Beruflich Dokumente
Kultur Dokumente
16116
presented by
Severin Wiilchli
Dip!. Masch.-Ing. ETH
born on the 13th June 1974
citizen of Brittnau (AG), Switzerland
2005
I
Danksagung
Herrn Dr. Marco Stampanoni von der Swiss Light Source (SLS) am Paul
Scherrer Institut danke ich fur die Hilfe bei den Vorbereitungen und den
grossen Einsatz bei den langen XTM-Messungen.
Die Herren Oliver Kdiuchi, Patrick Reichen und Marcel Strotz, welche im
Rahmen einer Semester- oder Diplomarbeit meine Arbeit unterstutzten,
danke ich fur ihr Engagement und ihre kritischen Fragen und Ideen.
Fur die tolle Zeit mit unvergesslichen Momenten danke ich allen ak-
tuellen und ehemaligen Kollegen am Institut fur Verfahrenstechnik. Ich
werde das sehr kollegiale Klima und die motivierenden und kritischen
Stunden vermissen. Meinen beiden Nachfolgern, Franz Trachsel und Do-
nata Banowski, danke ich fur die Mithilfe zum Schluss meiner Arbeit
und wunsche ihnen viel Spass und Vergnugen bei der Weiterfuhrung der
spannenden Arbeit.
Ganz besonders danken mochte ich naturlich meinen beiden langjahrigen
Schlieremer Kollegen Adrian Wegmann und Nils Kruse. Long live the
PowerLab!
Ein ganz besonderer Dank gilt meinen Eltern und meiner Familie, fur
ihre Unterstutzung wahrend meiner langen Ausbildungszeit.
Grossten Dank gilt meiner Freundin Sibylle Frei. Sibylle hat seit Beginn
meines Studiums viel Verstandnis fur meine Arbeit gezeigt. Danke, dass
Du immer da warst.
III
Summary
Z usammenfassung
Table of Contents
Danksagung I
Summary III
Zusammenfassung V
Nomenclature XIII
1 Introduction 1
4 Flow regime 53
5 Pressure drop 79
5.1 Pressure drop in microchannels 79
5.2 Comparison with correlations . 82
5.2.1 Homogeneous pressure drop model 87
5.2.1.1 Main equations of the homogeneous pres-
sure drop model . . . . . . . . . . . 87
5.2.1.2 Comparison with experimental data 88
5.2.2 Friction pressure drop by Lockhart-Martinelli 89
5.2.2.1 Theory of the Lockhart-Martinelli method 89
5.2.2.2 Comparison with experimental data 91
5.2.3 Friction pressure drop by Chawla . . . . . . . 95
5.2.3.1 Empirical pressure drop model by Chawla 95
5.2.3.2 Comparison with experimental data 97
5.2.4 Friction pressure drop correlation by Friedel . 99
5.2.4.1 Basics................. 99
5;2.4.2 Comparison with experimental data 101
5.2.5 Friction pressure drop method by Muller-
Steinhagen and Heck 102
Table of Contents XI
References 164
Nomenclature
Latin letters
A m2 (cross-sectional) area
a,A constant value
A 21 Einstein coefficient for the transition
b m channel width
b,B constant value
c constant value
C constant
C constant
C constant factor (Blasius equation)
Cc contraction coefficient
D m diameter
D m 2 /s diffusion coefficient
d m channel depth
d constant value
Dh m hydraulic diameter
De m effective particle diameter
Dp m particle diameter
DB m diameter of diffraction-limited
point spread function
D 2 -D 2-dimensional local velocity gradient tensor
e m smallest resolvable distance
e constant value
el,e2 empirical parameters (Chawla)
XIV Nomenclature
f constant value
ffric friction factor
F(T) fraction of total population
Fr Froude number
9 m/s2 gravitational acceleration
G kg/(m 2 s) liquid mass flux
G constant parameter (Muller-Steinhagen)
g(w) line shape factor
I laser intensity
'l,J pixel locations
J m/s superficial velocity
K constant value
ks m surface roughness
K/d surface roughness
i,L m length
M kg mass
M magnification
M kg/s mass flow rate
m kg/(m2 s) mass flux
N,n number of realizations
n index of refraction
NT absorber total number density
NA numerical aperture
Nu Nusselt number
P Pa, bar pressure
P constant value
Pfric Pa, bar friction pressure
Pace Pa, bar acceleration pressure
Pred Pa, bar reduction pressure
Pgrav Pa, bar gravitational pressure
Pl,P2 parameters (Mishima)
q constant value
Q electronic quenching rate
r,R cross-correlation factor
r m radius
r constant value
R m spatial resolution
Nomenclature xv
Re Reynolds number
s slip
s m diffusion distance
s -] constant value
SF fluorescence signal intensity
t,T s time
T K , C temperature
t constant value
u m/s velocity
u constant value
v constant value
V m3 volume
V m 3 /s volume flow rate
w m/s velocity
w constant value
We Weber number
x m Eulerian Cartensian coordinate
x mass fraction
x mass transport fraction
X two-phase multiplier (Lockhart-Martinelli)
z m length
z m thickness of scintillator
Greek letters
A m wavelength
A dimensionless parameter
Ar real eigenvalue
Acr complex conjugate eigenvalue (real)
Ad swirling strength,
complex conjugate eigenvalue (complex)
v m 2 /s kinematic viscosity
w l/s vorticity
I> 0
inclination angle
I> proportionality factor (Lockhart-Martinelli)
IT characteristic number
W surface tension parameter
W dimensionless parameter
W fluid properties ratio (Chawla)
{} kg/m3 fluid density
a N/m surface tension, interfacial tension
T N/m2 shear stress
s particle relaxation time
,
Tp
~ relative error
two-phase flow parameter (Chawla)
Abbreviations
abs absorbed
abs absolute
avg average
Ar Armand
B Brownian motion
Cha Chawla
Col Coleman
emit emitted
enr energy
eth ethanol
exp experimental
G gas
glyc glycerol
H homogeneous model
horn homogeneous
Ka Kawahara
L liquid
Ma Massena
XVIII Nomenclature
Chapter 1
Introduction
or for synthesis, the detailed knowledge of the mass and heat transfer
processes are essential. Many of the microreactor concepts were originally
based on single-phase gas or liquid systems (Ehrfeld et al., 2000) or on
two miscible or immiscible phases (Jensen, 1999). Only recently, gas-
liquid microreactors have been introduced in a systematic fashion (Losey
et al., 2001; de Mas et al., 2003; Jahnisch et al., 2000). For multiphase
reactions, the rate and selectivity of the chemical process can be greatly
influenced by the transport of a. single reagent within and across phases.
The enhanced heat transfer observed in microreactors may be exploited
for highly exothermic reactions due to the removal of heat generated and
the suppression of hot spot formation. Another aspect is the optimiza-
tion of slow reactions getting close to the thermodynamic equilibrium.
An example for this was demonstrated by TeGrotenhuis et al. (1999)
for the water-gas~shift reaction, which is investigated as a gas purifi-
cation reaction in the scope of a future application of microtechnology
for reforming applications. In microreactors, heat-exchanging channels
may be introduced, which allow for temperature profiling of the reactor
according to the specific needs of the reactions.
Jensen (2001, 1999) pointed out the advantages of designing and manu-
facturing chemical reactors similar to micro-electro-mechanical systems
(MEMS): heaters and sensors may be integrated and short response
times for heat exchanging tasks are found due to the thin walls applied.
He propagates the application of aerosol and technologies similar to the
inkjet method besides wet preparation and thin film technologies.
The specific surface area is generally high for microstructured reactors.
This increased surface-to-volume ratio which allows for the suppression
of homogeneous reactions (Veser, 2001), however, may also promote un-
desired follow-up reactions, such as total oxidations, at active, but unse-
lective construction materials, such as stainless steel at high temperature.
Microreactors allow safe processing in otherwise explosive regimes. Many
partial oxidations can thus be carried out using pure oxygen and at
elevated pressures. By this, space-time yields can be increased, as e.g.
demonstrated for the ethylene oxide synthesis (Kestenbaum et al., 2002).
A very instructive example for safe processing is given by the hydrogen-
oxygen reaction, which, according to all experience gained so far with
1.1 Motivation for the presented work 3
microreactors, can be handled very safely when operated within the ex-
plosive regime (Veser, 2001). In this content, intrinsic saftey is an often
quoted and recently theoretically confirmed term for such microreactor
processing (Veser, 2001).
Finally, microreactors allow for easier scale-up, as the geometry of the
"unit cell" (the microchannel) does not need to be changed. However,
when numbering-up of microstructured reactors is done, it needs to be
taken into consideration that the housing material has to be minimized to
reduce heat-losses, start-up energy demand and time demand. Therefore,
the practical needs might result in more macro-sized reactors carrying
microchannels as far as scale-up is concerned.
Chapter 2
gas - liquid
gas - solid
liquid - liquid
liquid - solid
solid - solid
The characteristic of multiphase flows is that the two phases can show
different flow velocities. Generally, the gas velocity is higher than the li-
quid velocity. This phenomena, called phase slip, requires the differentia-
tion between mass fraction and mass transport fraction, and the volume
fraction and volume transport fraction, respectively. This is equivalent
to a differentiation between the ratio of the amounts of fluid in a part
of a channel (volumetric content or void fraction) and the ratio of the
amounts of fluid flowing through a cross-section of the channel.
The following definitions are commonly used to describe the phase frac-
tions in two-phase gas-liquid flows:
10 2. Basics and state of the art
. M~
Xi=---- with (2.1)
Ma+ML
with (2.2)
with (2.3)
with (2.4)
The phase slip, s, is defined as the ratio of the actual velocities of the
two considered fluid phases.
Wa
s=- (2.5)
WL
Wi D
Reynolds number Rei = - - (2.6)
Vi
2.1 Multiphase flow 11
w? rh?
F'roude number Fri = _1,_ = 1, (2.7)
gD {l~gD
QD
N usselt number NUi= - - (2.8)
A"1,
The written indexes, i = (L, G), can be used for both, gas and liquid
phase. Furthermore, rhi = lVIi/A stands for the mass flux [kg/m 2 s], a
is the heat transfer coefficient [W/m 2 K], Ai is the heat conductivity
[W/ mK], and a denotes the interfacial tension between the two phases
[N/m]. D stands for the pipe diameter. In case of rectangular channels,
the hydraulic diameter D h is considered,
4A
Dh = d 2 b = 150 ... 266.67 /-lm, (2.10)
2 + .
where A denotes the cross-sectional area of the channel, d is the channel
depth and b the channel width.
For the formation of the above mentioned dimensionless numbers, either
the phase velocities, Wi, and therewith the mass flux, rh i , are considered
using the effective channel cross-sectional area flown through by the fluid
(which is smaller than the total cross-sectional area), or using the total
cross-sectional area. In the latter case, the superficial velocities ji are
considered.
. Vi
Ji = - (2.11)
A
The pha.."e distribution of a multiphase flow is influenced by several forces
as
gravity
12 2. Basics and state of the art
pressure forces
buoyancy forces
interfacial tension forces
friction forces.
The interfacial tension, a, between the two phases tempts to form the gas
phase into spherical shape, and the liquid droplets into round droplets
counteracted by spatial pressure differences, friction forces, or buoyancy
forces.
outweigh the gravitational force. In this case, flow patterns are observed
that can also be found in vertical flows.
In the intermittent flow regime (slug flow and plug flow) the gas phase
is concentrated in elongated bubbles taking the bigger part of the cross-
sectional area. Intermittent flow is characterized by alternating liquid
and gas segments flowing in the channel.
At high gas velocities, the flow minimizes the energy dissipation due
to friction pressure drop and forms an annular flow pattern. The liquid
phase is pushed towards the channel walls, the gas phase is flowing at
high velocities in the liquid core.
In the case of small hydraulic diameters, as found in microchannels, the
ratio of channel surface to channel volume increases. This leads to a re-
lative increase of the surface forces (chap. 2.1.2) compared to gravity. In
channels with diameters below 1 mm only bubbly (relatively large bub-
bles), intermittent and annular flow are observed for a water-air system.
To predict the flow pattern of a fluidic system, flow regime maps are used.
In these maps, the expected flow regime is plotted a.." a function of the
gas and liquid mass fluxes, gas and liquid superficial velocities or other
flow constants. Commonly used flow regime maps for large diameter
pipes are constructed using a water-air flow at low pressures and fully
developed flows. Especially for fluids differing strongly from the phy-
sical properties of water and air, for large differences in the hydraulic
diameter, and for high pressure applications, the prediction of the flow
pattern results in substantial errors. The lines in the flow reginle maps
indicate mean values for the transition region between two flow regimes.
A sudden change in regimes for small parameter variation does not take
place.
In the early 1950's, Baker (1954) constructed a flow pattern map for
two-pha.."e gas-liquid flow in large diameter pipes.
In this diagram, the value Ma/().:rrD 2 /4) is plotted against the dimen-
sionless ratio MLAW / Ma, where
A = [( fla) (~)]
1.2 1000
1/2 (2.12)
16 2. Basics and state of the art
100
6
4
BUbbly
2
-
'b
~
t::
10
6
4
-
1$
Cl
:::i!:
2
6
4 Stratified
2
0.1
2 468 2 466 2 468
10 100 1000
(A. M L ' I M G)
Figure 2.1: Flow regime map by Baker (1954) for horizontal pipes.
