Chemical Engineering Science 101 (2013) 69–80

Contents lists available at ScienceDirect

Chemical Engineering Science

journal homepage: www.elsevier.com/locate/ces

Mass transfer characteristics of gas–liquid absorption during

Taylor flow in mini/microchannel reactors
Harish Ganapathy a,b, Ebrahim Al-Hajri b, Michael Ohadi a,n
a
Smart and Small Thermal Systems Laboratory, Center for Environmental Energy Engineering, University of Maryland, College Park, MD 20742, USA
b
Department of Mechanical Engineering, The Petroleum Institute, P.O. Box 2533, Abu Dhabi, United Arab Emirates

H I G H L I G H T S

Numerical simulation of mass transfer in Taylor flow microchannel reactors.

Purely theoretical approach that models channel length and inlet mixing region.

Reaction system of absorption of carbon dioxide in aqueous sodium hydroxide.

Effect of length, wettability, inlet mixing, phase concentrations, temperature.

Predicted results compare favorably with experimental data in the literature.

art ic l e i nf o a b s t r a c t

Article history: This paper reports a numerical study of the mass transfer characteristics during Taylor flow in mini/
Received 29 September 2012 microchannel reactors. A finite-element implementation of the phase field method was used to predict the
Received in revised form hydrodynamics of the two-phase flow. The phase distribution thus obtained was used to define the
16 April 2013
computational domain to model the reactive mass transfer. The reaction system of the absorption of CO2 into
Accepted 2 June 2013
aqueous NaOH solution was considered. Channels with characteristic dimensions ranging from 100 μm to
Available online 11 June 2013
750 μm were modeled with cross-flow and flow-focusing inlet configurations. The effect of channel length
Keywords: was studied by varying the residence time in the transient simulation. The results indicated that channels
Mass transfer having a small characteristic dimension could yield reductions in the residence time, and therefore the
Absorption
reactor size, by as much as 85%. This reduction was further enhanced by higher concentration levels of the
Numerical analysis
liquid reactant and increased temperatures. The inlet mixing region was found to have a significant influence
Multiphase flow
Microstructure on the total mass transfer. The channel wall wettability was found to affect the mass transfer characteristics
Taylor flow negligibly. The predictions from the currently proposed model were compared with available experimental
data, as well as with predictions of an earlier unit cell-based model, and a good agreement was obtained.
& 2013 Elsevier Ltd. All rights reserved.

1. Introduction designed, the inherent higher surface area to volume ratio of these
systems provides substantial enhancements in the heat and mass
Microscale technologies are quickly penetrating new application transfer performance, while keeping the pressure drops at moderate
areas in diverse engineering applications due to their proven levels. A widely accepted classification system for microscale systems
potential for process enhancement, improved process control, was proposed by Kandlikar and Grande (2003) which is based on the
increased safety and overall size reduction (Ehrfeld et al., 2000; hydraulic diameter, D. It was proposed that conventional channels
Guangwen et al., 2008). Specifically, multiphase microreactors are of have D43 mm, minichannels have 200 μmoDo3 mm, and micro-
application to various engineering processes and when optimally channels have 200 μmoDo10 μm.
Taylor flow (Davies and Taylor, 1950) is a two-phase regime
that is dominated by surface tension forces and is characterized by
an alternating liquid slug and gas bubble formation with a thin
n
Corresponding author. Center for Environmental Energy Engineering (CEEE), film of liquid separating the gas phase from the channel wall.
Thermal Management Faculty, CALCE Electronic Products and Systems, 4164C Further, convective mixing is enhanced by the recirculating
Glenn L. Martin Hall, Department of Mechanical Engineering, University
velocity components and hence this flow regime has been of
of Maryland, College Park, MD 20742, USA. Tel.: +1 301 405 5263;
fax: +1 301 405 2025. significant interest for heat and mass transport processes and is
E-mail address: [email protected] (M. Ohadi). the focus of the present work.

0009-2509/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ces.2013.06.005
70 H. Ganapathy et al. / Chemical Engineering Science 101 (2013) 69–80

Berčič and Pintar (1997) studied the absorption of methane in flux across the interface were overcome by effecting appropriate
water during Taylor flow in minichannels and subsequently modifications to the cell face diffusivities.
proposed an empirical correlation for the mass transfer coefficient. Shao et al. (2010) noted that some previous modeling
However, as has been reviewed and remarked upon in many approaches are not well suited for applications involving chemical
subsequent works, the notable absence of any functional depen- reactions for which a highly accurate prediction of the interfacial
dence on channel diameter in their correlation has led to a concentration is required. These include approaches that consider
relatively poor predictive accuracy of the expression for alternative only the bubble boundary and apply a tracer concentration at the
systems. interface. They proposed a numerical model which considered
Fundamental approaches to predict the mass transfer coeffi- both phases and simultaneously solved for the concentration
cient were developed by Irandoust and Andersson (1989) by change in each phase. The gas concentration at the interface was
separately considering the various contributions to the overall continuously updated in a method very similar to that of Onea
mass transfer from the gas bubble through the liquid film to the et al. (2009), by using Henry's law to obtain a continuous
channel wall, from the gas bubble to the liquid slugs through the concentration profile. As in most previous studies, the Taylor flow
bubble caps, and from the liquid slugs to the channel wall. regime was modeled using the unit cell approach, with mass
Likewise, van Baten and Krishna (2004) modeled a single unit cell transfer occurring across the interface to both the liquid slugs and
of Taylor flow in minichannels of varying diameters and accounted the thin liquid film region between the gas bubble and channel
for the mass transfer contributions from the two bubble caps and wall. They favorably compared their predictions against experi-
the film region. Modifications to the same in order to consider long mental data while the disagreement with existing empirical
contact times (or large unit cell lengths) were suggested by Vandu correlations was noted.
et al. (2005) in order to match their predictions with experimental From the brief review above, which was largely limited to mass
data for the absorption of oxygen into water, wherein the film transfer modeling approaches in mini/microchannel reactors, the
contribution to mass transfer was shown to be the dominant following points can be inferred. First, in spite of the common
factor. interest in the development of empirical correlations, these
Yue et al. (2007) reported reactive mass transfer experiments in correlations are generally specific to certain conditions only and
a rectangular minichannel and proposed an empirical correlation their universal applicability as yet remains to be established.
for the mass transfer coefficient. Subsequently, Yue et al. (2009) Second, relatively few studies have focused on microscale Taylor
reported another empirical correlation for non-reactive mass flows involving reactive mass transfer as compared to non-
transfer in square minichannels to improve the inaccurate predic- reacting flows. In this regard, the numerical model of Shao et al.
tions obtained by using certain previous mass transfer correla- (2010) can be considered a definite advancement, although it is
tions, such as that of van Baten and Krishna (2004), for their clearly limited by its need for experimental data for defining the
square minichannel application. Likewise, Su et al. (2010) and Tan unit cell domain. Unfortunately, the use of empirical correlations
et al. (2012a) have also proposed empirical expressions for Sher- for obtaining this data limits the universal applicability of the
wood number and mass transfer coefficient, respectively. model. In other words, the computational domain for mass
While most of the works reviewed above have been based on transfer would vary depending on the correlation being used, as
empirical approaches, these methods are typically limited to was demonstrated by Shao et al. (2008), while comparing experi-
specific systems only and are not universal in nature. For this mentally measured Taylor bubble lengths against multiple empiri-
reason, certain studies, including the present one, have focused on cal correlations in literature. While this variation can certainly be
the development of numerical approaches. avoided by developing purely numerical approaches, one must
Salman et al. (2004) modeled a unit cell of the Taylor flow take into consideration the associated increase in the computa-
regime using a finite volume method-based code. An initial tracer tional requirements. Third, the unit-cell based approaches have
concentration was applied to the cell, and changes to its concen- been quite popular on account of the periodic nature of the Taylor
tration were tracked. The model was reported to be suitable for flow regime. While these approaches do reduce the computational
low Bodenstein numbers only, and a simplified analytical solution time, they are grossly limited by their inability to model and thus
was proposed for higher Bodenstein numbers, typically greater account for the mass transfer that occurs in the inlet mixing region
than 10. A similar approach was presented by van Baten and present in most microchannel reactors. The significance of the
Krishna (2004) considering only the liquid slug and treating the same limitation for reacting flows requires investigation, espe-
Taylor bubble as a void with a constant initial tracer concentration cially considering that enhancement factors as high as 1.6 were
at the interface. reported by Fries and von Rohr (2009) depending on the inlet
Raimondi et al. (2008) performed direct numerical simulation mixing region. Likewise, Tan et al. (2012b) demonstrated that up
(DNS) of liquid–liquid mass transfer during Taylor flow in square to 30% of the total mass transfer could occur during the formation
microchannels. The hydrodynamics were obtained by the volume stage of Taylor flow in the inlet region. Lastly, it is worth noting
of fluid (VOF) method. The concentration field was governed by that one of the primary goals of an accurate mass transfer model is
using transformational parameters to modify the conventional its potential use in the design of scaled-up microchannel reactors
convention–diffusion equations in order to derive equations that having complex channel geometries and optimized channel
were compatible with the one-fluid approach used for predicting dimensions. In view of these limitations, the following points are
the hydrodynamics of the flow. It is worth noting that Raimondi addressed in the present paper.
et al. (2008) decoupled the hydrodynamics from the mass transfer A numerical approach to simulate gas–liquid mass transfer
and therefore their method is accurate only when the transport of with chemical reactions during Taylor flow in a mini/microchannel
species negligibly affects the hydrodynamics of the flow. In other reactor is proposed in this paper. The modeling methodology that
words, interfacial deformations due to mass transfer are assumed governs the hydrodynamics of two-phase flow and the mass
to be negligible here. These issues were outlined by Onea et al. transfer phenomena is based on purely theoretical (non-empirical)
(2009), who reported a qualitative study of the mass transfer approaches. The hydrodynamics of the flow are decoupled from
across deforming interfaces. The model was based on the VOF the mass transport phenomena. The computational domain con-
approach and considered a continuous numerical concentration siders an entire microreactor that includes the channel length as
across the interface by implementing Henry's law on one side and well as the inlet mixing region. The numerical predictions are
a reference concentration on the other. The discontinuities in mass validated with experimental and theoretical data reported in the
 H. Ganapathy et al. / Chemical Engineering Science 101 (2013) 69–80 71

