You are on page 1of 9

Ind. Eng. Chem. Res.

2008, 47, 7447–7455 7447

Laminar to Turbulent Transition and Heat Transfer in a Microreactor:


Mathematical Modeling and Experiments
Andrej Pohar and Igor Plazl*
UniVersity of Ljubljana, Faculty of Chemistry and Chemical Technology, AskerceVa 5, 1000 Ljubljana, SloVenia

Heat transfer of aqueous flows was studied in a Y-shaped microreactor at different flow rates. In order to
analyze the experimental data and to forecast the microreactor performance, a three-dimensional mathematical
model with convection and conduction was developed, considering the velocity profile for laminar flow at
steady state. The dependence of temperature on the thermophysical properties of water was implemented into
the mathematical model. The microreactor investigated consists of a rectangular microchannel, which is divided
into two inlet channels, a central channel, and two outlet channels. The average temperatures of water outflows
were monitored. Very good agreement with the model calculations and the experimental data was achieved
without any fitting procedure. In addition, the transition from laminar to turbulent flow was studied for different
microchannel geometries, and the results showed that the channel aspect ratio and the angle of merging of
two inlet channels substantially influence the critical Reynolds number.

1. Introduction hydraulic diameters of 133-367 µm were studied. They reported


that the transition to turbulence occurs at Reynolds numbers as
Microprocess engineering has been developing rapidly in
low as 200-700. Furthermore, they showed that the transition
recent years and has had a growing role in the chemical,
Reynolds numbers decrease as the hydraulic diameter decreases,
biochemical, electronic, biotechnological, and pharmaceutical
and that this holds true only for liquid flow in microchannels
industries. The use of microreactors has been proven to have
with identical microchannel dimensions. Wu and Little4 found
some significant advantages over regular reactors. Microreactors
have a high surface to volume ratio and consequently highly that the transition from laminar to turbulent flow in microchan-
efficient heat transfer. The residence times of reactants are very nels with hydraulic diameters ranging from 50 to 80 µm
small, which results in increased safety in dealing with occurred much earlier than predicted by the classical macroscale
potentially dangerous substances and reactions. They allow theory, at Reynolds numbers of about 400-900 for various
precise control of process parameters and thus the means of tested configurations.
achieving optimal conditions for a specific chemical transforma- Xu et al.5 considered liquid flow within microchannels of
tion. For many applications, isothermal or near-isothermal hydraulic diameters between 30 and 344 µm. Their results
operation is favorable, especially for the elimination of hot spots agreed with the conventional macroscale behavior predicted by
due to localized reaction zones.1 Due to these unique charac- the Navier-Stokes equation. They concluded that, for liquid
teristics, they are especially suitable for chemical reactions that flow in microchannels, any special microeffect due to the very
are highly exothermic, explosive, or toxic. They provide the small dimensions would not be so significant as to make
means for industrial innovation as well as optimization of conventional theories unsuitable for the prediction of flow
existing processes. Some reactions that were previously unfea- characteristics in channels with hydraulic diameters bigger than
sible are now made possible, but successful operation of such 30 µm. They also suggested that deviations in earlier studies
reactions demands deep understanding of heat transfer and fluid may have resulted from measurement errors and were not due
flow inside the microreactor system. to novel phenomena at the microscale level.
1.1. Laminar to Turbulent Transition. Laminar flow is a The transition from laminar to turbulent flow in glass
determining characteristic of fluid flow in a microreactor and microtubes with diameters ranging from 50 to 247 µm was
as such should be carefully investigated. Since microchannels studied by Sharp and Adrian.6 Their experiments consisted of
in microreactors are mostly not circular in shape due to different the observation of flow resistance, pressure drop, and velocity
fabrication processes, the evaluation of the flow characteristics fluctuations measured by microparticle image velocimetry
with regard to the shape of the microchannels should be (micro-PIV). The results conclusively showed that the transition
considered, as well as the effect of the channel aspect ratio. A to turbulence begins in virtually the same Reynolds number
microreactor typically consists of inlet channels which feed range as that found for macroscale flow (1800-2300). The
reactants into a central (reaction) channel, and the influence of behavior of the flow in microtubes, down to 50 µm in diameter,
the angle of merging of inlet channels is also expected to was proven to be consistent with the classical theory. Similarly,
influence the critical Reynolds number. Choi et al.7 found that the transition to turbulent flow for glass
There have been many studies concerning the transition from microtubes of 53 and 81 µm in diameter occurred at Re ) 2000.
laminar to turbulent regime in microchannels which suggested They also found that the value decreases for smaller microtubes,
a significant lowering of the critical Reynolds number. Peng being 500 for 9.7 and 6.9 µm in diameter.
and Peterson2 studied rectangular microchannels and reported Wibel and Ehrhard8 conducted experiments on the laminar
the upper laminar flow border at a Reynolds number of 400. In to turbulent transition of liquid flows in rectangular microchan-
another work by Peng and Peterson,3 microchannels with nels. Three different aspect ratios (1:1, 1:2, 1:5) with the
* To whom correspondence should be addressed. Tel.: +386-1-2419- hydraulic diameter of 133 µm were used as well as two different
512. E-mail: igor.plazl@fkkt.uni-lj.si. surface roughnesses (1-2 and 25 µm). Their conclusions were
10.1021/ie8001765 CCC: $40.75  2008 American Chemical Society
Published on Web 08/28/2008
7448 Ind. Eng. Chem. Res., Vol. 47, No. 19, 2008

