the previously
identified dead volumes, which can be seen in Figure 4. The TB system exhibits the highest
accumulation level of NPs in the corners, which can be attributed to inefficient mixture of
reagents. The SB system shows more efficient mixing as evidenced by the absence of dead
volumes. Finally, the 3DB system, although it has several dead volume zones, the abrupt
changes in the flow direction appears to promote a better interaction between the reacting
species. It is important to note that all 3 configurations arrive to similar concentrations of
NPs (i.e., SB= 23.22 mM, TB= 22.43 mM and 3DB= 23.18 mM). This suggests that the flow
rates, dimensions, and channel features are adequate to produce the magnetite NPs.
13
Figure 6. Fe3O4 NPs concentration after the reaction for each of the studied micromixers. A) Serpentine-based mixer, B)
Triangular-based mixer, C) 3D-based mixer.
The magnetite NPs synthesis process can be studied using a thermodynamic nucleation
model [47][48]. Nucleation is the first step to obtain a new thermodynamic (solid) phase.
The classical theory of nucleation (CNT) is based on the changes of Gibbs free energy during
the reaction (Eq.11). Where ∆𝐺 is a Gibbs free energy of the cluster, 𝑟 is a radio of the cluster,
|∆𝐺𝑣| the difference in Gibbs bulk free energy per unit volume and 𝛾 the surface energy per
unit area [47]. relating the favorable binding and formation energy of the NP cluster per unit
of volume (bulk free energy) and surface energies that disfavored the stability of such cluster
[48].
∆𝐺 = −4
3𝜋𝑟3|∆𝐺𝑣| + 4𝜋𝑟2𝛾 Eq. 11
When a critical size is reached in the cluster after overcoming the energy barrier, the
assembly becomes a stable crystal and the growth process begins by aggregating more
precursors from the solution [47]. If this energy barrier is not overcame, it is most likely that
the cluster fails to form a stable crystal, and eventually disintegrates into its precursors [47].
This suggests that by increasing the shear rate during the synthesis reaction, their nucleation
process is affected, and therefore, their final size. One probable explanation for this
phenomenon is that, by increasing the energy barrier, we prevent cluster formation very
quickly, which would lead to a process of growth by inducing aggregation [48][47]. There
14
are a few studies where this phenomenon has been already described, but in some of them
it has been suggested that the shear rate is directly related to the aggregation rates [48].
According to this notion, only particles that manage to overcome this energy barrier to
remain stable in the suspension.
3.3. Torus Microreactor
Figure S4 in supplementary information shows the convergence plot for a one-loop
microreactor. In this case, the system reaches convergence with 60,000 mesh elements for
the fluid domain. To preserve the convergence in the other simulation domains, the element
size was kept constant (0.18 mm).
The velocity profiles and magnetic field fluxes for the 3 proposed devices are shown in
Figure 7 and Figure 8. Dead zones of low velocities (blue color) are observable in the 3
configurations, however, the two-vertical-loop microreactor shows a larger dead zone
compared with the other 2 devices. This can be explained by the relatively important
changes in height as the fluid passes through the loops. In addition, the fluid decelerates as
a result of changes in the cross-sectional area of the microchannels.
Figure 7. Velocity profile for the three different configurations under study. A) One-loop, B) Two-horizontal-loop
and C) Two-vertical-loop.
The magnetic field flux results show a uniform distribution around the magnet for the case
of the single loop (Figure 8.A). For the two magnets arranged in a horizontal configuration
(Figure 8.B), there is no interaction between the field lines produced by the magnets. Finally,
the configuration of two-vertical-loop (Figure 8.C) showed the greatest field interaction.
Figure 8. Magnetic field flux for the three different configurations under study. A) One-loop, B) Two-horizontal-
loop and C) Two-vertical-loop.
The time evolution of the particle's transport within the one-loop microreactor was studied
under varying intensities of the applied magnetic field (i.e., 0, 50, 100, 200, 300, 350, 500, and
15
1000 mT). The simulations were ran for a total time of 30 seconds with data collection at 1,
5, 15 and 25 s (See ¡Error! No se encuentra el origen de la referencia.9). For these
simulations, each particle trajectory is random and driven by the exerted drag and
magnetophoretic forces. As the particles are attracted to the magnet and due to the interplay
of involved forces in the loop, they follow a parabolic trajectory until they finally stick to
wall of the microreactor channel. For some particles, the magnetic field will not be enough
to retain them within the loop. This unique trajectories have been exploited by others for the
separation and manipulation of magnetic particles within microchannels [49][50]. In our
case, the movement is enough to perturb the laminar flow, thereby generating mixing
patterns that are useful to promote the intimate interaction of the nanoparticles with other
components present in solution. This was also evidenced by a better suspension of the
nanoparticles as the magnetic field was increased.
