Chemical Engineering Science 61 (2006) 7609 – 7625

www.elsevier.com/locate/ces

Numerical study on gas and liquid slugs for Taylor flow

in a T-junction microchannel
Dongying Qian ∗ , Adeniyi Lawal
New Jersey Center for Micro Chemical Systems, Department of Chemical, Biomedical and Materials Engineering, Stevens Institute of Technology,
Castle Point on Hudson, Hoboken, NJ 07030, USA

Received 4 May 2006; received in revised form 26 July 2006; accepted 23 August 2006
Available online 14 September 2006

Abstract
The rapid development of microfabrication techniques creates new opportunities for applications of microchannel reactor technology in
chemical reaction engineering. The extremely large surface-to-volume ratio and the short transport path in microchannels enhance heat and
mass transfer dramatically, and hence provide many potential opportunities in chemical process development and intensification. Multiphase
reactions involving gas/liquid reactants with a solid as a catalyst are ubiquitous in chemical and pharmaceutical industries. The hydrodynamics
of the flow affects the reactor performance significantly; therefore it plays a prominent role in reactor design. For gas/liquid two-phase flow
in a microchannel, the Taylor slug flow regime is the most commonly encountered flow pattern. The present study deals with the numerical
simulation of the Taylor flow in a microchannel, particularly on gas and liquid slugs. A T-junction empty microchannel with varying cross-
sectional width (0.25, 0.5, 0.75, 1, 2 and 3 mm) served as the model micro-reactor, and a finite volume based commercial computational fluid
dynamics (CFD) package, FLUENT, was adopted for the numerical simulation. The gas and liquid slug lengths at various operating and fluid
conditions were obtained and found to be in good agreement with the literature data. Several correlations in the T-junction microchannel were
developed based on the simulation results. The slug flows for other geometries and inlet conditions were also studied.
䉷 2006 Elsevier Ltd. All rights reserved.

Keywords: Fluid mechanics; Multiphase flow; Hydrodynamics; Numerical simulation; Microchannel

1. Introduction packed bed microreactor, single phase microreactor and multi-

Microchannel reactors have many advantages over conven-
tional macroreactors. With its small transverse dimensions
(submillimeter), microchannel reactors possess extremely high
surface to volume ratios, and consequently exhibits enhanced
heat and mass transfer rates. Scale-up for high throughput is
achieved by simple replication of microreactor units; which
eliminates costly reactor redesign and pilot plant experiments,
thus shortening the development time from laboratory to com-
mercial production. Microchannel reaction technology has
numerous potential applications in chemical engineering, and
different types of reactor are often required for various re-
actions, i.e., thin-wall microreactor, membrane microreactor,
∗ Corresponding author. Tel.: +1 201 216 5539; fax: +1 201 216 8306.
E-mail address: dqian@stevens.edu (D. Qian).
0009-2509/$ - see front matter 䉷 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2006.08.073
phase microreactor (Jensen, 2001).
Multiphase reaction in microchannels finds good applications
in many industrially important chemical processes, e.g. direct
synthesis of hydrogen peroxide, gas–liquid hydrogenation and
immiscible liquid–liquid nitration. The hydrodynamic charac-
teristics of the flow plays a very important role in the reactor
design. Since the flow in a microchannel is typically laminar,
the mass transfer between the phases is mainly dominated by
diffusion, and turbulent convective flow normally relied upon
for rapid and effective mixing in macroreactors is non-existent.
Mixing is therefore a critical issue in the design of multi-
phase microreactors. Many investigators have characterized
gas–liquid flows in a microchannel (Chen et al., 2002; Chung
and Kawaji, 2004; Triplett et al., 1999; Yang and Shieh, 2001;
Fukano and Kariyasaki, 1993; Coleman and Garimella, 1999;
Akbar et al., 2003). Bubbly, slug, churn and annular flow are
among the observed flow regimes. However, over a wide range
7610 D. Qian, A. Lawal / Chemical Engineering Science 61 (2006) 7609 – 7625
of operating conditions, the flow in a multiphase microchannel
is typically the so-called Taylor flow regime (Heiszwolf et al.,
2001). The Taylor flow is a special case of slug flow where the
bullet-shaped bubbles (Taylor bubbles) are separated by free-
gas-entrained liquid slugs. The elongated bubble has a charac-
teristic capsular shape with equivalent diameter larger than the
channel width. There is a very thin film between the gas bub-
ble and the channel wall. Taylor slug flow has been shown to
increase transverse heat and mass transfer compared to single
phase laminar flow because of the recirculation within the liq-
uid slugs and the reduction of axial mixing between the liquid
slugs (Bercic and Pintar, 1997; Kreutzer et al., 2001; Irandoust
and Andersson, 1989a; Thulasidas et al., 1997).
A number of studies on Taylor flow in a microchannel have
been carried out both experimentally and numerically. Kreutzer
et al. (2005c) have summarized the work from the literature.
Bretherton (1961) theoretically studied the liquid film thick-
ness by applying lubrication theory. Irandoust and Andersson
(1989b) measured the film thickness
in a capillary with spec-
trometric method and obtained an empirical formula:

/d = 0.18[1 − exp(−3.1Ca 0.54 )], Ca = L UT P /, (1)
where d is diameter of the capillary, and Ca is Capillary num-
ber. Thulasidas et al. (1995a) measured the bubble size and
shape, bubble velocity with flow visualization method. Taha
and Cui (2004, 2006a,b) performed CFD studies on bubble
size and shape, bubble velocity and velocity field inside the
slugs in the microchannels by applying the VOF model in
FLUENT package and found good agreement with the pub-
lished experimental data. For slug flows in capillaries, at very
low Ca, the film thickness tends to be zero, and the bubble
velocity is approximately equal to the sum of the liquid su-
perficial velocity and the gas superficial velocity. The veloc-
ity inside the liquid slug has a parabolic profile with respect
to the solid wall except for very short liquid slugs. There is a
clear recirculation relative to the bubble movement. The mass
transfer for the Taylor flow in a microchannel has been inves-
tigated intensively in the literature. Bercic and Pintar (1997)
proposed the following mass transfer correlations for gas to
liquid and liquid to solid after measuring the physical absorp-
tion of methane in water and dissolution of benzoic acid into
water:
kGL aGL = 0.111UT1.19 0.57
P /LL , (2)
kLS aLS = 0.069UT0.63
P /(LL − 0.105LG )
0.44
, (3)
where kGL aGL , kLS aLS are mass transfer coefficients, LL and
LG are slug lengths, and UT P is unit cell velocity. For the
mass transfer of gas to solid through the liquid film, Kreutzer
et al. (2001) considered a stagnant layer and provided the mass
transfer coefficient kGS aGS , as
kGS aGS = D/
, (4)
where D is diffusion coefficient. Van Baten and Krishna (2004,
2005) performed CFD studies for gas to liquid and liquid to
solid mass transfer using CFX package. Vandu et al. (2005)
obtained a dimensionless expression for liquid to solid mass
transfer after experimentally studying oxygen absorption dy-
namics. Kreutzer et al. (2005b) measured the pressure drop in
a capillary and also performed CFD studies with FIDAP code.
They correlated the apparent friction factor f with the following
expression:

  
16 d Re 1/3  UT P d
f= 1 + 0.07 , Re = L , (5)
Re LL Ca L
where Re is Reynolds number. This empirical correlation holds
for Weber number W e = L UT2 P d/ > 0.1. The constant 0.17
should be used instead of 0.07 if there are impurities in the
experiment, because the impurities give rise to surface tension
gradient, thus resulting in Marangoni effect. Liu et al. (2005)
presented another form of correlation for the friction faction
from their experiments. Bercic and Pintar (1997), Thulasidas
et al. (1995b, 1999), and Kreutzer et al. (2005a) measured res-
ident time distribution (RTD) in microchannels and found a
dependence of RTD on the slug length.
All the results available in the literature indicate that mass
transfer, pressure drop and RTD for slug flows in microchan-
nels are highly liquid film and slug length dependent. Unlike
other flow parameters such as superficial velocities and fluid
properties, the liquid film thickness and the slug length can-
not be determined a priori. Flow visualization, conductivity,
or other advanced tomography techniques like magnetic reso-
nance imaging (MRI) are commonly used to determine these
quantities. While the film thickness has been extensively stud-
ied experimentally and numerically, there are limited studies
on the slug length. Laborie et al. (1999) visualized the slug
flows in a 1–4 mm capillary for water, glycerol, ethanol—air
systems, and obtained a dimensionless expression for the gas
and liquid slug lengths:
LG Re0.63  Vb d  d 2g
= 0.0878 V1.26 , ReV = L , Eo = L , (6)
d Eo L 
 1.27
LL 1  UG d
= 3451 , ReG = L , (7)
d ReG Eo L
where Eo is Eotvos number. Liu et al. (2005) and Vandu
et al. (2004) attempted to correlate their experimental data from
a 0.9–3.0 mm circular or square channel for water, ethanol–air
systems by adopting an approximate form of Eq. (2):
L = 0.088ReG ReL ,
UT P /L0.5 ReG = G UG d/G ,
0.72 0.19
ReL = L UL d/L . (8)
However, all the data parities for Eqs. (6)–(8) are not satis-
factory. Mantle et al. (2002) measured the gas slug lengths in
1.2 mm square channels within a ceramic monolith using MRI
method, and found a wide range of bubble size distributions.
Heiszwolf et al. (2001) and Kreutzer et al. (2005d) detected
slug frequency using conductivity probes in 1.56 mm square
channels within a monolith, from which they calculated the gas
and liquid slug lengths. They found that the slug length was
 D. Qian, A. Lawal / Chemical Engineering Science 61 (2006) 7609 – 7625
normally between 2 and 10 times the channel diameter. They
reported that the slug length was highly dependent on the
geometry of the inlet distributors. There was a considerable
variation of the gas and liquid slug lengths, even for the same
experimental setup; 30% standard deviation was not uncom-
mon. They proposed to measure the pressure drop and then
compute the slug length for the short slugs.
The slug length is an important parameter for Taylor flows,
however, there is a significant inconsistency in the literature
studies, therefore, a systematic study on the gas and liquid
slug lengths in microchannels is warranted. As Kreutzer et al.
(2005c) pointed out in their review paper, CFD studies will
provide the mechanistic insight for the slug flow in microchan-
nels. Current CFD studies mainly focus on one unit slug cell,
which presume that the gas and liquid slug lengths are known.
Researchers either used single phase model with void gas
(Van Baten and Krishna, 2004, 2005; Kreutzer et al., 2005b;
Irandoust and Andersson, 1989a) or implemented two phase
VOF model (Taha and Cui, 2004, 2006a,b) to investigate
bubble shape, bubble velocity, film thickness, mass transfer,
pressure drop and velocity profile inside the liquid slug. Some
commercial software CFD packages such as FLUENT, FIDAP
and CFX have proven to be appropriate to perform these stud-
ies. The aim of this paper is to study numerically the gas and
liquid slugs for Taylor flow in microchannels. A T-junction
microchannel serves as a model geometry. The development of
the gas and liquid slugs in the model geometry with different
inlet configurations at various operating and fluid conditions is
investigated.
2. Analysis
2.1. Model geometry
The hydrodynamic characteristics of the flow in microchan-
nels are different from those encountered in ordinarily large-
size channels. The threshold of hydraulic channel diameter is
about 1 mm with fluid properties similar to air and water (Akbar
et al., 2003). In microchannels, the buoyancy effect is negligi-
ble in comparison to surface tension, which in effect renders
the flow characteristic independent of channel orientation with
respect to gravity (Triplett et al., 1999). In the studies of chem-
ical reaction in microchannels, a T-junction microreactor with
millimeter to submillimeter range is typically chosen to inves-
tigate the hydrodynamics or kinetics for multiphase reactions.
The present work considers an empty T-junction microchannel
reactor (not packed with catalyst) with cross-sectional width
of 0.25, 0.5, 0.75, 1, 2 and 3 mm for the studies of gas–liquid
hydrodynamics. This type of reactor is applicable to homoge-
neous reaction or surface reaction with thin-film catalyst coated
on the wall.
Fig. 1 is the model T-junction microchannel reactor used in
our simulation. The microchannel reactor comprises a vertical
inlet mixing zone and a horizontal reaction zone. A stream of
water and a stream of air are fed separately into the two inlets
of the mixing zone, and then enter the reaction zone. The cross-
sectional dimensions of the channel are the same, which are
7611
water
d
6d
60 d
air
y
0 x
Fig. 1. Model T-junction microreactor used in the simulation.
represented by d. The mixing zone has a length of 6d while the
reactor zone has a length of 60d. “Cold” flow without any chem-
ical reaction is considered. The whole system is maintained at
room temperature, and the pressure is atmospheric at the exit.
The superficial gas velocity as well as superficial liquid veloc-
ity varies from 0.01 to 0.25 m/s. According to the flow regime
maps by Chung and Kawaji (2004), Triplett et al. (1999), Yang
and Shieh (2001), Fukano and Kariyasaki (1993), Coleman and
Garimella (1999) and Akbar et al. (2003), at these operating
conditions, the flow falls within the Taylor slug regime in mi-
crochannels.
2.2. Governing equations
A finite volume based commercial CFD package, FLUENT
(Release 6.1.22, 2003), was used to perform the numerical sim-
ulations. In order to track the interface between the gas and
liquid slugs, one of the general multiphase models in FLUENT,
the volume of fluid (VOF) model, was adopted to simulate the
Taylor slug flows in the T-junction microchannel. Among sev-
eral general multiphase models currently in use, the VOF model
is the only multiphase model that enables identification of the
interface clearly. The VOF model in FLUENT package has been
successfully applied in various applications: drop ejection from
the nozzle of a printhead in an ink jet printer, the shape of free
surface change due to a sudden change of gravity, three phase
free surface flow over a weir, flow of a gas bubble moving in
the surrounding liquid, etc. (FLUENT application briefs and
technical notes, 2002a–d). Taha and Cui (2004, 2006a,b) have
also demonstrated that the VOF model in FLUENT correctly
predicts the Taylor flow behavior in a microchannel.
The governing equations of the VOF formulations on multi-
phase flows are as follows:
Equation of continuity:
j
+ ∇ · (
v ) = 0. (9)
jt
Equation of motion:
j(v)
+ ∇ · (
v v) = − ∇p + ∇ · [(∇ v + ∇ vT )]
jt
g + F .
+  (10)
7612 D. Qian, A. Lawal / Chemical Engineering Science 61 (2006) 7609 – 7625