that the laminar to turbulent transition for smooth channels was


the same for all aspect ratios and in agreement with findings
for macroscopic channels (Re ) 1900-2200). Rough channel
walls, on the other hand, had a substantial effect on the critical
Reynolds number of the microchannel with the 1:5 aspect ratio,
lowering the value to 900.
Guo and Li9 studied the size effect on microscale single-
phase flow and heat transfer. They showed that, for a given
relative roughness, the number of roughness elements per unit
channel length is much larger for microchannels, which gener-
ates more frequent disturbances of the channel flow, resulting
in an increased friction factor. The disturbances cause an early
transition from laminar to turbulent flow when the Reynolds
number is high enough. Papautsky et al.,10 in a review of laminar
Figure 1. (a) Scheme of the microreactor used for studying heat transfer,
flow in microchannels, reported that no reliable range of the (b) shape of the microchannels used for studying laminar to turbulent
Reynolds number for turbulence transition has been reported. transition, and (c) 150 × 50 µm channel with a 90° angle of merging of
They expressed the need for additional experimental investiga- inlet channels.
tions using microchannels of various dimensions and surface studied thermal management in catalytic microreactors. They
roughnesses. found that the assumption of thermal uniformity inside a
1.2. Heat Transfer. Similar heat transfer studies were microreactor is not valid, but is highly dependent on the reactor
conducted on microheat sinks, used for the cooling of Micro- material, the reactor geometry, and the flow rate.
Electro Mechanical Systems, but a number of discrepancies exist In this study, six different microreactors were analyzed for
between the results of various investigations. The experimental the critical Reynolds number. Five microreactors with rectan-
results vary with different researchers and the derived empirical gular cross sections in the dimensions 110 × 50 µm, 220 × 50
correlations are only applicable for a particular channel aspect µm, 440 × 100 µm, 660 × 100 µm, and 1000 × 100 µm had
ratio and fabrication process.11,12 a parallel merging of the inlet channels. The sixth microreactor,
Qu and Mudawar13 studied three-dimensional heat transfer with a 150 × 50 µm cross section, had the two inlet channels
in microchannel heat sinks. Their results provided a detailed merging at an angle of 90°. The roughness of the channel walls
description of heat transfer characteristics, but with the assump- was approximately 1 µm for the 50 µm deep channels and 2
tion of constant fluid thermophysical properties. The classical µm for the 100 µm deep channels, due to the fabrication process
fin model was used for the numerical investigation; however, (HF etching). HF etching is one of the most commonly used
some key assumptions applied in the model deviate significantly techniques for microchannel production. The relative roughness
from the real situation, which reduces the accuracy of such ranged from 1.46% (110 × 50 µm channel) to 1.10% (1000 ×
results. Peng and Peterson2 analyzed convective heat transfer 100 µm channel). Fluid flow was analyzed using colored water
of water in microchannels and found that the cross-sectional and uncolored water, which were pumped simultaneously into
shape of channels influences heat transfer characteristics. the microreactor, one in each inlet channel. A very small
Li et al.11 conducted a three-dimensional analysis of heat concentration of a strong coloring agent (amido blue) was used
transfer in a silicon-based microheat sink with single-phase flow. for the coloring of water, which ensured a negligible change of
Their results validated the assumption of hydrodynamic, fully water properties. The experiments were observed under the
developed laminar flow. They also suggested that the thermo- microscope.
physical properties of the liquid can significantly influence both Additionally, heat transfer in microchannels was analyzed
the flow and heat transfer in the microchannel heat sink. Lee et numerically and experimentally. A mathematical model was
al.12 found that numerical simulations based on a conventional developed, which considered the heat transfer and hydrodynamic
Navier-Stokes analysis accurately predict the thermal behavior characteristics on the macroscale level, to predict the outlet
in single-phase flow through rectangular microchannels. The temperatures of water on the microscale level. Two instances
numerical predictions compared to the experimental results were considered, one with constant boundary conditions and
showed an average deviation of only 5%. They also expressed one with dynamic boundary conditions, which took into account
the need to carefully match the entrance and boundary conditions the cooling of the inner wall of the microchannel. The study
for the simulation with an experiment for each specific case. did not include a reaction, but focused on microreactor heat
The average fluid temperature of the inlet and outlet was used transfer capabilities.
for the calculation of thermal properties, which were assumed
constant over the range of 20-70 °C. Furthermore, the 2. Theoretical Background
conduction effect in the solid substrate was neglected. The microreactor investigated for the heat transfer study was
Federov and Viskanta14 developed a three-dimensional model silicon-based with rectangular microchannels 110 µm wide (B)
to investigate flow and conjugate heat transfer in a microchannel- and 50 µm deep (H), which corresponds to a hydraulic diameter
based heat sink for electronic packaging applications. Their of 68.75 µm. Two inlet channels join together to form a central
study provided an insight into the heat flow pattern due to channel, which at the end splits into two outlet channels, as
combined convection-conduction effects and with the assump- can be seen in Figure 1. The central channel was of equal
tion of incompressible Navier-Stokes equations of motion. The dimensions as the inlet and outlet channels and the fluid velocity
model predictions proved to be consistent with the measure- there was consequently twice as high. Deionized water was used
ments. Their method of solution was also the classical fin as the testing fluid. The total path of the fluid in the microreactor
analysis, which only provides a simplified means to modeling was 43.46 mm (see Table 1).
heat transfer. One of the key assumptions is that the temperature 2.1. Incompressible Fluid Flow in a Microchannel. The
is uniform at any given channel cross section. Norton et al.1 length of the hydrodynamic developing region was assessed with
Ind. Eng. Chem. Res., Vol. 47, No. 19, 2008 7449
Table 1. Dimensions of the Y-Shaped Microreactor for Heat
Transfer Study
dimension length [mm]
B (width) 0.110
H (depth) 0.050
D 1.8
L1 30
L2 6.73
L1 + 2L2 43.46