Total retention of the particles was achieved for the fields of 500 and 1000 mT, as the particles
remain statically attached to the walls of the microreactor. To support this result, the
particle's loss and retention ratio were analyzed by counting the particles leaving the system
through the microreactor's outlet during the 30 seconds of simulation (See Figure 1010). The
experimentally applied magnetic field by the magnet was of 349.23 mT, which is close to the
one simulated at 350 mT. In this case, the particles retention approached 96.83% while the
loss ratio was 3.17%. The actual particles loss ratio obtained experimentally was between 13
- 20%, which was closer to the results obtained in silico at 300 mT where the retention ratio
was 82.67% and the loss ratio approached 17.33%. The reduction in the actual strength of
the magnetic field can be explained by the marked difference in the medium surrounding
the fluid computationally (i.e., air) and the actual medium (i.e., PMMA walls), which
attenuates the applied magnetic field.
16
Figure 9. Particle distribution in the One-Loop device at a magnetic field intensity: A) 0mT, B) 50mT, C) 100mT, D)
200mT, E) 300mT, F) 350mT, G) 500mT and H) 1000mT.
17
Figure 10. Particles (■) loss and (■) retention ratio analysis at the microreactor's outlet after 30 seconds of
simulation.
4. Conclusions
The use of multiphysics simulations is a very useful tool for rapid prototyping of
microsystems. Given the impossibility of manufacturing complex and expensive devices,
mathematical models allow experimentalists to give a prediction regarding performance
that is close to reality. This is accomplished by coupling several physics that account for the
interplay of different physical phenomena occurring simultaneously. The success of the
simulation, however, depends on the availability of physicochemical properties, numerical
method selected, mesh size and convergence, the proper definition of boundary and initial
conditions, among many others. The prepared video series accompanying this document
presents several study cases where the physical situations are properly modeled and
simulated in Comsol. The obtained results are also plotted according to the investigated
variable interactions to withdraw compelling conclusions and make sound predictions.
Here we illustrate the use of Multiphysics simulations to study the performance of several
microfluidic devices in silico prior to their prototyping. According to our studies, on the one
hand, we identified the presence of dead volumes in geometries with sharp edges, however,
the achieved shear rate values appear sufficient for the synthesis of nanoparticles with
homogeneous size distributions and morphologies. On the other hand, the torus
microreactor configuration equipped with a magnetic field source prolongs theresidence
time of the nanoparticles within the system without inducing detrimental perturbations in
the mixing patterns. This leads to a longer contact time between the nanoparticles and other
substances present in the medium. Even though the predictions might be closer to reality,
often the complexity of the real system adds extra phenomena that are not necessarily
addressed by the Multiphysics modeling. For this reason, it is always advisable to calibrate
the models with experimental data prior to the next round of prototype manufacture and
testing. Likewise, other aspects must also be considered when manufacturing the
18
microsystems such as their cost, ease of manufacture and instrumenting, and manufacturing
precision required.
5. Acknowledgement
The authors would like to thank the Clean Room laboratory of the Department of Electrical
and Electronic Engineering, for access to their facilities laboratories and license of the
software is gratefully acknowledged.
6. Annexes
6.1. Comsol Introduction.
6.2. Interface.
6.3. Pos-processing.
6.4. CFD modeling focusing on microfluidics.
6.5. Mass transfer model focusing on microfluidics.
6.6. Heat transfer model focusing on microfluidics.
6.7. Supplementary information.
7. References
[1] N. Convery and N. Gadegaard, “30 years of microfluidics,” Micro and Nano Engineering,
vol. 2. Elsevier B.V., pp. 76–91, Mar. 01, 2019, doi: 10.1016/j.mne.2019.01.003.
[2] M. Tokeshi, Ed., Applications of Microfluidic Systems in Biology and Medicine, vol. 7.
Singapore: Springer Singapore, 2019.
[3] M. Aa, A. Sa, and Adali T, “Mekonen et al Microfluidics Devices Manufacturing and
Biomedical Applications,” J Biosens Bioelectron, vol. 10, p. 1, 2019, doi: 10.4172/2155-
6210.1000265.
[4] H. A. Santos, D. Liu, and H. Zhang, Microfluidics for pharmaceutical applications: From
nano/micro systems fabrication to controlled drug delivery. Elsevier, 2018.
