Vous êtes sur la page 1sur 11

CHERD-1483; No.

of Pages 11
chemical engineering research and design x x x ( 2 0 1 4 ) xxx–xxx

Contents lists available at ScienceDirect

Chemical Engineering Research and Design

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

Mass transfer from nanofluid drops in a pulsed

liquid–liquid extraction column

Amir Bahmanyar a, Nafiseh Khoobi a,b, Mostafa Mohammad Ali Moharrer a,

Hossein Bahmanyar a,∗
a School of Chemical Engineering, College of Engineering, University of Tehran, Tehran, Iran
b Research Institute of Petroleum Industry, Tehran, Iran

a b s t r a c t

Mass transfer in gas–liquid systems has been significantly enhanced by recent developments in nanotechnology.
However, the influence of nanoparticles in liquid–liquid systems has received much less attention. In the present
study, both experimental and theoretical works were performed to investigate the influence of nanoparticles on the
mass transfer behaviour of drops inside a pulsed liquid–liquid extraction column (PLLEC). The chemical system of
kerosene–acetic acid–water was used, and the drops were organic nanofluids containing hydrophobic SiO2 nanopar-
ticles at concentrations of 0.01, 0.05, and 0.1 vol%. The experimental results indicate that the addition of 0.1 vol%
nanoparticles to the base fluid improves the mass transfer performance by up to 60%. The increase in mass trans-
fer with increased nanoparticle content was more apparent for lower pulsation intensities (0.3–1.3 cm/s). At high
pulsation intensities, the Sauter mean diameter (d32 ) decreased to smaller sizes (1.1–2.2 mm), leading to decreased
Brownian motion in the nanoparticles. Using an analogy for heat and mass transfer, an approach for determining
the mass diffusion coefficient was suggested. A new predictive correlation was proposed to calculate the effective
diffusivity and mass transfer coefficient in terms of the nanoparticle volume fraction, Reynolds number, and Schmidt
number. Finally, model predictions were directly compared with the experimental results for different nanofluids.
The absolute average relative error (%AARE) of the proposed correlation for the mass transfer coefficient and effective
diffusivity were 5.3% and 5.4%, respectively.
© 2014 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

Keywords: Pulsed column; Nanofluid drops; Mass transfer model; Effective mass diffusivity; Extraction efficiency

1. Introduction mixing leads to mass transfer faster than that predicted by

simple diffusion-based theories. The enhanced diffusion in
Nanofluids are stable suspensions of nanometre-sized par- nanofluids can therefore be utilised to improve mass transfer
ticles in conventional liquids. One advantage of nanofluids in unit operations.
is their improved heat transfer: it has been experimentally To determine the degree of thermal conductivity enhance-
shown in many instances that nanofluids have higher ther- ment in nanofluids, Krishnamurthy et al. (2006) experimen-
mal conductivities and improved convective heat transfer in tally observed the diffusion of a dye droplet in a water-based
comparison to host liquids (Chen et al., 2008; Ding et al., 2010; nanofluid and calculated the effective mass diffusivity of the
Garg et al., 2009; Li and Peterson, 2006; Wen et al., 2006). In dye in both the water and nanofluid. The investigation showed
addition to enhancing heat transfer, suspended nanoparticles that the suspended nanoparticles remarkably increase the
have been found to improve mass transfer processes inside mass transfer of the dye. Olle et al. (2006) measured the oxy-
binary nanofluids. Because convection and mass transfer are gen absorption rate in the presence of colloidal dispersions of
similar processes, several authors (Feng et al., 2005; Wen et al., magnetite (Fe3 O4 ) nanoparticles coated with oleic acid. Olle
2005; Olle et al., 2006) have proposed models that postulate et al. determined that this nanofluid improves the gas–liquid
that nanoscale convection induced by the Brownian motion oxygen mass transfer coefficient by more than 1.6-fold and
of the nanoparticles causes enhanced mixing. This increased the kL a value by up to 6-fold at nanoparticle concentrations

Corresponding author. Tel.: +98 21 61112213; fax: +98 2166967788.
E-mail address: hbahmany@ut.ac.ir (H. Bahmanyar).
Received 30 June 2013; Received in revised form 17 December 2013; Accepted 18 January 2014
0263-8762/$ – see front matter © 2014 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