Langhaar’s equation, which holds true for Reynolds numbers Figure 2. The studied system.
of 2000 and smaller. The entrance length is defined as the
distance at which the center-line velocity reaches 99% of the velocity. In the limit of ∆x, ∆y, ∆z, and ∆t approaching 0,
velocity in a fully developed Hagen-Poiseuille profile. and after the introduction of the following dimensionless
variables:
Le
) 0.0575Re (1) y x z
Dh ψ) , ω) , ξ) (7)
H H L
For the assumption of a hydrodynamic, fully developed and thermal diffusivity
laminar flow, the entrance length must be less than 5% of the
total length.11 In the case of the fluid flow of 2000 µL/min, the R)
λ
(8)
highest value of the entrance length to the total length ratio is Fcp
4%. This calculation validates the assumption of a fully
the heat balance equation in steady state can be written in the
developed laminar flow in our heat transfer investigation.
following form:
The flow in microchannels is governed by the incompress-

( )
RL ∂2T
( )
RL ∂2T
ible Navier-Stokes equations. If the compressibility and the ∂T
gravitational force are neglected, the continuity and momen- Vz(ω, ψ) ) 2 2
+ 2 (9)
∂ξ H ∂ω H ∂ψ2
tum equations are
In eq 9, R is temperature dependent and Vz is a function of
∇ · v)0 (2) the position in the x-y plane of the microchannel. Figure 2
shows the microchannel parameters. Water at room temperature
F(∇ · v)v ) - ∇ P + η∇ v 2
(3) T0 enters the microchannel at ξ ) 0, and T∞ represents the inner
For a steady, fully developed Poiseuille-type flow, the wall temperature. The depth (ω) and the length (ξ) of the
z-momentum equation can be simplified to microchannel in dimensionless variables range from 0 to 1, while

[ ]
the width ranges from 0 to B/H.
∂P ∂2Vz ∂2Vz Two cases with different sets of boundary conditions were
0)- +η + 2 (4) considered. In the first case, the temperature of the inner wall
∂z ∂x2 ∂y of the microchannel was presumed constant and equal to the
with the associated boundary conditions: temperature of the water bath. This represents an ideal situation
for efficient temperature regulation. The boundary conditions
Vz(x, B) ) Vz(x, 0) ) 0; 0 e x e H are T(ω,ψ,0) ) T0, T(ω,0,ξ) ) T∞, T(ω,B/H,ξ) ) T∞, T(0,ψ,ξ)
(5)
Vz(H, y) ) Vz(0, y) ) 0; 0 e y e B ) T∞, and T(1,ψ,ξ) ) T∞.
In the second case, dynamic boundary conditions were
where x and y are coordinates in the direction of the channel presumed. In this case, the inner wall is cooled due to the
width and height and Vz is the z-directional velocity of water. entrance of cold water. The heat flow from the water bath has
A constant pressure gradient was applied along the length L of to overcome the heat resistance of the borosilicate glass before
the microchannel, and therefore -∂P/∂z can be simplified to entering the fluid. The boundary conditions can be written
∆P/L. The velocity profile calculated from Navier-Stokes
∂T
equations for Newtonian fluids of constant density (eqs 4 and -λ ) U(T(ω,0,ξ) - T∞)
5) was then used in heat transfer simulations, considering the H ∂ ψ (ω,0,ξ)
assumption of negligible density change.11 The velocity field ∂T
-λ ) U(T(ω,B⁄H,ξ) - T∞)
in this study was approximated by the following mathematical H ∂ ψ (ω,B⁄H,ξ)
expression as well, which is a reasonable approximation of the ∂T (10)
-λ ) U(T(0,ψ,ξ) - T∞)
actual state in the microchannel.15 The effect of temperature H ∂ ω (0,ψ,ξ)
change on the velocity profile was not implemented. The result ∂T
-λ ) U(T(1,ψ,ξ) - T∞)
was compared to the Navier-Stokes solution. H ∂ ω (1,ψ,ξ)
The dependence of the physical properties of water on
9 2x 2 2y 2