[5] S. He, N. Joseph, S. Feng, M. Jellicoe, and C. L. Raston, “Application of microfluidic
technology in food processing,” Food and Function, vol. 11, no. 7. Royal Society of
Chemistry, pp. 5726–5737, Jul. 01, 2020, doi: 10.1039/d0fo01278e.
[6] R. M. Guijt and A. Manz, “Miniaturised total chemical-analysis systems (ΜTAS) that
periodically convert chemical into electronic information,” Sensors and Actuators, B:
Chemical, vol. 273. Elsevier B.V., pp. 1334–1345, Nov. 10, 2018, doi:
10.1016/j.snb.2018.06.054.
19
[7] H. Kimura, Y. Sakai, and T. Fujii, “Organ/body-on-a-chip based on microfluidic technology
for drug discovery,” Drug Metabolism and Pharmacokinetics, vol. 33, no. 1. Japanese
Society for the Study of Xenobiotics, pp. 43–48, Feb. 01, 2018, doi:
10.1016/j.dmpk.2017.11.003.
[8] J. R. Mejía-Salazar, K. Rodrigues Cruz, E. M. Materón Vásques, and O. Novais de Oliveira
Jr., “Microfluidic Point-of-Care Devices: New Trends and Future Prospects for eHealth
Diagnostics,” Sensors, vol. 20, no. 7, p. 1951, Mar. 2020, doi: 10.3390/s20071951.
[9] S. Kim et al., “High-throughput automated microfluidic sample preparation for accurate
microbial genomics,” Nat. Commun., vol. 8, no. 1, pp. 1–10, Jan. 2017, doi:
10.1038/ncomms13919.
[10] K. Kant et al., “Microfluidic devices for sample preparation and rapid detection of
foodborne pathogens,” Biotechnology Advances, vol. 36, no. 4. Elsevier Inc., pp. 1003–
1024, Jul. 01, 2018, doi: 10.1016/j.biotechadv.2018.03.002.
[11] J. Bemetz et al., “Microfluidic-Based Synthesis of Magnetic Nanoparticles Coupled with
Miniaturized NMR for Online Relaxation Studies,” Anal. Chem., vol. 90, no. 16, pp. 9975–
9982, Aug. 2018, doi: 10.1021/acs.analchem.8b02374.
[12] L. Zhang, Q. Chen, Y. Ma, and J. Sun, “Microfluidic Methods for Fabrication and
Engineering of Nanoparticle Drug Delivery Systems,” ACS Appl. Bio Mater., vol. 3, no. 1,
pp. 107–120, Jan. 2020, doi: 10.1021/acsabm.9b00853.
[13] A. Liga, J. A. S. Morton, and M. Kersaudy-Kerhoas, “Safe and cost-effective rapid-
prototyping of multilayer PMMA microfluidic devices,” Microfluid. Nanofluidics, vol. 20,
no. 12, p. 164, Dec. 2016, doi: 10.1007/s10404-016-1823-1.
[14] K. T. L. Trinh, D. A. Thai, W. R. Chae, and N. Y. Lee, “Rapid Fabrication of Poly(methyl
methacrylate) Devices for Lab-ona-Chip Applications Using Acetic Acid and UV
Treatment,” ACS Omega, vol. 5, no. 28, pp. 17396–17404, Jul. 2020, doi:
10.1021/acsomega.0c01770.
[15] H. Bruus, Theoretical microfluidics. Oxford University Press, 2008.
[16] S. Choi, M. G. Lee, and J. K. Parka, “Microfluidic parallel circuit for measurement of
hydraulic resistance,” Biomicrofluidics, vol. 4, no. 3, 2010, doi: 10.1063/1.3486609.
[17] J. Wang, N. Zhang, J. Chen, V. G. J. Rodgers, P. Brisk, and W. H. Grover, “Finding the
optimal design of a passive microfluidic mixer,” Lab Chip, vol. 19, no. 21, pp. 3618–3627,
Nov. 2019, doi: 10.1039/c9lc00546c.
[18] C. Y. Lee, W. T. Wang, C. C. Liu, and L. M. Fu, “Passive mixers in microfluidic systems: A
review,” Chemical Engineering Journal, vol. 288. Elsevier, pp. 146–160, Mar. 15, 2016,
doi: 10.1016/j.cej.2015.10.122.
[19] S. B. Suh, M. A. Traore, and B. Behkam, “Bacterial chemotaxis-enabled autonomous sorting
of nanoparticles of comparable sizes,” Lab Chip, vol. 16, no. 7, pp. 1254–1260, Apr. 2016,
doi: 10.1039/c6lc00059b.