(2.13)
is composed of
dPfric
--- (2.15)
dz
dpacc
--- = -
d ( (]L' 2 (1- EO) + (]o wo'
wL'
2EO) (2.16)
dz dz '
18 2. Basics and state of the art
- dPgrav
dz = 9 ( (
{}L' 1 - EG
)
+ {}G EG ). s'l.n<I>.
. (2.17)
For large-scale vertical flows, the gravitational pressure drop is the major
factor in the total pressure drop. For horizontal flows on the other hand
(inclination angle <I> = 0), this portion can be neglected.
Friction causes a pressure decrease along the direction of flow. If a gas-
phase is present, it will expand due to the pressure dependence of the gas
density. The following increase in velocity, according to the conservation
of mass, leads to an additional pressure drop due to acceleration. In
most cases, the pressure drop due to acceleration is a lot smaller than
the frictional pressure drop and can be neglected.
By far the largest contribution to the total pressure drop is provided by
the frictional pressure drop. For multiphase flow in large diameter pipes,
several methods for the estimation of the friction pressure drop exist.
In most cases, the phase distribution and the flow pattern are neglected
and the pressure drop results are therefore inaccurate. Among the most
established calculation methods are those by Lockhart and Martinelli
(1949); Chawla (1968); Friedel (1978), and Muller-Steinhagen and Heck
(1986).
The mostly heterogeneous catalysts can be divided into two types: those
used in fixed-bed processing, in which the catalyst is stationary and the
reactants pass over it, and those used in fluidized-bed or slurry process-
ing. Many metals and metal oxides have general hydrogenation activity,
but palladium, platinum, rhodium, and ruthenium make exceptionally
active hydrogenation catalysts and are used frequently.
0+
nitrobenzene
C6HsN02
3 H2
-
Pd/C5%
6
aniline
C6H7N
+ 2Hf)
Chapter 3
3.1.1 Microchannels
3.1.1.1 Manufacturing methods
inhibitor, and resin (e.g. positive photo resist) and used mainly as a mask
for transferring a pattern into an underlying layer by wet or dry etching.
The masks are written using e-beam (or laser beam) exposure machines,
so-called pattern generators. The layout data is digitized and drives the
pattern generator, that directly transfers the patterns to the photosen-
sitized masks. 5-inch wafer masks with little data, large structures, and
limited accuracy (as required in the present experiments) are very cost-
intensive. A cheaper alternative is the use of exposed transparency film.
For the formation of microchannels (bulk micromachining) different fab-
rication methods are available, wet etching techniques and dry etch-
ing methods, respectively. Wet etching techniques are mostly isotropic.
Special etching solutions (e.g. Potassium Hydroxide (KOH), Ethylene
Diamine Pyrocatechel (EDP) or Tetramethyl Ammonium Hydroxide
(TMAH)) etch single crystalline silicon with different etch rates along
different crystal directions. Microchannels fabricated using a wet etch-
ing method often feature triangular or semi-triangular channel cross-
sections. The width to depth ratio is limited by the crystalline composi-
tion of the silicon. The Reactive-Ion Etching (RIE) technique is the most
established dry etching method. Reactive ions are generated in a plasma
discharge. An additional physical activation energy supply assists the
chemical etching process. The anisotropy of this process can be adjusted
by an electrode bias. RIE provides etch rates between 100-1000 nm/min.
An expansion of the RIE technique is the Deep reactive-ion etching
(DRIE) method. DRIE is advantageous for anisotropic dry etching of
silicon for deep structures up to 1000 J.Lm. This is enabled by a high
density (inductively coupled) plasma source and an alternating process
of etching and protective deposition (multiplexing, Fig. 3.1B). Typical
performance data are etching rates between 1 - 3 J.Lm/min, aspect ra-
tios up to 30 : 1, a maximum etch depth of approximately 1 mm. Main
advantages compared to the wet etching techniques is the anisotropy,
which is independent of the crystal orientation, and the possibility of
manufacturing straight sidewalls with high aspect ratios (Fig. 3.1A). In
contrast, DRIE is a single wafer process, which means the process is time
consuming.
All the microchannels and microreactors presented in this work are fabri-
cated in the clean room facilities of the Institute of Mechanical Systems
24 3. Experimental setup and mea..'3urement methods
i id: B
Figure 3.1: SEM image showing the rectangular channel cross-
section with straight sidewalls (A). Due to the relatively high aspect
ratio, the channel ground in the center is etched slightly deeper than
close to the sidewalls (b:.d ~ 10%). On the channel sidewalls, the
single etching steps of ~ 1 pm lead to streamwise trenches (B).
3.1 Experimental setup 25
Flow characterization
The most simple setup for a two-phase gas-liquid microfluidic applica-
tion is aT-shaped channel. The two fluids enter the microchannel, meet
at the T-junction, and form a multiphase flow pattern in the long com-
mon channel section towards the multiphase outlet. Test experiments
with this first configuration showed pulsating, unsteady flow, especially
for wide channels and slow superficial velocities. This disadvantage was
improved by the implementation of a static mixer at the mixing zone of
the T-shaped channel. A small number of round pins with 50 J.Lm in di-
ameter were included (Fig. 3.2A), slightly increasing the pressure drop.
26 3. Experimental setup and measurement methods
\ ,
oL
oor
000
000
00
38 mm
I I
-
,\
, \c"
\
B
38 mm
I
oor
00
000
000
I
00
/)
Figure 3.2: Schematic of the used channel geometries and enlarge-
ment of the two-phase mixing zone. T-shaped channel with static
mixer to avoid pulsating flow (A), and V-shaped final design of the
microchannels (B).
depths (100, 150, and 200 J-lm). This results in hydraulic diameters
4A
Dh = 2 d +2 b = 120 ... 266.67 J-lm. (3.1)
l
D = 142.5 ... 316.7. (3.2)
h
Hydrogenation reaction
The chemical reaction (catalytic hydrogenation of nitrobenzene) makes
great demands on the design of the microreactor. Even tough the mass
and heat transfer rates are higher than in commercial batch reactors, the
reaction process requires a certain residence time. This can be realized by
slower velocities or longer channels. As slower fluid velocities are assumed
to result in less internal recirculation, we concentrated on the alternative
of longer reactors. This is realized by meandering channels.
The hydrogenation of nitrobenzene requires the presence of Palladium
as catalyst. Two different possibilities for the contacting of the educts
with the catalyst are considered, the packed-bed method and the sur-
face coating option. For the packed-bed method (see Fig. 3.3A), activated
carbon particles, covered with Palladium, are inserted directly in the mi-
croreactor. Thus the nitrobenzene and the hydrogen will flow through
this catalyst fill. To prevent the catalyst particles from leaving the mi-
croreactor or from flowing back into the inlet tubing, obstacles at the
beginning of the multiphase part of the reactor and in front of the outlet
are included in the channel design (Fig. 3.3B). To enable the heating
of the flow (reaction requires 50C) and to improve the fluid mixing,
part of the inlet is equipped with mixing pins (as in the characteriza-
tion version of the microchannels, Fig. 3.2B and C). The presence of the
packed-bed of the particles causes a high pressure drop over the reac-
tion channel. Hence the length is limited by the fluid delivery equipment
and the channel sealing. The decreased residence time due to the limited
channel length is compensated by the extraordinary fluid mixing and
28 3. Experimental setup and measurement methods
i\. ~~
Educt \.I rf/ //
,
1'/ [ """".
.' ~ Ul \ A
c:=::J
c:=:J
--
c==J
c:J
c==J
c:::::J
-----,Ij
c:::::J
- B c
UqU~~ inlet Hea~?g zone Product outlet
[(umum;/
: .
.''
~.:;,.: ~::Jt-~~
Gas inlet D
To feed the microchannels described in chap. :3.1.1 with the two fluids
of the gas-liquid multiphase flow, a connection using tubing was needed.
The tubing and especially the connection to the microchannel has to
fulfill several n~quirements, such as tightness, chemical resistance, or the
possibility to be cleaned. The most common method (e.g. Losey et al.
(2001); de Mas et al. (2003)) is to glue the tubing air tight to the channel.
Main advantage of this connection is the easy production and the tight-
ness. The cleaning of the tubing, especially from seeding particles (see
chap. 3.:).2), can be difficult. Furthermore, if a large number of different
channel layouts are considered, the glued connection may be unhandy.
For this reason wc developed a channel holder.
The microchanm-)l can be placed inside the lower part of the channel
holder. An optical access is enabled by the opening in the bottom side of
the hokler. As scaling, a GoreTex or a Neopren layer with holes for
the fluidic connections is used. Both materials are largely chemically re-
sistant and feature an excellent tightness up to 6 bar fluidic pressure. The
channel is tightened by screwing the upper part to the bottom (compres-
sion scaling). Due to the hardness of the GoreTex sealing, the chance of
breaking a microchannel when tightening the screws is higher than when
3.1 Experimental setup 31
Fluidic connections
Sealing
Microchannell -reactor
Figure 3.6: Channel holder with sealing and connections for the
tubing and the tightening screws.
using the Neopren sealing. The upper part of the holder features three
fluidic connections, one for the gas inlet, one for the liquid inlet, and one
for the multiphase flow outlet. The connections are fully compatible to
the Upchruch tubing connections. For the chemical reaction, an addi-
tional connection was applied for introducing the catalyst particles (see
chap. 8).
The channel holder's dimensions (length and width) correspond to a
commonly used microscope slide. Hence it is possible to clamp the holder
into the corresponding position of the microscope.
A c
Figure 3.8: Experimental setup with the Zeiss Axiovert 200 micro-
scope (left) and the Nd:YAG laser with the light connection to the
microscope (right).
and reproducibility within 0.05%). The volumetric flow rate of the liquid
phase, VL, is varied between 0.05 and 5 ml/min for the flow characteri-
zation experiments. Gas-phase container was a commonly used 200 bar
pressurized bottle using pressure reducing valves. The volumetric flow
rate, VG, was controlled by a mass flow controller (Bronkhorst, accuracy
0.5% of reading plus 0.1% full scale) and is varied between 0.5 and
20 ml/min. The fluids are connected to the channel (channel holder) by
flexible tubing (Upchurch, PEEK, 1/16 inch outer diameter and 0.5 mm
inner diameter).
Pressures are measured, at the inlet and outlet of the microchannel
(single phase inlets, multiphase outlet), by pressure sensors (Endress +
Hauser, accuracy 0.5% of full scale). The multiphase flow at the outlet
is fed into a sample container at ambient pressure.
'\-O-----L_.....1 Micrjannel I
r ,..1
.....D<II----JL..-----L-
j
-~-~:-:=~::j'.;-I
I I
I I
LJ.. . 1
I
I
3.1.4 Fluids
water, ethanol and aqueous glycerol solutions (10% and 20% glycerol
mass fraction) as liquid phase and nitrogen as gas-phase were used. All
the characterization experiments were carried out at room temperature
(T = 25C). Water and ethanol have similar values for the viscosity,
but differ by a factor of approximately 3 in surface tension towards ni-
trogen (see Tab. 3.1). The glycerol solutions on the other hand features
a surface tension close to the one of water, but has differences in the
viscosity.
Storage bottle
H2
The channel holder (sealing) was found to be insufficient for high pres-
I:mre experiments. Another way to feed the fiuids without leaking to the
microreactor had to be considered. The channel was covered with thin
layers of chromium (Cr), nickel (Ni), copper (Cu), and gold (Ag) byelec-
tron beam evaporation. Ag provides complete wetting with silver solder.
Cu prevents the formation of an oxide layer on the nickel layer, which it-
self acts as diffusion barrier. Cl' is the actual undercoating. Stainless steel
capillaries are soldered airtight to the reactor (Fig. 3.11). Experiments
showed that the current reactor and fiuidic connections can withstand
pressures up to 120 bar and temperatures up to 90C.
Different ways for heating the reaction zone to the necessary reaction
start-up temperature were considered, a heat-exchanging device, opera-
ted in counter flow mode, heating the complete microreactor including
the channel holder, heating by means of electric resistance, and the im-
plementation of microchannel heat exchanger integrated in the silicon
36 3. Experirnental setup and measurement methods
chip, respectively. The latter was abandoned due to the limited space
for connecting the tubing in the channel holder. Due to the high spe-
cific heating power, the control of the required temperature, and the
small number of experimental adaptations (no additional sealing or fit-
ting devices), we decided the use of an active heating foil (Minco AG,
6.4 x 6.4 mm 2 , heating power 2.4 W). A foil thermometer and a DC
temperature control system (Minco AG) were used for the control of the
heating temperature.
(3.4)
38 3. Experimental setup and measurement methods
In contrast to large scale applications using a thin laser sheet for the
illumination of the flow field, laser-induced fluorelicence in microfluidic
experiments uses whole field illumination. LIF can therefore no more be
regarded as a 2-dimensional measurement method. The recorded reemit-
ted light is the sum of the intensities over the whole channel depth.
Hence, in the presented experiments, the LIF images are 2-dimensionally
averaged over the direction of observation.
In the present experiments, Rhodamine B (Fluka) was added to the li-
quid phase (0.5 ml of 25% Rhodamine B solution into 50 mlliquid, 0.25%
by volume). Rhodamine B emits red light (peak excitation wavelength
550 nm, peak emission wavelength 605 nm) when excited by 532 nm
laser light. On the recorded grayscale images, the liquid therefore is rep-
resented by bright colors, the gas-phase has no fluorelicent signal and is
therefore represented by black colors.