4 ( (
Vz(x, y) ) Vmean 1 -
H
-1 1-) )( ( B
-1 )) (6) temperature was included in the model, as well as the parabolic
velocity profile calculated from eqs 4 and 5. The input heat
2.2. The Mathematical Model. A fully developed laminar rate was treated as being constant on all wall surfaces. The
flow at steady-state conditions was assumed. The energy compressibility, gravitational forces, and heat dissipation caused
balance equation included heat convection in the z direction by the viscosity of the liquid were neglected, as suggested by
and heat conduction in the x and y directions, with z Li et al.11 Koo and Kleinstreuer16 studied viscous dissipation
corresponding to the length of the microchannel, x to the effects in microtubes and microchannels. They suggested that
depth, and y to the width. The contribution of heat conduction viscous dissipation effects have influences on fluids with low
in the z direction was neglected on account of very high fluid specific heat capacities, with high viscosities, or when the
7450 Ind. Eng. Chem. Res., Vol. 47, No. 19, 2008

The flow rates used in the study ranged from 100 to 2000
µL/min on each of the two high-performance syringe pumps
(PHD 4400 Hpsi Syringe Pump Series, Harvard Apparatus,
Holliston, MA). The Reynolds numbers in these cases were
below 2000, which indicated laminar flow. The residence times
of the fluid in the microreactor were very small, ranging from
0.0047 s at 100 µL/min to 0.0939 s at 2000 µL/min.

