Article

pubs.acs.org/IECR

OpenFOAM Computational Fluid Dynamic Simulations of Two-Phase

Flow and Mass Transfer in an Advanced-Flow Reactor
Mara Jose Nieves-Remacha, Lu Yang, and Klavs F. Jensen*
Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, United States
*
S Supporting Information

ABSTRACT: Advanced-ow reactor (AFR) technology is an alternative to scale-up continuous ow chemistries from micro to
milli scales, while retaining mass and heat transfer performance. Here we conduct two-phase computational uid dynamic (CFD)
simulations using the open source software OpenFOAM in order to predict hydrodynamic parameters in the AFR for dierent
operating conditions. After modication and validation of the interFoam solver based on the volume-of-uid method to account
for mass transfer across immiscible interfaces, it is applied to the AFR to predict specic interfacial areas (a) and individual mass
transfer coecients (kL) to yield overall mass transfer coecients (kLa). The results are in good agreement with semiempirical
values and the surface renewal theory of Danckwerts, except at the largest ow rates for which numerical coalescence is observed.
A study of the inuence of uid properties yields the following conclusions. The contact angle is the variable that aects the ow
patterns the most (and more specically, the interfacial area); varying the contact angle can change the ow regime from bubbly
to stratied ow. Decreasing surface tension decreases droplet size, but in order to achieve large specic interfacial areas and have
a positive impact on the mass transfer process, a larger dispersed phase ow rate with increased holdup is also required. Viscosity
of the continuous phase does not have a signicant eect on mass transfer. The eect of reactor design was also seen to not be
signicant for the AFR designs and ow rates tested, with overall mass transfer coecients varying within 17%.

INTRODUCTION
Microreactor technology has been demonstrated by numerous
authors to be very useful on the laboratory scale for kinetic
studies, elucidation of reaction mechanisms, catalyst screening,
optimization of reaction conditions, synthesis of new
compounds, and multiphase chemistries. Microreactors oer
advantages over batch reactors, including fast and ecient
mixing, safer operation, more precise reaction control, and high
mass and heat transfer rates.15 However, the limitation of
microreactors is the low throughput that they provide (mg/
min), and thus the main challenge becomes how to increase the
production rate without losing mass and heat transfer eciency
so that they can be used by the pharmaceutical and ne
chemicals industry for production purposes (kg/min).
Although parallelization of microreaction units is not generally
a feasible alternative, a combination of the scale-up and scale-
out6 concepts becomes a possible solution. Here we consider
the advanced-ow reactor (AFR) technology manufactured by
Corning Inc., which has a minimum channel width of 1 mm
and can operate at ow rates in the range 10100 mL/min.
These devices have a modular structure, so that tens of units
can be easily parallelized to achieve production rates on the
order of kg/min.79
The excellent mass transfer performance of the AFR has
been demonstrated experimentally for dierent systems.10,11
However, a general simulation tool able to model two-phase
ow would allow determination of the mass transfer perform-
ance for any system and could be further used to introduce
modications into the reactor design for optimization purposes.
Eorts over the years have been directed toward the
simulation of multiphase ow with the aim of capturing the
dynamic events that are observed experimentally.12,13 Captur-
ing detachment, deformation, breakup, and coalescence of
bubbles and drops becomes a challenging task. An additional
challenge is to capture accurately capillary eects, which are
important on small scales. Although accuracy is a desired
characteristic in the simulation, in order to achieve high
accuracy very large computation times are normally required.
Over time, dierent approaches have been developed, each one
providing specic features and advantages over the others.
Among the continuum methods used to model two-phase
ow, there are two main approaches:613 the Lagrangian and
the Eulerian. The use of a moving (unstructured) mesh is
included within the rst approach with a Lagrangian movement
of the interface. The advantage of this method is that sharp
interfaces are obtained. However, strong topological changes
may cause extreme deformations of the grid cells making the
computation very complex. The front-tracking methods also fall
within the Lagrangian group. The interface position is
described by the motion of Lagrangian particles or markers in
the structured mesh. The disadvantage is the complexity of the
method. Additional markers need to be added when the
interface stretches or eliminated when the interface grows.
Other problems arise in three-dimensional computations and
merging of interfaces within one grid cell.
The second group (the Eulerian approach) includes the
level-set (LS), volume-of-uid (VOF), and phase-eld
methods.1420 The LS method uses a distance function to
describe the location of the interface.15 In the VOF method, the
Received: February 4, 2015
Revised: June 1, 2015
Accepted: June 9, 2015
Published: June 23, 2015