literature. Parametric studies varying the phase concentration, equation (Eq. (2)):
channel size, wall contact angle and temperature are performed
∇⋅U ¼ 0 ð1Þ
and discussed.
∂U
ρ þ ρðU:∇ÞU ¼ ∇:½−PI þ μð∇U þ ∇U′Þ þ Fg þ FST ð2Þ
∂t
2. Numerical model The Eötvös number criterion of Brauner and Moalem-Maron
(1992), Eö o(2π)2, was used to verify that for any given condition,
The presently proposed approach decouples the hydrody- the surface tension forces were indeed dominant over gravita-
namics of two-phase flow from the mass transfer phenomena. tional forces and hence the latter could be neglected. Surface
Accordingly, the governing equations for each process were tension was modeled as a body force (FST) in the momentum
separately solved. A diffuse interface model, the phase field equation. Two second-order partial differential equations are
method, was used to simulate the fluid flow, as reported in given by Eqs. (3) and (4), wherein ϕ is the phase field variable
Section 2.1. For the reaction system detailed in Section 2.2, the that governs the interface. As a convention, in the present study ϕ
mass transfer across the interface of the Taylor bubbles was was chosen to equal −1 in the pure gas phase and +1 in pure liquid
simulated using the formulation reported in Section 2.3. phase:
∂ϕ γλ
þ U:∇ϕ ¼ ∇: 2 ∇ψ ð3Þ
2.1. Modeling of two-phase flow ∂t ξ
!
Various models have been developed for multiphase systems ξ2 ∂f ext
ψ ¼ −∇:ξ2 ∇ϕ þ ðϕ2 −1Þϕ þ ð4Þ
and have been applied towards two-phase flow in both mini- λ ∂ϕ
channels and microchannels. The VOF method (Hirt and Nichols,
1981), which has proven to be the most popular, and the level set In the above equations, λ is the magnitude of mixing energy
method (Osher and Sethian, 1988; Sussman et al., 1994), belong to and ξ scales with the thickness of the interface. The time scale of
the class of diffuse-interface modeling approaches. They are the Cahn–Hilliard diffusion is determined by γ as defined by
primarily based on the approximation of surface tension forces Eq. (5). It is a function of two user-defined parameters: mobility, χ,
in the interfacial region as a body force, using the continuum and interface thickness, ξː
surface force (CSF) formulation of Brackbill et al. (1992). These γ ¼ χξ2 ð5Þ
approaches have been implemented in numerous studies includ-
ing Taha and Cui (2004), Qian and Lawal (2006), Liu andWang Suitable values for the mobility and interface thickness para-
(2008), Kumar et al. (2007), Carlson et al. (2008), Fang et al. meters, χ and ξ, respectively, were chosen following the recom-
(2008), Lakhehal et al. (2008), Chen et al. (2009), Gupta et al. mendations of Ganapathy et al. (2013a) to obtain a suitable time
(2009), Fang et al. (2010), Krishnan et al. (2010), Santos and Kawaji scale of diffusion that did not excessively damp the convective
(2010), Asadolahi et al. (2011), Jeon et al. (2011), Asadolahi et al. motion and yet retained a constant interfacial thickness (Jacqmin,
(2012), and Ganapathy et al. (2013c). The phase-field method also 1999). Likewise, sufficient diffusion was recommended to be
belongs to the class of diffuse interface techniques but governs the present to resist straining flows that thin or thicken the interface.
interface based on the fluid free energy. The free energy density is The effect of surface tension was modeled as a body force, FST,
composed of the gradient energy and the bulk energy density (van in the equation for conservation of momentum (Eq. (2)). It was
der Waals, 1893). Cahn and Hilliard (1959) defined the phase-field computed as the product of the chemical potential of the system,
variable, ϕ, which is considered as a measure of phase. It assumes G, and the gradient of the phase-field variable, ϕ:
two distinct values in either phase and undergoes a rapid smooth F ST ¼ G:∇ϕ ð6Þ
change across the interface. In the interfacial region, the two
" #
phases are considered to be mixed and are thereby associated with ϕðϕ2 −1Þ
a mixing energy. G ¼ λ −∇2 ϕ þ ð7Þ
ξ2
Parasitic or spurious currents are the unphysical velocity
components at the interface (Lafaurie et al., 1994). These currents The scaling relation given by Eq. (8) was used to obtain the
are an important issue in the modeling of multiphase flow, volume fraction of either phase, which takes values from 0 to 1.
especially while using the CSF formulation which forms the basis The effective fluid properties such as density and viscosity were
for the VOF model. The currents are generated in the near- computed as the average value of the two phases weighed by their
interface regions on account of local variations in the CSF body respective volume fractions obtained from the following equation:
force, which is essentially an approximation of the surface tension 8
forces in this region. At higher magnitudes, the parasitic currents < 1−ϕ : Gasphase
2
ε ¼ 1þϕ ð8Þ
can potentially destroy the interface and thereby adversely affect : 2 : Liquidphase
the hydrodynamics of the flow. Recent works by De Menech
(2006), He and Kasagi (2008) and Ganapathy et al. (2013b) have A schematic of the computational domain is reported in Fig. 1.
demonstrated that phase field modeling can significantly reduce A 2-D geometry was used in the present study on account of its
the magnitude of parasitic currents. This has been attributed to the reduced computational cost as compared to a 3-D geometry, and
appropriate transfer of the initially generated energy due to the the same is justified as follows. Qian and Lawal (2006) compared
discretization error, between the kinetic energy and the surface Taylor flow simulations obtained using 2-D and 3-D domains. The
tension energy, which is further dissipated as viscosity. On account gas bubble and liquid slug lengths were compared and negligible
of this improved accuracy, the phase field model has been utilized differences were reported. Subsequently, Krishnan et al. (2010)
in the present work for modeling two-phase flow and the and Ganapathy et al. (2013a) have also utilized 2-D domains in
governing equations for the same, as derived by Yue et al. their simulations of Taylor flow. However, 3-D computational
(2004), are reviewed below. domains yield improved accuracy, particularly with respect to
The hydrodynamics of two-phase flow are described by the the pressure and velocity distribution in the channel which
equation for continuity (Eq. (1)) and incompressible Navier–Stokes influences the heat/mass transport phenomena. For instance,
72 H. Ganapathy et al. / Chemical Engineering Science 101 (2013) 69–80

Fig. 1. Schematic of computational domain.