Please cite this article in press as: Bahmanyar, A., et al., Mass transfer from nanofluid drops in a pulsed liquid–liquid extraction column. Chem.
Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.01.024
CHERD-1483; No. of Pages 11
2 chemical engineering research and design x x x ( 2 0 1 4 ) xxx–xxx

a interfacial area, m2 /m3 c continuous phase
A pulsation amplitude, cm d dispersed phase
C solute concentration in the dispersed phase, p particle
kg/m3 s solid
C0 initial concentration of solute in the dispersed ad auxiliary diffusivity
phase, kg/m3 bf base fluid
C* equilibrium concentration of solute in the dis- nf nanofluid
persed phase, kg/m3
d droplet diameter, m
d32 Sauter mean diameter of droplets, m below 1 vol% in an agitated, sparged reactor. Interestingly, the
di the ith group droplet diameter, m enhancement in kL a levelled off at a nanoparticle volume frac-
Dd molecular diffusivity of transferred component tion of approximately 1% (v/v).
in the dispersed phase, m2 /s Although the mechanism of the enhanced mass transport
DE effective diffusivity in Handlos–Baron equa- from the suspended nanoparticles is still not well understood,
tion, Eddy diffusivity in the Temos equation, some existing work assumed that the irregular Brownian
m2 /s motion of the nanoparticles is one of the main factors con-
DOE overall effective diffusivity, m2 /s tributing to the enhancement of mass transport (Xuan et al.,
E mass transfer enhancement 2003; Prasher et al., 2005).
E0 Eotvus number, 1 s–1 In another study, Lee et al. (2010) found that the absorption
E0c Eotvus number when the droplet has critical rate of ammonia by nanofluids is higher than that of conven-
diameter pulsation frequency, 1 s–1 tional fluids. Lee et al. also showed that despite the higher
f pulsation frequency, s−1 thermal conductivity of a CNT-nanofluid compared to a Al2 O3 -
fV fractional segmental volume of drop which is nanofluid, the absorption performance of the CNT-nanofluid
stagnant is not better than that of the Al2 O3 -nanofluid. Lee et al. spec-
g gravity acceleration, m2 /s ulated that this behaviour was due to the higher aspect ratio
H effective height of the column, cm of CNTs, which leads to less Brownian motion in comparison
k the dimensionless number in Eq. (14), =E0c /6 to spherical Al2 O3 nanoparticles.
kd dispersed phase mass transfer coefficient, m/s To date, the use of nanofluids to enhance mass trans-
kH empirical constant of Eq. (12), which varies fer has dealt only with gas–liquid systems. However, little
between 0 and 1 is known regarding the effects of nanoparticles on the mass
KB Boltzmann constant, =1.3807 × 10−23 J/K transfer behaviour of droplets in liquid–liquid systems. Such
L column height, m phenomena are examined both experimentally and theo-
ni the number of the ith group droplets retically in this study using hydrophobic SiO2 nanoparticles
Nu Nusselt number, dimensionless with droplet flows in a pulsed liquid–liquid extraction col-
PI pulsation intensity, cm/s umn (PLLEC). The decision to use a pulsed column in this
Pr Prandtl number, dimensionless study was based on the column’s ability to offer a more robust
Qc continuous phase flow rate, m3 /s medium for the investigation of nanoparticle influence on
Qd dispersed phase flow rate, m3 /s droplet behaviour. Experimental and computational investi-
 the modifying coefficient of the transferred gations were conducted to study the effect of nanoparticles on
component molecular diffusivity due to the the mass transfer characteristics of nanofluid drops. Employ-
internal circulations ing the analogy between heat and mass transfer, an approach
Re Reynolds number, dimensionless for determining the effective mass diffusivity and mass trans-
Sc Schmidt number, dimensionless fer coefficient was suggested and discussed. The final aim of
Sh Sherwood number, dimensionless this study was to clarify how nanoparticles can be utilised
t resident time of the dispersed phase in column, to improve mass transfer in liquid–liquid unit operations.
contact time, s Numerical analysis supports the concept that the Brownian
T temperature, K motion of the nanoparticles increases the internal circulation
Vt terminal velocity of a droplet, m/s of droplets. In this study, the enhanced diffusion coefficient
has been predicted, and its effect on mass transfer has been
Greek symbol calculated. These findings have direct implications for sep-
 interfacial tension, N/m aration technology and are promising for the application of
 nanoparticle volume fraction nanofluids in other liquid–liquid contact devices, such as
 density, kg/m3 rotary disc contractors and packed columns.
˛ thermal diffusivity, m2 /s
 dynamic viscosity, Ns/m2 2. Experiments
v kinematic viscosity, m2 /s
 ratio of the dispersed phase viscosity to contin- A schematic diagram of the experimental setup used in this
uous phase viscosity study is presented in Fig. 1. The column consisted of a 70 cm
ϕ dispersed phase hold-up long vertical Pyrex tube with an inner diameter of 90 mm.
The column contained 10 perforated stainless steel plates
regularly spaced 5 cm apart and was supported on an 8 mm