Volume fraction equation:
jG
+ v · ∇G = 0, (11)
jt
where
 = L L + G G ,  = L L + G G . (12)
The flow is treated as incompressible since the pressure drop
along the reactor is small. Compared to single phase flow sim-
ulations, the VOF method accomplishes interface tracking by
solving additional continuity-like equation(s) for the volume
fraction (n − 1 equations with n as the number of phases)
(Hirt and Nichols, 1981). For gas–liquid two phase flows, the
volume fraction of gas G is obtained by numerically solving
Eq. (11), and the volume fraction of liquid L is simply com-
puted from 1 − G . However, the implementation of the VOF
model is numerically challenging. First, both the density  and
viscosity  in Eqs. (9) and (10) are mixture properties, which
vary within the flow domain and are computed by the volume
fraction weighted average in Eq. (12). Second, the interface
needs to be constructed based on the computed value of the vol-
ume fraction with the application of interpolation schemes for
the identification of the interface. Third, surface tension along
the interface between each pair of phases, and wall adhesion
between the phases and the walls become very important for
some cases, especially when the gravitational effect is negli-
gible. All these complexities make VOF multiphase flow sim-
ulation computationally expensive, and convergence difficult
to achieve when compared to its single phase counterpart. In
fact, multiphase flow is perhaps the most difficult topic in the
CFD simulation. Theory and simulations agree with experiment
with a degree of accuracy rarely found in multiphase reactor
engineering (Kreutzer et al., 2005c).
2.3. Numerical approach
where G is the curvature computed from the divergence of the
unit surface normal, and  is the surface tension.
The model geometry was meshed with the preprocessor
GAMBIT software (part of FLUENT package), then imported
into processor FLUENT for calculation. The simulation results
were either analyzed by FLUENT integrated postprocessor or
exported into data files for other postprocessor packages. A
segregated time dependent unsteady solver in FLUENT was
used. The boundary conditions were: velocity inlet for the gas
and liquid feeds, specified as uniform entrance velocity, and
outflow for the reactor outlet which treats the flow as fully de-
veloped at the exit. The other options selected in FLUENT were
as follows: the PRESTO! (pressure staggering option) scheme
for the pressure interpolation, the PISO (pressure-implicit
with splitting of operators) scheme for the pressure–velocity
coupling, second-order up-wind differencing scheme for the
momentum equation, the geometric reconstruction scheme for
the interface interpolation, implicit body force treatment for
the body force formulation, and Courant number 0.25 for the
volume fraction calculation. In addition, the time step, the max-
imum number of iterations per time step, and the relaxation
factors need to be carefully adjusted to ensure convergence.
3. Results and discussion
3.1. Slug flow development
The development of the slug flow was initially investigated
in the model T-junction microchannel as well as in a straight
microchannel. For the straight channel, the feed was a mathe-
matically well pre-mixed gas/liquid mixture. The microchannel
was first filled with water, and then the streams were introduced
at time zero. Figs. 2 and 3 show the slug flow development at
different time steps in the T-junction geometry and the straight
microchannel respectively. It can be seen that, the gas and

The flow inside the microchannel is essentially laminar. Typ-

ically, surface tension force dominates over the gravitational
effect, and wall adhesion force is also important. In FLUENT,
the surface tension model is the continuum surface force (CSF)
model proposed by Brackbill et al. (1992). With this model,
the addition of surface tension to the VOF calculation results
in a source term in the momentum Eq. (10) (FLUENT, 2003).
In the case of wall adhesion, a contact angle needs to be spec-
ified. Rather than impose the boundary condition at the wall
itself, the contact angle that the fluid is assumed to make with
the wall is used to adjust the surface normal in cells near the
wall. This so-called dynamic boundary condition results in the
adjustment of the curvature of the surface near the wall, and
this curvature is then used to adjust the body force term in the
surface tension calculation (FLUENT, 2003). For gas–liquid
two phase flows, the source term in Eq. (10) that arises from
surface tension can be computed from the following:
2G ∇G
F =  , (13) Fig. 2. Slug flow development in the model geometry (red——gas slug,
(G + L ) blue——liquid slug, d = 0.5 mm, UG = UL = 0.02 m/s).
 D. Qian, A. Lawal / Chemical Engineering Science 61 (2006) 7609 – 7625 7613

Fig. 3. Slug flow development in a straight microchannel (red——gas slug,

blue——liquid slug, d = 0.5 mm, UG = UL = 0.02 m/s).