the Laplace pressure drop is higher in 3-D domain-based results, Table 1

as compared to 2-D domain-based results on account of the Values of parameters in Eq. (15) (Schumpe, 1993).
different radii of curvature. However, this was not an issue in the
i Component hi (m3/mol) (  10−3)
present mass transfer results since the hydrodynamics were
decoupled from the mass transfer. 1 Na+ 0.1171
The entire 2-D domain was discretized with a homogeneous 2 OH− 0.0756
unstructured grid consisting of triangular elements. To maintain a 3 HCO3− 0.1372
4 CO32− 0.1666
constant resolution of the interface throughout the channel, the G CO2(L) −0.0183
entire domain was discretized using the similar sized elements
making it impossible to use block-structuring to obtain a high
resolution near the channel walls. By monitoring the gas bubble Modeling of the mass transfer in the above system requires the
and liquid slug lengths, a non-dimensional element size of D/10, expression for the rate of the reaction to be determined from the
wherein D is the characteristic dimension of the channel, was reaction kinetics. While the major points are outlined here, the
found to yield grid-independent results. It is noted that grid- reader is referred to Zanfir et al. (2005) for a complete description.
independence analysis with the phase field method requires The first step, the dissolution of gaseous CO2 into the liquid
consideration of the element size, j, and the interface thickness solution, has a high process rate and hence the equilibrium at
parameter, ξ. This was investigated by Ganapathy et al. (2013a) for the interface is described by Henry's law (Danckwerts, 1970). The
a similar case of Taylor flow, based on the j/ξ ratio. A value of unity value of Henry's constant in Eq. (14) pertains to the solubility of
for this ratio was found to yield grid independent results and CO2 in aqueous NaOH solution. It is obtained from Eq. (15) which
accordingly, the same has been used in the present study as well. relates the Henry's law constant in the solution to that in water at
The inlets were modeled with a constant velocity boundary the same temperature. The summation in Eq. (15) is over all the
condition with a single phase being admitted through each inlet. A ions, i, present in the solution. The parameters hi and hG refer to
pressure-based outflow boundary condition was specified at the the ions and the absorbed gas in the liquid phase respectively and
channel outlet, which applied a Dirichlet condition on the pres- their values, as given by Schumpe (1993), are reported in Table 1.
sure. A no-slip condition was specified on the channel walls. The equilibrium solubility of CO2 in water was reported by
Additionally, the effect of contact angle was studied in Section Crovetto (1991):
3.5 by specifying an angle, θ, in the following equation which alters
the interface normal at the channel wall based on the value of C nCO2ðLÞ ¼ H:pCO2 ð14Þ
contact angle specified in the boundary condition:
π    
H
n:ξ2 ∇ϕ ¼ ξ2 tan −θ  ∇ϕ−ðn:∇ϕÞnj ð9Þ log 10 ¼ −∑ðhi þ hG ÞC i ð15Þ
2 H water i

The wetted-wall condition given by Eq. (9) specifies no-slip In the above reaction system, Eq. (10) is almost instantaneous,
(zero velocity) at the wall. For those simulation runs where the and Eq. (12), an ionic reaction, also has a significantly high rate of
wetting behavior was not modeled, the interface was assumed to reaction. Thus, the overall rate of reaction is governed by Eq. (11)
be normal to the channel wall and hence for θ equaling 901, the and follows second-order kinetics. The rate equation for the
right-hand side of Eq. (9) becomes zero. process is given by Eq. (16), in which the equilibrium constant
kOH−, as given by Pohorecki and Moniuk (1988), is given by Eq. (17).
2.2. Chemical reaction system The ionic strength of the solution, I, is given by Eq. (18) and is
obtained by considering the concentration and valence of the
The chemical reactions considered are between a gaseous various ions present in the solution:
mixture of CO2 and N2 with an aqueous solution of NaOH and
R ¼ kOH− C OH− C CO2 ð16Þ
follow the following steps:
CO2ðGÞ 2CO2ðLÞ ð10Þ
2382
log 10 ðkOH − Þ ¼ 11:916− þ ð0:221  IÞ−ð0:016  I 2 Þ ð17Þ
− − T
CO2ðLÞ þ OH 2HCO3 ð11Þ

HCO−3 þ OH− 2CO2− I ¼ 0:5∑ C i zi 2 ð18Þ

3 þ H2 O ð12Þ
i

The overall reaction for the above system is given as

The reaction rate was obtained by solving Eqs. (14)–(18)
CO2 þ 2NaOH-Na2 CO3 þ H2 O ð13Þ simultaneously (Zanfir et al., 2005; Shao et al., 2010).
 H. Ganapathy et al. / Chemical Engineering Science 101 (2013) 69–80 73

2.3. Mass transfer model and 10% (Chambers et al., 2001), in order to control the highly
exothermic nature of the reaction.
The numerical model considered the mass transfer occurring The above processes, under certain conditions, can serve as
between the gas bubble and the liquid slug. The study by Tan et al. potential applications for the presently developed numerical
(2012a) which reported transient measurements of gas bubble model. However, it must be noted that the ranges reported here
length and volume indicated a decrease in these quantities over serve merely as reference levels. Individual instances may involve
time as a result of the absorption process. However, the present higher or lower levels of gas phase concentration, and the present
model did not account for interfacial deformations due to mass modeling assumption becomes less valid for higher gas concen-
transfer. In other words, the change in bubble volume was trations. For this reason, relatively low CO2 concentration levels
assumed to be negligible, as suggested by Shao et al. (2010). The ranging from 1% to 10% were used in the parametric studies
assumption is justified only in mass transfer systems involving a (Sections 3.2–3.6). Likewise, Shao et al. (2010) utilized a CO2
small fraction of the gas being transferred out of the bubble, across concentration of 5% in their parametric studies as well. Hence,
the interface, into the liquid phase. This limits the application of the modeling assumption that the hydrodynamics of the flow is
the present model to processes involving low gas phase concen- negligibly affected by the mass transfer process is justified for
trations, some of which are reviewed subsequently. such cases.
Sweetening of natural gas is a process that involves the The computational domain for the mass transfer was obtained
absorption of CO2 and H2S from a sour gas mixture, and the by the following procedure. First, the exact location of the inter-
concentration of the above components depends upon the source face was obtained from the two-phase flow patterns simulated in
of the gas. The CO2 concentration in natural gas can range between Section 2.1. The interface was fixed at a phase volume fraction of
0.7–8% (Kohl and Nielsen, 1997), 8.3% (Leal et al., 1996), 13% 0.5, as suggested by He and Kasagi (2008). The flow pattern was
(Alonso, 2010), and 5% (Socolow et al., 2011), while the H2S processed based on the phase field variable, ϕ, to create separate
concentration can range between 2.1–16.5% (Kohl and Nielsen, regions (sub-domains) for each phase. The boundary between the
1997) and 10% (ConocoPhillips, 2012). Post-combustion capture two regions was the interface, and the mass transfer was modeled
systems involve the absorption of CO2 from flue gas streams to occur across this boundary line. Since the two-phase flow model
produced by the burning of fossil fuels. The CO2 concentration in did not accurately capture the thin liquid film between the gas
the gas stream varies depending upon various factors including bubble and channel wall, the channel boundary was slightly offset
the operating conditions and the nature of the fuel being used. from the gas bubble in order to create this thin film. The offset was
This has been estimated to range between 12% and 14% for coal quantified such that the resulting film thickness was in accordance
fired boilers and integrated gasification combined cycle, 11–13% for with the expression by Aussilous and Quere (2000), given by
oil fired boilers, 3–4% for gas turbines, and 7–10% for natural gas Eq. (19). The resulting change in the reactor volume was negligible
fired boilers (IPCC, 2005). Likewise, Katzer et al. (2007) and since the thickness of the liquid film was of the order of only a few
Socolow et al. (2011) estimated that between 10% and 15% CO2 is micrometers in length. A schematic of the Taylor flow results
present in flue gas produced by the combustion of coal. Natural obtained from the phase field method is reported in Fig. 2a,
gas is a relatively cleaner fuel for which the CO2 concentration in wherein there is no liquid film at the channel wall. Subsequently,
the flue gas is between 3% and 5% (Socolow et al., 2011). Direct Fig. 2b reports the result obtained after implementing the channel
capture of carbon dioxide from air is associated with a low gas wall offset to create the liquid film. It is noted that this offset
concentration of about 0.04% (Socolow et al., 2011). procedure does not alter the angle θ:
Several industrial gas streams require processing to remove
0:66Ca2=3
volatile organic compounds (VOCs) and other polluting gases that δ¼ D ð19Þ
1 þ 3:33Ca2=3
are present in relatively low, yet harmful concentrations. Examples
include benzene, carbon tetrachloride, methanol, and toluene, The above-described procedure, when implemented, resulted
each ranging between 0.05% and 0.5% concentration (Ozturk and in a computational domain for modeling the mass transfer. The
Yilmaz, 2006), H2S or mercaptans ranging between 0.006% domain consisted of an entire channel, including the inlet mixing
and0.01% concentration (Blaszczak et al., 1995), and NOX ranging region, with several gas bubbles and liquid slugs at intermittent
between 0.02% and 0.2% concentration (Yasuda et al., 2011). Lastly, intervals. Having obtained fully formed Taylor bubbles, as well as
the direct fluorination of aromatic compounds such as toluene is emerging Taylor bubbles in the inlet region, we modeled the mass
often performed using low concentrations of fluorine gas mixed transfer following an approach similar to Shao et al. (2010) which
with nitrogen at levels ranging from 1% to 10% (de Mas et al., 2001) is described below.