To eliminate the uneven laser and background intensity, the experimental
image with signal intensity SF is calibrated uliing the fluorelicent liignal of
a microchannel completely filled with liquid, SF,L, and the corresponding
signal of an empty channel (gas phase only), SF,G, according to Eq. 3.5.
SF - SFG
SF,cal = S S' (3.5)
F,L - F,G
"oo~"~
.....
~ .. ~
~
~
(3.6)
K,. T
D=----- (3.7)
3 . 7r . J.1 . D p
~x = u ~t. (3.8)
42 3. Experimental setup and measurement methods
(3.9)
This error establishes a lower limit on the desired measurement time in-
terval, ~t. In the present experiments, the characteristic liquid velocities
are in the order of u rv 10 cm/s and ~t rv 500 ns. This yields a relative
error due to Brownian motion of approximatively 1%. Since this error is
a result of diffusion, it is unbiased and can be substantially reduced by
averaging over several particle images in a single interrogation window
and by ensemble averaging over several realizations. According to San-
tiago et al. (1998), the diffusive uncertainty is proportional to !;B/v'"N,
where N is the total number of independent realizations. In the present
experiments we used N rv 200 realizations to ensemble-average the vec-
tor field. So the error due to Brownian motion reduces to less than 5%0
and therefore can be neglected.
In all of the recorded images, the region of best conformance with the
master had to be found. For this, the well known cross-correlation tech-
nique is used. Cross-correlation can be understood most straightfor-
wardly as an operation in the spatial domain. If the shape of a target
is made into a kernel whose values are multiplied by the pixel values
at every location in the image and normalized for the absolute values
of the pixels, the largest response will result when the image contains
the same pattern of grey scale values. A direct cross-correlation in the
spatial domain is computationally very expensive. Therefore, the opera-
tion is more efficiently performed in the Fourier domain by multiplying
(remembering that these are complex numbers) the transforms of the
target and the image together and performing an inverse transformation
on the result. With either implementation, the locations in the image
that match the target are found efficiently even in the presence of noise,
shading, or partial obscllration of the target (Fig. 3.15).
The cross-correlation images (response) are normalized with the maxi-
mum response, which is obviously obtained from the image, where the
master is taken from. For this image, the cross-correlation factor, r, is
maximal and the normalized r[i, j, Nmaster]max equals 1. For a further
44 3. Experimental setup and measurement methods
..
" - .' , , ". . ".'.".'.0'.'0."., ,._., _.,.~,,
--
~
_._~-,-, .. ,'-.-
__ _ . , _ . , _ . , _ , _ _ _
"" .. ,.""" .
,~.
. ..
-
_.~,~,.'
.~.~
_~,J.J
~.-
-.
-- _. __ ....
-. - .
,,~.,,~.,.,
_~,,_.,-,._
, ,
,..... .
.- , . -
,~,_
Figure 3.15: Schematic of the process steps used for the velocity
measurements of unsteady intermittent gas-liquid flow. The liquid
plug of each image is searched by cross-correlation in the Fouri(:~r
space (F{A} . F {B} * ), delivering the position of the best confor-
mance. With this information the raw velocity vectors of the liquid
plug is extracted.
46 3. Experimental setup and measurement methods
Objective Scintillator
~--+----+---=--"7:\
Sample
,~)
Relay mirror ~\
liqul~
PEEK tubing
/ Connectors =
j' [If
3
Ri nln ill
1_:_] .
1
1
I11
:
I J ~~~JJ
Microchannel Channel holder
A B
B C
The accuracy of the syringe pump (accuracy within 0.35%) and the gas
mass flow controller (accuracy 0.5% of reading plus 0.1 % full-scale),
respectively, lead to a mean error in the determination of the superficial
velocities.
All the presented experiments were conducted at room temperature and
atmospheric pressure at the discharge of the test section. To ensure a
steady state flow condition, the measurement location was chosen in
the last quarter of the microchannel. Therefore, the actual pressure at
the test section is close to the atmospheric pressure, and the volumetric
gas flow rate is assumed to be equal to the standard volume flow rate
(Nml/min). The influence of temperature fluctuations on the fluid prop-
erties are neglected, as a maximum change in temperature of 5C was
observed, leading to property differences below 2%.
For the presented reasons, the error due to the flow feed system and the
ambiance, exp, can assumed to be in the range of exp :::::: 5%.
fluorescent additive in the working fluid, and the reduction of the illumi-
nating laser intensity along the illumination path inside the microchan-
nel. The intensity fluctuations in the recorded images using liquids en-
riched with Rhodamine B only (no gas flow present), was determined
to be 5%. This uncertainty originated mainly from the fluctuations of
the laser intensity, as we assumed the fluorescent dye to be completely
dissolved in the liquid.
Due to the small depths of the used microchannels (d = 100 ... 200 jJ,m),
and the high laser intensities, the assumption of uniform illuminating
intensity was made. Theoretically, the recorded light intensity value is
the integral of the emitted intensities over the full depth of the chan-
nel. A reduction of the emitted intensity along the illuminating path
is expected to be present at high liquid content locations only. The
influence of off-focus recording is neglected. The accuracy of the void
fraction measurements is therefore determined to be in the range of
~voidfraction :::::::: 5 - 10%.
A
D s = 2.44 . M 2 . NA' (3.10)
(3.11)
n A n e
8z= NA2 + M.NA (3.12)
52 3. Experimental setup and measurement methods
Chapter 4
Flow regime
C
Figure 4.1: Images of two-phase flow in a microchannel at low
gas and high liquid flow rates (bubbly flow, A), at flow rates yield-
ing intermittent flow (B), and at high gas superficial velocities and
moderate liquid flow rates, leading to an annular flow regime (C).
Images recorded using Laser Induced Fluorescence (LIF, Rhodamine
B added to the liquid phase).
I:J.
!i.
/:;.
!:J.
A
li
li
/1
o
0
0
o 0 0
~
S. o oo 00 00 o o 0 0
4 0.1
o o 0 0
0.1
0 0
o o 0 0
o o 0 0
o
o 0
o o 0 0
o o
0
0 0
o o 0 0 0
o 0
0
0 o o 0 0
o 0
o 0
o
o o 0 0
o 0 0
o 0 0 0
0.01-1--.............---r.~........
, ....-.."T""~-..,..,- ......---r.........-.!'""'!",........+ 0.01+-.,..--,-...,....+--...,..,..---..,.....-,.---....,."". + "!,""!.....
1 10 10
Jolm/s) A D
o 0 o 0 0
o 0 0 /1 /1 /1 o 0 0
o 0 0
o o 0 0 0
0 0
o 0 0 0
o 0 0 o 0 0 0
o 0 0
[] D 0 D
0.1
o o 0 0.1
o o 0 0
o o 0
o 0 ODD 0
o
o
0
0
0
0
o o 0 r:J Cl D D
o 0 D 0 D 0
o 0 0 o o 0 o 0 0 0
o o 0 0 0
0.01+_................................,., . -......_....--....--....... .,.,+ 0.01-1-_....................-.h-,......... .---..,..,-....................-.r-T,-.."".+
4 I!i IIi ,. .0' " 15 7 ,
1 10 1 10
j o lml8]
B JOlm/s] E
lJ 0 0 0
6. I:J. o 0
o
o
0
0
o
lJ
0
lJ lJ
0 0
lJ
o
o
0 lJ
0 0
0
0
o
0 0 0
0.1
0
lJ o lJ 0
o o 0 0
o 0
0
0 0o
o
o o 0 0
0
0
o 0 0
1 1
JL jL .llg 2 Dh 2
II 1 = A
1
_1 D -- 1 1
I..J.g 2. h 2. (j2 (j2
II2 = JG
A
I..J.g _12 . D h --1... 1
(72
1
VL VL .llg 2
II3 = ---::-----:--- (4.1)
A
I..J.g _.1
2.
D.1 1
h 2 (72
1
Dh 2'(72
1
VG
II4 = ---::-----:---:-
A_1 D 1 1
I..J.{) 2. h 2 '(72
4.1 Flow regimes in microchannels 59
2
2 jL . D.g. Dh
IT 1 = (J
= WeL
2
2 jG . D.g . D h
IT 2 = (J
= WeG
-1 jL' Dh
IT 1 . IT3 = = Re L (4.2)
VL
-1 jG' Dh
IT 2 . II4 = = ReG
VG
ks
ITs = -
Dh
where the Reynolds number
jL,G' Dh
R eLG = (4.3)
, VL,a
(4.4)
is used to include the influence of the fluid densities and the surface
tension.
The Buckingham pi-theorem states, that there is a function f valid for
the regime transition boundaries (tb), where
(4.5)
60 4. Flow regime
(4.6)
(4.7)
4.1 Flow regimes in microchannels 61
0 0
0 0
"'0--.. "'-
IS
0 0 0
0 0 0 0
.ot...-'"
0 0 0 0
0
d~ 0 0 0
;:: 0.\ 0
0.1 0 0 0
0
:l . 0 0 0
ofi' 0 0 0 ~ 0 0 0 0
~rr.
0 0 0
~~ 0 0 0
~ : 0
0
0
0
0
0
0
0
0
0
0
0
0.01
,..
0 0
.. e l'
0
t
10
0 0 0
0.01
.,..
Reo0.2 Wft
o
IU
A Reo" Weo B
0 0
"'- 0 0 0 ~t? 0
IS 0 0 0
0 0
..I t
0
0
0
0
0
0 ....-'" 0
0
0
0
0
0
"( ( 01
D.I
" " " " " "
"
"ot
""
"0
0
0 o~
. " 0 0
rr.
~~
: " "" "" ~rr.
~
0
0
0
0
0
0
0" " " 0 0 0"
"" "" ""
0,01
" " " 0,01
0
" "
~ 0' IjI
" "
"'2' "" "" "'0--.. " 0 0 0
d~
"
"
0
" "
" " .-
,,~
,2
""
0
""
0
"" "
0
0
0
0
0
0
:lot
0.1
" " " "
"
""
0
.. Re:2 Weo"
..
0
10
0 0 0
ReoO,2 We:
E F
Figure 4.3: Flow pattern maps using the universal pattern tran-
sition boundaries for a rectangular microchannel for the same flow
and geometry conditions as in Fig. 4.2 (A: water, Dh 187.5 Mm;
B: ethanol, 187.5 Mm; C: glycerol, 187.5 Mm; D: water, 218 Mm; E:
ethanol, 218 Mm; F: glycerol, 218 Mm).
62 4. Flow regime
For all the presented experimental data, the universal flow pattern map
shows a satisfying agreement. The two boundary lines are located in the
transition regions for all the experimental data. Therefore the universal
flow pattern map is capable of making a prediction of the expected flow
pattern for a large variety of fluid and geometrical conditions. The fact
that additional important influencing factors as temperature, the cross-
sectional shape or the channel material and surface roughness, were not
varied or neglected in the presented flow regime map, shows the indis-
pensable need for the experimental verification of the flow pattern for
the detailed design of a chemical microreactor.
1008
6 Misty
4
N -:;;:-
0
l'e
4
-
6-
'':Ji.l:}
2
1
8
6
4
0.1
2 468 24682468
10 100 1000
(I. M'L'I' I M'G)
The flow regime map of Zhao and Bi (2001) (Fig. 4.5) is experimentally
derived using an air-water mixture in triangular microchannels with a
hydraulic diameter of Dh = 866 J.1>m. This hydraulic diameter is still
larger than the diameters in the present study by a factor of approxi-
mately 4. The agreement between the present data and the regime map
for the microchannel by Zhao is accurate for the flow pattern change be-
tween intermittent (plug) and annular flow. Due to the larger hydraulic
diameters and the different channel geometry, bubbly flow is predicted
only for very low superficial gas velocities.
64 4. Flow regime
2
'" '" '" '"
Intennittent
~
.. 0
.~
0.1
0 0 0
6 0 0 0 0
5
4 0 0 0 0
3 0 0 0 0 0
2 0 0 0 0 0
0 0 0 0 0 0 0
0.01
2 3 4 5 6 7 69 2 3 4 (; 6 7 89
1 10
iG [m/sI
The results of Triplett et al. (1999b) (Fig. 4.6) are based on experi-
ments using air and water as fluids in semi-triangular microchannels of
hydraulic diameters D h = 1.09 mm. Even tough the hydraulic diame-
ter and the cross-sectional shape of the channels are similar to the ones
in the experiments by Zhao, the flow regime map by Triplett predicts
the flow patterns of the two-phase flow in microchannels of rectangular
cross-section and hydraulic diameters smaller by a factor of 5 with high
accuracy. Again, no bubbly flow was predicted.
4.2 Comparison with correlations 65
2
bubbl~ chum
~ t::. t::. t::. t::. t::.
6
5 t::. t::.
t::.
t::.
t::.
t::.
t::.
t::.
t::.
t::.
0 0
0
0
4
3 t::. t::. t::.
slug
t::. t::. 0 0 0
2
~
oS 0 0
slug annular
0
.~
0.1
0 0 0
6 0 0 0
5
4 0 0 0
3 0 0 0 0
2 0 0 0
annular.