2015 American Chemical Society 6649 DOI: 10.1021/acs.iecr.5b00480

Ind. Eng. Chem. Res. 2015, 54, 66496659
Industrial & Engineering Chemistry Research
interface is described by an indicator function, oftentimes the
volume fraction of one of the phases. A geometric
reconstruction of the interface based on the evaluation of
uxes across cell faces usually yields better accuracy than the
more simple algorithmic approaches based on additional
compression steps. The phase eld method treats the interface
with nite thickness over which an order parameter varies
continuously.20 Here we use the volume-of-uid method and
the open source software OpenFOAM to study the inuence of
ow rates and uid properties on hydrodynamic parameters
and overall mass transfer coecients in the AFR.
There is a limited number of publications in the literature
regarding two-phase ow modeling in microchannels using
OpenFOAM. Most of the work is done using commercial
simulation software, such as Fluent, STAR-CCM, or
COMSOL. The advantage of the OpenFOAM software is
that being open source the original code is available for
consultation and modications to improve the current
performance of solvers or create new ones to satisfy specic
modeling requirements. Indeed, there are continuous eorts in
both the academic and industrial communities to broaden the
applicability of OpenFOAM to solve dierent problems.
TWO-PHASE FLOW MODELING
The OpenFOAM solver developed for modeling two-phase
ows called interFoam2123 is based on the color function
volume-of-uid (CF-VOF) method. Its implementation is
based on an Eulerian description of each phase on a xed
mesh (eq 1) and description of the interface between the two
phases using a transport equation for an indicator function (eq
2), which in OpenFOAM implementation is the local volume
fraction of one phase, . This software uses the multidimen-
sional universal limiter with explicit solution (MULES)
algorithm to solve the advection equation for the volume
fraction.
v
+ v v = P + g + 2 v
t (1)
+ (v ) = 0
t (2)
With this denition, each grid cell is described by a velocity
vector, pressure, and volume fraction. Those cells fully occupied
by phase 1 have a value of = 1, whereas cells fully occupied by
phase 2 adopt a value of = 0. Cells that contain the interface
have a value of between 0 and 1, depending on the volume
fraction of each phase. The approach of the VOF method is to
describe the entire domain with a single momentum
conservation equation with volume averaged values for the
uid properties at the interface: density (eq 3) and viscosity (eq
4).
= 1 + 2 (1 ) (3)
= 1 + 2 (1 ) (4)
The surface tension is a two-dimensional (2D) eect that in a
classical problem formulation appears in the stress boundary
condition at the interface, both in the normal and tangential
components. However, in the VOF approach the exact position
of the interface is not known and it is described by a number of
cells with nite volume. Thus, the surface tension eect (2D)
needs to be converted into a volume eect (3D) that can be
6650
Article
incorporated into the main momentum equation valid for the
entire domain (eq 5).
v
+ v v = P + g + 2 v + Fsv
t (5)
Figure 1. Schematic of the continuum surface force (CSF) approach.
The surface tension two-dimensional eect is converted into a three-
dimensional approach to be inserted into the main momentum
conservation equation.
The OpenFOAM solver designed to solve two-phase ows of
incompressible uids with no phase change called interFoam
follows the approach described by Brackbill et al.24 for the
surface tension volume force. The continuum surface force
(CSF) interprets surface tension as a continuous, three-
dimensional eect across an interface. For a volume force
Fsv(x) that provides the correct surface tension force per unit
interfacial area, Fsa(xs), for an interface thickness approaching
zero, and being zero outside of the interface:
lim
h0
V Fsv(x)dV = A Fsa(xs)dA (6)
where the area integral over the portion A of the interface
within the small volume of integration V. The original
discontinuous indicator function is dened by c(x) using eq 7:
phase 1 c(x) = c1
c(x) = interface c(x) = (c1 + c 2)/2
phase 2
c(x) = c 2 (7)
Within the thickness of the interface (h), called transition
region, an indicator function that varies smoothly from phase 1
to phase 2 is dened, c(x). For a very small interface thickness,
the expression given by eq 8 can be used to approximate the
surface tension eect as a body force that can be inserted in the
momentum conservation equation.
c ( x)
Fsv(x) = (x)
c 2 c1 (8)
An additional advection equation for the transport of is
dened (eq 9) and corrected with a compression term that
compensates for the numerical diusion of the interface.
+ (v ) + (vr ((1 )) = 0
t (9)
where vr is the relative velocity between phases and is calculated
using eq 10:
DOI: 10.1021/acs.iecr.5b00480
Ind. Eng. Chem. Res. 2015, 54, 66496659
Industrial & Engineering Chemistry Research
| |
C | |
vr = n f min r , max
|Sf | |Sf |
(10)
This compression factor is only active at the interface and
vanishes at those cells that are fully occupied by phase 1 or
phase 2. The amount of compression can be adjusted by
specifying the cAlpha value, Cr (usually ranges from 0 to 4) in
the system/fvSolution le.
The local surface curvature () is calculated using eq 11,
where n is the normal vector perpendicular to the interface,
calculated with eq 12.
= n
n = /||
(11)
(12)
The wall adhesion force is considered by specifying the three-
phase contact angle:
n = n w cos(w) + tw sin(w)
(13)
where n and tw are the normal and tangential vectors to the
wall, respectively. In the subsequent simulations, the contact
angle is considered to be constant and independent of the
velocity and the direction of the contact line movement.
The publications that report the use of the interFoam solver
for two-phase incompressible ow include the work performed
by Deshpande et al.,25 Hoang et al.,26 Raees et al.,27 and Raeini
et al.28 Deshpande et al. focused on a detail evaluation of the
performance of the interFoam solver using a variety of test
validation cases trying to cover a wide range of applications: (a)
pure advection; (b) dynamics for inertia dominated ows; (c)
dynamics for surface tension dominated ows. It is concluded
in this work that the performance of interFoam is comparable
to other VOF existing algorithms, but the geometric
reconstruction schemes, such as the one available in Fluent,
perform better. It was also shown that for inertia-dominated
ows with large density ratios interFoam is able to perform
with excellent agreement with experimental and analytical
values. In the case of surface tension dominated ows,
interFoam guarantees a consistent formulation of pressure
and surface tension. However, 10% disagreement for the
computation of curvature with respect to analytical values is
observed. Further improvements need to be done in this
regard.
A common problem when simulating two-phase ow at small
scales are articial velocities appearing at the interface. This
problem has been reported by many authors using VOF
methods2939 in their simulations. This is attributed to an
imbalance of forces caused by the introduction of the surface
tension body force as approximation for the surface eects into
the single-conservation momentum equation and/or incorrect
method for the computation of the interface curvature. This
eect is especially important at low or zero velocities and strong
capillary forces. The articial velocities created are often called
spurious or parasitic currents. To overcome the problem of
spurious currents in interFoam, Deshpande et al. suggest a
criterion based on the analysis of Galusinski and Vigneaux35 for
the time step, which depends on the uid properties (, , ).
However, other authors introduce modications into the main
solver. For instance, Hoang et al.39 use a Laplacian smoother
