Journal of Membrane Science 261 (2005) 114–120

Shell side mass transfer in a transverse flow hollow

fiber membrane contactor
Ju-Meng Zheng a,b , Zheng-Wei Dai b , Fook-Sin Wong a , Zhi-Kang Xu b,∗
a Institute of Environmental Science and Engineering, Nanyang Technological University, Innovation Center, Block 2, Unit 237,
18 Nanyang Drive, Singapore 637723, Singapore
b Institute of Polymer Science, Zhejiang University, Zheda Lu 38, Hangzhou 310027, PR China

Received 27 February 2004; received in revised form 16 February 2005; accepted 16 February 2005
Available online 20 April 2005

Abstract

The free surface model was introduced to describe the shell side fluid flow in a transverse flow hollow fiber membrane contactor, and a new
method was developed to calculate the shell side hydraulic diameter, the effective average velocity, and the Reynolds number. An empirical
shell side mass transfer correlation was presented for commercial Liqui-Cel® Extra-Flow contactors on the basis of the experimental data
reported by Sengupta et al. The data were correlated very well with maximum discrepancies of ±10% between the predicted and observed
results.
© 2005 Elsevier B.V. All rights reserved.

Keywords: Membrane contactors; Transverse flow; Mass transfer; Hydrodynamics; Free surface model

1. Introduction Flow module commercialized by CELGARD LLC (Char-

lotte, USA). Celgard® microporous polypropylene hollow
A membrane contactor is a device that achieves gas/liquid fiber membranes used in this module have been woven into
or liquid/liquid mass transfer without dispersion of one phase fabric to allow more uniform fiber spacing, which in turn
within another. This is accomplished by passing the fluids on leads to high mass transfer coefficient. The Extra-Flow mod-
the opposite sides of a microporous membrane. Through the ule contains a central shell side baffle, a feature that offers
careful control of the pressure difference between the fluids, two advantages: (1) the baffle can improve the mass transfer
one of the fluids is immobilized in the pores of the membrane efficiency by minimizing shell side by-passing; (2) it provides
so that the fluid/fluid interface is located at the mouth of each a component of velocity normal to the membrane surfaces,
pore [1]. Usually, two types of modules, called parallel flow which results in a higher mass transfer coefficient than that
and transverse or cross flow are used. It has been reported that achieved with strictly parallel flow [2].
the transverse flow module has a number of advantages such Generally, mass transfer in a hollow fiber contactor can
as a larger shell side mass transfer coefficient, minimal flow be described using a resistance-in-series model [1]. The tube
channeling, better scale-up characteristics and more precise side mass transfer can be described with the Lévêque equation
performance prediction [2]. The main features of the trans- and the membrane resistance can be calculated from known
verse flow module have been summarized by Gabelman and membrane parameters such as membrane thickness, tortuos-
Hwang [1] and Sengupta et al. [2]. ity and porosity. However, mass transfer correlation for shell
The most well-known transverse flow module, which is side fluid flow has not been well established up to now. A
schematically shown in Fig. 1, is the Liqui-Cel® Extra- conventional approach is to use the empirical correlation of
the following form:
∗ Corresponding author. Tel.: +86 571 879 52605; fax: +86 571 879 51773.
E-mail address: [email protected] (Z.-K. Xu). Sh = αReβ Sc0.33 (1)

0376-7388/$ – see front matter © 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.memsci.2005.02.035
 J.-M. Zheng et al. / Journal of Membrane Science 261 (2005) 114–120 115

Fig. 1. Sketch of the Liqui-Cel® Extra-Flow membrane contactor (redrawn from Ref. [2]).