0 0 0 0 0 0 0
0.01
2 3 4 5 6 7 69 2 3 4 5 6 7 69
1 10
iGlm/sj
t:.
6
5 /';. t:.
t:.
t:.
0
4
3 t:. /';. t:. /';.
slug - ring
/';.
2
~ 0 0 0
..,=t 0.1
0 o 0 0
6 0 0 0 0
5 0 0 0 0
4
3 0 0 0 0 0
2 0 0 0 0 0
semi annular
0 o 0 0 o 0 0
0.01
2 3 4 5 6 7 69 2 3 4 5 6 7 89
1 10
ja [m/sI
,I ,I
10
B
B
4
Bubbly
- - ~isPersed
2
Intennittent
-
B
!:>. !:>. !:>.!:>. !:>. !:>.!:>. -
B
4
!:>. IJ.IJ.IJ.
IJ. IJ.IJ. IJ.IJ.IJ. Wavy
00
Annular
.~ 2
!:>. !:>.!:>. !:>..!:>.
00 annular
o 00 b
0.1 - -
B 0 000
B
4
0
0
000
000
2
0
0
0
0
000
000
0.01
000
.,
0 000
'I
2 4 6 B 2 4 6 B 2 4 6 B
0.1 1 10 100
jG [mls]
A
2
2
/';.
/';.
/';.
/';.
/';.
/';.
/';.
/';.
/';. /';.
/';. o
Wavy annular
0 0
i 0 o 0 0
.~ 0.1
B o 0 0
6 o
o
0
0
0
0
4 Stratified
0 Wavy
0
o
o
0
0
0
0
2 0
0 o 0 0 o 0 0
0.01
2 3 4 5 6 7 B9 2 3 4 5 6 789
1 10
jG [m/s]
B
Figure 4.8: Two-phase flow regime maps of Coleman (A: circular
channels, Dh = 1.30 mm; B: rectangular channels, Dh = 5.36 mm)
using water as working fluid compared with the present experimen-
tal data of a rectangular microchannel with a hydraulic diameter
D h = 210 J.Lm. Experimental data: .: intermittent, 0: annular, 6.:
bubbly.
68 4. Flow regime
...,,
0.5
,;1
.
I
.~ 0.4
II
Slug
<DD
... 1:i
,,
,'" ""
~
'8 0
. . ... ,,,
,
.. ..
0.3
0 <D
,, " "
CD II III a ,,
.
<><>
>
"C 0 III a ,, Annular
'S "
g 0.2 ,O'll III a,a e 3 " <><> "
~ D DI 0
m
'0
lug-AI1J1ular,'
If all .De~.'" "'92" "
(I) 0.1 D Slug flow
C- " " 0 Annular flow
..
:::J
.' " 1
C/) "
0.0 +-~"'--T--Y---OI,...--,----r---'-I----r----r----r--r--Ir---'1""--'---'
Present study
Figure 4.9: Flow regime map for the rectangular microchannel used
in this experiments (250 x 100 j1,m 2 , D h = 143 j1,m).
In the first experiment, the Reynolds numbers are Ree = 21.4 and
ReL = 4.7, respectively. The XTM image in the direction of the flow
(in the direction of the channel axis, Fig. 4.10A) clearly shows the ex-
pected annular flow pattern inside the microchannel. For a better vis-
70 4. Flow regime
(A) (B)
The darker rays coming from the edges of the microchannel are artifacts,
resulting from the deflection of the X-ray beam at the sharp corners of
the silicon or glass.
As can be seen, after the reconstruction of the 1000 projections into the
3-dimensional image, the microchannel appears to be not exactly rect-
angular but slightly distorted. Due to the adaptation of the fluid feeding
system to the experimental S(~tup of the XTM measurement station,
longer and less flexible PEEK tubing had to be used (Fig. :3.17). This
4.3 3-dimensional flow pattern visualization using XTM 71
(A) (B)
(A) (B)
In the cross sectional image of this configuration (Fig. 4.12B), the inter-
facial boundary is no longer marked in clearly bright colors and therefore
hard to detect.
Despite the higher liquid volumetric flow rate, the volumetric gas content
inside the channel is increased only to EG = 0.72. Again, this has to
be compensated by an even higher liquid velocity. Therefore the phase
interface is unstable. No stationary flow pattern is observed.
2'50 ~
200 ~
1'30 :::;
100 :;;;
Q~
50 3-
=T==--O ~'
~ ~ ~ ~ 1= 1~ ~
Channel Length [piXl!I]
When looking at the optical reference image~ the assumption of the fluc-
tuating gas-liquid interface is confirmed. Fig. 4.1:) shows an instanta-
neous acquisition of the corresponding flow. If a large number of such
single images are averaged, stable annular flow seems to be present as in
the other experiments presented so far and as the XTM measurements
of this configuration suggest.
Starting from the configuration of experiment 2, we decrease the gas
volumetric flow rate from VG = 5.0 ml/min to Vc = 3.0 ml/min, while
the liquid flow rate remains at VL = 0.2 ml/min (ReG = 21.4 and
ReL = 19.0). 'l'his configuration, according to the previously presented
flow regime map, is in the range between slug flow and annular flow. In
the temporally resolved optical measurements, both of the flow patterns
could be detected. The XTM measurements (Fig. 4.14A), however, still
show annular flow. An interfacial boundary in the narrow corners of the
channel is difficult to detect, tough. The same effect can be seen in the
cross section view (Fig. 4.14B).
In additional experim(~nts we analyzecl the differences in the :)-
dimensional flow pattern for the two fluidic systems water - nitrogen and
74 4. Flow regime
(A) (8)
100 flm
120 flm
lized in the present work, a detailed set of flow regime maps was derived,
showing the differences in the flow conditions for varying liquid proper-
ties and channel hydraulic diameters. The individual flow patterns show
large transition regions, where multiple flow patterns coexist. Sharply
detached flow regimes are not detectable. In contrast to macro-scale
gas-liquid pipe flow, intermittent flow is dominant in microchannel flow.
Some flow patterns, such as the stratified, the wavy, or the dispersed
flow found in large diameter channels, were not observed for the applied
flow conditions in our microchannels. A low surface tension between the
two fluids favors the annular flow regime, the transition region to inter-
mittent flow is shifted towards lower superficial gas velocities. The same
effect can be found when increasing the channel hydraulic diameter.
The universal flow pattern map presents the regime boundaries for a
wide range of flow and fluid conditions. The flow regime map using di-
mensionless parameters instead of the commonly used superficial gas and
liquid velocities is able to predict sufficiently the apparent flow pattern
in microchannels of rectangular cross-sections. Nevertheless, one has to
consider that important influencing factors as the temperature (affecting
directly the fluid properties), different channel cross-sections, the chan-
nel material, and the surface roughness were not varied or have been
neglected in the flow regime map.
The comparison of the experimental data with flow pattern maps ob-
tained for large scale pipes as well as for microfluidic geometries shows
mostly insufficient agreement. It was shown that the best agreement is
achieved with flow pattern maps constructed for channels with similar
cross-sectional shape (rectangular cross-section), even for large differ-
ences in the hydraulic diameters (Dh,map/ Dh,exp ~ 5).
As a basic principle, an experimental validation of the occurring flow pat-
tern is indispensable for the design of any microchemical or microfluidic
device.
Chapter 5
Pressure drop
0.80 MO
0.78 Oh -120 J.ll11 0.45
-{}- i L =0.28 m1s
0.10
-0- iL =0.42 m1s
0.40
.. ...
---..
.
;::'
ac.
0.88
0.80
~ iL - 0.56 m1s
ic.
0.3&
0.30 0-
--0-
--0
0.55 0
e e
"0 0.25
"!!:> 0.50
0.45
!!
:> -n
0.20 -.::r
'"'"!! 0.40
'"~
0.. 0.1& Dh -133.3 "m
0..
0.35
0.10
=
-{}- iL 0.25 m/s
0.30 -0- iL " 0.38 m/s
0.25
0.0& =
-&- iL 0.50 m/s
0.20 0.00 I
2 3 4 & & 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.8 4.0 4.8 8.0 s.& M
Superficilll glls velocity i e [mls) Superficial gas velocity iG [m/s)
0.80 040
"'.."
030
~ iL =0.24 m/s
B-
0.35
"'"
.. 025
c. 0.30 c.
e
"0
e
"0
0.25 0.20
!! !!
iil
~
0.20 c--
0.15
--0 0
010
005
0.00
0.0 0.4 0.8 1.2 1.6 2.0 2.4 2.8
Superficial gas velocity i e [m/a]
For all the presented channel diameters D h , it can be seen in Fig. 5.1 that
higher superficial liquid velocities, iL, lead to larger pressure drops. For
5.1 Pressure drop in microchannels 81
for low gas flow rates. The pressure drop is decreasing with increasing
gas flow rates. For higher supprficial liquid vclocitiPH, the pressure drop
maximum at the change of flow regimes is less distinct than for the
D h = 150 J-Lm channel.
0.4"
0.35,
e.. 0.3 ..
a.
e
~ 0.25.
'"
!!:'"'" 0.2 ..
0.15.
".A
0.26,
0.24"
~ 0.22,
eo
g. 0.2.
-0
e 0.18,
::I
e0.16"
ID
a.
0.14 ..
0.12,
~Pacc =
(5.1)
~Pacc =(JL
( w L2 (1 - cc;) + (JC; Wc;2 cc; ) . (5.2)
where 'Y is the ratio of the cross-sectional areas of the multiphase part of
the microchannel and the inlet cross-sections, and Cc is the contraction
coefficient, which is a function of 'Y
Achannel
'Y= (5.4)
Ainlet
1
(5.5)
Cc = 0.639(1 - 'Y)O.5 + 1.
Not taken into consideration are the pressure drops over armatures and
valves and the gravitational pressure drop.
Figure 5.5 shows the contributions of the friction pressure drop, the
pressure drop due to acceleration, and the pressure drop caused by a
reduction in diameter to the total measured pressure drop for the two
different hydraulic channel diameters Dh = 120 J.lm and Dh = 160 J.lm.
Even though the pressure drop due to area reduction increases with
increasing superficial gas velocity, the contribution to the total pressure
drop remains negligible. The friction pressure drop to total pressure drop
ratio, tlPfric/ tlPtot, varies between 99.5% for large hydraulic diameters
and low superficial gas and liquid velocities, and 87.4% for small diameter
channel using high superficial velocities.
86 5. Pressure drop
100 16
98 o~
14
96
12
'Q. t>
94 '" '0
~
AI
10 JJ
-
B 92
!
a.~
-
<I
a.~
<I
90
88
;;r _.
.. -
__ f;l-
8
S
t>
'0
i
86 0
t:.
~Pt1c (friction) (left axis)
~p... (acceleration) (right axis) ..
84 'l ~Prod (reduction) (right axis)
v .. ia '" 0.20 mls 2
82 - JG '" 0.40 m1s
80 0
0 2 4 6 8
Superficial liquid velocity k[m/s]
A
100 16
98
14
96
94
12 .g
'1*
AI
10 JJ
-
92 0 ~lric (friction) (left axis)
~ ~P... (acceleration) (right axis)
!
-
<I
a.~
<I
90
88
'l ~P..d (reduction) (right axis)
.... ia '" 0.10 m1s
- ia '" 0.20 mls
8
6
t>
'0
i
86
4
84
2
82
80 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Superficial liquid velocity jL[m/s]
B
Figure 5.5: Pressure drop due to friction (0, left axis), acceleration
(6.), and reduction (V') (both right axis) in percent of the total
pressure drop for two different superficial gas velocities against the
superficial liquid velocities for hydraulic channel diameters D h
120 /-Lm (top) and D h = 160 J.Lm (bottom).
5.2 Comparison with correlations 87
homogeneous density
XG 1- XG)-l
(lH,hom = (- + --- (5.7)
(lC (lL
momentum density
(5.8)
energy density
{!H,enr =
(5.9)
Both the momentum density model (Eq. 5.8) and the energy density
model (Eq. 5.9) require the use of a void fraction (EG) model (see chap.
6.1).
The friction factor, frric, is calculated using the Blasius equation,
c (5.10)
frric = R m'
eH
88 5. Pressure drop
where C and m are constant factors according to the present flow con-
dition::;. For laminar flows, m is equal to unity, m = 1, and C is 64. For
two-phase flows the friction factor is calculated using the homogeneous
Reynolds number
rh D h
ReH = - - - (5.11)
TJH
Again, for the homogeneous viscosity, TJH, several approaches can be used
(Kawahara et al., 2002):
TJH,5
(XG + 1 - XG) -1
(5.16)
TJG TJL
TJL TJG
TJH,6 (5.17)
TJG + x~4 (TJL - TJG)
In Tab. 5.1 the root mean square errors in percent between the different
possible pressure drop calculations and the experimental data is shown.
It can be seen that the homogeneous density (Eq. 5.7) is not suitable for
microfluidic applications except for the combination with the viscosity
calculated using Eq. 5.14. Surprisingly, this density approach does not
only give the two predictions with the worst conformance (version 1 and
2) but also the version with the best conformance with the experimental
results (version 3, using Eqs. 5.7 and 5.14).