for the volume fraction and recommend a selection of
parameters, such as the compression factor Cr, for best
performance.
6651
Article
n
f = 1 f Sf
p = n
f = 1 Sf (14)
Raeini et al.28 follow a more complicated procedure,
incorporating a semisharp surface model (SSF) to reduce the
articial diusion of the interface, and two new ltering
methods (FSF) to correct for capillary forces and reduce
spurious currents. This solver is implemented after reformula-
tion of the NavierStokes equation by dening a dynamic
pressure and a body force that lumps together the eects of
gravity, capillary forces, and capillary pressure gradients.
MASS TRANSFER MODELING
The original interFoam solver has been modied to model mass
transfer across immiscible interfaces. Investigations about
modeling mass transfer using the VOF method are very few,
and the rst ones had the limitation of being applicable only for
constant concentrations across the interface. Ohta and Suzuki40
rst studied mass transfer from a drop in a solvent extraction
process. Sato et al.41 simulated in 3D the mass transfer process
of liquid carbon dioxide single rising droplets in the continuous
phase water using a front capturing method, Davidson and
Rudman42 developed a method to simulate in 2D heat and
mass transfer between interfaces, and only for continuous
equilibrium concentrations between phases; but it was Bothe et
al.43 who in 2003 started modeling mass transfer of oxygen
from single bubbles and bubble chains rising in aqueous
solutions using the VOF method and considering equilibrium
of concentrations across interfaces based on Henrys law. In
2006, Onea et al.44 developed a method based on the VOF
approach also based on Henrys law to simulate conjugate mass
transfer of dilute species accounting for the concentration jump
and applied it to investigate qualitative investigation of mass
transfer in upward bubble train ow within square and
rectangular channels is performed.
Another approach is that followed by Haroun et al.45,46 that
combines VOF approach with transport of species across
interfaces with application to stable liquid lms and structured
packing. More recently, Marschall et al.47 developed an
algorithm to solve for mass transfer of species across interfaces
in free-surface ows using OpenFOAM. To capture the
concentration jump at the interface, both approaches consider
that the solubility is determined by Henrys law with a constant
coecient. Here we present the implementation of Haroun et
al. approach in OpenFOAM and compare it with the
implementation of Marschall et al. using simple validation
cases (in Supporting Information). This solver is later applied
to the advanced-ow reactor for determination of mass transfer
coecients.
The HarounTransportFoam solver is based on the interFoam
solver with the additional species conservation equation, as
given by eq 15.
Ci
+ (vC
i) = (DiCi)
t (15)
The boundary conditions that need to be satised at the
interface are described by eqs 16 and 17:
He = CiL /CiG (16)
DiL CiL = DiGCiG (17)
DOI: 10.1021/acs.iecr.5b00480
Ind. Eng. Chem. Res. 2015, 54, 66496659
Industrial & Engineering Chemistry Research
The approach followed by Haroun et al. considers one-uid
variables in the interface region for concentration and diusion
coecients dened by eqs 18 and 19 and a concentration ux
dened by eq 20.
Ci = CiL + (1 )CiG (18)
Di = DiL + (1 )DiG (19)
JiL = JiG (20)
The concentration ux expression eq 21 is dierent from the
conventional one after the denition of these variables at the
interface. The second term Di(CLi CGi ) is zero within the
domain of each phase separately (where = 0), and is
dierent from zero only at the interface (where 0).
Ji = Di(CiL + (1 )CiG)
= DiCi + Di(CiL CiG) (21)
Reformulating in terms of the Henry constant, eq 22 is
obtained. The second term vanishes as well for a Henry
constant equal to 1.
Ci(1 He)
Ji = DiCi + Di
+ He(1 ) (22)
An additional study showed that the harmonic average of the
diusion coecient (eq 23) behaved better than the conven-
tional linear average based on the volume phase fraction .
DiL DiG
Di =
DiL + (1 )DiG (23)
Details on the validation of the implemented HarounTrans-
portFoam solver for mass transfer modeling are included in the
Supporting Information.
APPLICATION TO THE ADVANCED-FLOW
REACTOR
The motivation for modeling two-phase ow in the advanced-
ow reactor (AFR) is to be able to predict overall mass transfer
coecients (k La) for dierent uid systems prior to
experimentation and evaluate the mass transfer eciency in
the AFR. The overall mass transfer coecient has two dierent
contributions: (a) the individual mass transfer coecient, kL,
and (b) the specic interfacial area, a. Although there are
theoretical approximations to estimate kL (such as penetration
theory,48 surface renewal theory49), the specic interfacial area
remains frequently unknown. For devices such as the AFR
where the specic interfacial areas highly depend on the
operating conditions and cannot be known a priori, a CFD tool
such as OpenFOAM can be useful as a preliminary step to
predict interfacial areas. The interFoam solver itself can predict
this variable. However, incorporation of mass transfer of species
across interfaces allows also the prediction of mass transfer
coecients, kL. Furthermore, incorporation of reaction is
possible, which increases the potential application of the
CFD tool to predict reactor performance for actual reactions
coupled with mass transfer occurring in any multiphase reactor.
In this work, the results of applying the HarounTransport-
Foam solver in the AFR to predict ow patterns, average drop
size, specic interfacial areas, dispersed phase holdup, and mass
transfer coecients are presented and compared with published
6652
Article
experimental data.10,11 These experiments were performed for
hexane/water at dierent ow rates in the AFR ranging from 10
to 80 mL/min of each phase at atmospheric conditions. A
detailed study on the drop size distributions, specic surface
areas, dispersed phase holdups, and overall mass transfer
coecients were obtained experimentally. It was demonstrated
that kLa values were on the order of 1 s1 for both hexane/
water and carbon dioxide/water systems and depend on the
ow rates of each phase, being larger for larger phase ow rates.
Due to constraints in the computation time, only two heart
cells of the AFR are modeled. The mesh used in the simulations
is structured and three-dimensional, composed of 251 320 cells
and 10 cells in the z-direction. To save computation time, each
portion of the AFR where each uid ows separately has been
modeled using the simpleFoam solver, which solves incom-
pressible single-phase steady-state ows. Then, the output of
the CFD simulation for single-phase ow for each phase is used
as an input for the simulation of two-phase ow.
The estimation of the overall mass transfer coecient is
performed by computation of two variables: kL and a. Both
variables depend on the uid properties of the system of
interest and the operating conditions (temperature, pressure,
ow rates). In our simulations, temperature and pressure
remain constant and at ambient conditions. The individual
mass transfer coecient, kL, is computed from the CFD results
according to eq 24, from the ux of concentration of
component across the interface. Notice here that the
concentration of ux given by Ficks law directly does not
account for the additional term (eq 25) that is needed to
correct the ux by the presence of jump concentration.
(DC + )n
kL,local =
C L,local (24)
Cj(1 Hej)
= Dj
+ (1 )Hej (25)
In addition, calculation of the individual mass transfer
coecient based on the surface renewal theory48 is performed
and the results are compared with the obtained through CFD.
The surface renewal theory is an extension of the penetration
theory and considers that the liquid elements do not stay at the
interface the same time (there is a distribution). The inverse