Fig. 2. Schematic diagram of Taylor flow: (a) phase field simulation result without thin liquid film and (b) computational domain for mass transfer after implementing
channel wall offset to account for thickness of liquid film, δ.
74 H. Ganapathy et al. / Chemical Engineering Science 101 (2013) 69–80

The convection–diffusion mass transfer equations following the concentration at the interface and resulted in a realistic prediction
general form given by the following equation was solved for both of the mass transfer.
phases for the reactants CO2 and OH−:

∂C
þ ∇  ð−d  ∇CÞ ¼ R−U  ∇C ð20Þ 3. Results and discussion
∂t
The boundary conditions applied for the two reacting species, 3.1. Validation of model
CO2 and OH−, were different. For CO2, at the interface in the gas
bubbles, the concentration of the gas as obtained at each time step The modeling methodology adopted in the present study tied
was specified. In the liquid slugs, Henry's law was applied, which the predictive accuracy of the mass transfer model to that of the
accounted for the discontinuous concentrations of CO2 in the gas two-phase flow model. Both were validated as described below.
and liquid phases (Shao et al., 2010). For OH−, a convective flux Two-phase flow simulations of Taylor flow were performed
boundary condition was assumed along the interface in both the with nitrogen as the gas phase and water as the liquid phase. The
gas and liquid phase, which assumed that the mass flux of this gas void fraction averaged over an entire bubble formation cycle
specie due to diffusion is zero. The wall was insulated from the was estimated for channels having a characteristic dimension of
transport of species. 100, 250 and 500 μm. The same was compared with the predic-
Domain discretization was performed using an unstructured tions of the homogeneous model (Eq. (21)) and the Armand
grid consisting of triangular elements. Successive levels of mesh correlation (Eq. (22)) (Armand and Treschev, 1946).
refinement indicated that the mass transfer characteristics were
UG
less sensitive to element size in the interiors of the liquid slugs and εG ¼ β G ¼ ð21Þ
UG þ UL
gas bubbles, as compared to the interfacial regions. Grid indepen-
dent results were obtained by resolving the interior regions with
εG ¼ 0:833βG ð22Þ
non-dimensional element sizes of D/10 wherein D refers to the
characteristic dimension of the channel. However, the interfacial The results of gas void fraction are reported in Fig. 3 and a good
regions required resolution levels between D/100 and D/200, agreement was obtained with the homogeneous flow model and
which is at least an order of magnitude higher than the rest of the Armand and Treschev (1946) expression. Similar findings have
the computational domain. Similar high interfacial grid resolutions been reported in numerous earlier works as well (Zhao and
were also reported by Shao et al. (2010) for their unit cell model. Bi, 2001; Serizawa et al., 2002; Chung and Kawaji, 2004; He and
It is worth noting that the mass transfer model did not involve Kasagi, 2008; Yue et al., 2008).
an interface tracking algorithm and as a result, it was possible to The results of the mass transfer model were validated using the
use varying element sizes in the channel and therefore minimize experimental data of Shao et al. (2010) for a similar system and the
the computational time. This is believed to be a primary advantage predictions of their unit cell model. The inlet mixing region was
of developing decoupled approaches that separate the hydrody- modified from the cross-flow configuration shown in Fig. 1 to a
namics of the multiphase flow from the mass transfer because the flow-focusing inlet configuration, as used by Shao et al. (2010),
use of interface tracking algorithms necessitates the use of a wherein a single gas phase inlet was smoothly flanked between
constant element size in order to maintain a constant interface two liquid phase inlets. The gas phase was composed of CO2 gas at
resolution throughout the computational domain. In other words, a concentration of 23% by volume in a mixture with nitrogen and
the currently proposed modeling methodology is a computation- the liquid phase was 0.2 M NaOH.
ally inexpensive, purely theoretical (non-empirical) approach for The comparison was made with respect to the gas absorption
simulating reactive mass transfer characteristics in microchannel fraction, as defined by Eq. (23). Further, the residence time defined
geometries. by Eq. (24) is related to the length of the channel, L, and bubble
velocity, UB. The latter was obtained from Eq. (25) (Liu et al., 2005).
Clearly, for the same flow conditions, a longer channel is
2.4. Solver settings
1
The numerical model was implemented using a commercial Present study, D = 500 μm
Present study, D = 250 μm
CFD code, COMSOL Multiphysics (COMSOL Inc., Burlington, MA), Present study, D = 100 μm
version 3.5a, which is based on the finite element method. An Homogeneous model
0.8
implicit time-stepping algorithm was used with residual tolerance Armand and Treschev (1946)
levels of 10−4. This was verified to be sufficient, since like solutions
Gas void fraction, εG (-)

were obtained at lower tolerance levels too. All simulations were

carried out on a Dell workstation comprising two Intel Xeon 0.6
processors with eight cores and 4 GB of RAM.
For modeling of the two-phase flow, the phase-field variables
were initialized by selectively solving for ϕ and ψ. This transient 0.4
initialization run was used to define the initial interface. Since the
channel was assumed to be initially filled with quiescent gaseous
phase, the initial interface was formed at the liquid inlet. It was 0.2
ensured for every simulation that the interface created by the
initialization procedure was monotonic and the phase field vari-
able, ϕ, was always within the allowed range of −1 to +1. The final
0
transient simulation was solved for all the governing equations
0 0.2 0.4 0.6 0.8 1
using the initial values obtained above.
Volumetric flow ratio , βG (-)
For modeling the mass transfer phenomena, the governing
equations for both phases were solved simultaneously as sug- Fig. 3. Comparison of numerically predicted gas void fraction with correlations in
gested by Shao et al. (2010). This continuously updated the species the literature for nitrogen–water Taylor flow (D: 100, 250 and 500 μm).
 H. Ganapathy et al. / Chemical Engineering Science 101 (2013) 69–80 75

associated with a larger residence time: velocities of the liquid and gaseous phases were both 0.1 m/s.
Three sets of studies were conducted.
C CO2 ;in −C CO2 ;out
X¼  100% ð23Þ First, the initial concentration of the gaseous phase was varied
C CO2 ;out
from 1% to 10% by volume of CO2 with constant initial concentra-
L tion of NaOH equaling 0.2 M. The results reported in Fig. 6 indicate
t res ¼ ð24Þ that with higher initial gas concentration levels, the absorption
UB
fraction decreases for a certain residence time (or channel length).
1 This is a result of the increased consumption of the NaOH at the
U B ¼ U TP ð25Þ interface at higher gas concentration levels which decreases the
1−0:61Ca0:33
potential for mass transfer.
As shown in Fig. 4, the numerical predictions from the current
Second, the molar concentration of the NaOH solution was
model had a mean absolute error of 11.9% which was considered a
varied from 0.05 M to 1 M with a constant initial gas phase
strong indicator of its predictive accuracy. It is worth noting that
concentration equaling 5% by volume, for all cases. As seen in
the present predictions of absorption fraction could be slightly
Fig. 7, the absorption fraction achieved for a certain residence time
higher than the experimental data which was obtained with a CO2
(or channel length) steadily increases with increasing levels of
concentration of 23%. This is due to the nature of the present
liquid reactant concentration. This is because the increased liquid
modeling methodology which assumes negligible decrease in the
phase concentration, in addition to influencing the reaction rate
bubble volume, and thereby the interfacial area, over the residence
(Eq. (16)), also affects the ionic strength of the solution and
time. As previously noted in Section 2.3, the validity of this
therefore the rate constant (Eqs. (17) and (18)).
assumption decreases at higher gas phase concentration levels
and accordingly, interpretations of the numerical predictions
should consider this aspect.
A characteristic unit cell obtained from the present simulations 100
is shown in Fig. 5 from which the recirculating velocity compo-
nents, which are a characteristic feature of the Taylor flow regime,
can be observed. Subsequently, parametric studies were per-
80
formed with the validated model which are reported and
Absorption fraction (%)

discussed below.