Density
Visc.
llH,hom llH,mom llH,enr
17H,l 1080% 211% 41%
17H,2 1050% 205% 41%
rlH,3 21.5% 66% 83%
17H,4 181% 30% 59%
17H,5 450% 64% 33%
17H,6 805% 152% 21.6%
( tl
P
)
tll TP
= <P~ (tltllP ) G = P
<Pt (tl )
tll L
(5.18)
90 5. Pressure drop
<I>b = 1 + ex + x 2 (5.20)
(5.21)
C = 21 (1 - e-O.319.Vh) (5.22)
C = A . Aq . wr . Reto (5.23)
2
A- 1]L W = 1]L(jG + jL) , (5.24)
- (}L (J Dh ' (J
5.2 Comparison with correlations 91
Lee and Lee (2001) defined the parameters A, q, T, and s empirically for
their channels and laminar-Iaminar flow conditions as A = 6.833 X 10- 8 ,
q = -1.317, T = 0.719, and s = 0.557. Depending on the choice of the
factor C, the model by Lockhart and Martinelli (1949) can be adapted
to small-scale channel flow.
The pressure drop model by Lockhart and Martinelli (1949) was defined
for large diameter pipes, even though no area of validity is mentioned.
The data calculated using the original approach for strictly laminar flow
(C = 5) overestimates the pressure drop with a mean deviation of t1p =
0.55 bar and a standard deviation (j = 0.40 bar.
2.0 - t - -.........- - - ' - - - ' - - - . . l . - - - ' - - - - : : : - ' - - - - - - L - - t _
Ri 1.8
0
So
a; 1.6 0
0
0
a.~
0
0
1.4
<I 0
0
e
a. 1.2 0
(\\)
o 0
'0
~ 1.0
::J
:g 0.8
~
a.
'0 0.6
~ 0.4
:c
~
a.. 0.2
Knowing the fact that the model is intended for two-phase flow in large
diameter pipes, one could assume that the agreement with the experi-
mental data of the microchannels with larger hydraulic diameter is bet-
92 5. Pressure drop
ter than for smaller diameters. Figure 5.7 shows a large dispersion in the
pressure drop for all the channel diameters.
If the data is varied by the ratio of the superficial gas to liquid velocities,
it can be seen in Fig. 5.8 that the agreement with the model is better for
flow conditions using smaller ratios jG/jL. This trend can be noticed for
all pressure drop models originally designed for large diameter channels
or pipes. Those models were mainly developed by the oil industries for
predicting the pressure drop in pipelines. Consequently low gas flow rates
were considered.
2.0 -1-_--L-_---L._----''--_..L..-_-'------,,---'-_--I-_ __+_
ffi 1.8
0
~
11 1.6
g 1.4 0
0
0
Q.
<l 0 e:J
eQ. 1.2 0
o
'0
e 1.0
:::J Oh = 120 /lm
l2 0.6 o = 133.3 /lm
e
Q. ~ ~ A
Oh
Oh = 150 /lm ~
'0 0.6 'J Oh = 155.6 /lm
~ 0.4
Oh =160 /lm
'0
e
Q. 0.2
- - - +/- 20%
0.0 ...joooI::=----.----r---,r---.....---,--~-~-_+_
With the variation of the parameter C, Mishima and Hibiki (1996) and
Lee and Lee (2001) tried to adapt the Lockhart-Martinelli model to
the special conditions of microsystem two-phase flows. Both methods
use empirical parameters (Fig. 5.9). The two models show better agree-
ment with the present experimental data. However, still a large scatter
is present.
5.2 Comparison with correlations 93
la 1.8 ...
a.
li 1.6
"' ...
a.~
...
1.4
<J
a.
l::! 1.2 ...
"C o
...
(\)
~
1.0
to fit the models to the present data does not lead to a significant further
improvement of the pressure drop predictions. The disadvantage that the
obtained set of parameters is valid only for a specific channel geometry
or flow condition outweighs the advantage of the slightly better results.
According to the constant factor C = 5 proposed by Chisholm (1967)
for strictly laminar flow, Kawahara et al. (2002) used the experimen-
tally determined average of the factor C as basis for their prediction
model. As the data is extracted from the experimental results, using
C = 0.24 yielded good agreement of the predicted and the measured
pressure drops. However, the consistency of our experimental data with
the Kawahara prediction is less accurate than with the data obtained
using the correlation by Mishima and Hibiki (1996).
Finally, a comparison of the two-phase frictional pressure gradient data
with the predictions of the Lockhart-Martinelli correlation using different
94 5. Pressure drop
0.8 +-_....L-_...J-_---L..._----L.._--L_--L_.,......l._----::>I-
~
o
! 0.7
"
a.
I 0.6
<J
a. 0.5
o...
"C
o
e 0.4
:::I
:z
...a.
Cl) 0.3
~ 0.2
U i
'6
e
ll.
0.1
0.0 -F----.----,----.------r------,----.----r-__;__
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Measured pressure drop ~Pexp [bar]
"'-l
- C 5 (Chisholm)
- - - C 0.0009 (Mishima) A
..
e
,~ 100
C .0.0113 (Lee)
_. - C .0.24 (Kawahara)
0.
'';::
o experimental data
"3
E
iii
g 10
13
:E
Q)
Ul
RI
.
6
~
0.1 +---r--r'''T''T'"","r--r-'''T""'I''''''''I''nT-T'''''''I''""T'''IrTTI'IT"""'''''''""''T'"T'T"l".,.,f-
01 1 10 100 1000
Lockhart-Martinelli parameter X
- C 5 (Chisholm)
- _. C 0.0009 (Mishima)
B
----- C .. 0.0113 (Lee)
- . - c 0.24 (Kawahara)
o experimental data
.,4--.....-.........
7 ,
----"""T'"---.---.----.r---r-.....-..--,......j.
6 7 8 !I
10
Lockhart-Martinelli parameter X
The pressure drop of the gas core flow is calculated using the Blasius
equation. The main equation of this model is
1"_(-3+
~ - El E
_3)-1/3 (5.27)
2
with e,3 = exp [ ( 0.1675 - 0.055ln (~) ) In( I]i) - 0.67], (5.29)
where Kid is the surface roughness of the channel, iJ a ratio of the ft.uid
densities and viscosities
(5.30)
with (5.31)
with (5.32)
5.2 Comparison with correlations 97
As seen in Eqs. 5.28 - 5.32, the model suggested by Chawla (1968) uses
several empirical paranleters for the definition of the two-phase flow pa-
rameter ,. With this large number of influencing variables, a parameter
variation can provide acceptable agreement of the calculated pressure
drop with the experimental data. Unfortunately, a single set of parame-
ters is valid only for a specific condition (channel geometry, flow condi-
tions, and fluid properties).
In Fig. 5.11, the present experimental data are compared with the pres-
sure drop model by Chawla (1968). The main equation for the two-phase
flow parameter (is used as presented in Eq. 5.27. A remarkable spread of
the calculated values can be found. The correlation shows no tendency in
over- or underestimating the actual pressure drop. The mean deviation
of the model data is /:::,.p = 0.52 bar with a large standard deviation of
a = 1.31 bar. Thus, the original pressure drop model by Chawla (1968)
is not capable of predicting the pressure drop in microchannels.
In the Eqs. 5.28 - 5.30 Chawla introduced several empirical parameters.
Using the lea..~t squares method, the best fit for our channel geometries
and flow conditions was determined (Tab. 5.2). With this new parameter
set, the mean error of the pressure drop prediction can be minimized to
approximately 30% (Fig. 5.12).
An important factor of this prediction method is the surface roughness
K / d. The influence of K / d on the pressure drop prediction quality is
presented in Tab. 5.3. The mean error of the best parameter fit increases
with decreasing surface roughness.
The original pressure drop prediction method by Chawla (1968) is not
suitable for microfluidic applications. With the help of a parameter vari-
ation for the empirical values introduced, the mean prediction error can
be minimized, but remaining faults on the order of 30% and no distin-
guishable basic over- or underestimation of the present pressure drop
show that this method is inapplicable for small-scale channels.
98 5. Pressure drop
3.0
0
';:'
2.8
1-- - +/0 20 %1
.[ 2.6
0
1i 2.4
~
Cl.
2.2
0
<l 2.0 0
a.
e
u
1.6
1.8
0
0
~ 1.4 0
:;, 0
VI 0
VI 1.2
~
Cl. 1.0
uQ) 0.8
1)
0.6
'5
~ 0.4
a. 0.2
0.0 -+-~~~~Dap.l~--=---,-_~---,....,._-+
0.0 0.1 0.2 0.3 0.4 0.5 0.8 0.7 0.8
Measured pressure drop ~Pexp [bar]
0.8
~
J!1li
0.7
0.
~ 0.6
<.1
0. 0.5
e
"l!? 0.4 ---
0
0
:J
~
...
Cl>
0.
0.3
"~ 0.2
0
'0
l!? 0.1
a..
0.0 ~--J;L=;::-----r----r----.--.------r----r---+
5.2.4.1 l3asics
( ~p)
~l TP -
- <1>2
L
(~p)
~l L
(5.33)
A -- (1 - XG
.)2 + XG
2 {}L ffric,G
(5.35)
{}G frric,L
The single-phase friction factors are calculated using Eq. 5.10. The two-
phase Froude and Weber numbers are defined as
X
'2
x'2 D h
FrTP = D 2 WeTP = 2' (5.36)
9 h {}m a {}m
For {Jm, the homogeneous densities {JH (Eqs. 5.7 - 5.9) or the liquid
density {JL can be used.
Even for large diameter pipes, different values for the parameters B, 5,
t, U, v, w, p, and r can be found (Friedel, 1978; Triplett et al., 1999a).
This indicates that this pressure drop correlation is able to predict the
pressure drop for a single channel geometry and flow condition with high
accuracy, but the same parameters can yield wrong results for any other
fluidic system.
5.2 Comparison with correlations 101
region of validity
Gas mass fraction Xc < 1 [-]
Total mass flux rh 5 -10330 [kg/m 2 s]
Pressure p 1-171 [bar]
Density ratio eL/ec 6 - 1194 [-]
Viscosity ratio 'r/L/'r/G 6 - 5991 [-]
Hydraulic diameter Dh 1 - 257.4 [mm]
Surface tension a 15 - 80.10- 3 [N/m]
The pressure drop model by Friedel (1978) is valid for the flow, geometry
and fluid conditions found in Tab. 5.4.
Obviously, the agreement of the predicted pressure drop with the exper-
imentally determined pressure gradient is not expected to be good, as
the diameters of the present microchannels are smaller by a factor of
10 than the required ones.
f'V
a.~
o
1.4 o
<1 o
ea. 1.2
0 0 o
"0
~ 1.0 CO::> 0
:;, 0
I/) 0
I/) 0.8 0 0
0 0
~ 0
13
a. 0
"0 0.6 00
~
'6 0.4
~
0.. 0.2
0.0 ....jooo=::;;;;.,.-r-----.-----,.-----....----.-----.--~--+
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Measured pressure drop 6p"xp [bar]
Muller-Steinhagen and Heck (1986) made the observation that for a gas-
liquid mixtures with a volumetric gas content E = 0.5, the friction pres-
sure drop is similar to the pressure drop of a single-phase gas flow (E = 1),
which features the same mass flux as the two-phase flow.
2
t:J..p) rh (5.37)
A = ( t:J..l L = !fric,L 2 flL Dh
2
t:J..p) rh (5.38)
B = ( t:J..l G = frric,G 2 {!G Dh
5.2 Comparison with correlations 103
't:"
as
So 0.7
CL
i 0.6
<l
CL 0.5
...o
"0
~ 0.4
:::I
11)
~ 0.3
CL
~ 0.2
:a~ 0.1
a.
0.0 ~---r-----r----.-----.--r---~--r---+
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Measured pressure drop 6Pex:p [bar]
Again, the friction factors are calculated using Eq. 5.10. From Eqs. 5.37,
5.38, and
; [ 3.5 o
1-- -+/.20%1
o 0
I
0.
3.0
<I 00 o
c.. 2.5 o 0
e
-0
~ 2.0 o
~
o
UJ
UJ
~ 1.5 o o
0. o
o
~ 1.0 o
1:)
ij
~ 0.5
a.
0.0 -+-~~~:;~-.---___.-__r-__r-___,.-_I_
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Measured pressure drop .1.Pexp [bar]
The pressure drop for intermittent flow is significantly higher than for
annular flow patterns (at moderate flow velocities). Increasing the su-
perficial gas velocities, the pressure drop features a local maximum just
before changing the flow pattern from intermittent to annular flow. For
larger hydraulic channel diameters, the pressure drop at low superficial
liquid velocities is proven to be nearly independent of the gas flow rate.
As in macro-scale applications, the major contribution to the total pres-
sure drop in microchannels is shown to be the friction pressure drop. For
the considered hydraulic channel diameters, the ratio of friction pressure
drop to total pressure drop is determined to be in the range of 85 - 99%.
A comparison with the various friction pressure drop models presented
in the literature can be made. The prediction methods originally devel-
oped for large scale pipe flow (homogeneous pressure drop model, and
correlations by Lockhart and Martinelli (1949); Chawla (1968); Friedel
(1978); Muller-Steinhagen and Heck (1986)) substantially overestimate
the pressure drop for channel sizes below 1 mm. Extrapolating the pres-
sure drop from the larger channel sizes seems to be a wrong approach.