residence time of the liquid at the interface gives an estimation
of the rate of surface renewal (s). The individual mass transfer
coecient (kL) is proportional to the square root of the
diusion coecient times the rate of surface renewal:
kL = DL s (26)
The specic interfacial area is extracted from the ow
patterns obtained in each simulation. A Matlab script was
written to measure the contour of the droplets. Depending on
the droplet size with respect to the reactor height, the specic
area was computed as a sphere or as a planar droplet
sandwiched between the two reactor walls. The results were
computed from the transient simulation starting at time zero
until the steady state was reached. Additionally, the bulk
concentration of species in the dispersed phase is monitored
over time. Comparison with experimental observations10,11 is
performed hereafter by comparing average drop size, hexane
holdup, specic interfacial areas, and mass transfer coecients.
In the following sections, the eect of ow rates, uid
properties, and reactor design on mass transfer and hydro-
DOI: 10.1021/acs.iecr.5b00480
Ind. Eng. Chem. Res. 2015, 54, 66496659
Industrial & Engineering Chemistry Research
dynamic properties are presented. Details of the simulation
results are presented only for the rst variable studied (eect of
water ow rate) in the form of graphs for the dierent
measured variables over time. For the following cases, only
steady-state values are reported here. Additional details are
included in the Supporting Information.
Eect of Flow Rates. The dierent ow patterns obtained
at constant hexane ow rate and increasing water ow rate are
shown in Figures 2 and 3. The transient evolution of the
Figure 2. Flow patterns for hexane/water system using Haroun-
TransportFoam (OpenFOAM) for dierent ow rates: (a) Qw = 10
mL/min, Qh = 10 mL/min; (b) Qw = 20 mL/min, Qh = 10 mL/min;
(c) Qw = 40 mL/min, Qh = 10 mL/min; (d) Qw = 80 mL/min, Qh =
10 mL/min; cross section view at middle plane perpendicular to z
coordinate. Legend: = 0 (hexane); = 1 (water).
dierent variables with time is shown in Figure 3. It is clear that
the main eect of water ow rate is to decrease the drop size,
which is also observed experimentally in the AFR.10,11
The mass transfer coecients shown in Figure 3e,f have been
computed based on the surface renewal theory, using the
residence time of the dispersed phase as the contact time and
the CFD hexane holdup. It is seen that the values of kLa
decrease with increasing water ow rate. Here two factors aect
kLa in dierent directions: (a) increasing the velocity of the
continuous phase increases the shear rate and slip velocity, thus
increasing the individual mass transfer coecient kL; (b) the
specic interfacial area decreases with water ow rate. Although
smaller droplet sizes increase the specic interfacial area per
droplet, higher hexane holdups are needed for an overall
increase in interfacial surface. In this particular case, the highest
water ow rate does not yield the highest mass transfer
coecient for this reason. Increasing the hexane ow rate from
10 mL/min would denitely increase the hexane holdup and
thus, the specic interfacial area while keeping the droplet size
small thanks to the high shear rates.
Table 1 includes a comparison between the holdup
computed from the CFD simulations and the experimental
holdup in the rst two hearts of the AFR. There is a good
agreement between the simulations and the experiments,
considering that although the CFD results correspond to
steady-state averaged values, the experimental results have been
measured at steady-state from a single image using visualization
techniques. To have an estimate of the mass transfer eciency
in the AFR, the concentration of acetone was measured in the
dispersed phase over time. It is observed that the concentration
of acetone at steady-state is lower than the maximum
concentration that could be achieved if the mass transfer
process was perfect. This yields values for the mass transfer
eciency that range from 68 to 84% for hexane ow rates of 10
mL/min and considering only two heart cells. The mass
transfer eciency may increase by considering all heart cells
6653
Article
composing the reactor, since the contact time between phases
also increases. The eciency can be increased by increasing the
hexane ow rate as both kL and a would increase. This is in
agreement with the recommendations by Corning to operate
the AFR at high ow rates (200 mL/min).
A comparison of the CFD individual mass transfer
coecients and specic interfacial areas with experimental
results is included in Table 2. As it is seen from the results, the
kL values obtained by the dierent methods (surface renewal
theory and from the concentration gradient) are in good
agreement with the experimental results also based on surface
renewal theory. The largest dierences are encountered in the
specic interfacial areas. In both simulations and experimental
results, it is found that the specic interfacial area decreases
when increasing the continuous phase ow rate due to the
decrease in hexane holdup, despite of the decrease in droplet
size. This shows the importance of having a correct VOF
method that is able to capture accurately the formation of
droplets and bubbles.
Table 3 shows that the kLa values estimated using the surface
renewal theory of Danckwerts and from the concentration
gradient based on the CFD results are in good agreement, with
dierences below 25%. They are also in fair agreement with the
semiempirical values with some discrepancies, especially at the
largest ow rates, mainly due to dierences in the prediction of
the specic interfacial area.
The eect of total ow rate for constant water to hexane ratio
is shown in Figure 4. The details on the transient simulation
and evolution of the dierent properties over time are shown in
the Supporting Information. Here only values at steady-state
are shown. While in the experimental work already a breakup is
observed in the rst heart with formation of small drops,
breakup hardly occurs in CFD simulations at 40 and 80 mL/
min of each phase. At large ow rates, higher mesh resolution
may be needed in order to capture all small drops formed.
Because the water-to-hexane ratio is one for these computa-
tions, the proximity of drops causes an articial coalescence
eect. The current VOF method does not prevent the
numerical coalescence of bubbles/drops that are within one
grid cell distance apart. The method requires very ne mesh
resolution in order to describe the interface sharpness
accurately and avoid coalescence eects that are not present
in reality. This is a problem common in VOF methods
normally present in liquidliquid CFD simulations and it is
certainly a topic that needs to be further studied, but its
implementation is out of the scope of this work.
Table 4 includes information about the hexane holdup
obtained experimentally and its comparison with CFD
simulations. Dierences below 20% are observed for all ow
rates. In addition, the concentrations of species in the dispersed
phase yield mass transfer eciencies in the range of 4671%.
The decrease with increasing total ow rate is due to the
decrease in specic interfacial due to articial coalescence. In
reality, we expect the mass transfer rates to increase with total
ow rates enhanced by the shear rates, shown by the larger
overall mass transfer coecients obtained experimentally.
Table 5 presents a comparison of CFD results and
experiments for specic interfacial areas and individual mass
transfer coecients at dierent total ow rates. Again, there is a
good agreement among individual mass transfer coecients
computed by surface renewal theory and from the CFD
concentration gradient. In any case, the largest discrepancies are
observed at 80 and 160 mL/min total ow rate, as expected by
DOI: 10.1021/acs.iecr.5b00480
Ind. Eng. Chem. Res. 2015, 54, 66496659
Industrial & Engineering Chemistry Research Article