60
3.2. Effect of phase concentration

The effect of initial concentration of the gas and liquid phase

was studied for a channel having a characteristic dimension of 40
500 μm and cross-flow inlet configuration. The superficial
1% CO2
100 20 2.5% CO2
Numerically predicted absorption fraction (%)

5% CO2
10% CO2
90
0
0 0.05 0.1 0.15 0.2 0.25 0.3
MAE: 11.9% Residence time (s)
80
Fig. 6. Effect of gas phase concentration on absorption fraction (D, 500 μm; UG,
0.1 m/s; UL, 0.1 m/s; CNaOH, 0.2 M; T, 298 K).
70

100
60

80
50 Parity line
Absorption fraction (%)

Present model
Unit cell model (Shao et al., 2010)
40 60
40 50 60 70 80 90 100
Experimental absorption fraction (%)
Water
Fig. 4. Comparison of numerical predictions of absorption fraction with the 40 0.05 M NaOH
experimental data of Shao et al. (2010). 0.1 M NaOH
0.2 M NaOH
0.5 M NaOH
20 1.0 M NaOH

0
0 0.05 0.1 0.15 0.2 0.25 0.3
Fig. 5. Characteristic unit cell obtained from simulations with relative velocity Residence time (s)
vectors and streamlines (contours of gas phase volume fraction are in red).
(For interpretation of the references to color in this figure caption, the reader is Fig. 7. Effect of liquid phase concentration on absorption fraction (D, 500 μm;
referred to the web version of this article.) UG, 0.1 m/s; UL, 0.1 m/s; CCO2, 5% by volume; T, 298 K).
76 H. Ganapathy et al. / Chemical Engineering Science 101 (2013) 69–80

Third, the gas to liquid concentration ratio was maintained at a 60% and 85% for the 500, 300 and 100 μm channels, respectively.
constant value while the individual concentration levels were Clearly, any reduction in the residence time translates into a linear
simultaneously varied. Characteristic results reported in Fig. 8 for decrease in the channel length and therefore the overall size of the
a ratio of 10 indicate that higher concentration levels of the liquid microreactor.
reactant were found to yield improved mass transfer character- For the same flow conditions, the smaller channels were found
istics. Similar findings were also reported by Zanfir et al. (2005) in to have longer bubbles than the larger channels. This is consistent
a comparable experimental analysis of a falling film microreactor. with the physics of the flow, since the gas void fraction does not
change with the characteristic dimension of the channel (Armand
3.3. Effect of channel size correlation, Eq. (22)). Hence the enhanced mass transfer charac-
teristics of the smaller channels are attributed to the larger
The effect of channel size was studied for characteristic interfacial areas present as a result of the longer bubbles. It should
dimensions ranging from 100 to 750 μm in channels having cross be noted that these results are not in contradiction with those of
flow inlet configuration. The superficial velocities of the liquid and Shao et al. (2010) in which an increase in the bubble length was
gaseous phases were both 0.1 m/s. The concentration of the gas found to adversely affect the absorption fraction, since their study
phase was 5% by volume CO2 and that of the liquid phase was was seemingly performed by arbitrarily varying the bubble length
0.2 M NaOH. The results shown in Fig. 9 indicate that shorter while maintaining a constant channel size. This would have led to
residence times (or channel lengths) were required to achieve a an increase in the gas void fraction and hence resulted in
certain level of absorption fraction as the channel size decreases. deteriorated absorption performance.
In comparison to a 750 μm channel, the residence time required to
attain a 95% absorption fraction level, was lower by around 40%, 3.4. Effect of modeling the inlet mixing region

100 A main advantage of the present model is its ability to account

for the mass transfer that occurs in the inlet mixing region.
However, it was necessary to quantify the mass transfer that
occurs in the region in detail in order to better understand its
80
significance.
Absorption fraction (%)

It was shown in Fig. 6 that the initial concentration of CO2 in

the gas bubble affected the predicted value of the absorption
60 fraction at any residence time. Hence, prior to modeling the mass
transfer in the gas bubble after it detached itself from the free
stream flow at the inlet we attempted to quantify the mass
40 transfer that occurred during the bubble formation period in the
inlet mixing region. This mass transfer was considered to occur
0.1 M NaOH
over the duration of a bubble formation cycle, tcycle. The value tcycle was
0.2 M NaOH obtained from the two-phase flow model detailed in Section 2.1. The
20
0.5 M NaOH change in concentration was accounted for at intervals of 1 ms such
that each transient simulation run yielded a reduced level of CO2
1.0 M NaOH concentration in the gas bubble. The value obtained at the end of each
0 transient simulation run was used to define the concentration in the
0 0.05 0.1 0.15 0.2 0.25 0.3
subsequent, further evolved gas bubble. Since the mass transfer in the
Residence time (s) inlet mixing region was modeled for duration of tcycle, the reaction
Fig. 8. Variation of absorption fraction with liquid phase concentration and time in the channel length, tchannel was quantified as the difference
constant gas–liquid concentration ratio (D, 500 μm; Cgas/Cliquid ¼ 10; T, 298 K). between the total residence time in the reactor, tres, and bubble
formation time, tcycle:

100 t channel ¼ t res −t cycle ð26Þ

Figs. 10 and 11 show the results obtained by the above

procedure, denoted as “Inlet modeled.” The curves denoted as
80 “Inlet not modeled” were obtained using the conventional method
followed in the unit cell-based approaches wherein the bubble
Absorption fraction (%)

evolution in the inlet mixing region is completely neglected.

60 Rather, the mass transfer from the detached bubble was assumed
to occur over the entire residence time.
Fig. 10 compares the numerically predicted absorption fraction
obtained by the two methods over a residence time tcycle. This
40
corresponded to the time spent by a bubble in the channel prior to
D = 100 μm detachment. The initial concentration of the gas phase was 5% by
volume CO2 and that of the liquid phase was 0.2 M NaOH. For the
20 D = 300 μm
100 μm channel, the simulation performed without considering
D = 500 μm the inlet mixing region was found to over predict the absorption
D = 750 μm fraction by nearly 16%. In the cases with characteristic dimensions
0 of 300 μm and 500 μm, the absorption fraction was over-predicted
0 0.1 0.2 0.3 0.4 0.5 0.6 by nearly 9% and 8% respectively. In terms of CO2 concentration,
Residence time (s) this corresponded to differences of nearly 19%, 10% and 8.5% for
Fig. 9. Effect of channel size on absorption fraction (UG, 0.1 m/s; UL, 0.1 m/s; CCO2, the 100, 300 and 500 μm channels, respectively. These discrepan-
5% by volume; CNaOH, 0.2 M; T, 298 K). cies can be explained as follows. The cases that did not consider
 H. Ganapathy et al. / Chemical Engineering Science 101 (2013) 69–80 77

the inlet mixing region were associated with large, constant value interfacial area for mass transfer, gradually increases as the bubble
of bubble length for the entire duration of tcycle, thereby resulting develops and emerges. Further, the discrepancies are more domi-
in a higher value of absorption fraction due to the higher inter- nant for smaller channels since they are associated with longer
facial area. On the other hand, with the cases that did model the bubbles and therefore larger interfacial areas.
inlet mixing region, the value of bubble length, and therefore This study described above, which was performed for a
relatively short residence time, tcycle, demonstrated the importance
of modeling the inlet mixing region, especially for smaller dia-
50 meter channels. Subsequently, the effect of modeling the inlet
D = 100 μm; Inlet not modeled
D = 100 μm; Inlet modeled mixing region was analyzed for a longer residence time. As can be
D = 300 μm; Inlet not modeled seen in Fig. 11, there is a continuous discrepancy in absorption
D = 300 μm; Inlet modeled fraction throughout the channel length for the 500 μm channel.
40 D = 500 μm; Inlet not modeled
D = 500 μm; Inlet modeled This suggests that even for long residence times, it is essential to
Absorption fraction (%)

accurately account for the mass transfer that occurs in the inlet
mixing region. It is worth noting that the relatively smaller
30
discrepancies observed with smaller diameter channels in Fig. 11
can be attributed to the improved mass transfer characteristics
associated with smaller channels.
20 Based on the above discussion, it is clear that the present
observations are in agreement with earlier findings in the litera-
ture (Fries and von Rohr, 2009; Tan et al., 2012b) and accordingly,
10 it is recommended that numerical modeling of gas–liquid mass
transport phenomena in mini/microchannels should account for
the mass transfer that occurs in the inlet mixing region, in addition
0 to that which occurs in the channel length. It is worth noting that
0 0.005 0.01 0.015 0.02 the present study used a simplified 2-D domain, as against a more
Residence time (s) realistic 3-D domain. Nonetheless, based on the present results, it
Fig. 10. Effect of modeling inlet mixing region on absorption fraction for residence
is believed that even with a 3-D domain, numerical simulations
time, tcycle (D, 500 μm; UG, 0.1 m/s; UL, 0.1 m/s; CCO2, 5% by volume; CNaOH, 0.2 M; that do not model the inlet region will result in erroneous
T, 298 K). predictions. However, the magnitude of discrepancy associated
with 3-D domain-based results could be different from that
associated with 2-D domain-based predictions.
100
3.5. Effect of wall contact angle