In most cases, the pressure drop prediction for small ratios of superficial
gas to superficial liquid velocities (je/ iL, high volumetric liquid con-
tent flows) are significantly more accurate than those for high gas void
fractions. This tendency is not surprising, thinking of the origin of the
prediction models (flow of oil in large-scale pipelines, low gas content).
In contrast to the velocity ratio, the hydraulic diameter does not have a
major influence on the prediction quality. By means of adaptations to the
models for microfluidic applications (Mishima and Hibiki, 1996; Lee and
Lee, 2001), the prediction quality can be slightly improved. A parameter
variation for the models using empirical parameters, yields better agree-
ment with the experimental data, loosing however the general validity
of the method.
The pressure drop, and therefore the dimensioning of the pumping sys-
tem of any microchemical device (lab-on-a-chip) is often the crucial com-
ponent. As seen for the prediction of the existing flow pattern, an experi-
mental validation of the occurring pressure drop is inevitable, designing a
novel microreactor design, as the known pressure drop prediction models
are not able to predict the pressure drop accurately.
106
Chapter 6
Void fraction
1.0-r-...I-_...I-_-L--.....L._--&._-J;===::;1-
" 0.7
C'"
1Il
C 0.6
8
III 0.5
III
0'1
o 0.4
'1:
Q)
E 0.3
;:)
0.9 B
I 0.8
'"
C " 0.7
AI
l5 0.6
o
IQ 0.5
0'1
.g 0.4
~;:)
0.3
(5 -0- I ia - 0.44 mls
> 0.2 -0- I JG - 0.67 mla
--l!r-- I JG 0.89 m/a
0.1
-SJ- I JG - 2.22 mla
0.0 -l-r--r---r--......-r-...........,.-::.:::;::;;:::;:::::;::;;:::;:::::;::;~-r--;r-r--r--;r-r--.-t-
0.00 0.04 0.08 0.12 0.16 0.20 0.24
Superficial liquid velocity i L [m/s]
the larger channel diameter, all the flow regimes are either in the tran-
110 6. Void fraction
100
90
A
eo
~ 70
","
c:
.2 eo
U
,g 50
"0
'g 40
13
..2
30
-0- iG 0.74 mlo. ",.0.67
20 ooQ- i G 1.11 mlo. ",0.69
...... iG 1.49 mlo, ",.0.72
10 ...... k. 3.70 mlo, ",0,$7
0
0.0 0.1 0.2 0.3 0.4 0.5 0,6 0.7 0.8 0.9 1.0
normalized channel position xlb [-]
100
90
B
90
~ 70
<:>
'" eo
<:
~ 50
Jg
"0
'0 40
;>
-<>-- jlJ 0.44 m/. s~ 0.89
~
:11:
30 &"
0
0,0 0.1 0.2 0,3 0.4 0,5 0.6 0.7 0.8 0.9 1.0
normalized channel position xlb [-]
Increasing the volumetric liquid flow rate to 1IL = 0.3 ml/min (yielding
6.2 Comparison with correlations 111
superficial liquid velocities jL,150 = 0.2 m/s, and jL,187.5 = 0.13 m/s,
respectively), the mean void fraction decreases (Fig. 6.3). In contrast to
the low superficial velocity in Fig. 6.2, the variation in the superficial
gas velocity yields differences in the local void fraction distribution due
to different plug lengths. For a hydraulic channel diameter of D h =
187.5 j.Lm, the same phenomena a9 for lower superficial liquid velocities
(jL = 0.02 m/s, Fig. 6.2B) can be observed. The film thickness at the
channel sidewalls increases with increasing jc.
Xc . (2L
cC horn = . (.) = E c (6.3)
, Xc {2L + 1- Xc . {2L
100 -j---JL-----l-_...L-----l_-L_..L-----L_--l-_.L--+
80
~ 70
<:l
<.>
<: 80
.~
g 50
"0
"g 40
~
30
-0-- k.. 0.74 mlS. EO" 0.45
20 .-0- lo 1.11 mlS. EO" 0.5&
...... Jo .. 1.48 mls, Eo .. 0.57
10 ..... io " 3.70 mls, Eo" 0.63
0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.6 0.9 1.0
normalized channel position x/b H
100
90
80
~ 70
<:l
'"
<: 80
j 60
~ 40
o....!IjpI~r--..,__-r-~___._____y-.---~~
O~ O~ ~2 O~ ~4 ~5 ~6 ~7 O~ o. ,~
normalized channel position x/b H
fraction based only on the volume transport fraction eG, neglecting any
mass flow data.
For eG < 0.9 this method is suitable for large diameter pipes. For large
6.2 Comparison with correlations 113
1.0
.....
...!..
0.9
00
.c
6 0.8 0
ll)
0.1
c::: 000
0
13 0.6 ,-
'0 0.5
"g 0.4
'0
Cb
0.3
U
'6
I!? 0.2
a..
0.1
1---+'-20%1
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.1 0.8 0.9 1.0
Measured void fraction Eexp [-]
volume transport fractions Eo > 0.9, Massena (1960) proposed the fol-
lowing extension
Figure 6.5 clarifies that this correlation mainly overestimates the void
fraction for microf:l.uidic applications. All the predicted volumetric gas
contents for the channel with the smaller hydraulic diameter are larger
than the mea..",ured data. Especially for high superficial gas velocities,
this trend is confirmed. For D h = 150 J.lm the predictions for both fluids,
water and ethanol, do not differ significantly.
The experimental data correlate remarkably well with the predictions
using the microchannel with D h = 187.5 J.lm and ethanol as working
liquid.
114 6. Void fraction
...... 0.9
..!.. e "" 0
" c!1 ~ ..t.~~ Do 8 &@> "
J::
0
o C e elt:J d5'" 0 0 d; .........
De J.
.-- .... "'" c:..00 ...... 0'"
c:
:fi 0.6
eo.
~ ~... "0,,'"
0.5
'0 o Oh = 150 Ilm. ethanol
'g 0.4
Oh = 150 Ilm. water
-g 0.3 o Oh" 187.5 Ilm. ethanol
U e D." 187.5Ilm. water
'5 0.2
~ " " ,," ---+1-20%
a. 0.1
0.0 *-.,........--,-----.---.---.---r----r-.....,...-~_+
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Measured void fraction &exp [-]
(6.6)
where
defined as
(6.8)
The comparison between the experimental data and the predictions using
Eqs. 6.6 - 6.8 does show a large scatter. All the predicted void fractions
are in the range of 0.4 to 0.7, whereas the experimental data show gas
contents up to ~ 0.9. The root mean square of the deviation values is
0" = 0.14 (rv 24%). In contrast to the above illustrated correlation by
Armand and Treschew (1946), the current prediction method consid-
ers not only the gas volume transport fraction, Sa, and the gas mass
transport fraction, xa, but also the fluids' properties. The empirical pa-
rameters of this correlation model are based on water - air experiments
for large diameter pipes. The comparison with the experimental data of
our microchannels (ratio of hydraulic diameters Dh,expl Dh,model :=::::: 100)
shows satisfying agreement using water as working fluid with a stan-
dard deviation of 0" = 0.07 (13%). The agreement of the experimen-
tal data using ethanol with the prediction, on the other hand, is worse
(0" = 0.22, 33%). Especially the void fractions for the configuration
(D h = 187.5 J-Lm I ethanol), yielding the best results using the model by
Armand and Treschew (1946), are not useful for further considerations
(0" = 0.32, 40%).
0.0 -F--,---,---r--r---,---,-----,----,---r--+
0.0 0.1 0.2 0.3 004 0.5 0.6 0.7 0.8 0.9 1.0
Measured void fraction &exp [-]
1.0 +--'---'----'----'----1..---'---'""'---'-.....,.....--'---"'71-
.....
...!,.. 0.9 o D. =150 "m, ethanol
Oh =150 "m, water
~I 0.8 o Oh'" 187.5 "m, ethanol
I
(,)
0.7
O. = 187.5 "m, water
- - - + 1- 20%
,,
CJ:I
c: 0.6
o
13 0.5
"0 004
'0
>
"0 0.3
~
'6 0.2
~
a.. 0.1
0.0 -F-...,...--,---,---r----r----r---,-----,----,-__+_
0.0 0.1 0.2 0.3 0.4 0.5 0.8 0.7 0.8 0.9 1.0
Measured void fraction &exp [-]
"7" 0.9
ca
~ 0.8
~ 0.7 o
.0.
w
5 0.6
ts
~ 0.5
'0
'0 0.4 o On = 150 ",rn, ethanol
> =
On 150 ",rn, water
'0 0.3 "" o =
On 187.5 I-lrn, ethanol
~ =
On 187.5 ",rn, water
'6 0.2
~ ---+/-20%
a. 0.1
0.0 --I"'=--~---r----.---.---r------,r----r---r---.----+-
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Measured void fraction &exp [-]
1
EG,Tripplet = ( ) q ( ) r
(6.10)
Ae~X)p ~ ~ +1
where the empirical parameters A, P, q, and r are supposed to be set to
the values in Tab. 6.1 (version 1).
The correlation by Triplett et al. (1999a) is based on experimental data
using air and water in circular microchannels with 1.1 and 1.45 mm
inner diameter and in microchannels with semi-triangular (triangular
with one corner smoothed) cross-sections with hydraulic diameters 1.09
and 1.49 mm. The superficial gas and liquid velocities were of the same
magnitude as in the present experiments. The four empirical parameters
6.2 Comparison with correlations 119
Version 1 (Triplett) 2 3 4
A 0.28 0.32 0.01 0.09
p 0.64 0.39 0.42 0.50
q 0.36 0.37 0.31 0.01
r 0.07 0.41 0.90 0.01
of the fluids' viscosities. In the present study, the impact of the fluid
superficial velocities is increased in favor of the viscosity difference. A
possible explanation is the fact, that 'friplett et al. (1999a) used a much
larger range of superficial velocities (Triplctt: 0.02 m/s < jG < 80 m/s
and 0.02 m/s < jL < 8 m/si present study: 0.44 m/s < jG < 3.70 m/s
and 0.02 m/s < jL < 0.37 m/s). No similarity can be found for the pa-
rameter settings for ethanol experiments only, and the settings for all
the experimental data.
1.0
0.9
..
~
...!..
t 0.8
.0. ~~
~.
0
~
...
~
0 0
'1:
.... ~~ .~)-'i
.D'0
09~
~ 0.7
~~c. o~~~o 0
l::
0 0.8
0 ~
~
0
n .g~. ~..,
~ ~ ~
~
0
~
0
0
~ 0.5 ~
~
~
<0
0
'0
.~ 0.4
~
~
~
0
'0 0 Oh " 150 ",rn, ethanol
~
0.3 Oh =150 ",rn. waler
'is 0.2
0 Oh =187.5 ",rn, ethanol
~
a.
0.1
Oh =187.5 ",rn, water
0.0 ~r----r-....------,r-----r--""::::::;~::;::=::;::::=-+
0,0 0.1 0.2 0.3 0.4 0.5 0.8 0.7 0.8 0.9 1.0
Measured void fraction Eexp [-]
A further void fraction model for the estimation of the volumetric gas
content in two-phase flow in microchannels is presented by Coleman and
6.2 Comparison with correlations 121
, ,,
,,
...!.. 0.9
~;1:1 0.8
. ...
. , ~
,,
,
10.7
'C
l-
t.)
C
o
0.6 .
. .'.. ....,.,;
,;
,'
;11'.
.,. .,'
,'.
is 0.5
,.... !!I...... "
't:I 0.4 , III '
'0
>
't:I 0.3
(\) Oh = 150 J.lm
n 0.2 Oh" 187.5 J.lm
'0
!
a. 0.1 ~ ~ - +/-20%
0.0 -fC--..------r---r--r---r-----,r------r---r--~__+_
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Measured void fraction &exp [-I
Krause (2004). This correlation, based on Huq (1998), does not make
use of any empirical parameter.
2 (1 - XC)2
cc, Col = 1- ---------'------'--------.",..0.""'"5. (6.11)
1 - 2 Xc + [1 + 4 Xc (1 - xc) (~ - 1) ]
The model shows agreement in the prediction of the void fraction within
30% (0- = 0.18). The void fraction is generally overestimated. Good
agreement with the experimentally determined gas contents are detected
only for the ethanol-nitrogen configuration in the microchannel with
D h = 187.5 J.1m (0- = 0.08, 10%).
::I: 0.9
~ 0.8 l: ~ 0
~ o l:
(J 0.7
lol
.0.
C
o 0.6 De
13
~ 0.5
"0
.~ 0.4
o Oh = 150 llm, ethanol
"0 0.3 Oh =150 llm, water
~ o Oh =187.5 llm. ethanol
'0 0.2 Oh =187.5 llm, water
~
a. 0.1 +/-20%
0.0 ~--r------'----r"--r--......----,.--------r---r--.,...--+
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Measured void fraction &BJ(p [-]
in round microchannels.
0.03E~l
6'G,Ka = 05' (6.12)
1- 0.97Ed
::c 0.9
o Oh =150 llrn, ethanol
Oh =150 llrn, water , " "
l!
o =187.5 llrn, ethanol
.
.l! 0.8 Oh
=187.5 "rn, water
!
Oh
,,'DiI
~
a
13
'0
0.7
0.6
0.5
- - - +/-20%
6
...
......"