Please cite this article in press as: Bahmanyar, A., et al., Mass transfer from nanofluid drops in a pulsed liquid–liquid extraction column. Chem.
Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.01.024
CHERD-1483; No. of Pages 11
chemical engineering research and design x x x ( 2 0 1 4 ) xxx–xxx 3

Fig. 1 – A schematic diagram of the pulsed liquid–liquid extraction column used in this study.

diameter central rod. Two separating chambers were present Evaluation of the nanofluids’ stability was carried out by
on either side of the column. Holes 3 mm in diameter were the sedimentation method using UV–vis spectrometry, which
arranged in a 6 mm triangular pitch at each plate, providing a proved that the nanofluids remained stable for several days
nominal free fraction area of 0.20 (Khoobi et al., 2013). Pulsa- (Khoobi et al., 2013). Despite this observation, nanofluids were
tion was obtained via a newly designed pulsator, consisting of applied to the pulsed column shortly after their preparation.
an air compressor; a 3-way, 2-position direct-acting solenoid At the end of each test run, the absence of the nanoparticles
valve; and a microcontroller (AVR-8051) as a programmable in the continuous phase was ensured using UV–vis spectrom-
controller unit (PCU). The PCU was capable of periodically etry. The amorphous fumed hydrophobic silica powders are
energising the solenoid valve to allow flow of the compressed highly pure, with nanoparticles in the range of 5–30 nm and a
air for an adjustable period of time, pulsating the flow along density of 2200 kg/m3 . Their surfaces have been modified by
the column with an intensity of 0.3–2.3 cm/s. The PCU pro- OSi(CH3 )2 groups. The densities of water and kerosene are
vided an adjustable time-off in the range of 0.020–65.535 s. 996 and 800 kg/m3 , respectively. In addition, the viscosities of
This pulsator, which was specifically designed and built for the mentioned liquids are 1.0 and 1.67 mPa s, respectively.
this type of research, offers the advantage of 1 ms accuracy
in pulsation time, does not require maintenance typical of
3. Analysis
mechanical pump pulsators, and does not suffer from the lim-
itations of airlift pulsators. Two flowmeters were employed
to supply and monitor the flow rates of the continuous and
3.1. Nanofluid properties in modelling
dispersed phases.
The availability of correlations for various nanofluid prop-
Both the continuous and dispersed phases were mutu-
erties is critical when solving the governing conservation
ally saturated before being used in the experiments. At the
equations. However, generating a specific correlation for each
beginning of each test run, the continuous phase was ini-
type of nanofluid is cumbersome. To address this issue, Vajjha
tially allowed into the column from the top, which was filled
and Das (2010) carefully analysed all of the data from Namburu
to the specified height (valve 4, H = 44.5 cm). The dispersed
et al. (2007a,b) and Sahoo et al. (2009) to develop a gen-
phase was then fed into the column via a glassy nozzle (inner
eral correlation for the nanofluid viscosity. A correlation was
diameter 2 mm) from the bottom. The flowmeters were then
derived which expressed the viscosity in a non-dimensional
fixed to the specified ratio (Qc /Qd = 1.2). The pulsation ampli-
form, valid for nanofluids containing CuO, Al2 O3 , and/or SiO2
tude and frequency were adjusted to the desired values by
setting the time-on and time-off periods in the PCU. Acetic
acid (5 vol%) was added to the saturated kerosene to pro-
duce the base fluid for all of the systems under investigation, nf
= A1 e(A2 ) (1)
as listed in Table 1. The nanofluids were prepared by dis- bf
persing SiO2 nanoparticles supplied by the Wacker–Chemie
Company (HDK H18 with hydrophobicity of 18) into the base In the generalised correlation above, A1 and A2 are constants
fluid. All of the nanofluid samples were subjected to ultra- and are not functions of nanoparticle volume fraction (),
sonication (Heilscher ultrasound generator) for 1 h to ensure unlike previous correlations (Sahoo et al., 2009). The values of
proper nanoparticle dispersion. constants A1 and A2 are given for SiO2 nanoparticles in Table 2.

Please cite this article in press as: Bahmanyar, A., et al., Mass transfer from nanofluid drops in a pulsed liquid–liquid extraction column. Chem.
Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.01.024
CHERD-1483; No. of Pages 11
4 chemical engineering research and design x x x ( 2 0 1 4 ) xxx–xxx

Table 1 – Description of the chemical systems used.