Figure 3. Eect of water ow rate at 10 mL/min hexane ow rate on (a) concentration in hexane; (b) drop size; (c) specic interfacial area; (d)
hexane holdup; (e) individual mass transfer coecient; (f) overall mass transfer coecient. Water ow rate (mL/min): , 10; - - -, 20; , 40; red
, 80.

Table 1. Mass Transfer Eciency, Concentration, and Hexane Holdup for Dierent Water Flow Rates
hexane (mL/min) water (mL/min) experimental holdup () CFD holdup () CFD Cfinal (mol/m3) CFD Cmax (mol/m3) eciency (%)
10 10 0.36 0.43 0.39 0.55 70.9
10 20 0.37 0.36 0.44 0.65 68.3
10 40 0.22 0.22 0.65 0.78 83.3
10 80 0.14 0.09 0.65 0.92 70.7

Table 2. Comparison of Experimental and CFD kL and a for Dierent Water Flow Rates
hexane (mL/min) water (mL/min) aCFD (m2/m3) kL CFD Ia 104 (m/s) kL CFD IIb 104 (m/s) aexp (m2/m3) kL exp 104 (m/s)
10 10 1250 1.11 0.84 1028 0.91
10 20 1120 1.07 0.91 1260 1.20
10 40 940 1.11 1.15 1001 1.16
10 80 450 1.75 1.96 930 1.43
a
kL CFD I: from concentration gradient. bkL CFD II: from surface renewal theory.

the poor prediction of ow patterns with the current mesh
resolution and the problem of numerical coalescence observed.
Overall mass transfer coecients obtained by three dierent
approaches are shown in Table 6. The largest discrepancies are
6654
seen for these simulations for which the hexane/water ratio is
unity, especially at the largest ow rates. The discrepancies are
explained from the diculty in predicting the correct ow
patterns when two cells corresponding to two dierent droplets
DOI: 10.1021/acs.iecr.5b00480
Ind. Eng. Chem. Res. 2015, 54, 66496659
Industrial & Engineering Chemistry Research Article

Table 3. Comparison of Experimental and CFD kLa for about 0.3 mm, which, together with a decrease in hexane
Dierent Water Flow Rates holdup, decreases the specic interfacial area from 805 to 700
m2/m3. This makes the nal concentration achieved in the case
hexane water kLaCFD I kLaCFD II kLaexp
(mL/min) (mL/min) (1/s) (1/s) (1/s) of highest viscosity to be smaller than in the original case (0.55
10 10 0.139 0.105 0.094
versus 0.65 mol/m3), which represents 67% eciency in mass
10 20 0.119 0.102 0.152
transfer with respect to the maximum concentration possible.
10 40 0.105 0.093 0.116
In consequence, the overall mass transfer coecient is also
10 80 0.092 0.070 0.133 slightly smaller for the higher uid viscosity. However, the
dierence is not signicant for such a change in viscosity.
Surface Tension. Surface tension is another variable
important in two-phase ow that aects ow patterns. As it
can be observed from Figure 6, the main eect of surface
tension is to decrease the average drop size for decreasing
surface tension. However, the decrease in drop size is
accompanied by a decrease in hexane hold up as well. This
aects negatively to the specic interfacial area, which also
decreases with decreasing surface area.
The eect of contact angle on the ow patterns for 40 mL/
min water and 10 mL/min hexane is shown in Figure 7. It is
observed that the shape and size of the droplets are completely
determined by the contact angle. For 75, water partially wets
the reactor walls, while at a contact angle of 0, water wets
Figure 4. Flow patterns for hexane/water system using Haroun- completely the wall. When comparing the images from the
TransportFoam (OpenFOAM) for dierent total ow rates and a
constant water/hexane ratio: (a) Qw = 10 mL/min, Qh = 10 mL/min;
visualization experiments and the simulation, it is seen that
(b) Qw = 40 mL/min, Qh = 40 mL/min; (c) Qw = 80 mL/min, Qh = contact angles of 0 provide droplet shapes and sizes that are in
80 mL/min; cross section view at middle plane perpendicular to z agreement with the experimental results, as would be expected
coordinate. since the AFR material is known to be hydrophilic.
The contact angle completely changes the ow patterns, as
are very close to each other, enhancing the articial coalescence opposed to other variables. In other simulations it was
phenomenon, aecting the interfacial area. demonstrated that the eect of viscosity or surface tension
Eect of Fluid Properties. Being able to predict the was mainly on the droplet size, but the main characteristics of
reactor behavior with a computational tool that can avoid or the ow patterns remain unchanged. Although kL values remain
reduce the need of performing experiments is especially approximately the same, the main eect of the contact angle is
important when working with dangerous or very expensive on the specic interfacial area. The largest kLa values are
materials. Simulations can help us decide what materials are obtained for a contact angle of 0 (water wets the walls totally),
more appropriate if dierent alternatives are available depend- whereas in the two other cases, due to the no complete
ing on the specic requirements of the system. In previous wettability of the walls, the interfacial area is smaller and the
sections, it was concluded that, for xed material properties, the overall mass transfer coecients are lower. In conclusion, for
inlet phase ow rates determined the drop size distribution, better mass transfer eciency, the best scenario is to operate in
specic interfacial areas, and mass transfer coecients. Here a bubbly ow with smaller bubbles and very high holdup. This is
comparison between simulations for dierent uid properties is achieved with a hydrophilic phase wetting the walls of the
presented, based on ow patterns, specic interfacial areas, and reactor, high continuous phase ow rate to decrease the overall
mass transfer coecients. droplet size, and high dispersed phase ow rate to increase
Viscosity. It was already shown in the CFD simulation of holdup.
two-phase ow in T-junctions that viscosity does not have a Eect of Reactor Design. Corning Inc. provides AFR
strong eect on the slug size.50 The eect, however, may be designs that have slight modications within the heart cell
dierent in the AFR, and therefore is studied here. Here we (Figure 8). The original AFR design which has two obstacles
present the results for constant surface tension, contact angle, (dot design or Gen I), a second design with only one curved
and density of the uids, but variable viscosity of the obstacle (no dot design or Gen II), and a third design (low-
continuous phase (water). The results of the ow patterns ow reactor or LFR design) with a single obstacle that
are shown in Figure 5. It can be observed that increasing the removes the zone between the two posts in the original design.
viscosity of the continuous phase aects negatively on the mass The eect of reactor design on ow patterns was studied for
transfer process. First, the average drop size is decreased by 40 mL/min water and 10 mL/min hexane and is shown in