[20] W. Han, X. Chen, Z. Wu, and Y. Zheng, “Three-dimensional numerical simulation of
droplet formation in a microfluidic flow-focusing device,” J. Brazilian Soc. Mech. Sci. Eng.,
vol. 41, no. 6, p. 265, Jun. 2019, doi: 10.1007/s40430-019-1767-y.
20
[21] K. Plevniak, M. Campbell, and M. He, “3D printed microfluidic mixer for point-of-care
diagnosis of anemia,” in Proceedings of the Annual International Conference of the IEEE
Engineering in Medicine and Biology Society, EMBS, Oct. 2016, vol. 2016-October, pp.
267–270, doi: 10.1109/EMBC.2016.7590691.
[22] P. Yin et al., “Simulation and practice of particle inertial focusing in 3D-printed serpentine
microfluidic chips: Via commercial 3D-printers,” Soft Matter, vol. 16, no. 12, pp. 3096–
3105, Mar. 2020, doi: 10.1039/d0sm00084a.
[23] J. Peiró and S. Sherwin, “Finite Difference, Finite Element and Finite Volume Methods for
Partial Differential Equations,” in Handbook of Materials Modeling, Springer Netherlands,
2005, pp. 2415–2446.
[24] “Detailed Explanation of the Finite Element Method (FEM).”
https://www.comsol.com/multiphysics/finite-element-method?parent=physics-pdes-
numerical-042-62 (accessed Nov. 22, 2020).
[25] A. Shamloo and M. Boodaghi, “Design and simulation of a microfluidic device for acoustic
cell separation,” Ultrasonics, vol. 84, pp. 234–243, Mar. 2018, doi:
10.1016/j.ultras.2017.11.009.
[26] K. Plevniak, M. Campbell, T. Myers, A. Hodges, and M. He, “3D printed auto-mixing chip
enables rapid smartphone diagnosis of anemia,” Biomicrofluidics, vol. 10, no. 5, p. 054113,
Sep. 2016, doi: 10.1063/1.4964499.
[27] A. Shamloo and F. Y. Parast, “Simulation of blood particle separation in a trapezoidal
microfluidic device by acoustic force,” IEEE Trans. Electron Devices, vol. 66, no. 3, pp.
1495–1503, Mar. 2019, doi: 10.1109/TED.2018.2889912.
[28] A. L. Campaña et al., “Enzyme-based electrochemical biosensors for microfluidic platforms
to detect pharmaceutical residues in wastewater,” Biosensors, vol. 9, no. 1. MDPI AG, p. 41,
Mar. 15, 2019, doi: 10.3390/bios9010041.
[29] W. Han, X. Chen, Z. Wu, and Y. Zheng, “Three-dimensional numerical simulation of
droplet formation in a microfluidic flow-focusing device,” J. Brazilian Soc. Mech. Sci. Eng.,
vol. 41, no. 6, p. 265, Jun. 2019, doi: 10.1007/s40430-019-1767-y.
[30] Y. Zhang and X. Chen, “Dielectrophoretic microfluidic device for separation of red blood
cells and platelets: a model-based study,” J. Brazilian Soc. Mech. Sci. Eng., vol. 42, no. 2, p.
89, Feb. 2020, doi: 10.1007/s40430-020-2169-x.
[31] A. Datta and V. Rakesh, An Introduction to Modeling of Transport Processes. Cambridge
University Press, 2009.
[32] “Innovative Food Processing Technologies: Advances in Multiphysics Simulation | Wiley.”
https://www.wiley.com/en-
us/Innovative+Food+Processing+Technologies%3A+Advances+in+Multiphysics+Simulatio
n-p-9780813817545 (accessed Nov. 30, 2020).
[33] D. Zou and D. Cui, “Advances in isolation and detection of circulating tumor cells based on
microfluidics,” Cancer Biology and Medicine, vol. 15, no. 4. Cancer Biology and Medicine,
pp. 335–353, Nov. 01, 2018, doi: 10.20892/j.issn.2095-3941.2018.0256.
[34] T. Salafi, K. K. Zeming, and Y. Zhang, “Advancements in microfluidics for nanoparticle
21
separation,” Lab Chip, vol. 17, no. 1, pp. 11–33, Jan. 2017, doi: 10.1039/C6LC01045H.
[35] M. Antfolk and T. Laurell, “Continuous flow microfluidic separation and processing of rare
cells and bioparticles found in blood – A review,” Analytica Chimica Acta, vol. 965.
Elsevier B.V., pp. 9–35, May 01, 2017, doi: 10.1016/j.aca.2017.02.017.