where the constants α and β are determined from experimen- one of the rods in the assemblage and the outer cylinder of a
tal results. Based on this equation, some empirical correla- fluid enveloped with a free surface [18]. In previous studies
tions have been proposed for transverse flow modules [3–14]. [19–21], the free surface model was adapted to describe the
However, there are only three correlations have been found shell side fluid flow in a parallel flow hollow fiber membrane
applicable for commercial Liqui-Cel® Extra-Flow membrane contactor. In fact, the free surface model can also be applied
contactors, as listed in Table 1. Among them, the correlations to the case of transverse flow [18]. The only difference is
reported by Schöner et al. [10] and Baudot et al. [11] were that, for the transverse flow, the boundary condition is that
derived directly from liquid–liquid extraction experiments the fluid radial direction velocity (ur ), the fluid angular direc-
using a 2.5 in. × 8 in. Liqui-Cel® Extra-Flow contactor. The tion velocity (uθ ) and the free surface velocity (uc ) hold the
correlation suggested by Kreith and Black [12], which was following relationship at the free surface [18]:
originally for a closely packed shell-and-tube heat exchanger,
u2 = u2r + u2θ , ur = u cos θ, uθ = −u sin θ (5)
was found to give a good prediction of the shell side mass
transfer coefficient in the Liqui-Cel® contactors in some stud- The present work extends the idea of free surface model
ies [13,14]. Sengupta et al. [2] studied the shell side mass to the case of a transverse flow module. It is expected to de-
transfer in the large-scale application of membrane contactor rive expressions for the calculation of the shell side effective
for gas stripping with Liqui-Cel® modules. These investiga- velocity and the hydraulic diameter, which can then be used
tions indicated that the shell side mass transfer coefficient to calculate the Reynolds number. It is also aimed to find an
is proportional to Q0.38–0.45 (Q is the feed flow rate into the applicable correlation to predict the shell side mass transfer
module) [2]. However, they did not present a correlation in coefficients in the commercial Liqui-Cel® Extra-Flow mem-
the form of Eq. (1). brane contactors. This part of the work is based on the exper-
Besides, different methods were used for the calculation imental results reported by Sengupta et al. [2].
of the effective shell side velocity, the hydraulic diameter, and
thus the Reynolds number [10–11,13–16]. This is due to the
fact that there is no fundamental mathematical description of 2. Theory
the shell side flow in a transverse flow contactor. Regarding to
Seibert and Fair’s work [17], the shell side flow was assumed In order to study the shell side fluid flow in a transverse
to be perfectly mixed. However, as commented by Baudot et flow module, the fiber bundle was firstly divided into small,
al. [11], this is not true for fluid flow through the fiber bundles. equally spaced cells with one fiber in each cell. And a free
In the literature, a method describing viscous flow relative surface was presented at the imaginary cell boundary. It was
to the arrays of solid rods is Happel’s free surface model [18]. assumed these flow cells are regularly arranged in layers be-
This model was developed on the basics that two concentric tween the center feed tube and the wall of the module. This
cylinders can serve as the model for fluid moving through means that the fiber number in the layer gradually increases
an assemblage of cylinders. The inner cylinder consists of from the layer near the center feed tube to the layer near the

Table 1
Mass transfer correlations for transverse flow modules
Correlation no. Correlation Brief conditionsa Reference
(2) Sh = 1.76Re0.82 Sc0.33 Feed flow rate 0.2–200 × 10−6 m3 /s [10]
(3) Sh = 0.56Re0.62 Sc0.33 Feed flow rate 0.2–200 × 10−6 m3 /s [11]
(4) Sh = 0.39Re0.59 Sc0.33b Feed flow rate 10–50 × 10−6 m3 /s [13]
Feed flow rate 33 × 10−6 m3 /s [14]
a Applied for Celgard Liqui-Cel® Extra-Flow 2.5 in. × 8 in. contactor.
b Kreith and Black equation, originally for a closely packed shell-and-tube heat exchanger [12].
116 J.-M. Zheng et al. / Journal of Membrane Science 261 (2005) 114–120

(dt + dc )2 − dt2
Nmin = ε (11)
df2
For the flow cell, when the velocity in the free surface (uc,m ) is
determined, we can calculate the effective cell velocity (also
can be considered the effective shell side velocity, ue ) using
the method proposed in Refs. [10,14]. It has:




Q 1 uc,m rc
ue dr = dr = dr (12)
Nlm 2π(leff /2)jr r
Integral Eq. (12) from fiber radius (rf ) to cell radius (rc ), the
following relationship can be obtained as:
Fig. 2. Schematic representation for the flow cell: (a) in the cross-section
division, (b) of module with free surface. dc ln(df /dc )
ue = uc,m (13)
df − d c
module wall. The division of module cross-section and the
flow cell with free surface are sketched in Fig. 2. Combining Eqs. (8)–(13), the effective cell velocity can be
The cell diameter (dc ) can be determined by assuming that expressed as:
the cell and the module have the same packing fraction: ln ε ln[(2dm − dc )/(2dt + dc )] Q
ue = √ (14)
df 4( ε − 1) (dm − dt − dc ) π(leff /2)
dc = √ (6)
ε In order to calculate the Reynolds number, it still needs to
where df is the outer diameter of the hollow fiber, ε is module know the hydraulic diameter (dh ). For the flow cell, the hy-
packing fraction which can be calculated as: draulic diameter is defined as the ratio of four times the cross-
sectional area to the wetted perimeter as usually done:
Ndf2
ε= (7) four cell void volumes d 2 − df2 1−ε
dm
2 − d2
t dh = = c = df (15)
wetted surface of the cell df ε
where dm is the inner diameter of the module, dt is the outer
diameter of the center feed tube, and N is the total number of Using Eqs. (14) and (15), the Reynolds number and the Sher-
the hollow fibers. wood number can be calculated in term of the following def-
In a transverse flows module, the feed fluid flows from the initions:
porous center tube passes through the fiber layer around the dh ue ρ
Re = (16)
tube to the layer near the module wall. In each fiber layer, µ
the total fluid flow is equally distributed between the cells.
As a result, the local cell free surface velocity in each fiber kdh 1 − ε kdf
Sh = = (17)
layer is dependent on the number of fibers number in the D ε D
layer. Therefore, the average cell free surface velocity (uc,m ) Eqs. (14)–(17) incorporate all the parameters of the module
is defined as the free surface velocity in a layer, which has (diameter of the module, diameter of the center feed tube,
the log-mean number of the fibers. It has: diameter of the fiber, effective length of the fiber and the
Q packing fraction) that affect the shell side hydrodynamics.
uc,m = (8) These will be helpful in developing a desirable empirical
Nlm πdc (leff /2)
correlation to describe the shell side mass transfer coefficient
here Q is the feed flow, leff /2 is effective length of the mod- in the transverse flow module.
ule (the module is divided into two chambers by the baffle,
Fig. 1); Nlm is the log-mean number of fibers in the fiber layer,
defined as: 3. Results and discussion
Nmax − Nmin
Nlm = (9) 3.1. Methods for the calculation of shell side effective
ln(Nmax /Nmin )
velocity
where Nmax is the number of fibers in the layer nearest to the
module wall, and Nmin is the number of fibers in the layer In the literatures, several methods have been reported for
around the center feed tube: the calculation of hydraulic diameter and shell side effective
velocity in the Liqui-Cel® Extra-Flow hollow fiber mem-
2 − (d − d )2
dm m c
Nmax = ε (10) brane contactors. There are mainly two equations have been
df2 used to calculate the effective shell side velocity. The first
 J.-M. Zheng et al. / Journal of Membrane Science 261 (2005) 114–120 117