Table 4. Mass Transfer Eciency, Concentration, and Hexane Holdup for Dierent Total Flow Rates at Constant Water/
Hexane Ratio

hexane (mL/min) water (mL/min) experimental holdup () CFD holdup () CFD Cfinal (mol/m3) CFD Cmax (mol/m3) eciency (%)
10 10 0.36 0.43 0.40 0.56 71.4
20 20 0.37 0.42 0.38 0.56 67.9
40 40 0.36 0.32 0.36 0.69 52.2
80 80 0.31 0.30 0.32 0.69 46.4

6655 DOI: 10.1021/acs.iecr.5b00480

Ind. Eng. Chem. Res. 2015, 54, 66496659
Industrial & Engineering Chemistry Research Article

Table 5. Comparison of Experimental and CFD kL and a for Dierent Total Flow Rates
hexane (mL/min) water (mL/min) aCFD (m2/m3) kL CFD Ia 104 (m/s) kL CFD IIb 104 (m/s) aexp (m2/m3) kL exp 104 (m/s)
10 10 1250 1.10 0.84 1028 0.91
20 20 1230 1.09 1.19 1320 1.25
40 40 1170 1.22 1.87 1135 1.81
80 80 850 1.32 2.78 1640 2.21
a
kL CFD I: from concentration gradient. bkL CFD II: from surface renewal theory.

Table 6. Comparison of Experimental and CFD kLa for designs. The largest drop size is obtained in the AFR, followed
Unity Flow Rate Ratio by the no dot and the LFR-like designs. While the AFR and
the no dot designs have similar specic interfacial areas and
hexane water kLa CFD I kLa CFD II kLa exp
(mL/min) (mL/min) (1/s) (1/s) (1/s) larger than the LFR design, all overall mass transfer
coecients fall within the range 0.0750.09 s1, being in the
10 10 0.138 0.105 0.094
upper limit for the AFR design.

20 20 0.134 0.146 0.165
40 40 0.142 0.219 0.205
80 80 0.112 0.236 0.364 CONCLUSIONS
A modied version of the interFoam solver has been developed
based on the approaches of Haroun and Marschall incorporat-
ing mass transfer of species across immiscible interfaces. With
this solver, ow patterns for hexane/water have been predicted
and are in good agreement with the experimental results at low
phase ow rates. Larger discrepancies were encountered for
ow rates above 40 mL/min due to articial numerical
coalescence observed in the simulations. This eect was more
important in liquid/liquid than in gas/liquid systems, where the
degree of coalescence in experiments is signicantly large.
Although rening the mesh could potentially solve this
Figure 5. Eect of viscosity on ow patterns at 40 mL/min water and problem, the increase in computation time limits this
10 mL/min hexane, for (a) / = 105 m2/s; (b) / = 106 m2/s; possibility. An alternative solution is to develop an algorithm
cross section view at middle plane perpendicular to z coordinate. that is able to prevent numerical coalescence while keeping
Legend: = 0 (hexane); = 1 (water). reasonable computation times.
Despite of the mentioned diculties encountered, the
modied solver is able to predict with reasonable accuracy
the specic interfacial areas, individual mass transfer coef-
cients, and overall mass transfer coecients (kLa) in the AFR,
being in good agreement with semi empirical results based on
the surface renewal theory of Danckwerts and visualization
experiments. On the basis of these results, a study of the
inuence of uid properties on the mass transfer eciency was
performed. It was concluded that the variable that aected the
ow patterns the most (and more specially, the specic
Figure 6. Eect of surface tension on ow patterns at 40 mL/min interfacial area) was the contact angle: by changing the wetting
water and 10 mL/min hexane, for (N/m): (a) 0.052; (b) 0.043; (c) properties of the material, the ow regime can change from
0.035; (d) 0.0068. Legend: = 0 (hexane); = 1 (water). bubbly to stratied. Another important variable to consider was
the interfacial tension between phases. It was seen that
decreasing surface tension decreased droplet size. However,
in order to achieve large specic interfacial areas and have a
positive impact on the mass transfer process, an increase in the
dispersed phase ow rate to increase holdup was required. On
the other hand, the viscosity of the continuous phase did not
have a signicant eect on mass transfer. The eect of reactor
design was also not signicant for the designs and ow rates
tested. Although slight dierences were observed in the ow
characteristics, the overall mass transfer coecients were within
17% dierence.
Figure 7. Eect of contact angle on ow patterns at 40 mL/min water All the studies performed in this work were based on two
and 10 mL/min hexane, for (a) = 0; (b) = 75; (c) 150. Legend: heart cells. From the visualization experiments performed for
= 0 (hexane); = 1 (water).
hexane/water and carbon dioxide/water, it is known that the
droplet/bubble size decreases from the inlet toward the outlet
Figure 9. The main ow pattern remains unchanged (bubbly of the reactor. Thus, a larger interfacial area is expected
ow type), but the droplet size is dierent among the dierent downstream as a consequence of smaller droplet size, and
6656 DOI: 10.1021/acs.iecr.5b00480
Ind. Eng. Chem. Res. 2015, 54, 66496659
Industrial & Engineering Chemistry Research Article

Figure 8. Geometry for three AFR designs studied.

Fsv = volume force (N)

h = interface thickness (m)
He = Henry constant ()
Ji = ux of species concentration (mol m2 s1)
kL = individual mass transfer coecient (m s1)
kLa = overall mass transfer coecient (s1)
n = normal vector to interface ()
nw = normal vector to wall ()
P = pressure (Pa)
Qi = ow rate of phase i (mL min)
Figure 9. Eect of contact angle at 40 mL/min water and 10 mL/min t = time (s)
hexane on ow patterns; cross section view at middle plane tw = tangential vector to wall ()
perpendicular to z coordinate. Legend: = 0 (hexane); = 1 (water). v = velocity vector (m s1)
vr = relative velocity vector between phases (m s1)
therefore, larger values for kLa assuming that kL remains V = volume (m3)
constant along the reactor ow path.