Table 1
Slug length in a 3D tube and a 2D duct with a pre-well-mixed inlet stream

d UG UL LG LG LL LL
(mm) (m/s) (m/s) (mm) (mm) (mm) (mm)
Fig. 4. Laplace pressure in a 3D tube and a 2D duct.
(3D) (2D) (3D) (2D)

0.25 0.05 0.05 0.53 0.54 0.37 0.47

0.50 0.05 0.05 0.96 0.90 0.62 0.80
0.75 0.05 0.05 1.57 1.55 1.34 1.40 that of a 2D case since it involves both radii of curvature with
1.0 0.05 0.05 2.10 2.10 1.85 1.80 a spherical bubble. This Laplace pressure term, which results
2.0 0.05 0.05 4.98 4.73 3.60 3.70 from the surface tension, was computed from the average pres-
3.0 0.05 0.05 7.70 7.45 6.00 5.85 sure drop across the front interface and the back interface, with
0.5 0.02 0.08 0.39 0.36 1.56 1.78
0.5 0.08 0.02 2.57 2.40 0.33 0.31
several randomly selected points for each tube size. From the
figure, one can see that the Laplace pressure from the simu-
lations satisfies the Young–Laplace equation quite well. Some
discrepancy is caused by the difference between the tube size
liquid slugs develop and propagate either at the Tee intersec- and the real bubble shape and size since the resolution in sim-
tion or near the entrance of the channel. After a slug develops ulations is not able to capture adequately the bubble shape and
and detaches, it travels downstream. The slug is generally sta- film thickness.
ble; it seldom breaks or coalesces or collapses unless it is a very In the following discussions, 2D simulation results are shown
short gas slug. Both a uniform velocity inlet and a parabolic because it takes an enormous amount of time to perform a single
velocity inlet have the same slug lengths for these geometries. 3D run in order to collect sufficient statistic data in comparison
It is also worth noting that even though the flow conditions are to a 2D run. The focus of this work is the slug length, and there
the same for these two cases, the slug lengths are different on is not much difference between a 3D tube and a 2D duct based
account of the difference of the inlet geometry. on the simulation results.
In the simulations, the Ca is on the order of 10−3 , so the film
3.2. Slug flow parameters thickness is about 0.01d, and the bubble velocity is about 3%
faster than the bulk superficial velocity due to the slip. Thus
Both 3D simulations for a circular tube and 2D simulations the gas slug velocity Vb will be in use approximately equal to
for a duct at the same operating conditions were first performed. UG + UL in the following discussions. From the simulation re-
The purpose of this exercise was to determine how the third di- sults, the gas slug velocity was computed and the approximation
rection affects the slug length. In this simulation, a straight tube was confirmed to be valid. The velocity across a sectional line
and a microchannel with a pre-well-mixed inlet stream were inside the bulk slug has a parabolic profile, with a slightly big-
selected. The results of the slug length are listed in Table1. One ger gas velocity than the liquid velocity; this is always true in
observes that between 3D and 2D geometries, there is not much the simulation as long as the slug is not too short. Flow recircu-
difference in terms of slug length. The little difference between lation inside the liquid slug has been observed on the reference
the liquid slug lengths especially for smaller size channels is frame of the gas slug. Fig. 5 shows the pressure profile along
probably due to the lower resolution of 3D simulations, that is, the axial direction. The upper lines are in the gas slugs while
the impact of artificially finite gas/liquid interface is stronger the lower ones are in the liquid slugs. The pressure drop can be
for the smaller size slugs. The Laplace pressure with respect decomposed into a viscous term due to friction and a Laplace
to the tube size is plotted in Fig. 4, where the lines represent term due to surface tension. The pressure difference between
the Laplace theory. The Laplace pressure of a 3D case is twice the adjacent gas and liquid slug is the Laplace pressure. As one
7614 D. Qian, A. Lawal / Chemical Engineering Science 61 (2006) 7609 – 7625

Fig. 6. Slugs at different mesh resolutions (d =0.5 mm, UG =UL =0.02 m/s).
Number of quadrilateral cells (a) 1665 (b) 3087 (c) 6600 (d) 13202 (e) 26400.

Fig. 5. Pressure profile along axial direction (d=0.5 mm, UG =UL =0.02 m/s).

can see, the pressure drop across the back interface is always
higher than that across the front interface, so one can expect
that the shape of the interface is different at the bubble head and
the bubble end, with the bubble end having a bigger curvature.

3.3. Comparison of experimental data with numerical

predictions

The simulations were carried out on Linux PC cluster with

2.8 GHz Pentium IV or Xeon processors. A few thousand to ten
thousand time steps were required for each run. Twenty to a
hundred slugs were collected, from which the sample averages
and other statistical information were computed. Unless oth-
erwise stated, the data presented in the following sections are
for an empty bed with surface tension of 0.072 N/m, a liquid
viscosity of 0.001 kg/ms and wall contact angle of liquid of 0◦ .
The study of mesh dependence on the simulations was ini-
tially performed. A uniform rectangular grid was used for 2D
simulations. In the beginning, a series of mesh cells with dif-
ferent fineness were chosen until the results did not show sig-
nificant changes. Fig. 6 shows the slug contours at different
mesh resolutions. The significant difference on these contour
plots occurs only on the interface. The higher the resolution,
the sharper the interface is. When the resolution is too low, part
of the interface is twisted so the shape is not always a circular.
The data presented in this work were based on 6600 quadrilat-
eral cells in the model geometry.
A number of code validation exercises have been carried out.
The experimental data are taken from the literature. As ear-
lier stated, most of the literature studies characterized the flow
patterns, and presented the flow regime maps inside the mi-
crochannels. Although there were a few pictures of the slug
flows in these studies (Chen et al., 2002; Chung and Kawaji,
2004; Triplett et al., 1999; Yang and Shieh, 2001; Fukano and
Fig. 7. (f Re) as a function of the dimensionless group (LL /d)∗ (Ca/Re)0.33 .
Kariyasaki, 1993; Coleman and Garimella, 1999), no corre-
sponding slug length data were presented. Laborie et al. (1999)
presented some experimental data about the gas and liquid slug
lengths inside capillaries. However, the air was introduced into
the water through a porous membrane, rather than through a
Tee mixer. According to our simulation, the degree of premix-
ing of the gas and the liquid and the inlet geometry strongly
affect the slug length, while the length of the mixer or the reac-
tor has no effect on the slug length as long as it is long enough
to obtain a fully developed flow.
Fig. 7 plots the dimensionless apparent friction factor for
air–water and air–decane (surface tension 0.024 N/m) systems
compared with experimental data by Kreutzer et al. (2005b).
The pressure drop and liquid slug length were the average
values of randomly selected points. The simulations are in
good agreement with the experiments. The scattering of the
points are due to the non-uniformity of the slugs. Vandu et al.
(2004) conducted air–water flow visualization experiments in
 D. Qian, A. Lawal / Chemical Engineering Science 61 (2006) 7609 – 7625 7615

Table 2
Comparison between numerical simulations and experiments in a 3 mm vertical circular capillary

UG UL LG LGmin (mm) LGmax (mm) LL LLmin (mm) LLmax (mm)

(m/s) (m/s) (mm) LGmin (exp.) LGmax (exp.) (mm) LLmin (exp.) LLmax (exp.)