System’s name Continuous phase Dispersed phase c (kg/m3 ) d (kg/m3 ) c (mPa s) d (mPa s) Interfacial tension

W-AA-K SW SK + AA 996 812 0.87 1.70 47.0

W-NF1 SW SK + AA + 0.01 vol% HDK H18 996 826 0.87 1.97 47.2
W-NF2 SW SK + AA + 0.05 vol% HDK H18 996 881 0.87 2.50 48.1
W-NF3 SW SK + AA + 0.1 vol% HDK H18 996 951 0.87 3.37 49.0

SW, saturated water; SK, saturated kerosene; AA, 0.05 vol% acetic acid; and NF, nanofluid.

The effective mass density of the nanofluid eff is calcu- when the experimental data (ϕ, d32 , and Qd ) throughout the
lated according to mixing theory (Vajjha and Das, 2010) using height of column are available, as is the case in this work, the
the following equation: contact time between two phases can be obtained using Eq.
(5). The mass transfer coefficient (Kd ) along the column height
eff = (1 − )bf + s (2) can then be obtained using Eqs. (3) and (4).

where bf and s are the mass densities of the base fluid and 3.3. Determination of the hydrodynamic
the solid nanoparticles, respectively. characteristics

3.2. Determination of the mass transfer characteristics To determine the drop size and hold-up, the pulsation inten-
sity in the column was varied while the mass flux ratio was
The overall mass transfer of the dispersed or continuous phase kept constant. When performing these hydrodynamic exper-
is a fundamental parameter in liquid–liquid extractor design. iments, the system must be operated in a steady state. To
In this study, the mass transfer coefficients were calculated ensure a steady state condition, the column was operated in
using a semi-empirical method. The mass balance for a drop recycle mode until no further changes in the phase interface
may be described by: could be detected. The drop sizes in the column were then
determined by taking digital photos of the column contents.
 d The drop size was evaluated as the Sauter mean diameter as:
Kd = − ln(1 − E) (3)
6t  3
where d32 =  i i2 (6)
ni di
C0 − C
E= (4) where ni is the number of corresponding droplet diameters
C0 − C∗
and di is the measured droplet diameter.
C0 is the solute concentration in the primary drop (before con- Hold-up was determined by the displacement method. At
tact), C is the solute concentration at a specified height from the end of each test run, the inlet and outlet flows were
the column bottom (44.5 cm for the setup in this work taken stopped simultaneously, and the dispersion was allowed to
from valve 4), and C* is the solute concentration of in equi- coalesce at the interface. The hold-up was then determined by
librium with the continuous phase. The solute (acetic acid) measuring the total volume and the dispersed phase volume
concentrations in the collected droplets (C) and the continuous as:
phase were determined by titration with a 0.1 M NaOH solution
and phenolphthalein as the indicator. At least three titrations ϕ= (7)
Vd + Vc
with a sampling volume of 10 mL were performed, and the
average value of the titrations was recorded. The uncertain- where Vd and Vc are the volumes of the dispersed and contin-
ties of the tests, including instrument and personal errors, uous phases, respectively.
were omitted. The solute concentration in equilibrium with
the continuous phase (C*) was determined using liquid–liquid 4. Theoretical predictions of mass transfer
equilibrium data for the ternary acetic acid–water–kerosene coefficients
system. The relationship between contact time (t), dispersed
phase volumetric flow rate (Qd ), and hold-up (ϕ) can be The overall mass transfer coefficient of the dispersed or con-
described as follows (Treybal, 1990): tinuous phase is a fundamental parameter in pulsed column
design. Several equations have been presented in the liter-
Qd t
L= (5) ature for calculating the overall mass transfer coefficient of

dispersed phase. These equations are generally based on three
theoretical models as presented in Table 3.
Table 2 – Constants of the viscosity correlation for The three mechanisms commonly used to determine the
nanofluids containing SiO2 nanoparticles. mass transfer rates of solutes in drops are: molecular diffu-
A1 A2 Average particle Concentration sion in a stagnant spherical drop (Newman, Eq. (8)), laminar
size (nm) (%) diffusion with circulation induced by relative motion of a drop
and the continuous phase (Kronig and Brink, Eq. (9)), and eddy
1.092 5.954 20 0 <  < 0.1
diffusion between internal toroidal streamlines (Handlos and
0.9693 7.074 50 0 <  < 0.06
1.005 4.669 100 0 <  < 0.06
Baron, Eq. (10)). The equations governing these mechanisms
are summarised in Table 3. Although small drops are often