Greek Letters
= volume fraction ()
ASSOCIATED CONTENT
p = smoothed volume fraction ()
*
S Supporting Information
= density (kg/m3)
Formulation of mass transfer of species in the OpenFOAM = curvature (1/m)
framework, mass transfer solver validation, and dynamic w = contact angle (deg)
simulations of mass transfer using OpenFOAM. The = mass ux (mol m2 s1)
Supporting Information is available free of charge on the = surface tension (N/m)
ACS Publications website at DOI: 10.1021/acs.iecr.5b00480. = viscosity (Pa s)

AUTHOR INFORMATION
Corresponding Author
= additional ux concentration term (mol m2 s1)
Superscripts
*K. F. Jensen. E-mail: kfjensen@mit.edu. L = liquid phase
G = gas phase
Present Address
Subscripts
The Dow Chemical Company, 2301 North Brazosport Blvd.
Bldg B-1603, Freeport, TX 77541 1 = phase 1
Notes 2 = phase 2
The authors declare no competing nancial interest. 12 = interface

f = face of cell
ACKNOWLEDGMENTS h = hexane
The authors acknowledge the Novartis MIT Center for w = water
Continuous Manufacturing for the nancial support oered
to conduct this research. REFERENCES

NOTATION
A = area (m2)
(1) Ehrfeld, W.; Hessel, V.; Haverkamp, V. Microreactors. Institute
fur Mikrotechnik Mainz GmbH. Ullmanns Encyclopedia of Industrial
Chemistry; Wiley-VCH Verlag GmbH & Co. KGaA: Weinheim,
a = specic interfacial area (m2 m3) Germany, 2012.
c(x) = indicator function () (2) Sergey, M.; Nordon, A.; Littejohn, D.; Wiles, C.; Watts, P.;
c(x) = smoothed indicator function () Dallin, P.; Girking, J. M. Improved method for kinetic studies in
Ci = concentration of i (mol m3) microreactors using flow manipulation and noninvasive Raman
Cmax = maximum concentration (mol m3) spectrometry. J. Am. Chem. Soc. 2011, 133 (10), 36013608.
Cr = adjusting compression factor () (3) Kashid, M. N.; Kiwi-Minsker, L. Microstructured reactors for
Di = diusion coecient of species i (m2 s1) multiphase reactions: State of the art. Ind. Eng. Chem. Res. 2009, 48,
Fsa = surface force (N) 64656485.