0.05 0.02 15.38 13.95 16.65 7.28 6.30 8.10

13.00 15.00 4.95 9.45
0.05 0.05 12.80 12.15 13.50 12.47 11.25 13.50
12.00 13.00 12.00 13.00
0.05 0.10 6.22 3.60 9.45 9.39 3.15 18.90
3.50 9.50 6.00 18.50
0.05 0.15 4.56 2.25 9.45 9.10 2.25 20.25
3.00 8.50 3.00 25.00
0.05 0.20 3.30 1.35 8.10 7.48 1.35 22.05
2.00 9.50 1.50 38.00
0.10 0.02 19.19 6.75 32.85 3.98 1.80 7.65
20.50 31.00 3.00 5.00
0.10 0.05 15.39 9.45 18.90 5.64 2.25 8.55
17.00 18.50 6.50 7.50
0.10 0.10 5.55 4.50 13.50 3.22 1.35 8.10
10.00 13.50 7.50 11.00
0.10 0.15 4.98 3.60 10.35 4.50 2.25 9.90
3.50 13.00 2.00 15.50
0.10 0.20 4.93 2.25 8.55 6.42 1.35 12.60
3.00 9.00 1.50 24.00

a 3.0 mm vertical capillary through a T-junction mixer. From

their movies, the slug length information was obtained. Table 2
lists our simulation results and their experimental data. It can be
seen that the simulations are consistent with the experiments. In
Table 2, LG and LL are sample average values (time and space
mean values). Both simulation results and experimental data
indicate that the slug length is not uniform throughout the chan-
nel because of the time-dependent toroidal vortices generated
when the two streams meet at the mixing section, and which
subsequently propagate throughout the length of the channel.
This non-uniformity becomes more pronounced as the superfi-
cial gas or liquid velocity increases. Fig. 8 shows five snapshot
flow patterns at different liquid flow rates. The non-uniformity
of the slugs can be clearly seen in this figure. The simulations
confirm that, for all the above operating conditions, the flow
in each case is in the Taylor slug regime. Fig. 9 is the bar plot
of the gas and liquid slug length distributions corresponding to
these cases, where N is the number of slugs at a particular size
range, and N0 is the total number of slugs. As one observes,
the slug length variation becomes wider as the liquid flow rate
increases. Fig. 8. Contour plots of volume fraction of air in a 3 mm vertical capillary
(gravity direction: −x).
Heiszwolf et al. (2001) and Kreutzer et al. (2005b) studied
the gas–liquid flows in a monolith with 0.5–5 mm channels. The
superficial velocity is in the range of 0.02 m/s < UG < 0.3 m/s the dimensionless gas slug length LG /d and liquid slug length
and 0.02 m/s < UL < 0.2 m/s. They observed slug flows in LL /d as a function of liquid hold-up L . Here the liquid flow
these operating conditions, and measured the slug length with rate volume fraction UL /(UG + UL ) is approximately equal to
conductivity probes. Different flow distributors were used and the liquid hold-up L . As one observes, the simulation results
the experiments showed that the slug lengths were depen- have the same trend as, and agree favorably with the exper-
dent on the inlet geometry. They also reported strong non- imental data. The scattered data indicate that the slug length
uniformity of the slugs. Our simulations were compared with should also depend on other parameters such as flow rate and
their experiments on the coarse distributor because we think fluid properties. The data variations in big diameter channels
that the behavior of the Tee mixer should closely approximate or high flow rates are bigger because of the high variation
the coarse distributor. Fig. 10 depicts the relationship between of the slug length at these conditions. For the considerable
7616 D. Qian, A. Lawal / Chemical Engineering Science 61 (2006) 7609 – 7625

Fig. 9. Slug length distributions in a 3 mm vertical capillary.

 D. Qian, A. Lawal / Chemical Engineering Science 61 (2006) 7609 – 7625 7617

Fig. 10. Dimensionless slug length against liquid hold-up.

Fig. 11. Gas slugs and liquid slugs at different inlet mixing levels (d = 0.5 mm, UG = UL = 0.02 m/s).

range of the liquid hold-up (0.25 < L < 0.75), the dimension-
less liquid slug length is between 1 and 10, which is well in
agreement with the literature.
3.4. Slug length for different inlet geometries
As stated above, the slug length is highly dependent on the
inlet geometry. In this work, the slug flows with different inlet
configurations or different mixing levels have been investigated.
Fig. 11 shows the slugs in a straight channel at different inlet
mixing levels. This can be modeled as if there are alternating
multilaminated channels before the inlet zone. There are 1, 2,
4 and infinite laminating levels corresponding to these cases.
The result indicates that the degree of the inlet premixing alone
will influence the slug length: the better the inlet premixing,
the shorter the slug length.
Fig. 12 shows the contour of gas slugs and liquid slugs at
three different inlet Tee orientations. As one can see, there is
not much difference in the slug length between (a) and (b),
however, the slug length of (c) is much longer. It is thus better
to introduce gas and liquid feeds head to head or perpendicular
to each other with the liquid stream parallel to the microchan-
nel. Perpendicular feeds with the gas stream parallel to the mi-
crochannel should be avoided because this case has smoother
stream at the mixing section, and the slug is more difficult to
detach. Fig. 13 is the contour of gas slugs and liquid slugs for
three different inlet Tee sizes. Although the microchannel ge-
ometry and flow rate are same, the slug length is totally differ-
ent: the smaller the diameter of the mixing zone, the shorter is
the slug length. This is due to the fact that the smaller mixing
zone generates larger velocity, hence bigger toroidal vortices
are produced at this region and make it easier for the slug to
7618 D. Qian, A. Lawal / Chemical Engineering Science 61 (2006) 7609 – 7625

Fig. 12. Gas slugs and liquid slugs at different tee orientations (d = 0.5 mm, UG = UL = 0.02 m/s).

detach. This is a conclusion that is generally valid, i.e., the gen-

eration of big toroidal vortices in the mixing zone reduces the
slug length.