80
In the two-phase flow model, a wetted wall boundary condi-
Absorption fraction (%)

tion allowed a specification of a contact angle at the wall. Two

values of contact angle, 901 and 601, were compared as shown in
60 Fig. 12 for phase superficial velocities of 0.1 m/s. The latter,
according to Gupta et al. (2009), would result only in a more
realistic bubble appearance and would be a “purely numerical
40 artifact.” The validity of this statement was tested by comparing
D = 100 μm; Inlet not modeled
D = 100 μm; Inlet modeled the gas void fraction between the two cases which was found to be
D = 300 μm; Inlet not modeled within 0.8% of each other. Additionally, no significant change in
20 D = 300 μm; Inlet modeled the bubble and slug lengths was observed. While the above does
D = 500 μm; Inlet not modeled partially verify the statement by Gupta et al. (2009), it is worth
D = 500 μm; Inlet modeled noting that the bubble boundary was altered due to the rounded
0 bubble caps (Fig. 12). The effect of the altered interfacial profile on
0 0.05 0.1 0.15 0.2 0.25 0.3 the mass transfer performance was further investigated.
Residence time (s) The mass transfer characteristics for contact angles equaling
Fig. 11. Effect of modeling inlet mixing region on absorption fraction for residence
901 and 601 of the two cases were compared for concentrations of
time, tchannel (D, 500 μm; UG, 0.1 m/s; UL, 0.1 m/s; CCO2, 5% by volume; CNaOH, 0.2 M; 5% by volume of CO2 and 0.2 M NaOH. As shown in Fig. 13, our
T, 298 K). results generally agree with the findings of Gupta et al. (2009),

Fig. 12. Typical image of Taylor flow pattern for wall contact angle (a) 901 and (b) 601 (D, 500 μm; UG, 0.1 m/s; UL, 0.1 m/s; CCO2, 5% by volume; CNaOH, 0.2 M; T, 298 K;
contours of gas phase volume fraction are in red). (For interpretation of the references to color in this figure caption, the reader is referred to the web version of this article.)
78 H. Ganapathy et al. / Chemical Engineering Science 101 (2013) 69–80

100 400
0.1 M NaOH
0.2 M NaOH
0.5 M NaOH
80 1.0 M NaOH

Rate constant, k (m3/kmol s)

300
Absorption fraction (%)

60
D = 100 μm; θ = 90°
D = 100 μm; θ = 60° 200

40 D = 300 μm; θ = 90°

D = 300 μm; θ = 60°
D = 500 μm; θ = 90°
D = 500 μm; θ = 60° 100
20

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 275 300 325 350 375
Residence time (s) Temperature (K)
Fig. 13. Effect of contact angle on absorption fraction (D, 500 μm; UG, 0.1 m/s; Fig. 15. Variation of rate constant with temperature at different levels of liquid
UL, 0.1 m/s; CCO2, 5% by volume; CNaOH, 0.2 M; T, 298 K). phase concentration.

100 The enhancement is more prominent between an increase from

ambient conditions of 298 K and that of 323 K, as compared to
further increases in temperature. This observed enhancement is
80 attributed to the relationship between temperature and the
reaction rate constant (Fig. 15) as given by Eq. (17). The effect of
Absorption fraction (%)

temperature is more pronounced at higher liquid phase concen-

60 tration levels.

4. Conclusions
40
This paper reports a numerical model to predict the mass
298 K transfer characteristics during Taylor flow in mini/microchannel
20 323 K reactors. The current approach is distinct from existing approaches
343 K because it models the entire reactor, including the channel length
363 K as well as the inlet mixing region. The computational require-
0 ments were reduced by decoupling the hydrodynamics of the two-
0 0.05 0.1 0.15 0.2 0.25 0.3 phase flow from the mass transfer phenomena. Further, the model
Residence time (s) was derived from a purely theoretical basis in order to permit its
Fig. 14. Effect of temperature on absorption fraction (D, 500 μm; UG, 0.1 m/s; application towards the design of scaled-up microscale reactors
UL, 0.1 m/s; CCO2, 5% by volume; CNaOH, 0.2 M). having complex channel geometries and optimized dimensions.
A finite-element implementation of the phase field method was
especially for the smaller channel sizes. The marginal discrepan- used to predict the hydrodynamics of the two-phase flow. The
cies of up to 4% that were observed for the 500 μm channel were phase distribution obtained was used to define the computational
most likely due to the increasing mass transfer contributions from domain to model the mass transfer. The reaction system of
the bubble caps for larger channels, in which the bubble lengths absorption of CO2 into aqueous NaOH solution was considered.
are also correspondingly less than the smaller channels. Thus it is Channels with characteristic dimensions ranging from 100 μm to
concluded that while an appropriate definition of the wall contact 750 μm were modeled with flow-focusing and cross-flow types of
angle may not generally provide highly improved accuracy for inlet configurations. The effect of channel length was also studied
channels having a small characteristic dimension, it could be by varying the residence time in the transient simulation. The
essential for mass transfer in larger channels. numerical predictions were compared with experimental data in
the literature and a good agreement was obtained with a mean
3.6. Effect of temperature absolute error of 11.9%. Parametric studies on the effect of the gas
and liquid phases concentrations were conducted, and the trends
Shao commented on the sensitivity of the reaction rate to of absorption fraction were consistent with the reaction chemistry.
temperature, due to which they maintained their test section at The effect of modeling the inlet mixing region was found to be
isothermal conditions. Likewise, Zanfir et al. (2005), Yue et al. significant and simulations that did not account for the mass
(2007), and Su et al. (2010) performed their experiments at transfer that occurs in the inlet region during the bubble forma-
ambient temperatures only. In the present study, the temperature tion stages, were demonstrated to yield discrepancies in CO2
was found to have a significant effect on the mass transfer, as concentration as high as 19%. This observation is in agreement
shown in Fig. 14 for a 500 μm channel in which the phase with previous studies in the literature, and based on the same it is
superficial velocities were 0.1 m/s and concentrations were 5% by recommended that numerical modeling of gas–liquid mass trans-
volume of CO2 and 0.2 M NaOH. port phenomena in microchannels should account for the mass
 H. Ganapathy et al. / Chemical Engineering Science 101 (2013) 69–80 79