Please cite this article in press as: Bahmanyar, A., et al., Mass transfer from nanofluid drops in a pulsed liquid–liquid extraction column. Chem.
Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.01.024
CHERD-1483; No. of Pages 11
chemical engineering research and design x x x ( 2 0 1 4 ) xxx–xxx 5

Table 3 – Previous equations for predicting the mass transfer correlation Kumar and Hartland (1999).
Models Expressions Remarks

6 1 −4n2 2 Dd t
Newman Kod = ln exp (8) Molecular diffusion in stagnant
6t 2 n2 d2


3 n=∞
n=1  64Dd t
Kronig and Brink Kod = ln 2
An exp − n (9) Laminar diffusion with circulation
6t 8 n=1 d2
induced by relative motion of drop
 and continuous phase
−d − n Dd tPe
Handlos and Baron Kod = ln 2 An exp (10) Eddy diffusion between internal
6t 128d2
n toroidal stream lines

subject to molecular diffusion rate control and large drops may where
exhibit Handlos–Baron behaviour, a transition between these  V d
two mechanisms based on Reynolds, Weber, or other factors DE = 3.29 × 10−4 d i
does not appear to be reliable.
   V d    
Other investigators have presented various correlations
× 1 − exp −3.29 × 10−4 d i d
using the Newman equation with effective diffusivity instead d d
of molecular diffusivity (Dd instead of Dd ). These equations
are based on the Newman equation, but effective diffusivity Vi for Re  1 is given by:
is used to account for internal circulation of droplets. Effec-    
tive diffusivity contains the effects of all known and unknown 2 + 3 1.45
Vi = 1− 0.5
Vt (18)
parameters that play an important role in calculating the exact 1 + (d d /c c ) Re0.5
value of mass transfer coefficients. Some of the equations pre-
sented by these investigators for calculating  are described while other models exist, most of them cannot be applied to
briefly below: the present study because they have been determined for a
Steiner (1986) proposed calculating the modifying coef- specific column, solvent or other condition (Jie et al., 2005;
ficient as a function of Reynolds and Schmidt number as Koncsag and Barbulescu, 2008).
5. Results and discussion
1 0.89
0.043 0.23
R = 1 + 0.177Re Sc (11)
1+ 5.1. Enhancement of mass transfer

This correlation is applicable when Re < 10. Hydrodynamic studies suggest that the column behaviour
Davies (1966) proposed a correlation in which the internal under near flooding conditions is not stable: small changes
part of the droplet is divided into two sections: a section with to the operating conditions (e.g., feed flow rates or pulsation
internal circulation and a stagnant section with molecular dif- intensity) lead to an over- or under-flooded operation. In this
fusion: study, the experiments were designed for the PLLEC to operate


kH Vt d
under no flooding conditions in accordance with prescribed
R= = fV + (1 − fV ) (12) operating conditions recommended in the literature (Godfrey
Dd 2048(1 + )Dd
and Slater, 1994). In analysing the data obtained from a PLLEC,
where fV is the fractional segment stagnant volume and kH is conventional pulsation intensity (PI), defined as the product of
the impurity coefficient and varies from 0 to 1. When kH = 0, pulsation frequency (f) and stroke length or pulsation ampli-
the system is completely pure. Conversely, when kH = 1, the tude (A), was used. The units of PI are cm/s.
system is completely impure.
PI = A · f (19)
Vt d
DE = (13)
2048(1 + ) Fig. 2 shows the enhancement of the mass transfer rate
with a nanoparticle addition for a mass flux ratio of 1:2. The
0.098E0 −2.21 mass transfer rate significantly increases with an increas-
fV = 1 − exp − , 6k = E0c (14) ing concentration of SiO2 nanoparticles. It was found that
the mass transfer rates for the nanofluids were improved
E0c shows the start of internal circulation of droplets and is by 4–21%, 5–47% and 5–60% for 0.01, 0.05 and 0.1 vol%,
given by: respectively. This enhancement is likely due to turbulence
caused by Brownian motion of nanoparticles inside droplets
 (Krishnamurthy et al., 2006). In particular, this enhancement
E0c = (15) was more significant at a lower PI, an effect that may be

attributed to the larger drop size obtained at a lower PI. Con-
Temos et al. (1993) present the relationship between eddy versely, the decreased mass transfer enhancement observed
diffusivity and molecular diffusivity in a simple way: at a high PI could be explained by the reduced droplet size,
which leads to reduced turbulence.
0.44DE Moreover, a clear peak in the mass transfer rate was
R=1+ (16)
Dd observed for a pulsation intensity of 1.8 cm/s and a