3.5. Gravitational effect on slug flow

In conventional-size tubes as well as millimeter order tubes,

two-phase flow patterns are dominated in general by gravity
with surface tension effects relatively insignificant. On the other
hand, in microchannels of the order of a few microns to a few
tens of microns, two-phase flow is influenced mainly by sur-
face tension, viscous and inertial effects, with gravity playing
a minor role. In order to investigate the gravitational effect in
microchannels, two sets of runs were performed; one with grav-
itational effect included while in the other it is ignored. Fig. 14
shows the comparison of gas and liquid slug lengths with and
without gravity, where the subscript “no” represents no gravity
condition, and Bo is Bond number (Bo = L gd 2 /). One can
see that, as the Bo number increases, the difference between the
slug lengths with and without gravity grows bigger. Since most Fig. 13. Gas slugs and liquid slugs at different inlet tee sizes (d = 0.5 mm,
data here at high Bo number are from bigger channels, gravity UG = UL = 0.02 m/s).
has more influence in millimeter channels. The scatter in the
data in the same size channels demonstrates that the degree of
the gravitational effect also depends on other flow parameters superficial liquid velocity UL , surface tension , liquid viscos-
like Re and Ca. ity L and wall contact angle of liquid  have been performed.
The influence of gas and liquid superficial velocities on the
3.6. Influence of gas and liquid superficial velocities gas and liquid slug lengths was first studied. Fig. 15 shows
the relationship between the gas slug length and the super-
The previous studies show that the slug length depends upon ficial velocities, while Fig. 16 shows the corresponding be-
flow velocity, inlet geometry and fluid properties. Now the havior for the liquid slug length. The gas slug grows with
slug flows in the model T-junction microchannel are focused to increase in superficial gas velocity, and decrease in superfi-
study. A total of about 500 simulation runs with varying param- cial liquid velocity. Similar trend is observed for the liquid
eter values: channel diameter d, superficial gas velocity UG , slug length, i.e., it increases with increase in superficial liquid
 D. Qian, A. Lawal / Chemical Engineering Science 61 (2006) 7609 – 7625
Fig. 14. Comparison of gas and liquid slug lengths with/without gravity.
velocity, and decrease in superficial gas velocity. This trend
is consistent with experimental observations of Laborie et al.
(1999) who reported a linear relationship between LG and UG
in their experimental region. Here, power-law fits for the sam-
ple data were used, which are represented as the solid curves
in Figs. 15 and 16. For cross-sectional dimensions of 1 mm
and lower, the data are well described by one curve. For the
computational geometries considered in this study, the 1 mm
and submillimeter channels exhibit more uniform slugs than
millimeter-size channels. The simulation data show more scat-
ter for the 2 and 3 mm channels in comparison to the submil-
limeter channels. The effect of superficial liquid velocity on
gas slug length, and the effect of superficial gas velocity on
liquid slug length are very remarkable. The higher the ratio of
superficial gas velocity to superficial liquid velocity, the longer
are the gas slugs. Similar observation also applies to the liquid
slugs. Figs. 15 and 16 also show the influence of the channel
width on the gas and liquid slug lengths: at the same superfi-
cial gas and liquid velocities, a wider channel has longer gas
and liquid slugs.
3.7. Influence of fluid properties
A number of simulations with varying liquid properties in
submillimeter channels were performed. Fig. 17 shows the in-
fluence of surface tension, while Fig. 18 shows the influence
of liquid viscosity. In our computational region, the slug length
slightly increases with the increase of surface tension; however,
there is almost no influence of liquid viscosity. These results in-
dicate that the surface tension effect is stronger than the viscous
effect for this system. When the surface tension is very small, it
is difficult to generate slug flow. For example, when the surface
tension is close to zero, the reality flow is the stratified flow.
7619
3.8. Influence of wall surface adhesion
The influence of wall surface adhesion was also investigated.
As stated above, the wall surface adhesion plays an equally
important role as the surface tension. Here, the contact angle of
the liquid to the wall was varied from 0◦ to 180◦ , and the results
are shown in Fig. 19. It is seen that both the gas and liquid
slug lengths decrease slightly until they are almost the same
after the liquid becomes non-wetting. With the change of the
wettability of the fluid on the wall from fully wetting (contact
angle 0◦ ) to fully non-wetting (contact angle 180◦ ), the shape
of the gas–liquid interface changes from concave in to convex
out from the liquid side accordingly. It is speculated that there
is a thin film around the gas slug for the fully wetting cases.
Most literature sources report a thin liquid film around the gas
slug on the order of several to tens of microns. Unfortunately
our grid resolution is not fine enough to capture the liquid film.
For fully wetting fluid, one can expect the gas to liquid slug
length ratio to be slightly higher than the volume flow rate
ratio because of the existence of the thin film, and the fact that
the slug length was computed across the central line, with the
shape of the interface affecting the values. Our data show that
there is always a 20–30% difference between the slug length
ratio and volume flow rate ratio.
3.9. Correlations
Eqs. (9)–(11) can be rewritten in dimensionless form using
dimensional analysis.
Equation of continuity:
j∗
+ ∇ · (∗ v∗ ) = 0. (14)
jt ∗
7620 D. Qian, A. Lawal / Chemical Engineering Science 61 (2006) 7609 – 7625

Fig. 15. Influence of gas and liquid superficial velocities on gas slug length.

Equation of motion: Volume fraction equation:

j(∗ v∗ ) jG

+ ∇ · (∗ v∗ v∗ ) + v∗ · ∇G = 0, (16)
jt ∗ jt ∗
1
= −∇p ∗ + ∇ · [∗ (∇ v∗ + ∇ v∗ )]
T
where
Re
1 ∗ ∗ F ∗ ∗ = /L = G G /L + (1 − G ),
+  g + . (15)
Fr Re · Ca ∗ = /L = G G /L + (1 − G ), (17)
 D. Qian, A. Lawal / Chemical Engineering Science 61 (2006) 7609 – 7625 7621

Fig. 16. Influence of gas and liquid superficial velocities on liquid slug length.

tV b p g velocity Vb . For the conditions used in the simulation, Vb ≈

t∗ = , v ∗ = v/Vb , p∗ = , g∗ = , (18)
d L Vb2 g UG + UL and G ≈ UG /(UG + UL ), so UG = G Vb and UL =
L Vb = (1 − G )Vb . From Eqs. (14)–(16) and the correspond-
L Vb d L Vb Vb2
Re = , Ca = , Fr = . (19) ing dimensionless boundary conditions, one can solve for the
L  gd three dependent variables p∗ , v∗ , G as a function of the inde-
Here the dimensionless numbers are defined based on chan- pendent variables t ∗ , x∗ and the parameters Re(=L Vb d/L ),
nel width d, liquid properties L , L and gas slug moving Ca(=L Vb /), F r(=Vb2 /gd), G /L , G /L , G , , where
7622 D. Qian, A. Lawal / Chemical Engineering Science 61 (2006) 7609 – 7625

Fig. 17. Influence of surface tension on gas and liquid slug length (UG = 0.1 m/s, UL = 0.1 m/s).