"
"
0
"
0"
0
0
00
0
0
'0 0.4
o ". c. 0 0 0
>
'0 0.3 o 0
Q)
1:) .::J." 0 0
'5 0.2 .~: iOO
~ o
a.. 0.1
0.0 --toC--....---......---r-----r----,--.,...--....---r------.-~
QO Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Q9 1.0
Measured void fraction &exp [-I
The flow pattern map by Kawahara et al. (2002) (Fig. 4.7) clarifies that
in the experiments only intermittent flow regimes were detected, where
in our rectangular microchannels annular flow is observed. The annular
flow pattern features higher volumetric gas contents than intermittent
flow. The flow pattern (and therefore also the agreement of the void
fraction with the correlation by Kawahara et al. (2002)) seems to be more
dependent on the channel cross-section than on the hydraulic diameter.
124 6. Void fraction
1.0
0.9 - Homogeneous
..... Armand
......
.2... 8erizawa
0.8 _ Kawahara
Cl
-.!l
lA)
c
0.7
0.6
_ ... Coleman
- Triplett (original)
Triplett (Water)
o Experiment
8
fI) 0.5
CO
0)
-
0
.;:: 0.4
(J) ,
E 0.3
,
::::I
(5 ,
, ,
0.2 o
> .. ~.
0.1 ----_.--
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 O.~ 0.8 0.9 1.0
Volumetric transport fraction I; G [-]
The trend of overestimating the void fraction can be seen for the ho-
mogeneous model, and the theories by Armand and Treschew (1946);
Serizawa et al. (2002); Kawahara et al. (2002); Coleman and Garimella
(1999) and Triplett et al. (1999a). Especially for high volumetric gas
6.3 Void fraction measurements using XTM 125
6.3 Void fraction measurements uSIng
XTM
X-ray Tomographic Microscopy (XTM) is not only a suitable technique
to visualize the flow regime in microchannels (see chap. 4.3), but can
be used also to investigate the fluid distribution and therefore the void
fraction. However one has to consider that results obtained by XTM
measurements are temporally averaged.
With the help of standard image processing steps (image threshold, edge
detection, area sizing) the area taken by the liquid and the one taken
by the gas can be determined in the 2-dimensional cross-sectional XTM-
images of the two-phase flow (Fig. 6.14). The volumetric gas content is
then defined as
Table 6.2 shows the volumetric gas content as identified by X-ray To-
mographic Microscopy (XTM), and as determined using the method de-
scribed in chap. 6.1. The XTM void fractions appear to be significantly
higher than the experimental results of the optically derived void frac-
tion. The best results are obtained using ethanol as working fluid. The
hydraulic channel diameter does not seem to have much impact on the
consistency of the two measurement methods. Major differpnces between
ethanol and water are the surface tension (creth/crwater = 0.31), and less
significantly the density ({2eth/ (2water = 0.79), whereas the viscosity is
of the same magnitude (rJeth/rJwatcr = 1.19). Known from the analy-
6.4 Summary and conclusions 127
HiH of the flow pattern experiments, the lower surface tension of ethanol
causes temporally and spatially stable fluid interfaces in the annular flow
regime. This leads to sharper boundary lines in the XTM reconstruction
and therefore, thc reconstructed area assigned to the gas is more ac-
curate than for wavy surfaces. Fig. 6.15 illustrates this phenomena for
th(~ two configurations for ethanol, D h = 187.5 Mm, jL 0.04 rn/s,
.hi = 5.33 m/s (top), and for water, D h = 200 pm, jL = 0.04 m/s,
jc; = 2.97 m/s (bottom).
Table 6.2: Comparison of the obtained values for the void fraction
using the two different methods XTM and optical determination.
In contrast to the flow regime and the pressurP drop, which arc both
widely investigated for large scale applications, comparatively little in-
formation on the prediction of the void fraction can be found in the
literature. Thinking of the importance of knowing the void fraction, de-
termining the reHidence time or its infi1Hmcf~ on mass and heat transfer
processes, accurate measurements of the void fraction are desirable. We
present experimental results of void fraction measurements using Laser-
128 6. Void fraction
Figure 6.15: Explanation for the large differences in the void frac-
tion measurements between XTM and the optical measurements.
The more the phase interface is temporally and spatially stable, the
more distinct is the reproduced phase boundary in the XTM recon-
struction, and therefore a more accuntte void fraction is defined.
Chapter 7
Liquid velocity
measurements
At first sight, the velocities of all the analyzed flow conditions show the
parabolic distribution, typical for strictly laminar flows. More interesting
is the internal recirculation motion, that is the relative velocities
which are responsible for the mass transport from the bulk towards the
phase boundaries and the channel walls.
With the help of calculated streamlines of the relative velocities in the
liquid plug, the determination of a mass transfer over the center line of
the microchannel can be analyzed. The transverse fluid exchange can be
crucial for mixing, if a fluid component is injected into complex microre-
actor designs from one side of the channel only. This fluid exchange is
favored by stronger internal recirculation motion. The latter is, amongst
others, represented by high wall normal velocity components and the
strength of the occurring vortices.
w=\lxl1. (7.2)
Unfortunately, vorticity not only identifies vortex cores but also any
shearing motion present in the flow. These shearing motions can render
the task of computing reliable vortex statistics virtually impossible.
New methods of extracting underlying structures from velocity fields
which involve critical-point analysis of the local velocity gradient tensor
132 7. Liquid velocity measurements
and its corresponding eigenvalues are proposed (Chong et al., 1990; Dall-
man et al., 1991; Zhou et al., 1999, 1996). In three dimensions, the local
velocity gradient tensor will have one real eigenvalue (A r ) and a pair of
complex conjugate eigenvalues (A cr iAci), when the discriminant of its
characteristic equation is positive. When this is true, the particle tra-
jectories about the eigenvector corresponding to Ar exhibit a swirling,
spiral motion (Chong et al., 1990). A~/ represents the period required
for a particle to swirl once about the Ar-axis. If the flow is pure shear
flow, the particle orbits are infinitely-long ellipses and the orbit period
is also infinite, corresponding to Aci = O. Thus, Aci > 0 corresponds to
shorter, more circular ellipses, i.e. eddies. Zhou et al. (1999) show, that
the strength of any local swirling motion is quantified by Aci, which they
define as the swirling strength of the vortex. Vortex identification based
on swirling strength does not reveal regions which contain significant
vorticity but do not have of any local swirling motion (i.e. shear layers).
Since PIV fields are usually two-dimensional, the full local velocity gra-
dient tensor cannot be formed. However, an equivalent two-dimensional
form of this tensor can be computed in the plane in which the PIV data
are recorded; namely,
(7.3)
where (Xl, X2) and (UI' U2) are the streamwise and wall-normal direc-
tions and velocities, respectively. In this case, D 2-D will either have two
real eigenvalues or a pair of complex conjugate eigenvalues. Therefore,
vortices are easily identified by plotting iso-regions of Aci > O.
Fig. 7.1 shows the sum of all considered raw digital PIV images (A) of
a two-phase flow in a Dh = 150 jjm channel, and the according rela-
tive velocity distribution (B). The streamwise (C) and wall normal (D)
velocity component contour plot indicates two stagnation points on the
channel center line close to the bubbles' trailing and leading end. Note
that in all the velocity plots, the bulk velocity, Vhulk, was subtracted.
The corresponding swirling strength contour plot is shown in (E), and
7.1 PIV analysis for the liquid recirculation motion in intermittent flow 133
the contours of the vorticity in (F). Obviously the swirling strength in-
dicates the position of the two counter-rotating vortices whereas the
vorticity is dominated by the shearing motions.
For the analysis of the recirculation motion, for the determination of the
mass transfer over the center line, and the vortices (strength and posi-
tions) in the upcoming section, we therefore concentrate on the swirling
strength instead of the vorticity.
Flow 50llm
----- OAm/S
A_ _
c o
E F
[1/s2]
Since many low Reynolds number flows in microfluidic systems are lami-
nar, it is not necessary to determine instantaneous velocity information.
It is sufficient to measure only time-averaged or phase-averaged (peri-
odic flows) velocity fields. Average velocity fields can be obtained by
first measuring instantaneous velocities and then averaging them in ei-
ther space or time. A alternative method of estimating average velocity
fields directly from the correlation function is the average correlation
method (Meinhart et al., 2000).
Estimation of velocity-vector fields using PIV involves three primary
136 7. Liquid velocity mea..'iurements
steps:
+
N
1
AN
11
8N I --.. I R ANBN I --.. /
/
Image Image A Image 8 Correlation
Sequence R AB
Peak
Detection
B
1
I A1
11
81
I --.. I RA,B,
+
I
2
I A2
11
82
I --.. 1
R A2B2
+
I
3
I A3
11
83
I
--.. 1
R A3B3
I
+
+
N
1
AN
11 8N
1--"1 R ANBN
1
1
RAB
1
--.. /
Figure 7.3: Diagrams depicting the different ways in which the
average velocity can be estimated: (A) average velocity method,
(B) average correlation method.
.+
+
Figure 7.4: Instantaneous cross correlation functions that are aver-
aged together to produce an average correlation function (Meinhart
et al., 2(00). The average correlation function has a much higher
signal-to-noise ratio than the instantaneous correlation functions.
. - jL' D h = 7.45, .. _ ja D h
R,CL - R,Ca - = 4.31 (7.4)
VL va
Figure 7.6A shows the velocity distribution inside the liquid water plug
of the intermittent flow. The average streamwiHe liquid velocity (bulk
velocity) is 0.36 m/so To investigate the recirculation motion and the
mass transport towards the phase interphases, the relative velocities are
plotted in Fig. 7.6B (vrel = Vabs - Vbulk), where Vabs are the measured
absolute velocities and Vbulk is the mean velocity. Two counter-rotating
vortices are observable. In the center of the microchannel, low stream-
wise velocities are present (Fig. 7.6C), and hardly any mass transfer
over the center line is detectable. The velocity distribution is symmetric.
Two distinct vortex cores are illustrated using the swirling strength (Fig.
7.6E).
Ethanol has a similar viscosity, but a much lower surface tension than
water (see Tab. a.l). Using the same flow conditions as in the water case
(jL = 0.032 m/s and ja = 0.317 m/s), the liquid Reynolds number drops
to RCL = 4.94. The average streamwise velocity inside the plug increases
to Vbulk = 0.38 m/so The ethanol segment length is slightly smaller than
the water plug. The absolute velocity distribution (Fig. 7.7A) shows sim-
ilar characteristics as the one using water, whereas the relative velocities
(Fig. 7.7B) behave remarkably different. Comparatively large relative ve-
locities are measured throughout the liquid plug and especially along the
center line of the channel. The typical vortices are more distinctive, the
value of the swirling strength is slightly increased compared to the pre-
vious experiment using water as liquid phase (Aci,water,max = 1.93 l/s 2 ,
2
Aci,cth,max = 2.14 l/s ). The area of swirling in ethanol is much more
spatially concentrated, showing large gradients. No stagnation points in
the liquid phase close to the leading and the trailing end of the gas phase,
as seen in the water case, are found.
140 7. Liquid velocity measurements
0.55
[m/s]
.............
-~_
J _ ~_ _
..
...
-~"
-~., 1,1"'
I""'-~~"""""""""'''''''''''''
- - ...
-',,,. .. "''''
~ ". .. W'~
-_ -
--_ ...
...
o - , ... ,r ,r '"
.:
~~,
~~ ~ ~ ~:',
..... ~_~~
~ - ~ ~
r. _.".". ':.
..
~~~~
r_...........
... ..
. .....
~ ,
0.25 0 0.07
[m/s] [m/s]
-0.25 -0.07
E 2.0 F
Figure 7.6: Distribution of the absolute (A) and relative (B) ve-
locities, inside a water plug. The contours in (A) denote the veloc-
ity magnitude. The values of the streamwise (C) and wall normal
(D) relative velocities as well as the swirling strength (E) and the
streamlines (F) are shown.
. -, -.....-
-~
...
~....--
~~....."...-
w::.~:::::::=-::::
_
~r~r
-.~-
r.
::-
~,
,-_ ....
.. -
_
-_.-
M ....... . . - . . - , . . . . . . . _
...
[m/si
... _
-~--_
~._~~
----.-
-----.
_-- .....
....... - .....
~
~
" .....
. ............
,----~
~.~--~
. . . . __ r ~
~ ~-
~--
-
, ... ...............
' ~~ __
o ~~rL~~~
c 0.25 0 0.07
[m/si [m/si
~0.25 -0.07
E 2.0 F
Figure 7.7: Distribution of the absolute (A) and relative (B) ve-
locities, inside an ethanol plug. The contours in (A) denote the ve-
locity magnitude. The values of the streamwise (C) and wall normal
(D) relative velocities as well as the swirling strength (E) and the
streamlines (F) are shown.
and a value for the interfacial tension against nitrogen similar to water
(O"glyc = 70.50 mN/m, O"water = 71.99 mN/m). These properties yield a
liquid Reynolds number of ReL = 6.64. The average streamwise veloc-
ity of the liquid plug is lower than in the two cases presented above,
VtHllk = 0.22 m/so This can also be seen in the distribution of the abso-
lute velocity distribution in Fig. 7.8A. However, the distribution of the
relative velocities resembles the ones found using water as liquid. Low
relative velocities and two stagnation points at the center line are signif-
icant. The distribution of the swirling strength as well as the streamlines
show similar behavior as water. In contrast to the previous experiment
(ethanol - nitrogen), the streamlines and the relative vector plots of the
two configurations using watf~r and glycerol, indicate vortices of nearly
triangular shape, symmetrical not only with respect to the center line of
the microchannel, but also with respect to the wall normal liquid plug
center.