one is that proposed by Schöner et al. [10]: tightly packed (ε → ∞). It seems that only the influence of the
Q ln(dm /dt ) decrease in module void area, which caused by the increase
ue = (18) of module packing fraction, on the effective shell side ve-
π(leff /2) dm − dt
locity is taken into consideration in Eq. (20). Whereas, these
The second one, suggested by Mahmud et al. [14], is a cor- authors might overlook the resistance enhancement caused
rection incorporating the module packing fraction (ε): by the packing fraction increase.
1 Q ln(dm /dt ) The method developed in this work indicates that the ef-
ue = (19) fective shell side velocity decreases slightly with the increase
1 − ε π(leff /2) dm − dt
of the module packing fraction, as shown in Fig. 3. This can
In the present study, the effective shell side velocity was be contributed to the combination of two factors: (1) the de-
calculated with Eq. (14). Because the outer diameter of the crease of the module void area; and at the same time, (2) the
hollow fiber and the diameter of the free surface are much fast increase of the flow resistance caused by the increase
smaller than the inner diameter of the module (dm ) and the of the module packing fraction [18]. It also can be observed
outer diameter of the feed tube (dt ), Eq. (14) can be simplified from Fig. 3 that, the dimensionless effective shell side veloc-
as: ity is slightly higher than 1 and this should not be the true.
ln ε ln(dm /dt ) Q This result is due to that it was assumed that the cell diameter
ue = √ (20) become infinity for ε → 0 in model development (Eq. (6)).
4( ε − 1) (dm − dt ) π(leff /2)
Comparing Eqs. (15)–(17), it can be found that these three
3.2. Methods for the calculation of hydraulic diameter
equations describe in quite different ways for the influences
of the module packing fraction on the effective shell side
In the studies of the shell side mass transfer, the Reynolds
velocity. The discrepancy among these three equations is also
numbers have normally been calculated based on either the
illustrated in Fig. 3. In this figure, Y-coordinate is labeled as
outer diameter of the fiber (df ) [10,16] or the hydraulic diame-
η, a dimensionless parameter defined as:
ter (dh ) [10,11]. And the hydraulic diameter can be calculated
π(leff /2) dm − dt as:
η = ue (21)
Q ln(dm /dt ) 2 − d 2 − Nd 2
dm t
where ue is calculated according to Eqs. (18)–(20) respec- dh = f
(22)
Ndf
tively.
Eq. (19) was derived on the assumption that there is no As commented by Schöner et al. [10], the use of hydraulic
fiber in the module. As a result, it can be seen from Fig. 3 that, diameter (in this case, the fiber bundle was treated as a packed
the module packing fraction does not affect the shell side ef- bed of fibers) is more reasonable than the use of outer diame-
fective velocity at all. Therefore, it is difficult to imagine that ter. When introducing Eqs. (7) into (22), it can be found that
it (Eq. (19)) can be applied to the module packed with thou- Eq. (22) is the same as Eq. (15), which gives the definition of
sands upon thousands hollow fibers. On the other hand, Eq. hydraulic diameter by the free surface model. It indicates that
(20) indicates that the effective shell side velocity increases the method regarding the module as a packed bed of fibers
with the increase of module packing fraction, and the velocity and the free surface model deduce to the same definition of
will become infinity (η → ∞) when the module is extremely hydraulic diameter.

3.3. Shell side mass transfer correlation

Mass transfer in the shell side of transverse flow modules

was studied by Sengupta et al. [2]. In their study, oxygen
removal from water was performed in excess sweep condi-
tions through Liqui-Cel® Extra-Flow contactors with vari-
ous dimensions from lab scale to industrial using ones. It
was found that the shell side mass transfer coefficient was
proportional to Q0.38–0.45 [2]. However, they did not present
a mass transfer correlation. Here, the experimental results
obtained by Sengupta et al. [2] were adopted to develop an
empirical correlation for the shell side mass transfer in trans-
verse flow module. Detailed specifications for the contactors
used in their experiment work are listed in Table 2. The sepa-
ration characteristics (oxygen removal rate, E) under various
Fig. 3. Dimensionless shell side effective velocity defined with different operating conditions (shell side feed flow, Q) are cited in
methods. Table 3. And then the overall mass transfer coefficients were
118 J.-M. Zheng et al. / Journal of Membrane Science 261 (2005) 114–120

Table 2
Dimensional details for the Liqui-Cel® Extra-Flow contactorsa
Nominal contactor Average cartridge Average cartridge Effective fiber Fiber packing Total contactor area
size designator i.d. (m) o.d. (m) length (m) fraction (based on fiber i.d.) (m2 )
2.5 × 8 0.022 0.050 0.16 0.45 1.2
4 × 28 0.032 0.088 0.62 0.43 14.9
4 × 13 0.032 0.085 0.25 0.43 5.7
10 × 28 0.114 0.245 0.61 0.43 103.8
a Nominal fiber inside diameter (i.d.) = 240 × 10−6 m; nominal fiber outside diameter (o.d.) = 300 × 10−6 m. All data were reported by Sengupta et al. [2].

calculated by the following equation [2]:

 
Q 1
k= ln (23)
A 1−E
where A is the total contact area based on fiber inner diameter
(Table 2). For the oxygen removal process in excess sweeping
conditions, the main resistance to mass transfer is mainly
located in the shell side. In other words, it is reasonable to
assume the shell side mass transfer coefficient is equal to
the overall mass transfer coefficient calculated by Eq. (23).
Following this assumption, it is possible to get a shell side
mass transfer correlation.
On the other hand, the shell side Reynolds number was
calculated according to Eqs. (14)–(16) using the known data
of feed flow (Table 3) and module dimensions (Table 2). Sher-
wood number was calculated using Eqs. (15) and (17) when
the mass transfer coefficient was already calculated by Eq. Fig. 4. Comparison of the shell side mass transfer predicted by the pro-
posed correlation (Eq. (24)) to experimental data [2] for various Liqui-Cel®
(23). Results for the calculated Reynolds number and Sher-
contactors.
wood number are also summarized in Table 3.
Further examination of the relationship between the shows that this correlation relates with the experimental re-
Reynolds number and the Sherwood number reveals that the sults very well. It can be seen that the maximum discrepancy
data can be correlated as follows: between the correlation predictions and the experimental data
Sh = 2.15Re0.42 Sc0.33 (24) is within 10%. This indicates that mass transfer data reported
by Sengupta et al. [2] can be described by an empirical cor-
The correlation is valid in the Reynolds number from 0.8 relation that uses the definitions of the effective shell side ve-
to 20, subject to the experimental conditions (Table 3). Fig. 4 locity, hydraulic diameter, Reynolds number and Sherwood

Table 3
Experimental data [2] and calculated Reynolds number and Sherwood number
2.5 × 8 Contactor 4 × 28 Contactor 4 × 13 Contactor 10 × 28 Contactor

Water Oxygen Reb Shb Water Oxygen Reb Shb Water Oxygen Reb Shb Water Oxygen Reb Shb
flowa removala flowa removala flowa removala flowa removala
(×106 m/s) (%) (×106 m/s) (%) (×106 m/s) (%) (×106 m/s) (%)
32 96.14 0.842 14.85 252.3 99.68 1.162 18.72 126.2 98.48 1.410 17.20 3195 99.59 4.778 32.04
63 89.32 1.658 20.09 315.4 99.41 1.453 20.91 252.3 93.89 2.817 22.95 5665 97.90 8.471 39.92
126 79.12 3.317 28.15 504.7 98.24 2.324 26.32 378.5 89.68 4.226 27.98 7904 95.60 11.82 45.03
189 70.42 4.975 32.82 630.8 97.18 2.905 29.07 504.7 85.58 5.634 31.82 10168 93.39 15.21 50.39
757.0 95.96 3.487 31.36 630.8 81.67 7.042 34.85 11689 91.38 17.48 52.27
946.3 93.89 4.358 34.15 757.0 77.93 8.460 37.28 13279 89.54 19.86 54.69
1009.3 93.15 4.648 34.94 883.2 74.37 9.860 39.13
1261.7 90.10 5.811 37.67 1009.3 70.97 11.27 40.64
1514.0 87.06 6.972 39.98
1577.7 86.33 7.263 40.52
1766.3 84.27 8.134 42.19
2018.7 81.97 9.297 44.65
a Experimental data reported by Sengupta et al. [2].
b Calculated Reynolds number and Sherwood number with the correlations developed in this work.
 J.-M. Zheng et al. / Journal of Membrane Science 261 (2005) 114–120 119

[11,16]. Therefore, the correlation suggested by Schöner et

al. was corrected with the membrane surface porosity (γ) as
follows when it was plotted in Fig. 5:

Sh = 1.76γRe0.82 Sc0.33 (25)