Fig. 18. Influence of liquid viscosity on gas and liquid slug lengths (UG = 0.1 m/s, UL = 0.1 m/s).

G is introduced from the velocity inlet boundary conditions, value (time and space average), one can conclude that the
and  comes from the wall surface adhesion boundary condi- dimensionless slug length will have the following dependence:
tion. Then the slug length can be obtained from the gas vol-
ume fraction map. Since the simulated slug length is a mean L∗ = L∗ (Re, Ca, F r, G /L , G /L , G , ). (20)
 D. Qian, A. Lawal / Chemical Engineering Science 61 (2006) 7609 – 7625
Fig. 19. Influence of contact angle on gas and liquid slug lengths (d=0.50 mm,
UG = 0.1 m/s, UL = 0.1 m/s).
From the simulations, we have confirmed that gravity does
not have much of an effect in slug flow in microchannels, so
the Froude number can be ignored. The viscosity of the gas
is negligible compared to that of the liquid; hence G /L can
be dropped. We have observed this from the simulations. The
same argument applies to the density ratio; therefore, G /L is
ignored. Furthermore we do not consider the influence of the
contact angle, and finally Eq. (20) reduces to the form:
L∗ = L∗ (Re, Ca, G ).
Regression of the simulation data provides the following cor-
relations for the dimensionless slug length:
(LG + LL )/d = 1.637−0.893
G (1 − G )−1.05
× Re−0.075 Ca −0.0687 ,
LG /d = 1.6370.107
G (1 − G )−1.05 Re−0.075 Ca −0.0687 ,
LL /d = 1.637−0.893
G (1 − G )−0.05 Re−0.075 Ca −0.0687 .
A total of 148 sets of numerical simulation data were
correlated with channel width 1 mm and less for parame-
ter values in the range: 0.09 < G < 0.91, 15 < Re < 1500,
0.000278 < Ca < 0.01. Eq. (23) is obtained by multiplying
Eq. (22) by G , while Eq. (24) is obtained by multiplying
Eq. (22) by 1 − G , for LG /(LG + LL ) ≈ G and LL /(LG +
LL ) ≈ 1 − G in our simulation conditions. The average dif-
ference between the correlated data and the simulated data is
about 10%. These correlations show that the dimensionless slug
length is mainly determined by the phase hold-up, while the
Re and Ca have a slight effect. This result can explain why the
7623
dimensionless slug length falls within a narrow range even for
wide operating conditions.
4. Conclusions
The characteristics of Taylor slug flow in microchannels have
been elucidated using CFD simulations. The results show that
the slug length is not uniform throughout the channel especially
for channel cross-sectional dimensions exceeding 1 mm. When
the gas or liquid flow rate increases, the slug non-uniformity
becomes more pronounced. The slug length is also highly de-
pendent on the inlet configuration. From the studies of gas and
liquid slugs in a T-junction microchannel at various operating
conditions, the following conclusions are drawn:
• The gas slug length increases with increase in superficial gas
velocity, and decrease in superficial liquid velocity.
• The liquid slug length increases with increase of superficial
liquid velocity, and decrease of superficial gas velocity.
• The dimensionless slug length is mainly determined by the
phase hold-up, with a slight effect of Re and Ca. Therefore,
wider channels have longer slug length at the same superficial
gas and liquid velocities.
• Gravitational effects can be ignored in microchannels.
• The effects of fluid density and viscosity are also negligible.
• The surface tension and wall surface adhesion moderately
impact the slug lengths.
After a dimensionless analysis from the governing equa-
tions, correlations for the dimensionless unit slug length, gas
slug length and liquid slug length in the model T-junction mi-
crochannel have been obtained. These correlations can be used
to predict the slug length based on the operating condition,
(21) and subsequently some important transport parameters such as
pressure drop, heat transfer and mass transfer coefficients can
be estimated. Based on this numerical study of Taylor flow in
microchannels, we can begin to understand, and quantify heat
and mass transfer enhancement in microchannels, which will
be the subject of our next study.
(22)
Notation
(23)
a mass transfer surface area, m2 /m3
(24) d internal channel width, m
D diffusion coefficient, m2 /s
f apparent friction factor
F body force, N
g gravity, kg/m2
k mass transfer coefficient, m/s
L mean slug length, m
L̃ in-situ slug length, m
N number of slugs
N0 total number of slugs
p system pressure, Pa
t time, s
U superficial velocity, m/s
7624 D. Qian, A. Lawal / Chemical Engineering Science 61 (2006) 7609 – 7625

UT P unit cell velocity (=UG + UL ), m/s Coleman, J.W., Garimella, S., 1999. Characterization of two-phase flow
v velocity, m/s patterns in small diameter round and rectangular tubes. International
Journal of Heat and Mass Transfer 42, 2869–2881.
Vb gas slug velocity (≈ UG + UL ), m/s
FLUENT application briefs, 2002a. Drop ejection from a printhead nozzle,
EX147, Fluent Incorporated, Lebanon, New Hampshire.
Greek letters FLUENT application briefs, 2002b. Free surface change during free-fall,
EX168, Fluent Incorporated, Lebanon, New Hampshire.
 volume fraction FLUENT application briefs, 2002c. Three phase free surface flow, EX184,