4. Results and Discussion


The typical velocity profile at laminar flow conditions for
the flow rate of 1000 µL/min is presented in Figure 4a. Similar
results can be obtained with eq 6, as can be seen in Figure 4b.
The average fluid velocity is 3.03 m/s. Although the numerical
solution of the approximation yields results similar to the ones
obtained with the Navier-Stokes equations, the latter gives a
Figure 3. The experimental setup. more accurate description of fluid flow inside the microchannel.
For the determination of the transition from laminar to
hydraulic diameter is less than 50 µm. In the case of this turbulent regime, colored water and uncolored water were
investigation, viscous dissipation effects can reasonably be pumped into the two inlet channels and photos of the central
neglected. channel were taken near the exit of the microreactor (see Figure
Heat gain in the steady state by the fluid can be determined 5). Results were obtained with excellent repeatability.
from the energy balance: For the 110 × 50 µm channel, the flow at Re ) 1850 is still
laminar. Water flows in parallel layers, without any disruption
q ) FcpφV(Tout - Tin) (11) between the layers. Where the central channel splits into two
outlet channels, the two flows are fully separated. The channels
The density and the specific heat were calculated based on
were smooth with relative roughness (ε/Dh) of only 1.46%. The
the mean fluid temperature. The average heat transfer coefficient
transition from laminar to turbulent occurred around Re ) 1900,
was evaluated as
which is in agreement with conventional theory and has also
h ) q/A(Tw - Tm) (12) been reported in more recent studies.
In the case of the 220 × 50 µm microchannel, the transition
where Tw is the average wall temperature, Tm is the mean water to turbulence occurred much earlier, around Re ) 1200 (Figure
temperature, and A is the channel wall area; A ) 2L(B + H). 5b). The relative roughness of the channels was 1.23%, less
The average Nusselt number was calculated as Nu ) hDh/λ, than that of the previous channel. The aspect ratio of the channel
and the results were plotted as a function of the Reynolds walls, on the other hand, was twice as high (1:4.4 compared to
number. 1:2.2 of the previous microreactor).
2.3. Method of Solution. The method used for solving the The microreactor with the 440 × 100 µm channel had the
second-order partial differential equation (eq 9) was the alternat- same aspect ratio and relative roughness as the prior micro-
ing-direction implicit method or ADI. This finite difference channel. The transition to turbulence was found to be around
method is used for solving parabolic differential equations in Re ) 1100, which is in good agreement with the 220 × 50 µm
three dimensions with the use of tridiagonal matrices. The microchannel. Although the microchannel was twice the size,
numerical calculations were done using Matlab. at microscale the aspect ratio played the determining role in
The value of the heat conductivity coefficient of borosilicate lowering the critical Reynolds number. The onset of turbulence
glass, as found in the literature, is 1.14 W/m K.17 The thickness could be seen in great detail owing to the wider channel. Upon
of the microreactor wall was approximated by the average value the increase of the flow rate at a certain critical point, the border
of the upper and bottom wall thicknesses. The thermophysical fluctuated from perfect laminar and stream lines of the opposite
properties of water (thermal conductivity, density, viscosity, and color as the surrounding liquid were displaced on both sides of
specific heat) are all temperature dependent. Their values were the border. In an interval of about 2 s the flow would disrupt
found in Perry’s Chemical Engineers Handbook.17 The func- and then reassemble to laminar regime. With the increase of
tional dependence of thermal diffusivity on the temperature was flow rate, the disturbances of the flow would intensify.
determined as The results for the 660 × 100 µm and 1000 × 100 µm
microchannel showed a further lowering of the critical Reynolds
R ) -1.571 × 10-12T(°C)2 + 4.489 × 10-10T(°C) +
number. Transition for both cases occurred around Re ) 410.
1352 × 10-7(13) Since the 1000 µm wide channel (aspect ratio 1:10) is
significantly wider than the 660 µm wide channel, we can
3. Experimental Setup (Figure 3)
conclude that we have reached the lowest critical Reynolds
A laboratory-on-a-chip system (Micronit Microfluidics B.V., number with regard to the increase of the aspect ratio. The onset
Enschede, The Netherlands) was employed in the investigation. of turbulence was also very clearly seen in these two cases and
The microreactor was thermostatted in a water bath at 59 °C, is shown for the 1000 µm wide channel in Figure 5f-h.
and water at room temperature (21 °C) was pumped through Clearly the aspect ratio of rectangular microchannels signifi-
the two inlet microchannels. The temperatures at the two outlets cantly influences (lowers) the critical Reynolds number. On the
of the microreactor were monitored using type-J (iron-constantan) microscale, channels with equal hydraulic diameters and dif-
thermocouples, and a computer was used for data acquisition. ferent channel wall aspect ratios cannot be evaluated in the same
The experimental data were obtained as an average of numerous manner since the increase of the aspect ratio means that the
measurements. channel wall area is much larger, and surface roughness effects
Ind. Eng. Chem. Res., Vol. 47, No. 19, 2008 7451

Figure 4. Velocity field at the flow rate of 1000 µL/min calculated using (a) incompressible Navier-Stokes equations and (b) numerical approximation.

Figure 6. Critical Reynolds numbers for different channel aspect ratios.

to be fed separately into the microchannels. The aspect ratio is


1:3 and relative roughness is 1.33%. The transition to turbulent
flow was observed around Re ) 850 (Figure 5b). The higher
angle of merging lowered the critical Reynolds number in the
central channel as opposed to parallel merging.
The finite difference numerical code written in Matlab
provided detailed temperature and heat flux distributions in the
microchannel. A 31 × 31 × 5000 grid (channel width, depth,
and length, respectively) was used for the discretization of the
Figure 5. (a) Laminar flow at Re ) 835 and (b) transition to turbulence at
computational domain. Simulations with a more accurate grid
Re ) 870 for 150 × 50 µm microchannel with 90° merging of inlet channels;
(c) laminar flow at Re ) 1160 and (d) transition to turbulence at Re ) did not yield substantially different results. Out of the 5000 steps
1212 for 220 × 50 µm microchannel; (e) laminar flow at Re ) 317 for which comprise the length of the microreactor, 1-774 represent
1000 × 100 µm microchannel; (f-h) transition to turbulence at Re ) 412. the inlet channel, 775-4226 the central channel, and 4227-5000
The photos are not to scale. the outlet channel.
cause more disturbances in the fluid flow. This was observed For the simulation with constant boundary conditions, the
by comparing the 1:4.4 and 1:6.6 microchannels. The first has inner wall temperature is constant at 59 °C. At the microchannel
Dh ) 163 µm, while the second has Dh ) 174 µm. This would entrance the temperature of water is 21 °C. The simulation
suggest the onset of turbulence at a similar Reynolds number, showed that further along the flow direction water is heated,
but in fact there is a large difference between the two values quickly reaching 59 °C near the inner wall. In the center of the
(1100 and 410, respectively). The surface area, on the other microchannel, where the velocity is the greatest, the temperatures
hand, is 1.5-fold larger in the case of the 660 µm wide stay the lowest. A very fast temperature increase was shown to
microchannel than it is in the case of the 440 µm microchannel. be in the inlet channel, followed by a slightly slower increase
The results were evaluated in a plot of the critical Reynolds in the central channel. It is evident that heat transfer is extremely
number against the aspect ratio of the microchannel walls fast, with the average water temperature achieving 80% of the
(Figure 6). Conventional theory holds true for channels with total temperature difference in the inlet channel.
an aspect ratio around 1. As the aspect ratio increases, the critical In the second case, dynamic boundary conditions take into
Reynolds number decreases toward 410, when the aspect ratio account the heat resistance of the wall of the microreactor. Heat
is above 6. After that, the third dimension does not play a role flow from the water bath has to overcome the heat resistance
anymore and the symmetry approximation can be invoked. The of the borosilicate glass before entering the fluid. This simulation
results hold for smooth channels with relative roughness around was done for the comparison with the experimental results from
1%. Lower values reported by some researchers2-4 were most the studied microreactor system. Figure 7 represents the
likely due to a higher surface roughness of channels. The temperature distribution in the x-y plane at different longitudinal
average value of the critical Reynolds numbers provided by locations for the flow rate of 1000 µL/min on each of the two
Sharp and Adrian6 was added to the plot as well as the value pumps. The inner wall temperature is very low at the beginning
given by Harms et al.18 of the microreactor and slowly rises toward the outer wall
The 150 × 50 µm microreactor had a 90° merging of the temperature of 59 °C, along with the water inside the channel.
inlet channels, which is very common because reactants have The temperatures in the center of the microchannel are lower
7452 Ind. Eng. Chem. Res., Vol. 47, No. 19, 2008