It can be seen from Fig. 5 that, except the correlation

proposed by Schöner et al., the shell side mass transfer co-
efficients show similar Reynolds exponent dependency in
other three correlations. On the other hand, the correlations
of Schöner et al. [10], Baudot et al. [11] and Kreith and
Black [12] predict lower mass transfer coefficient than that
observed by experiments [2]. More recently, in the study of
liquid–liquid extraction with Liqui-Cel® Extra-Flow contac-
tor, Soldenhoff et al. [16] found that the experimental overall
mass transfer coefficients were much lower than the predicted
Fig. 5. Comparison of various correlations applied to the 2.5 in. × 8 in. values. The prediction was performed with the conventional
Liqui-Cel® Extra-Flow contactor. resistance-in-series model, where the shell side mass transfer
was estimated with the correlation proposed by Baudot et al.
number developed in present study. Such correlation will be [11], and the tube side mass transfer with Lévêque equation.
very helpful in the design of membrane contactor processes The authors stated that the difference is due to the contribution
using Liqui-Cel® modules. of extraction chemical kinetics at or very close to the interface
The Reynolds exponent in Eq. (24) is 0.42, which can well [16]. Whatever, we think the mass transfer in a liquid–liquid
fit the experimental data (k ∝ Q0.38–0.45 ) reported by Sengupta extraction is more complex than that in a gas stripping or ab-
et al. [2]. A value larger than 0.33 indicates that the transverse sorption process. In other words, it is much easier to reach the
flow module provides a component of velocity perpendicular condition that the overall mass transfer coefficient is equal to
to the hollow fiber surface, which results in a higher mass the shell side mass transfer in a gas stripping or absorption
transfer coefficient than that achieved with the parallel flow process and thus it is a suitable system for the study of the
module. On the other hand, the value is lower than 0.66. It in- complex shell side mass transfer.
fers that the shell side flow cannot reach the turbulence flow in
the tested Reynolds range (0.8–20). In the literatures, similar
Reynolds exponents were also reported by Wickramasinghe 4. Conclusions
et al. [3] (Sh ∝ Re0.47 ) and Wang et al. [4] (Sh ∝ Re0.46 )
for cross flow modules fabricated in their laboratories by Free surface model was applied to describe the shell side
wrapping hollow fiber fabric around a central pipe, as in flow in a transverse flow hollow fiber membrane contactor.
the case of the commercial Liqui-Cel® Extra-Flow module Based on this, a method was developed to calculate the shell
[16]. side hydraulic diameter, the effective shell side velocity and
The shell side mass transfer correlations listed in Table 1 the Reynolds number. An empirical correlation was proposed
were also used to fit the experimental data provided by Sen- to relate with the mass transfer data (oxygen stripping in ex-
gupta et al. [2]. The results are shown in Fig. 5. For com- cess sweep conditions) reported by Sengupta et al. [2]. The
parison, the predictions by Eq. (24) are also shown in this discrepancies between the correlation predictions and exper-
figure. imental data were within 10%. Additionally, it was shown
It should be noticed that, in Kreith and Black’s correlation, that the correlation come out from liquid–liquid extraction
the shell side Reynolds number and the Sherwood number predicts lower shell side mass transfer coefficients than that
were defined in term of the fiber outer diameter. The shell obtained by experiments. It indicates that gas stripping or ab-
side effective velocity was calculated by Eq. (19). In the cor- sorption experiment is a suitable system for the study of the
relations of Schöner et al. and Baudot et al., the shell side complex shell mass transfer, where we can easily reach the
effective velocity and the hydraulics diameter are calculated condition that the overall mass transfer coefficient is mainly
by Eqs. (18) and (22) respectively. As a result, in Fig. 5, X located in the shell side.
and Y coordinates were labeled as feed flow (Q) (but not the
Reynolds number), mass transfer coefficient (k) (but not the
Sherwood number) respectively. Besides, in Fig. 5, the equa- Acknowledgements
tion of Schöner et al. was modified from its original form,
because the authors used the pore area but not the whole The authors gratefully acknowledge the support from
membrane surface as the contact area, and it would lead to Agency of Science, Technology and Research of Singapore
high mass transfer coefficients as pointed in the literatures (A*STAR) to fund the IESE program.
120 J.-M. Zheng et al. / Journal of Membrane Science 261 (2005) 114–120