Please cite this article in press as: Bahmanyar, A., et al., Mass transfer from nanofluid drops in a pulsed liquid–liquid extraction column. Chem.
Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.01.024
CHERD-1483; No. of Pages 11
6 chemical engineering research and design x x x ( 2 0 1 4 ) xxx–xxx

Fig. 2 – Extraction efficiency for all systems at PI’s of (a) 0.3, (b) 0.7, (c) 1.3 and (d) 1.8 cm/s; mass flux ratio = 1.2.

nanoparticle concentration of 0.05 vol%, suggesting that the nanoparticles. Small drops are usually treated as rigid spheres
mass transfer rate may not improve above a certain PI (1.8 cm/s with negligible internal circulation; mass transfer inside the
in the present experiment). This peak could be due to the pro- drop is controlled by molecular diffusion (Newman model).
duction of small droplets at high PIs. Because small droplets Therefore, the average error of the Newman equation at
behave as rigid spheres, molecular diffusion governs the mass higher PIs, at which molecular diffusion governs the system,
transfer in the system. is less than the error at lower PIs.
Although the Handlos–Baron equation is more accurate
5.2. Comparison with previous models compared to the other available equations, an interesting
observation can be made when comparing different nanopar-
The mass transfer coefficient (Kd ) was calculated for the vari- ticle contents and PIs. Having established that the Kd s
ous systems tested using the theoretical equations presented calculated using the Newman and Kronig–Brink equations
in Section 4. These results were then compared with the are strongly affected by nanoparticle presence, we expected
experimentally measured mass transfer coefficients, and the a similar nanoparticle influence on the mass transfer coef-
absolute errors are presented in Table 4. ficient calculated by the Handlos–Baron equation. However,
The Newman and Kronig–Brink equations underestimate such an influence was not observed and may be explained
the mass transfer coefficient by 48% and 34%, respectively, by the theoretical background of the Handlos–Baron equation,
regardless of the nanoparticle concentration and pulsation which considers turbulent circulation in each drop in addition
intensity. This difference is likely because these models do to the circulation from Brownian motion of the nanoparti-
not consider circulation within the drop. Interestingly, the cles. In other words, the Handlos–Baron equation overpredicts
mass transfer coefficient obtained from the Handlos–Baron the data because it is based on fully turbulent internal circu-
equation is on average 24% higher than the experimentally lations, a condition which is not reached in practice: drops
measured value. The Handlos–Baron equation therefore yields experience a transition state between laminar and fully turbu-
a more precise and accurate estimate of the mass transfer lent flow. This suggests that the Newman equation is the most
coefficient, likely due to the model’s consideration of turbulent proper equation for the modelling of mass transfer coefficients
circulation within the drop. in the presence of nanoparticles.
It can also be observed that regardless of the chemical The mass transfer coefficient in drops is a strong function
system, the absolute errors calculated from the Newman of drop size; therefore, the comparison between experimen-
and Kronig–Brink equations are larger for lower PIs (0.7 and tal dispersed phase mass transfer coefficients and the results
1.3 cm/s) and lower at higher PIs (1.8 and 2.3 cm/s). The obtained from previous equations is shown in Fig. 3 as a func-
average absolute error calculated from the Newman model tion of drop diameter.
is approximately −66% at low PIs (PI ≤ 1.3 cm/s) and −24% at The results of Fig. 3 support the hypothesis that the
high PIs (PI ≥ 1.8 cm/s). In fact, higher PIs result in a smaller Brownian motion of the nanoparticles increases the internal
Sauter mean diameter and lesser Brownian motion of the circulation within drops. The experimental results are in good

Please cite this article in press as: Bahmanyar, A., et al., Mass transfer from nanofluid drops in a pulsed liquid–liquid extraction column. Chem.
Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.01.024
CHERD-1483; No. of Pages 11
chemical engineering research and design x x x ( 2 0 1 4 ) xxx–xxx 7

Table 4 – Calculated values for the mass transfer coefficients, Kd × 105 (m/s), and percentage of errors.
PI (cm/s) d32 (mm) Experimental Newman Kronig and Brink Handlos and Baron

Kd Kd Error% Kd Error% Kd Error%

W-AA-K system
0.3 3.7 7.10 2.64 −63 3.20 −55 8.67 23.3
0.7 3.37 3.15 1.26 −60 1.77 −44 3.91 24.2
1.3 3.17 2.71 1.59 −41 1.89 −30 3.36 24.3
1.8 2.11 1.48 1.07 −28 1.25 −16 1.84 24.5
2.3 1.61 1.07 0.81 −24 0.92 −14 1.33 24.5