transfer that occurs both in the inlet mixing region and the res residence
channel length. The effect of modeling a wall contact angle was S liquid slug
found to yield marginal discrepancies only for the larger channels. ST surface tension
Substantial reductions, as high as 85% in the residence time and TP two-phase
therefore the overall reactor size, were obtained as a result of
using channels having a reduced characteristic dimension. These
enhancements were found to be further enhanced by higher
concentration levels of the liquid reactant and increased tempera- Acknowledgements
tures. On the whole, the current model had a reasonably good
predictive accuracy and can be used for the designing scaled-up
The authors gratefully acknowledge the financial support
microchannel reactors for reacting two-phase flows.
provided by GASCO and ADGAS companies of ADNOC Group of
Companies, Abu Dhabi, United Arab Emirates. The authors are also
grateful to members of the Research and Development Gas Sub-
Nomenclature Committee of the ADNOC Group for their close technical interac-
tion and support.
C concentration (mol/m3)
Cn interfacial concentration (mol/m3)
Ca capillary number (dimensionless) References
CCO2, in concentration of CO2 in gas phase at channel inlet
(mol/m3) Alonso, T.V., 2010. High efficiency on CO2 removal in natural gas with UCARSOL
solvents. In: Proceedings of the Rio Oil & Gas Expo and Conference, IBP
CCO2, out concentration of CO2 in gas phase at channel outlet 2778_10.
(mol/m3) Armand, A.A., Treschev, G.G., 1946. The resistance during the movement of a two-
D hydraulic diameter or characteristic dimension (m) phase system in horizontal pipes. Izv. Vses. Teplotekh. Inst. 1, 16–23.
Asadolahi, A.N., Gupta, R., Fletcher, D.F., Haynes, B.S., 2011. CFD approaches for the
d diffusivity (m2/s)
simulation of hydrodynamics and heat transfer in Taylor flow. Chem. Eng. Sci. 66,
Eö Eötvös number (dimensionless) 5575–5584.
F force (N) Asadolahi, A.N., Gupta, R., Leung, S.S.Y., Fletcher, D.F., Haynes, B.S., 2012. Validation
fext external free energy density (J/m3) of a CFD model of Taylor flow hydrodynamics and heat transfer. Chem. Eng. Sci. 69,
541–552.
G chemical potential (N/m2) Aussilous, P., Quere, D., 2000. Quick deposition of a fluid on the wall of a tube. Phys.
H Henry's constant (mol/m3 Pa) Fluids 12 (10), 2367–2371.
h ion-specific parameter in Eq. (15) (m3/mol) Berčič, G., Pintar, A., 1997. The role of gas bubbles and liquid slug lengths on mass
transport in the Taylor flow through capillaries. Chem. Eng. Sci. 52, 3709–3719.
i ion Blaszczak, R.J., Zerbonia, R.A., Spivey, J.J., Agarwal, S.K., Damle, A.S., Sanford, C.W.,
I ionic strength (mol/m3) 1995. Survey of Control Technologies for Low Concentration Organic Vapor Gas
j mesh element size (m) Streams. U.S. Environmental Protection Agency EPA-456/R-95-003, pp. 93–103.
Brackbill, J.U., Kothe, D.B., Zemach, C., 1992. A continuum method for modeling
k reaction rate constant (m3/mol s) surface tension. J. Comput. Phys. 100 (2), 335–354.
L length (m) Brauner, N., Moalem-Maron, D., 1992. Identification of the range of 'small diameter'
p partial pressure (Pa) conduits regarding two-phase flow pattern transition. Int. Commun. Heat Mass
Transfer 19, 29–39.
P pressure (Pa) Cahn, J.W., Hilliard, J.E., 1959. Free energy of a non-uniform system. III. Nucleation
R reaction rate (mol/m3 s) in a two-component incompressible fluid. J. Chem. Phys. 31 (3), 688–699.
T temperature (K) Carlson, A., Kudinov, P., Narayanan, C., 2008. Prediction of two-phase flow in small
tubes: a systematic comparison of state-of-the-art CMFD codes. In: Proceedings
t time (s)
of 5th European Thermal-Sciences Conference, Eindhoven, The Netherlands.
U velocity (m/s) Chambers, R.D., Holling, D., Spink, R.C.H., Sandford, G., 2001. Elemental fluorine
X absorption fraction (%) part 13. Gas–liquid thin film microreactors for selective direct fluorination. Lab.
z valence of ion i (dimensionless) Chip. 1, 132–137.
Chen, Y., Kulenovic, R.., Mertz, R., 2009. Numerical study on the formation of Taylor
bubbles in capillary tubes. Int. J. Thermal Sci. 48, 234–242.
Greek symbols Chung, P.M.-Y., Kawaji, M., 2004. The effect of channel diameter on adiabatic two-
phase flow characteristics in microchannels. Int. J. Multiphase Flow 30,
735–761.
β volumetric flow ratio (dimensionless) ConocoPhillips, 2012. Natural gas, sour (Canada) 〈http://www.conocophillips.com/
γ time scale of Cahn–Hilliard diffusion (m3 s/kg) EN/products/safetydata/Documents/MSDS%20Canada/791785%20Natural%
20Gas,%20Sour%20_Canada_.pdf〉 (retrieved on 31st March 2013).
δ liquid film thickness (m) Crovetto, R., 1991. Evaluation of solubility data of the system CO2–H2O from 273 K
ε volume/void fraction (dimensionless) to the critical point of water. J. Phys. Chem. Reference Data 20, 575–589.
λ magnitude of mixing energy (N) Danckwerts, P.V., 1970. Gas–Liquid Reactions. McGraw Hill, New York.
Davies, R.M., Taylor, G., 1950. The mechanics of large bubbles rising through
μ dynamic viscosity (Pa s) extended liquids and through liquids in tubes. Proc. R. Soc. London Ser. A: Math.
ρ density (kg/m3) Phys. Sci. 200 (1062), 375–390.
χ mobility parameter (m s/kg) de Mas, N., Jackman, R.J., Schmidt, M.A., Jensen, K.F., 2001. Microchemical systems
for direct fluorination of aromatics. In: Microreaction Technology. IMRET 5:
ψ phase field help variable (dimensionless) Proceedings of the Fifth International Microreaction Technology, Springer,
ϕ phase field variable (dimensionless) Berlin, Heidelberg, pp. 60–67.
y contact angle (degrees) De Menech, M., 2006. Modeling of droplet breakup in a microfluidic T-shaped
junction with a phase field model. Phys. Rev. E 73, 031515.
ξ interface thickness parameter (m)
Ehrfeld, W., Hessel, V., Löwe, H., 2000. Microreactors: New Technology for Modern
Chemistry. Wiley-VCH, Weinheim.
Subscripts and superscripts Fang, C., Hidrovo, C., Wang, F., Eaton, J., Goodson, J., 2008. 3-D numerical simulation
of contact angle hysteresis for microscale two phase flow. Int. J. Multiphase
Flow 34, 690–705.
B bubble Fang, C., David, M., Rogacs, A., Goodson, K., 2010. Volume of fluid simulation of
G gas boiling two-phase flow in a vapor venting microchannel. Front. Heat Mass
Transfer 1, 013002.
g gravity Fries, D.M., von Rohr, P.R., 2009. Impact of inlet design on mass transfer in gas-
L liquid liquid rectangular microchannels. Microfluid Nanofluid 6, 27–35.
80 H. Ganapathy et al. / Chemical Engineering Science 101 (2013) 69–80