[36] K. Doufène, C. Tourné-Péteilh, P. Etienne, and A. Aubert-Pouëssel, “Microfluidic Systems
for Droplet Generation in Aqueous Continuous Phases: A Focus Review,” Langmuir, vol.
35, no. 39. American Chemical Society, pp. 12597–12612, Oct. 01, 2019, doi:
10.1021/acs.langmuir.9b02179.
[37] A. Lashkaripour, C. Rodriguez, L. Ortiz, and D. Densmore, “Performance tuning of
microfluidic flow-focusing droplet generators,” Lab Chip, vol. 19, no. 6, pp. 1041–1053,
Mar. 2019, doi: 10.1039/C8LC01253A.
[38] P. Zhu and L. Wang, “Passive and active droplet generation with microfluidics: a review,”
Lab Chip, vol. 17, no. 1, pp. 34–75, Jan. 2017, doi: 10.1039/C6LC01018K.
[39] Y. Zhang, D. Liu, H. Zhang, and H. A. Santos, “Microfluidic mixing and devices for
preparing nanoparticulate drug delivery systems,” in Microfluidics for Pharmaceutical
Applications: From Nano/Micro Systems Fabrication to Controlled Drug Delivery, Elsevier,
2018, pp. 155–177.
[40] C.-H. D. Tsai and X.-Y. Lin, “Experimental Study on Microfluidic Mixing with Different
Zigzag Angles,” Micromachines, vol. 10, no. 9, p. 583, Aug. 2019, doi:
10.3390/mi10090583.
[41] C. Y. Lee, C. L. Chang, Y. N. Wang, and L. M. Fu, “Microfluidic mixing: A review,”
International Journal of Molecular Sciences, vol. 12, no. 5. Int J Mol Sci, pp. 3263–3287,
May 2011, doi: 10.3390/ijms12053263.
[42] H. E. H. Meijer, M. K. Singh, T. G. Kang, J. M. J. den Toonder, and P. D. Anderson,
“Passive and Active Mixing in Microfluidic Devices,” Macromol. Symp., vol. 279, no. 1, pp.
201–209, May 2009, doi: 10.1002/masy.200950530.
[43] V. Hessel, H. Löwe, and F. Schönfeld, “Micromixers - A review on passive and active
mixing principles,” in Chemical Engineering Science, Apr. 2005, vol. 60, no. 8-9 SPEC.
ISS., pp. 2479–2501, doi: 10.1016/j.ces.2004.11.033.
[44] O. Schenk, K. Gärtner, W. Fichtner, and A. Stricker, “PARDISO: A high-performance serial
and parallel sparse linear solver in semiconductor device simulation,” Futur. Gener.
Comput. Syst., vol. 18, no. 1, pp. 69–78, Sep. 2001, doi: 10.1016/S0167-739X(00)00076-5.
[45] I. Pereiro, A. Fomitcheva Khartchenko, L. Petrini, and G. V. Kaigala, “Nip the bubble in the
bud: A guide to avoid gas nucleation in microfluidics,” Lab on a Chip, vol. 19, no. 14. Royal
Society of Chemistry, pp. 2296–2314, Jul. 21, 2019, doi: 10.1039/c9lc00211a.
[46] J. Liu, S. Li, and D. Mitra, “Multiphysical phenomenon of air bubble growth in
polydimethylsiloxane channel corners under microfluidic negative pressure-driven flow,”
Int. J. Heat Mass Transf., vol. 91, pp. 611–618, Aug. 2015, doi:
10.1016/j.ijheatmasstransfer.2015.08.006.
[47] J. Polte, “Fundamental growth principles of colloidal metal nanoparticles - a new
perspective,” CrystEngComm, vol. 17, no. 36, pp. 6809–6830, Jun. 2015, doi:
22
10.1039/c5ce01014d.
[48] F. Mura and A. Zaccone, “Effects of shear flow on phase nucleation and crystallization,”
Phys. Rev. E, vol. 93, no. 4, p. 042803, Apr. 2016, doi: 10.1103/PhysRevE.93.042803.
[49] X. Xuan, “Recent advances in continuous-flow particle manipulations using magnetic
fluids,” Micromachines, vol. 10, no. 11. MDPI AG, p. 744, Nov. 01, 2019, doi:
10.3390/mi10110744.
[50] Q. Cao, X. Han, and L. Li, “Configurations and control of magnetic fields for manipulating
magnetic particles in microfluidic applications: Magnet systems and manipulation
mechanisms,” Lab on a Chip, vol. 14, no. 15. Royal Society of Chemistry, pp. 2762–2777,
Aug. 07, 2014, doi: 10.1039/c4lc00367e.