channels quickly rises at the very beginning. In the central


channel, the increase of temperature is slower, since the fluid
velocity here is twice as high. The temperature rise in the outlet
channel is again faster, but much slower than in the inlet channel
due to a smaller temperature gradient. The same trend can be
observed at higher fluid flows, but the temperature rise here is
not as fast and the outlet temperature is consequently smaller.
The temperature rise along the flow direction in the solid and
fluid regions of the microchannel can be approximated as linear,
as already noted by Qu and Mudawar,13 but only at higher flow
rates.
In Figure 10, the mean outlet temperatures of the model are
compared to the experimental results. In the case of constant
boundary conditions, the outlet temperatures are very high,
which is due to highly efficient heat transfer inside the
microchannel. At all flow rates, except 2000 µL/min, water
temperature reached the inner wall temperature before exiting
the microreactor. Water does not reach the inner wall temper-
ature only at extremely small residence times (the residence
time for the flow rate of 2000 µL/min is only 0.0047 s). The
results show how well temperature can be regulated inside a
microreactor system.
With dynamic boundary conditions the heat resistance of the
borosilicate glass wall of the microreactor was taken into
account. The simulation shows that the temperature of water
inside the microchannel is almost the same as the inner wall
temperature at all times. The model accurately predicted the
outlet temperatures. The experimental measurements were done
with a small degree of experimental uncertainty at lower flow
rates, since the temperature of the outflow of water was difficult
to measure accurately. This shows at water flows less than 500
µL/min where the outlet temperatures were measured lower than
the actual state and the predictions of the model. The forming
drop at the outlet would lose some of its heat to the surroundings
(including the thermocouples) before the actual measurement.
On the basis of the good agreement between the outlet
temperatures predicted by the model with dynamic boundary
conditions and the temperatures measured, it can be concluded
that the model successfully describes the heat transfer phenom-
enon in the microchannel.
The experimental data from this study for the central channel
were compared to correlations proposed by previous investiga-
tors in Figure 11. The results are presented in terms of the
Nusselt number variation as a function of the Reynolds number.
The Hausen19 correlation was developed for long macrotubes
with laminar flow at constant wall temperature and is a function
of Reynolds number, Prandtl number, and hydraulic diameter.
The proposed correlation did not provide satisfactory predictions
of the average Nusselt numbers, which was already noted by
Lee et al.12 The correlation developed by Li et al.11 is also
inconsistent with the experimental results from this work. The
correlation was developed specifically for the water/silicon
system and for the specific geometry of the heat sink, as was
Figure 7. Local temperature distribution on the x-y plane at different explained by the authors. The Reynolds numbers considered in
locations along the flow direction for 1500 µL/min. (a) Immediately after
the work were less than 200.
the entrance of the microchannel, (b) at the entrance into the central channel,
and (c) at the exit of the microreactor. The best agreement was found by comparing results with the
work of Wibulswas20 for a rectangular channel with an aspect
due to higher fluid velocity. The temperature distribution on ratio of 4. The proposed correlations by Lee et al.11 and Lee
the cross section of the microreactor at half-depth is shown in and Garimella21 for thermally developing flow also show good
Figure 8. agreement. The correlation by Lee et al.11 was developed from
Figure 9 shows the temperature rise along the length of the the “thin wall” model, and is plotted for the smallest micro-
microreactor. The dashed line represents the inner channel wall, channel tested (884 × 194 µm, Dh ) 318 µm, R ) 4.56). The
while the continuous line represents the mean temperature of correlation of Lee and Garimella21 is shown for a 1250 × 229
the fluid. At lower flow rates the temperature in the inlet µm channel (Dh ) 387 µm, R ) 5.46).
Ind. Eng. Chem. Res., Vol. 47, No. 19, 2008 7453