A similar behavior can be found for a microchannel with a smaller hy-
draulic diameter of D h = 187.5 J-Lm. Using the same volumetric flow
rates CVL = 0.1 ml/min, VG = 1.0 ml/min), the superficial velocities
are increased to .iL = 0.044 m/s and ,iG = 0.44 m/s, respectively.
For this geometry, the liquid Reynolds numbers are ReL,water = 9.15,
RCL,eth = 6.06 and RCL,glyc = 5.00. The gas phase Reynolds number is
RCG = 5.:39 for all the analyzed fluid combinations. The average stream-
wise velocities are Vbulk = 0.46 m/s (water), 0.64 m/s (ethanol), and
0.43 m/s (glycerol), respectively.
The highest relative velocities were again measured in the ethanol case
(Fig. 7.9B). The two conditions "water - nitrogen" (Fig. 7.9A) and "glyc-
erol - nitrogen" (Fig. 7.9C) show similar characteristics, including low
velocities at the center line and two stagnation points in the vicinity of
the gas bubbles leading and trailing end. The maximum swirling strength
is measured for the ethanol - nitrogen flow. As a result of the lower bulk
velocity in the glycerol slug, an increased slug length is detected (con-
servation of mass).
In all the presented configurations, the gas bubble trailing end curvature
radius is found to be larger than the leading end curvature radius. For
the ethanol plug however, the leading edge is significantly more curved
than in the two other configurations. This leach; to an asymmetry of the
7.2 Liquid velocities in straight rectangular microchannels 143
A 0.55
"~"
,
........
~
.
- -
~~~--
-
...
~
--_ ,.......
.... ~ .. '
~.
r ,
,.. ,: :::::-::::: : : :
,"~ ~---~~~ ,
., '1,'" __ _ .
~--- -
~ ' ::
~""~~~_r
-- _ ... - - ... ,. ~
[rn/s]
---
__-, .~.w~
_._-~~~~,
~
~~~
~_"~r"'''~_'~''''''''''''''
...........
--_................. ...,., ..
... " " ,
.... -
, I'.
~
o -
r ~ -
.. ..
_ ~ ~...
... ~
..
,
~
- - . ..
~ ~
, , ... , ..
~
. ~ .
---_.......
..... ~ ~ ~ ~ r r r , .
~--
c 0.25 0 0.07
[rn/s] [rn/s]
~0.25 -0.07
E 2.0 F
Figure 7.8: Distribution of the absolute (A) and relative (B) ve-
locities, inside a aqueous glycerol plug. The contours in (A) denote
the velocity magnitude. The values of the streamwise (C) and wall
normal (D) relative velocities as well as the swirling strength (E)
and the streamlines (F) are shown.
--
A ~ \
,\"~---~~~~~
'\ \
\
'I,
~ ~ ~ w
_ .... _ ~<#'''''''.".
.".
.,,. ..
r
,
_
~ -~--
"~--
".
",."
.".
'"
,
..
,j
..
-
~
.....
.... , _., --
. . . ,, ,, ,, ,, . -
.
~ ~
, , . -
~ ~ '\ ~
- .,. - _
~.., \ > , ~
~ ~
,.. ."" .. -
~
~ ~~-_
"
_.-.,,~,--_
. . .. ---~
-_ ~~
~'''''''''''
'I. ..
,
'\
,
,\
\
~
~ ~
>.
.
... .. ~ ~ r ~ .
.. .. ---
".1"
-~
... J
B "._---------.".
.......... ~"
","
~
__ '
-_--
..... .... "".,,. _rv
............~~------~~~~... .",~
~,'~ ..... ~~~~-_, _ _ ~ .. " . " r
_.L ~_.,""
.......... ,"",.. - ~;"'''' ,
\-~
.~
-~.., ~
-, ...
,.
.... - ----.-----
_r
~
~
-" ,, ..-
,_
I
~ ......
- ~~-=:-":::;:::""--'-:=-:::::.--~- ,:"_"::.:
,I"r ~ ~
/f' _
1', .,,, " ....
,~. + , , J _ , ~ .. ~ ~ ~....,
~~ ~ - - , , I , , , __ ~~
:::-----------~-~~~~~~----
C -:::::::::::---::::::::::
" ~_ ~"''''~~-~-~~y
'
...... ... _.,r __
~
- _-_
.~, ~-----~~~~~~+.
_
-- .
& ~
~,
,~~------~~~~.,+
_~--
... .....
- +
".,
_-
- -- -" , ... -
---.... , ., _ N ~ ," ~ , _
-"""- ~ ~
-_. _ -
r~--""--"''''''''''
, I ,~~---
r __
_
~
_ , _ , ~ , , .. r _
,- "
-~--_ '."-
-_. ...
.....
-... r r".~ ... _+- _ _" " " " " ' ' ' ' ' ' ' ' ' ' ' ' ,
_~~-+----, ~ .. . "-, .
. -
...
~
~"
...... ~ .
,.~".,---~_
~~~_~~~
I ,
-.
-
-.....
,
~~~
~ ~
~~ __
,
~J~J ....
, _ ..
~
-.
0.5 rnls o 17.0 100 J..Lrn
Obviously, the surface tension affectH not only the Hhape of the gaH bubble
leading and trailing edge, and therefore the shape of the liquid segment,
but also the velocity distribution inside the liquid phase. Lower surface
tension values tend to increase the internal recirculation velocities and
therefore the mixing quality.
All the presented cases in a straight microchannel of rectangular cross-
Hection Hhow no Hignificant maHH transfer across the centerline of the
7.3 Recirculation in meandering channels 145
r - - - - - - - - - - - - - - - - - -,
:3 ,
,
,,
I
I
I
:1 2:
I I
,
,,,
I
I
I
I
,,
._-
I
channel turning are shown on the left, whereas the liquid plug leaving
the bend is found on the right hand side.
For all the presented results, water and nitrogen were used as working
fluids. For the large diameter channel (D h = 200 p,m) with a turning
radius T = 1200 Mm, and je = 0.37 m/s eVe = 1 ml/min), and jL =
0.04 m/s CVL = 0.1 ml/min) as fluid superficial velocities, the liquid
recirculation in the plug just before entering the channel curve (Fig.
7.11, top) shows the same characteristics as for the straight channels in
chap. 7.2. Two counter-rotating vortices symmetrical to the channel axis
can be found. For the liquid plug after the channel bend, however, the
velocities inside the liquid plug are more uniform, which implies lower
relative velocities. The vorticity (top) and the swirling strength (center)
7.3 Recirculation in meandering channels 147
[1/5]
-8
-------. flow direction -------. flow direction
1.2
0.0
- - 0.5 [m/sI
16
[1/s]
-16
12
100 Ilm
- . 0.5 [m/s]
16
[1/s]
-16
100 J.l.m
trailing end. High relative velocities and strong vorticity at the channel
outside are determined for high superficial velocities (Fig. 7.16).
The calculation of the streamlines and the swirling strength for the in-
stantaneous velocity distributions did not show reasonable information
7.3 Recirculation in meandering channels 151
-+ 0.5 [m/sI
16
[1/5]
-16
~F 1100~m
Figure 7.14: Characteristics of two liquid plugs entering (left) and
leaving (right) a microchannel bend (r = 600 /l-m) for a channel
with Dh = 150 /l-rn and a fiuid HyHtern with jc = 0.56 m/s (Vc =
0.75 ml/min) and jL = 0.04 m/s (VL = 0.05 ml/min). The relative
velocity vectors together with the liquid vorticity are shown on the
top comparison. The swirling strengths are depicted in the center ,
whpn~as thp strpamlines of the two confignrations can be seen at the
bottom.
Vorticity:
-16 [1/5] 16
Velocity:
- 0.5 [m/5]
velocity of the liquid plug. Note that the exact plug position (turning
angle) in the channel bend is difficult to measure in the raw images
((angle ~ 2). Due to the strong dependency of the average velocity
components on the turning angle, the accuracy is limited.
7.4 Summary and conclusions 153
Vorticity:
-16 16
Velocity:
...... 0.5 [m/s]
0.05
~
--
0
0
E
......
g'
0.04
0.03
>
III o
0.02 o
g' o
>111 o
0.01
~ o 0 Il 0 0
'0
0
0.00 ______ .~-----------------------------L-------
,
Q)
> 0.01
Q)
Cl
~ -0.02
,
~
co -0.03
Q)
>
:;:;
co -0.04
o
Q)
a::: 0.05
0 20 40 60 80 100 120 140 160 180
Turning angle rl
Figure 7.17: Average spanwise (wall normal, transfer motion) and
streamwise velocity component inside the liquid plug relative to the
velocities in a straight microchannel dependent on the liquid plug
position (turning angle).
within the vector file was possible. As the velocity distribution is ex-
pected to change as the fluid advances in curved meandering channels
(change in angular position), averaging is abandoned and the instanta-
neous recirculation motion is analyzed.
Of special interest concerning the mass transfer of liquid towards the
phase boundary or a catalyst material coated onto the reactor walls,
was the mixing q~ality and relative velocity distribution. Especially the
existence of mass transfer across the channel center line is desirable. The
computation of the streamlines and the analysis of the swirling strength
gave additional information on the internal recirculation.
For straight microchannels, an increase in relative velocities for low sur-
7.4 Summary and conclusions 155
Chapter 8
Hydrogenation of
nitrobenzene
Yeong et al. (2004) used a falling film microreactor for the hydrogenation
of nitrobenzene, featuring 64 straight, parallel channels (300 Mm wide,
100 Mm deep, and 78 mm long). Experiments were conducted at 60C,
varying the pressure from 1 to 6 bar. The liquid flow rates were 0.5 -
3 ml/min nitrobenzene solution. Depending on the applied pressure and
the liquid flow rate, conversion rate between 0.25 and 0.85 were achieved.
According to the results of Yeong et 1'11. (2004), we conducted the hy-
drogenation at different flow and boundary conditions. The microreactor
and the inlet tubing were heated to 60 - 75C and pressures between 5
and 10 bar were applied. The nitrobenzene was diluted in pure ethanol
(1:25 by volume). The liquid flow rate was kept as small as possible to
increase the residence time (5 to 10 MI/min). The gas flow rate was lim-
ited by the applied back pressure regulator, requiring a minimum flow
rate to ensure a specific pressure. Minimum flow rate ensuring a pres-
sure of 10 bar was 0.5 Nml/min hydrogen. The applied microreactor had
a hydraulic diameter of D h = 150 Mm (square cross-section) and a to-
tal two-phase length of 120 mm. Assuming a volumetric ga.-s content of
~ 0.8, the residence time is estimated to be in the order of 3.25 to 6.5 s.
158
Chapter 9
rapidly expanding field. However, for the proper design of actuators (e.g.
pumps), it is crucial to know the flow characteristics of the desired pro-
cess. It appears to be necessary to accumulate experimental data on flow
characteristics (which includes the flow regime, the pressure drop, and
the void fraction) to develop new equations and accurate models for the
prediction of these characteristics in microfluidic systems. In view of this,
the flow regime, the pressure drop, the void fraction, and the liquid veloc-
ity distribution (recirculation motion) have been studied in the present
research project for two-phase flow in small-diameter microchannels. All
the mentioned flow properties have been shown to behave significantly
different from the large-scale pipe flow. Therefore, the existing correla-
tions are not able to predict the phenomena of microchannel two-phase
flow, as the influence of the hydraulic diameter does not scale linearly.
Furthermore, the increasing importance of the surface forces (and con-
sequently the loss of importance of the body forces) causes novel phe-
nomena, implicating the necessity of further experimental investigation.
ecobalance. This is mainly due to the high energy consumption. All other
equipment contribute less than 0.5% to the total energy use. The relative
influence of the clean room maintenance can only be reduced by a higher
efficiency or load of the facility.
Besides the fabrication also the use of silicon based microreactors for
the production of aniline (hydrogenation of nitrobenzene) was analyzed.
It can be shown, that the fraction of of the lifetime of a microreactor
used for effective aniline production is in the range of 7%. Conventional
batch reactors are in the order of 50%. Not taken into consideration are
the chemicals and facilities for the reactor fabrication. Obviously eco-
logical reasons do not legitimate the use of microreactors for the aniline
production.
9.2 Outlook
Further research work advancing the present project is proposed, con-
tinuing the path to chemical engineering applications.
HPLC
Pump
/f
I
I
Substrate
onIlne sample
Analysl. method.: Preparation
Thermocouples
SubsequenlAnalysis
Pressure SenSOl'&
lIlPIV)
References
Senkan, S., Krantz, K., Ozturk, S., Zengin, V., and Onal, 1. (1999). High-
throughput testing of heterogeneous catalyst libraries using array mi-
croreactors and mass spectrometry. Angewandte Chemie-International
Edition, 38(18):2794-2799.
Curriculum Vitae
Severin Walchli
Date of birth: June 13th 1974
Place of birth: Brugg (AG), Switzerland
Citizen of: Brittnau (AG), Switzerland
Zurich, 2005