liquid film thickness, m Fluent Incorporated, Lebanon, New Hampshire.
 hold-up FLUENT technical notes, 2002d. Unsteady deformation and internal
 contact angle, ◦ circulation of a gas bubble in a zero gravity uniform flow. TN 282, Fluent
Incorporated, Lebanon, New Hampshire.
 curvature, 1/m
FLUENT 6.1 documentation, 2003. Fluent Incorporated, Lebanon, New
 molecular viscosity, kg/m s Hampshire.
 density, kg/m3 Fukano, T., Kariyasaki, A., 1993. Characteristics of gas–liquid two–phase
 surface tension, N/m flow in a capillary tube. Nuclear Engineering and Design 141, 59–68.
Heiszwolf, J.J., Kreutzer, M.T., van den Eijnden, M.G., Kapteijn, F., Moulijn,
Dimensionless numbers J.A., 2001. Gas–liquid mass transfer of aqueous Taylor flow in monoliths.
Catalysis Today 69, 51–55.
Bo Bond number (=d 2 g/) Hirt, C.W., Nichols, B.D., 1981. Volume of fluid (VOF) method for the
dynamics of free boundaries. Journal of Computational Physics 39,
Ca capillary number (=U/)
201–225.
Eo Eotvos number (=d 2 g/) Irandoust, S., Andersson, B., 1989a. Simulation of flow and mass-transfer
Fr Froude number (=U 2 /gd) in Taylor flow through a capillary. Computers Chemical Engineering 13
Re Reynolds number (=U d/) (4/5), 519–526.
We Weber number (=U 2 d/) Irandoust, S., Andersson, B., 1989b. Liquid film in Taylor flow through a
capillary. Industrial and Engineering Chemistry Research 28, 1684–1688.
Jensen, K.F., 2001. Microreaction engineering—is small better? Chemical
Subscripts and superscripts
Engineering Science 56, 293–303.
Kreutzer, M.T., Du, P., Heiszwolf, J.J., Kapteijn, F., Moulijn, J.A., 2001.
G gas Mass transfer characteristics of three-phase monolith reactors. Chemical
L liquid Engineering Science 56, 6015–6023.
S solid Kreutzer, M.T., Bakker, J.J.W., Kapteijn, F., Moulijn, J.A., Verheijen,
min minimum P.J.T., 2005a. Scaling-up multiphase monolith reactors: linking residence
max maximum time distribution and feed maldistribution using isobars. Industrial and
Engineering Chemistry Research 44 (14), 4898–4913.
exp experiment
Kreutzer, M.T., Kapteijn, F., Moulijn, J.A., Kleijn, C.R., Heiszwolf, J.J.,
* dimensionless value 2005b. Inertial and interfacial effects on pressure drop of Taylor flow in
capillaries. A.I.Ch.E. Journal 51 (9), 2428–2440.
Kreutzer, M.T., Kapteijn, F., Moulijn, J.A., Heiszwolf, J.J., 2005c. Multiphase
Acknowledgments monolith reactors: chemical reaction engineering of segmented flow in
microchannels. Chemical Engineering Science 60, 5895–5916.
The authors gratefully acknowledge the financial support Kreutzer, M.T., van der Eijnden, M.G., Kapteijn, F., Moulijn, J.A., Heiszwolf,
from Department of Energy-Industrial Technologies Program J.J., 2005d. The pressure drop experiment to determine slug lengths in
monoliths. Catalysis Today 105, 667–672.
(DOE-ITP) under Grant Nos. DE-FC36-021D14427 and DE-
Laborie, S., Cabassud, C., Durand-Bourlier, L., Laine, J.M., 1999.
FC36-03G013156. We also thank Dr. Kreutzer for providing Characterization of gas–liquid two-phase flow inside capillaries. Chemical
some raw experimental data. Engineering Science 54, 5723–5735.
Liu, H., Vandu, C.O., Krishna, R., 2005. Hydrodynamics of Taylor flow in
vertical capillaries: flow regimes, bubble rise velocity, liquid slug length,
References and pressure drop. Industrial and Engineering Chemistry Research 44 (14),
4884–4897.
Akbar, M.K., Plummer, D.A., Ghiaasiaan, S.M., 2003. On gas–liquid two- Mantle, M.D., Sederman, A.J., Gladden, L.F., 2002. Dynamic MRI
phase flow regimes in microchannels. International Journal of Multiphase visualization of two-phase flow in a ceramic monolith. A.I.Ch.E. Journal
Flow 29, 855–865. 48 (4), 909–912.
Bercic, G., Pintar, A., 1997. The role of gas bubbles and liquid slug lengths on Taha, T., Cui, Z.F., 2004. Hydrodynamics of slug flow inside capillaries.
mass transport in the Taylor flow through capillaries. Chemical Engineering Chemical Engineering Science 59, 1181–1190.
Science 52 (21/22), 3709–3719. Taha, T., Cui, Z.F., 2006a. CFD modelling of slug flow in vertical tubes.
Brackbill, J.U., Kothe, D.B., Zemach, C., 1992. A continuum method for Chemical Engineering Science 61, 676–687.
modeling surface tension. Journal of Computational Physics 100, 335–354. Taha, T., Cui, Z.F., 2006b. CFD modelling of slug flowinside square capillaries.
Bretherton, F.P., 1961. The motion of long bubbles in tubes. Journal of Fluid Chemical Engineering Science 61, 665–675.
Mechanics 10, 166–188. Thulasidas, T.C., Abraham, M.A., Cerro, R.L., 1995a. Bubble-train flow in
Chen, W.L., Twu, M.C., Pan, C., 2002. Gas–liquid two-phase flow in micro- capillaries of circular and square cross section. Chemical Engineering
channels. International Journal of Multiphase Flow 28, 1235–1247. Science 50 (2), 183–199.
Chung, P.M.-Y., Kawaji, M., 2004. The effect of channel diameter on adiabatic Thulasidas, T.C., Cerro, R.L., Abraham, M.A., 1995b. The monolith froth
two-phases flow characteristics in microchannels. International Journal of reactor: residence time modelling and analysis. Chemical Engineering
Multiphase Flow 30, 735–761. Research and Design 73, 314–319.
 D. Qian, A. Lawal / Chemical Engineering Science 61 (2006) 7609 – 7625
Thulasidas, T.C., Abraham, M.A., Cerro, R.L., 1997. Flow patterns in liquid
slugs during bubble-train flow inside capillaries. Chemical Engineering
Science 52 (17), 2947–2962.
Thulasidas, T.C., Abraham, M.A., Cerro, R.L., 1999. Dispersion during
bubble-train flow in capillaries. Chemical Engineering Science 54 (1),
61–76.
Triplett, K.A., Ghiaasiaan, S.M., Abdel-Khalik, S.I., Sadowski, D.L., 1999.
Gas–liquid two-phase flow in microchannels Part I: two-phase flow
patterns. 25, 377–394.
Van Baten, J.M., Krishna, R., 2004. CFD simulation of mass transfer from
Taylor bubbles rising in circular capillaries. Chemical Engineering Science
59, 2535–2545.
7625
Van Baten, J.M., Krishna, R., 2005. CFD simulation of wall mass transfer
for Taylor flow in circular capillaries. Chemical Engineering Science 60,
1117–1126.
Vandu, C.O., Liu, H., Krishna, R., 2004. Taylor bubble rise in circular and
square capillaries. University of Amsterdam, Amsterdam, The Netherlands,
the 20th of July 2004 http://ct-cr4.chem.uva.nl/singlecapillary/.
Vandu, C.O., Liu, H., Krishna, R., 2005. Mass transfer from Taylor
bubbles rising in single capillaries. Chemical Engineering Science 60,
6430–6437.
Yang, C.-Y., Shieh, C.-C., 2001. Flow pattern of air–water and two-phase
R-134a in small circular tubes. International Journal of Multiphase Flow
27, 1163–1177.