Ganapathy, H., Al-Hajri, E., Ohadi, M.M., 2013a. Phase field modeling of Taylor flow Pohorecki, R., Moniuk, W., 1988. Kinetics of reaction between carbon dioxide and
in mini/microchannels, part I: bubble formation mechanisms and phase field hydroxyl ions in aqueous electrolyte solutions. Chem. Eng. Sci. 43, 1677–1684.
parameters. Chem. Eng. Sci. 94, 138–149. Qian, D., Lawal, A., 2006. Numerical study on gas and liquid slugs for Taylor Flow in
Ganapathy, H., Al-Hajri, E., Ohadi, M.M., 2013b. Phase field modeling of Taylor flow a T-junction microchannel. Chem. Eng. Sci. 61, 7609–7625.
in mini/microchannels, part II: hydrodynamics of Taylor flow. Chem. Eng. Sci. Raimondi, N.D.M., Prat, L., Gourdon, C., Cognet, P., 2008. Direct numerical simula-
94, 156–165. tions of mass transfer in square microchannels for liquid-liquid slug flow.
Ganapathy, H., Shooshtari, A., Choo, K., Dessiatoun, S., Alshehhi, M., Ohadi, M., Chem. Eng. Sci. 53, 5522–5530.
2013c. Volume of fluid-based numerical modeling of condensation heat Salman, W., Gavriilidis, A., Angeli, P., 2004. A model for predicting axial mixing
transfer and fluid flow characteristics in microchannels. Int. J. Heat Mass during gas–liquid Taylor flow in microchannels at low Bodenstein numbers.
Transfer 65, 62-72. Chem. Eng. J. 101, 391–396.
Guangwen, C., Jun, Y., Quan, Y., 2008. Gas-liquid microreaction technology: recent Santos, R.M., Kawaji, M., 2010. Numerical modeling and experimental investigation
developments and future challenges. Chin. J. Chem. Eng. 16 (5), 663–669. of gas-liquid slug formation in a microchannel T-junction. Int. J. Multiphase
Gupta, R., Fletcher, D.F., Haynes, B.S., 2009. On the CFD modeling of Taylor flow in Flow 36, 314–323.
microchannels. Chem. Eng. Sci. 64, 2941–2950. Schumpe, A., 1993. The estimation of gas solubilities in salt solutions. Chem. Eng.
He, Q., Kasagi, N., 2008. Phase-field simulation of small capillary-number two- Sci. 48, 153–158.
phase flow in a microtube. Fluid Dyn. Res. 40, 497–509. Serizawa, A., Feng, Z., Kawara, Z., 2002. Two-phase flow in microchannels. Exp.
Hirt, C.W., Nichols, B.D., 1981. Volume of fluid (VOF) method for the dynamics of Thermal Fluid Sci. 26, 703–714.
free boundaries. J. Comput. Phys. 39, 201–225. Shao, N., Gavriilidis, A., Angeli, P., 2010. Mass transfer during Taylor flow in
IPCC, 2005. IPCC Special Report on Carbon Dioxide Capture and Storage. In: Metz, B., microchannels with and without chemical reaction. Chem. Eng. J. 160, 873–881.
Davidson, O., de Coninck, H., Loos, M., Meyer, L. (Eds.), Working Group III of the Shao, N., Salman, W., Gavriilidis, A., Angeli, P., 2008. CFD simulations of the effect of
Intergovernmental Panel on Climate Change. Cambridge University Press, Cam- inlet conditions on Taylor flow formation. Int. J. Heat Fluid Flow 29, 1603–1611.
bridge, United Kingdom and New York, NY, USA, pp. 75–104. Socolow, R., Desmond, M., Aines, R., Blackstock, J., Bolland, O., Kaarsberg, T., Lewis, N.,
Irandoust, S., Andersson, B., 1989. Simulation of flow and mass transfer in Taylor Mazzotti, M., Pfeffer, A., Sawyer, K., Siirola, J., Smit, B., Wilcox, J., 2011. Direct Air
flow through a capillary. Comput. Chem. Eng. 13 (4/5), 519–526. Capture of CO2 With Chemicals. American Physical Society, pp. 4–44.
Jacqmin, D., 1999. Calculation of two-phase Navier-Stokes flows using phase-field Su, H., Wang, S., Niu, H., Pan, L., Wang, A., Hu, Y., 2010. Mass transfer characteristics
modeling. J. Comput. Phys. 155, 96–127.
of H2S absorption from gaseous mixture into methyldiethanolamine solution in
Jeon, S.S., Kim, S.J., Park, G.C., 2011. Numerical study of condensing bubble in
a T-junction microchannel. Sep. Purif. Technol. 72, 326–334.
subcooled boiling flow using volume of fluid model. Chem. Eng. Sci. 66, 5899-
Sussman, M., Smereka, P., Osher, S., 1994. A level set approach for computing
5909.
solutions to incompressible two-phase flow. J. Comput. Phys. 114, 146–159.
Kandlikar, S.G., Grande, W.J., 2003. Evolution of microchannel flow passages-
Taha, T., Cui, Z.F., 2004. Hydrodynamics of slug flow inside capillaries. Chem. Eng.
thermohydraulic performance and fabrication technology. Heat Transfer Eng.
Sci. 59, 1181–1190.
24 (1), 3–17.
Tan, J., Lu, Y.C., Xu, J.H., Luo, G.S., 2012a. Mass transfer performance of gas-liquid
Katzer, J., Ansolabehere, S., Beer, J., Deutch, J., Ellerman, A.D., Friedmann, S.J.,
segmented flow in microchannels. Chem. Eng. J. 181 and 182, 229–235.
Herzog, H., Jacoby, H.D., Joskow, P.L., McRae, G., Lester, R., Moniz, E.J., Steinfeld,
Tan, J., Lu, Y.C., Xu, J.H., Luo, G.S., 2012b. Mass transfer characteristic in the
E., 2007. The Future of Coal, Massachusetts Institute of Technology, pp. 17–42.
formation stage of gas–liquid segmented flow in microchannel. Chem. Eng. J.
Kohl, A., Nielsen, R., 1997. Gas Purification. Gulf Publishing Company, pp. 40–186.
185–186, 314–320.
Krishnan, R.N., Vivek, S., Chatterjee, D., Das, S.K., 2010. Performance of numerical
van Baten, J.M., Krishna, R., 2004. CFD simulations of mass transfer from Taylor
schemes in the simulation of two-phase free flows and wall bounded mini
channel flows. Chem. Eng. Sci. 65, 5117–5136. bubbles rising in circular capillaries. Chem. Eng. Sci. 59, 2535–2545.
Kumar, V., Vashisth, S., Hoarau, Y., Nigam, K.D.P., 2007. Slug flow in curved van der Waals, J.D., 1893. The thermodynamic theory of capillarity flow under the
microreactors: hydrodynamic study. Chem. Eng. Sci. 62, 7494–7504. hypothesis of a continuous variation of density. J. Stat. Phys. 20, 197–244.
Lafaurie, B., Nardone, C., Scardovelli, R., Zaleski, S., Zanetti, G., 1994. Modelling Vandu, C.O., Liu, H., Krishna, R., 2005. Mass transfer from Taylor bubbles rising in
merging and fragmentation in multiphase flows with SURFER. J. Comput. Phys. single capillaries. Chem. Eng. Sci. 60, 6430–6437.
113, 134–147. Yasuda, M., Tsugita, N., Ito, K., Yamauchi, S., Glomm, W.R., Tsuji, I., Asano, H., 2011.
Lakhehal, D., Larrignon, G., Narayanan, C., 2008. Computational heat transfer and High-efficiency NOX absorption in water using equipment packed with a glass
two-phase flow topology in miniature tubes. Microfluid Nanofluid 4, 261–271. fiber filter. Environ. Sci. Technol. 45, 1840–1846.
Leal, O., Bolivar, C., Ovalles, C., Urbina, A., Revette, J., Garcia, J.J., 1996. Carbon dioxide Yue, J., Chen, G., Yuan, Q., Luo, L., Gonthier, Y., 2007. Hydrodynamics and mass
removal from natural gas using amine surface bonded adsorbents. In: Proceed- transfer characteristics in gas-liquid flow through a rectangular microchannel.
ings of Symposium on the Capture, Utilization and Disposal of CO2, American Chem. Eng. Sci. 62, 2096–2108.
Chemical Society, pp. 1332–1336. Yue, J., Luo, L., Gonthier, Y., Chen, G., Yuan, Q., 2009. An experimental study of air-
Liu, D., Wang, S., 2008. Hydrodynamics of Taylor flow in noncircular capillaries. water Taylor flow and mass transfer inside square microchannels. Chem. Eng.
Chem. Eng. Process.: Process Intensification 47, 2098–2106. Sci. 64, 3697–3708.
Liu, H., Vandu, C.O., Krishna, R., 2005. Hydrodynamics of Taylor flow in vertical Yue, J., Luo, L., Gonthier, Y., Chen, G., Yuan, Q., 2008. An experimental investigation
capillaries: flow regimes, bubble rise velocity, liquid slug length and pressure of gas–liquid two-phase flow in single microchannel contactors. Chem. Eng. Sci.
drop. Ind. Eng. Chem. Res. 44, 4884–4897. 63, 4189–4202.
Onea, A., Worner, M., Cacuci, D.G., 2009. A qualitative computational study of mass Yue, P., Feng, J.J., Liu, C., Shen, J., 2004. A diffuse-interface method for simulating
transfer in upward bubble train flow through square and rectangular mini- two-phase flows of complex fluids. J. Fluid Mech. 515, 293–317.
channels. Chem. Eng. Sci. 64, 1416–1435. Zanfir, M., Gavriilidis, A., Wille, Ch., Hessel, V., 2005. Carbon dioxide absorption in a
Osher, S., Sethian, J.A., 1988. Fronts propagating with curvature-dependent speed: falling film microstructured reactor: experiments and modeling. Ind. Eng.
algorithms based on Hamilton–Jacobi formulations. J. Comput. Phys. 79, 12–49. Chem. Res. 44, 1742–1751.
Ozturk, B., Yilmaz, D., 2006. Absorptive removal of volatile organic compounds Zhao, T.S., Bi, Q.C., 2001. Co-current air–water two-phase flow patterns in vertical
from flue gas streams. Process Saf. Environ. Prot. 84, 391–398. triangular microchannels. Int. J. Multiphase Flow 27, 765–782.