W-NF1 system
0.3 3.47 3.11 1.15 −63 1.76 −43 3.86 24.2
0.7 3.12 2.56 1.03 −60 1.54 −40 3.18 24.3
1.3 2.60 1.94 0.84 −57 1.28 −34 2.41 24.4
1.8 1.94 1.26 0.82 −35 0.94 −25 1.57 24.5
2.3 1.18 0.69 0.51 −27 0.53 −24 0.87 24.6

W-NF2 system
0.3 3.82 4.21 1.17 −72 1.65 −61 5.22 24.0
0.7 3.32 2.72 1.02 −62 1.42 −48 3.38 24.3
1.3 2.83 2.01 0.88 −56 1.21 −40 2.50 24.4
1.8 2.20 1.28 0.81 −37 0.95 −26 1.59 24.6
2.3 1.76 0.81 0.61 −25 0.69 −15 1.01 24.7

W-NF3 system
0.3 3.78 4.44 1.12 −75 1.58 −64 5.50 23.9
0.7 3.30 2.55 0.95 −63 1.30 −49 3.17 24.3
1.3 2.67 1.87 0.76 −59 1.19 −36 2.32 24.4
1.8 2.09 0.90 0.71 −21 0.77 −14 1.12 24.7
2.3 1.67 0.62 0.59 −5 0.65 6 0.77 24.7

agreement with the Handlos–Baron equation, while other cor- Similar to the parameters for the molecular diffusion, the
relations underpredict the mass transfer coefficients of the effective Prandtl number (Pr) and Schmidt number (Sc) are
nanofluids that were investigated in this study. There is no introduced for heat and mass transfer inside a nanofluid with
significant variation of mass transfer coefficient between the effective viscosity vnf
experimental data and the values obtained from previously
published correlations for smaller drops. This agreement can nf nf
be attributed to the fact that the internal circulation within Prnf = and Scnf = (21)
˛nf Dnf
very small drops is negligible (Re < 1), and the mass transfer
mechanism within these drops is molecular diffusion (New-
man model). On the other hand, for a larger drop diameter, in It is assumed that the analogous functional relationship
which turbulent circulation occurs within the drop, the New- between temperature and concentration in the differential
man and Kronig–Brink equations fail to predict the dispersed equations for heat and mass transfer allows for the heat trans-
phase mass transfer coefficient of the investigated nanofluids. fer equation to be converted to a mass transfer equation by
This may be because these correlations assume molecular dif- simply replacing the Prandtl number with the Schmidt num-
fusion and laminar circulation inside the drops while ignoring ber and the Nusselt number (Nu) with the Sherwood number
Brownian motion of the nanoparticles. (Sh) (Eckert et al., 2001). The heat and mass transfer analogy
means that the Nusselt number for heat transfer and the Sher-
wood number for mass transfer inside a binary nanofluid can
5.3. Predictive correlation for the effective diffusivity of
be described by Nunf = f (Renf , Prnf ) and Shnf = f (Renf , Scnf ).
nanofluids by means of a heat and mass transfer analogy
An appreciation for the similarity between heat and mass
transfer can be gained using dimensional analysis.
Although enhancement effects in binary nanofluids have been
For spherical nanoparticles, the Nusselt number may be
observed, no sophisticated theory has been established to
deduced from previous work (Xuan and Li, 2003):
describe mass transport process inside binary nanofluids. It
is now acknowledged that the stochastic Brownian motion of
 m m
suspended nanoparticles and induced microscopic convection Nunf = 2 + cm1 Renf2 Prnf3 (22)
of the fluids around the nanoparticles are the two most impor-
tant factors for the enhancement of mass transport processes
in binary nanofluids. where m3 is dependent upon the heat transfer mode of fluid
One possible approach to determine the mass diffusivity of cooling or heating. The coefficient c and the exponents m1 and
a binary nanofluid involves the heat and mass transfer anal- m2 can be determined from the experimental data by a proper
ogy. The Reynolds number of a nanoparticle can be expressed data-reduction procedure.
as (Nagy et al., 2007) The analogy principle between heat and mass transfer
leads to the Sherwood number:

vp dp 1 18KB T
Renf = = (20)
nf nf dp  Shnf = 2 + cm1 Rem 2 m3
nf Scnf (23)

Please cite this article in press as: Bahmanyar, A., et al., Mass transfer from nanofluid drops in a pulsed liquid–liquid extraction column. Chem.
Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.01.024
CHERD-1483; No. of Pages 11
8 chemical engineering research and design x x x ( 2 0 1 4 ) xxx–xxx

Fig. 3 – Dispersed phase mass transfer coefficient as a

function of droplet diameter for the (a) W-NF1 (0.01 vol%),
(b) W-NF2 (0.05 vol%) and W-NF3 (0.1 vol%) systems.