[3] S.R. Wickramasinghe, M.T. Semmens, E.L. Cussler, Mass transfer

Nomenclature in various hollow fiber geometries, J. Membr. Sci. 84 (1993) 1–14.
[4] K.L. Wang, E.L. Cussler, Baffled membrane modules made with
A total contact area offered by the membrane hollow fiber fabric, J. Membr. Sci. 85 (1993) 265–278.
contactor (m2 ) [5] M.-C. Yang, E.L. Cussler, Designing hollow fiber contactors, AIChE
dc diameter of the cell (m) J. 32 (1986) 1910–1915.
[6] P. Cota, T.-L. Bersillon, A. Hugard, Bubble-free aeration using mem-
df fiber outer diameter (m) brane: mass transfer analysis, J. Membr. Sci. 47 (1989) 91–106.
dh hydraulic diameter (m) [7] T. Ahmed, M.J. Semmens, Use of transverse flow hollow fiber
dm inside diameter of the cartridge (m) for bubbleless membrane aeration, Water Res. 30 (1996) 440–
dt outside diameter of the center feed tube (m) 446.
D diffusion coefficient (m2 /s) [8] D.W. Johnson, M.J. Semmens, J.S. Gulliver, Diffusive transport
across unconfined hollow fiber membrane, J. Membr. Sci. 128 (1997)
k shell side mass transfer coefficient (m/s) 67–81.
leff effective length of the fiber (m) [9] D. Bhaumik, S. Majumdar, K.K. Sirkar, Absorption of CO2 in a
N total number of the fibers transverse flow hollow fiber membrane module having a few wraps
Nlm log-mean number of the fibers in fibers layer of the fiber mat, J. Membr. Sci. 138 (1998) 77–82.
Nmax number of the fibers in the layer nearest to the [10] P. Schöner, P. Plucinske, W. Nitsch, U. Daiminger, Mass transfer in
the shell side of cross flow hollow fiber modules, Chem. Eng. Sci.
wall 53 (13) (1998) 2316–2319.
Nmin number of the fibers in the layer nearest to the [11] A. Baudot, J. Floury, H.E. Smoreburg, Liquid–liquid extraction of
feed tube aroma compound with hollow fiber contactor, AIChE J. 47 (2001)
Q volumetric flow rate of shell side (m3 /s) 1780–1793.
rc free surface radius (m) [12] F. Kreith, W.Z. Black, Basic Heat Transfer, Harper & Row, New
York, 1980.
rf fiber radius (m) [13] F.-X. Pierre, I. Souchon, V. Athes-Dutour, M. Marin, Membrane-
uc,m average cell free surface velocity (m/s) based solvent extraction of sulfur aroma compounds: influence of
ue effective shell side velocity (m/s) operating conditions on mass transfer coefficients in a hollow fiber
ur radial direction velocity (m/s) contactor, Desalination 148 (2002) 199–204.
uθ angular direction velocity (m/s) [14] H. Mahmud, A. Kumar, R.M. Narbaitz, T. Matsuura, A study of mass
transfer in the membrane air-stripping process using microporous
polypropylene hollow fibers, J. Membr. Sci. 179 (2000) 29–41.
Greek letters [15] M.J. Gonzalez-Munoz, S. Luque, J.R. Alvarez, J. Coca, Recovery
α,β constants in Eq. (1) of phenol from aqueous solutions using hollow fiber contactors, J.
γ membrane porosity Membr. Sci. 213 (2003) 181–193.
η dimensionless shell side effective velocity, Eq. [16] K. Soldenhoff, M. Shamieh, A. Manis, Liquid–liquid extraction
(21) of cobalt with hollow fiber contactor, J. Membr. Sci. 252 (2005)
183–194.
ε module packing fraction [17] A.F. Seibert, J.R. Fair, Scale-up of hollow fiber exactors, Sep. Sci.
µ viscosity of shell side liquid (Pa s) Technol. 32 (1997) 573–583.
ρ density of shell side liquid (kg/m3 ) [18] J. Happel, Viscous flow relative to arrays of cylinders, AIChE J. 5
(1959) 174–179.
[19] J.-M. Zheng, Y.-Y. Xu, Z.-K. Xu, Flow distribution in randomly
References packed hollow fiber membrane module, J. Membr. Sci. 211 (2003)
263–269.
[20] J.-M. Zheng, Y.-Y. Xu, Z.-K. Xu, Shell side mass transfer charac-
[1] A. Gabelman, S.-T. Hwang, Hollow fiber membrane contactors, J.
teristics in a parallel flow hollow fiber membrane module, Sep. Sci.
Membr. Sci. 159 (1999) 61–106.
Technol. 38 (2003) 1247–1267.
[2] A. Sengupta, P.A. Peterson, B.D. Miller, J. Schneider, C.W. Fulk
[21] S. Karoor, K.K. Sirkar, Gas absorption studies in microporous hollow
Jr., Large-scale application of membrane contactors for gas transfer
fiber membrane modules, Ind. Eng. Chem. Res. 32 (1993) 674–684.
from or to ultrapure water, Sep. Purif. Technol. 14 (1998) 189–200.