Figure 8. Cross section of the microreactor at H/2, representing the temperature distribution at the y-z plane. (a) For 1500 µL/min; (b) for 2000 µL/min.

a much higher surface to volume ratio in comparison to


conventional-sized channels. It has been shown that water flow
in smooth microtubes, down to Dh ) 50 µm, shows no
difference from macroscale flow as long as the relative surface
roughness is less than 1.5%.22 In this work, the microchannel
with a small channel wall aspect ratio (1:2.2) exhibited similar
behavior. Deviations from classical theory were observed at
higher channel aspect ratios. This can be explained by the fact
that the hydraulic diameter, used for the calculation of the
Reynolds number, does not account for the large increase of
the channel wall surface area in channels with a high aspect
ratio. In such microchannels, surface roughness has more effect
on the fluid flow which causes an early transition from laminar
to turbulent regime. The Reynolds number is interpreted as the
ratio of the inertia force to the viscous force. At low Reynolds
numbers, the viscous force is large compared to the inertia force
Figure 9. Temperature of the inner wall and mean temperature of water and small disturbances in the velocity field created by small
for 500 and 1500 µL/min. roughness elements on the surface are damped out. With the
increase of the Reynolds number, the effect of viscous damping
5. Conclusions becomes smaller and perturbations are allowed to grow. When
In conventional theory, the effect of internal wall roughness large variations in the velocity field can be maintained, the flow
on laminar flow characteristics can be ignored for relative becomes unstable and experiences an early transition to
roughness less than 5%. Microchannels, on the other hand, have turbulence.
7454 Ind. Eng. Chem. Res., Vol. 47, No. 19, 2008

critical Reynolds number in the central channel. A parallel


merging of channels is advised to avoid unexpected turbulence
inside a microreactor system.
A detailed numerical simulation of forced convection heat
transfer occurring in a microreactor was conducted using a three-
dimensional heat transfer model. The parabolic velocity profile
and the dependence of temperature on the thermophysical
properties of water were implemented into the mathematical
code and the ADI method used for solving the governing partial
differential equation. The model with dynamic boundary condi-
tions accurately predicted the temperatures at the outlet of the
studied microreactor at different flow rates. The numerical
calculations agreed with the traditional macroscale theory in
predicting the heat transfer for laminar flow in the microchannel
and provided an insight into the fluid flow characteristics. A
detailed temperature distribution of water inside the microchan-
nel was presented as well as the temperature of the channel
wall. For microchannels, which are not as smooth as the ones
in this work, one should also consider incorporating the
thermocapillary effect into the model to account for flow
instabilities due to the changes of the surface tension of water
at the interface with microchannel walls.
The heat transfer process in the microreactor is extremely
fast, as was observed in the model with constant boundary
conditions. The results show that the fluid can be heated or
cooled very quickly. With the use of an alternative material for
the microreactor, one with higher thermal conductivity (e.g.,
metals) and/or thinner walls, constant boundary conditions could
be achieved. The temperature needed for a specific chemical
transformation could be regulated even more easily and the
redundant heat from exothermic reactions removed even more
efficiently.

Nomenclature
Figure 10. Mean temperatures at the outlet of the microreactor at different
flow rates. A ) channel wall area (m2)
B ) microchannel width (m)
cp ) specific heat capacity (J/kg K)
D ) microreactor thickness (m)
Dh ) hydraulic diameter (m)
H ) microchannel height (m)
L ) microchannel length (m)
Le ) entrance length (m)
Re ) Reynolds number
Rec ) critical Reynolds number
T ) temperature (K)
Tm ) mean water temperature (K)
Tw ) average wall temperature (K)
T0 ) entering water temperature (K)
U ) overall coefficient of heat transfer (W/m2 K)
Vmean ) mean fluid velocity (m/s)
Vz ) z-directional velocity of water (m/s)
x ) coordinate in the direction of channel length
y ) coordinate in the direction of channel width
z ) coordinate in the direction of channel height