6657 DOI: 10.1021/acs.iecr.5b00480

Ind. Eng. Chem. Res. 2015, 54, 66496659
Industrial & Engineering Chemistry Research
(4) Lo we, H.; Ehrfeld, W. State-of-the-art in microreaction
technology: Concepts, manufacturing and applications. Electrochim.
Acta 1999, 44 (2122), 36793689.
(5) Ehrfeld, W.; Hessel, V.; Lowe, H. Microreactors: New Technology
for Modern Chemistry. State of the Art of Microreaction Technology;
Wiley-VCH Verlag GmbH & Co. KGaA: Weinheim, Germany, 2000.
ISBN-10: 3527295909.
(6) Amador, C.; Gavriilidis, A.; Angeli, P. Flow distribution in
different microreactor scale-out geometries and the effect of
manufacturing tolerances and channel blockage. Chem. Eng. J. 2004,
101 (13), 379390.
(7) Vizza, A. Corning Advanced-Flow Reactors: Engineered for
seamless scale-up CPAC/Atochemis Rome Workshop, Rome, Italy,
March 2527, 2013; Corning Reactor Technologies, Corning
European Technology Center: Avon, France. http://cpac.apl.
washington.edu/les/a_vizza_afrpresentationatcpac_2013.pdf (ac-
cessed 6/2/2014).
(8) Corning S.A.S. website; Corning Advanced Flow Reactors: Avon
Cedex, France. http://www.corning.com/products_services/afr/index.
aspx (accessed May 2014).
(9) Calabrese, G. S.; Pissavini, S. From batch to continuous flow
processing in chemicals manufacturing. AIChE J. 2011, 57 (4), 828
834.
(10) Nieves-Remacha, M. J.; Kulkarni, A. A.; Jensen, K. F.
Hydrodynamics of liquidliquid dispersion in an advanced-flow
reactor. Ind. Eng. Chem. Res. 2012, 51 (50), 1625116262.
(11) Nieves-Remacha, M. J.; Kulkarni, A. A.; Jensen, K. F. Gasliquid
flow and mass transfer in an advanced-flow reactor. Ind. Eng. Chem.
Res. 2013, 52 (26), 89969010.
(12) Worner, M. Numerical modeling of multiphase flows in
microfluidics and micro process engineering: A review of methods and
applications. Microfluid. Nanofluid. 2012, 12, 841886.
(13) Burns, A.; Eickenbusch, H.; Guilbert, P.; Yin, D. Application of
coupled solver technology to CFD modelling of multiphase flows with
CFX. Chem. Ing. Tech. 2001, 73 (6), 638.
(14) Trontin, P.; Vincent, S.; Estivalezes, J.-L.; Caltagirone, J.-P.
Detailed comparisons of front-capturing methods for turbulent two-
phase flow simulations. Int. J. Numer. Methods Fluids 2008, 56, 1543
1549.
(15) Changa, Y. C.; Houa, T. Y.; Merriman, B.; Osher, S. Level set
formulation of Eulerian interface capturing methods for incompres-
sible fluid flows. J. Comput. Phys. 1996, 124 (2), 449464.
(16) Unverdi, S. O.; Tryggvason, G. A front-tracking method for
viscous, incompressible, multi-fluid flows. J. Comput. Phys. 1992, 100
(1), 2537.
(17) Anderson, D. M.; McFadden, G. B.; Wheeler, A. A. Diffuse-
interface methods in fluid mechanics. Annu. Rev. Fluid Mech. 1998, 30,
13965.
(18) Denis Gueyffier, D.; Li, J.; Nadim, A.; Scardovelli, R.; Zaleski, S.
Volume-of-fluid interface tracking with smoothed surface stress
methods for three-dimensional flows. J. Comput. Phys. 1999, 152,
423456.
(19) Olsson, E.; Kreiss, G. A conservative level set method for two
phase flow. J. Comput. Phys. 2005, 210 (1), 225246.
(20) Jacqmin, D. Calculation of two-phase NavierStokes flows
using phase-field modeling. J. Comput. Phys. 1999, 155 (1), 96127.
(21) OpenFOAM website. www.openfoam.org (accessed 03/15/
2014).
(22) Jasak, H.; Weller, H. G. Interface Tracking Capabilities of the
Inter-gamma Dierencing Scheme; Department of Mechanical Engineer-
ing, Imperial College: London, 1995.
(23) Rusche, H. CFD of dispersed two-phase ows at high phase
fractions. Ph.D. Thesis, Imperial College, London, 2002.
(24) Brackbill, J. U.; Kothe, D. B.; Zemach, C. A continuum method
for modeling surface tension. J. Comput. Phys. 1992, 100, 335354.
(25) Deshpande, S. S.; Anumolu, L.; Trujillo, M. F. Evaluating the
performance of the two-phase flow solver interFoam. Comput. Sci.
Discovery 2012, 5 (1), 014016.
6658
Article
(26) Hoang, D. A.; Van Steijn, V.; Portela, L. M.; Kreutzer, M. T.;
Kleijn, C. R.; Simos, T. E.; Anastassi, Z. Modeling of low-capillary
number segmented flows in microchannels using OpenFOAM. Am.
Inst. Phys., Conf. Proc. 2012, 1479 (1), 8689.
(27) Raees, F., Van der Heul, D. R.; Vuik, C. Evaluation of the
Interface-Capturing Algorithm of OpenFoam for the Simulation of
Incompressible Immiscible Two-Phase Flow; Reports of the Department
of Applied Mathematical Analysis; REPORT 11-07; Delft University of
Technology: Delft, The Netherlands, 2011, ISSN 1389-6520.
(28) Raeini, A. Q.; Blunt, M. J.; Bijeljic, B. Modelling two-phase flow
in porous media at the pore scale using the volume-of-fluid method. J.
Comput. Phys. 2012, 231 (17), 56535668.
(29) Meier, M.; Yadigaroglu, G.; Smith, B. L. A novel technique for
including surface tension in PLIC-VOF methods. Eur. J. Mech. B/
Fluids 2002, 21, 6173.
(30) Francois, M. M.; Cummins, S. J.; Dendy, E. D.; Kothe, D. B.;
Sicilian, J. M.; Williams, M. W. A balanced-force algorithm for
continuous and sharp interfacial surface tension models within a
volume tracking framework. J. Comput. Phys. 2006, 213, 141173.
(31) Aulisa, E.; Manservisi, S.; Scardovelli, R. A novel representation
of the surface tension force for two-phase flow with reduced spurious
currents. Comput. Methods Appl. Mech. Eng. 2006, 195, 62396257.
(32) Tong, A. Y.; Wang, Z. A numerical method for capillarity-
dominant free surface flows. J. Comput. Phys. 2007, 221, 506523.
(33) Harvie, D. J. E.; Davidson, M. R.; Rudman, M. An analysis of
parasitic current generation in Volume of Fluid simulations. Appl.
Math. Modell. 2006, 30, 10561066.
(34) Gerlach, D.; Tomar, G.; Biswas, G.; Durst, F. Comparison of
volume-of-fluid methods for surface tension-dominant two-phase
flows. Int. J. Heat Mass Transfer 2006, 49, 740754.
(35) Galusinski, C.; Vigneaux, P. On stability condition for bifluid
flows with surface tension: Application to microfluidics. J. Comput.
Phys. 2008, 227, 61406164.
(36) Liovic, P.; Francois, M.; Murray Rudman, M.; Manasseh, R.
Efficient simulation of surface tension-dominated flows through
enhanced interface geometry interrogation. J. Comput. Phys. 2010,
229, 75207544.
(37) Albadawi, A.; Donoghue, D. B.; Robinson, A. J.; Murray, D. B.;
Delaure, Y. M. C. Influence of surface tension implementation in
Volume of Fluid and coupled Volume of Fluid with Level Set methods
for bubble growth and detachment. Int. J. Multiphase Flow 2013, 53,
1128.
(38) Renardy, Y.; Renardy, M. PROST: A parabolic reconstruction of
surface tension for the Volume-of-Fluid method. J. Comput. Phys.
2002, 183, 400421.
(39) Hoang, D. A.; van Steijn, V.; Portela, L. M.; Kreutzer, M. T.;
Kleijn, C. R. Benchmark numerical simulations of segmented two-
phase flows in microchannels using the Volume of Fluid method.
Comput. Fluids 2013, 86, 2836.
(40) Ohta, M.; Suzuki, M. Numerical analysis of mass transfer from a
free motion drop in a solvent extraction process. lvent Extr. Res. Dev.,
Jpn. 1996, 3, 138149.
(41) Sato, T.; Jung, R.-T.; Abe, S. Direct simulation of droplet flow
with mass transfer at interface. Trans. ASME 2000, 122, 510536.
(42) Davidson, M. R.; Rudman, M. Volume of fluid calculation of
heat or mass transfer across deforming interfaces in two-fluid flow.
Numerical Heat Transfer, Part B 2002, 41 (34), 291308.
(43) Bothe, D.; Koebe, M.; Wielage, K.; Warnecke, H.-J. VOF-
simulations of mass transfer from single bubbles and bubble chains
rising in aqueous solutions. ASME/JSME 4th Joint Fluids Summer
Engineering Conference, Honolulu, Hawai, 2003; Paper No.
FEDSM2003-45155, Vol. 2, pp 423429.
(44) Onea, A.; Worner, M.; Cacucib, D. G. A qualitative
computational study of mass transfer in upward bubble train flow
through square and rectangular mini-channels. Chem. Eng. Sci. 2009,
64 (7), 14161435.
(45) Haroun, Y.; Legendre, D.; Raynal, L. Volume of fluid method
for interfacial reactive mass transfer: Application to stable liquid film.
Chem. Eng. Sci. 2010, 65, 28962909.
DOI: 10.1021/acs.iecr.5b00480
Ind. Eng. Chem. Res. 2015, 54, 66496659
Industrial & Engineering Chemistry Research Article

(46) Haroun, Y.; Raynal, L.; Legendre, D. Mass transfer and liquid
hold-up determination in structured packing by CFD. Chem. Eng. Sci.
2012, 75, 342348.
(47) Marschall, H.; Hinterberger, K.; Schuler, C.; Habla, F.;
Hinrichsen, O. Numerical simulation of species transfer across fluid
interfaces in free-surface flows using OpenFOAM. Chem. Eng. Sci.
2012, 78, 111127.
(48) Higbie, R. The rate of absorption of a pure gas into a still liquid
during a short time of exposure. Trans. AIChE 1935, 31, 365389.
(49) Danckwerts, P. V. Significance of liquid-film coefficients in gas-
absorption. Ind. Eng. Chem. 1951, 43 (6), 14601467.
(50) Qian, D.; Lawal, A. Numerical study on gas and liquid slugs for
Taylor flow in a T-junction microchannel. Chem. Eng. Sci. 2006, 61,
76097625.

6659 DOI: 10.1021/acs.iecr.5b00480

Ind. Eng. Chem. Res. 2015, 54, 66496659