Greek Symbols
Figure 11. Experimental data compared against correlations.
R ) thermal diffusivity (m2/s)
The results show a fast decrease of the critical Reynolds ε ) channel roughness (m)
number, until the aspect ratio reaches a value of 6. After that, η ) water viscosity (Pa s)
the aspect ratio does not contribute to the transition to λ ) thermal conductivity (W/m K)
turbulence. The lowest critical Reynolds number was shown to ξ ) dimensionless independent variable ) z/L
be around 410. It was also shown that the angle of merging of ψ ) dimensionless independent variable ) y/H
inlet channels also influences the critical Reynolds number. Fluid F ) water density (kg/m3)
flows in channels with a higher angle of merging have a lower ω ) dimensionless independent variable ) x/H
Ind. Eng. Chem. Res., Vol. 47, No. 19, 2008 7455
Literature Cited (12) Lee, P. S.; Garimella, S. V.; Liu, D. Investigation of heat transfer
in rectangular microchannels. Int. J. Heat Mass Transfer 2005, 48, 1688–
(1) Norton, D. G.; Wetzel, E. D.; Vlachos, D. G. Thermal Management 1704.
in Catalytic Microreactors. Ind. Eng. Chem. Res. 2006, 45, 76–84. (13) Qu, W.; Mudawar, I. Analysis of three-dimensional heat transfer
(2) Peng, X. F.; Peterson, G. P. The effect of thermofluid and geometrical in micro-channel heat sinks. Int. J. Heat Mass Transfer 2002, 45, 3973–
parameters on convection of liquids through rectangular microchannels. Exp. 3985.
Heat Transfer 1994, 7, 249–264. (14) Fedorov, A. G.; Viskanta, R. Three-dimensional conjugate heat
(3) Peng, X. F.; Peterson, G. P. Convective heat transfer and flow friction transfer in the microchannel heat sink for electronic packaging. Int. J. Heat
for water flow in microchannel structures. Int. J. Heat Mass Transfer 1996, Mass Transfer 2000, 43, 399–415.
39, 2599–2608. (15) Bejan, A. ConVection Heat Transfer, 1st ed.; John Wiley & Sons:
(4) Wu, P.; Little, W. A. Measurements of friction factor for the flow New York, 1984.
of gases in very fine channels used for micro miniature Joule-Thompson (16) Koo, J.; Kleinstreuer, C. Viscous dissipation effects in microtubes
refrigerators. Cryogenics 1983, 23, 273–277. and microchannels. Int. J. Heat Mass Transfer 2004, 47, 3159–3169.
(5) Xu, B.; Ooi, K. T.; Wong, N. T.; Choi, W. K. Experimental (17) Perry’s Chemical Engineer’s Handbook, 7th ed.; McGraw-Hill:
investigation of flow friction for liquid flow in microchannels. Int. Commun. New York, 1999.
Heat Mass Transfer 2000, 27, 1165–1176. (18) Harms, T. M.; Kazmierczak, M. J.; Gerner, F. M. Developing
(6) Sharp, K. V.; Adrian, R. J. Transition from laminar to turbulent flow Convective Heat Transfer in Deep Rectangular Microchannels. Int. J. Heat
in liquid filled microtubes. Exp. Fluids 2004, 36, 741–747. Fluid Flow 1999, 20, 149–157.
(7) Choi, S.; Barron, R.; Warrington, R. Fluid flow and heat transfer in (19) Hausen, H. New equations for heat transfer with free or forced
microtubes. ASME Micromechan. Sens., Actuators, Syst. 1991, 32, 123– convection. Allg. Waermetech. 1959, 9, 75–79.
134. (20) Wibulswas, P. Laminar-flow heat-transfer in non-circular ducts.
(8) Wibel, W.; Ehrhard, P. Experiments on the Laminar/Turbulent Ph.D. Thesis, University of London, 1966.
Transition of Liquid Flows in Rectangular microchannels. Presented at (21) Lee, P. S.; Garimella, S. V. Thermally developing flow and heat
ASME, The Fifth International Conference on Nanochannels, Microchannels transfer in rectangular microchannels of different aspect ratios. Int. J. Heat
and Minichannels, 2007. Mass Transfer 2006, 49, 3060–3067.
(9) Guo, Z. Y.; Li, Z. X. Size effect on microscale single-phase flow (22) Li, Z.; He, Y. L.; Tang, G. H.; Tao, W. Q. Experimental and
and heat transfer. Int. J. Heat Mass Transfer 2003, 46, 149–159. numerical studies of liquid flow and heat transfer in microtubes. Int. J. Heat
(10) Papautsky, I.; Ameel, T.; Frazier, A. B. A review of laminar single- Mass Transfer 2007, 50, 3447–3460.
phase flow in microchannels. Proceeding of 2001 ASME International
Mechanical Engineering Congress and Exposition, New York, NY, ReceiVed for reView January 31, 2008
November 11-16, 2001. ReVised manuscript receiVed June 4, 2008
(11) Li, J.; Peterson, G. P.; Cheng, P. Three-dimensional analysis of Accepted July 16, 2008
heat transfer in a micro-heat sink with single phase flow. Int. J. Heat Mass
Transfer 2004, 47, 4215–4231. IE8001765

You might also like