A similar expression for the mass transfer coefficient

around a sphere in the laminar flow regime was described by
Bird et al. (2002).
With regard to heat transfer enhancement, the microcon-
vection of nanoparticles increases the macroscopic thermal
conductivity of nanofluids. Similar to Prasher’s formula
(Prasher et al., 2006), the apparent thermal conductivity of
binary nanofluids can be expressed as: Fig. 4 – Comparison of the experimental data with the
calculated values for the (a) W-NF1 (0.01 vol%), (b) W-NF2
(0.05 vol%) and W-NF3 (0.1 vol%) systems.
 m m
knf = kf (1 + cm1 Renf2 Prnf3 ) (24)
By introducing the auxiliary diffusivity Dad , the effective
From the analogy of mass and heat transfer, Eqs. (22) and mass diffusivity of the binary nanofluid can be expressed as:
(23), the apparent mass diffusion coefficient of binary nanoflu-
ids can be derived: Dnf = D0 + Dad (26)

where D0 is the molecular diffusivity of W-AA-K sys-

Dnf = Df (1 + cm1 Rem m3
nf Scnf )
2 (25)
tem (0.0 vol%) calculated at various PIs using the Newman

Please cite this article in press as: Bahmanyar, A., et al., Mass transfer from nanofluid drops in a pulsed liquid–liquid extraction column. Chem.
Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.01.024
CHERD-1483; No. of Pages 11
chemical engineering research and design x x x ( 2 0 1 4 ) xxx–xxx 9

Fig. 5 – Illustration of the influence of the pulsation intensity on the size of droplets.

equation, which is based on the continuity equation with As shown in Fig. 4, it is obvious that the previously
appropriate initial and boundary conditions. published correlations have relatively high average absolute
The main advantage of this approach is that the princi- errors (98%) when used to calculate mass transfer coefficients.
pal effects of external forces (e.g., PI) and Brownian motion The coefficients calculated from the correlation proposed by
from nanoparticles are taken into consideration in terms of D0 Steiner, Davis, and Temos et al. are much higher than the true
and Dad , respectively. Without these considerations, the exper- coefficients. The Handlos–Baron equation shows less average
imental data could not be fitted with acceptable accuracy. error than the other equations, but the average error (24%)
A comparison between Eqs. (25) and (26) yields of this equation is relatively high and unacceptable for pre-
cise design. In contrast, the figures indicate that the equation
proposed in this work can accurately estimate the dispersed
Dad = cn1 Rennf2 Scnf
D0 (27)
phase mass transfer coefficient. The dispersed phase mass
transfer coefficient calculated with this model reproduces the
To develop a new predictive correlation for the effective experimental data with an average error of only 5.3%. Table 5
mass diffusivity of nanofluids, experimental data were fit to shows the average error (%AARE) of the predicted Kd s for pre-
Eq. (27) to determine the curve fit coefficients c, n1 , n2 and n3 vious and present correlations. The %AARE for the N data is
using least squares, yielding Eq. (28): calculated as follows:

1   model − experiment 
Dad = 16500.203 Re0.039 Sc−1.064 (28)
 × 100 (29)
The calculated values of the effective diffusivity were then
used in Eq. (8) to calculate the mass transfer rate enhance- The effective mass diffusivity is also calculated by the cor-
ment. The overall mass transfer coefficients of the dispersed relation presented above, and the results are shown in Fig. 6 .
phase were then calculated using Eq. (3).
It should be noted that the Schmidt number physically
relates the relative thickness of the hydrodynamic layer and
the mass transfer boundary layer. The negative power for
the Schmidt number in Eq. (28) may be ascribed to liquid
molecular layering and suggests that the liquid–nanoparticle
interface decreases the mass boundary layer and lowers the
effective mass diffusivity. This view is in line with the works
of some investigators (Shenogin et al., 2004a,b; Nan et al.,
2003; Gao et al., 2007), who have ascribed the adverse effect of
liquid–nanoparticle interface to Kapita interfacial resistance.
In Fig. 4, the experimentally determined dispersed phase
overall mass transfer coefficients are compared with theo-
retical models and the proposed model for three chemical
systems with different nanoparticles contents of 0.01, 0.05 and
0.1 vol%, corresponding to W-NF1 , W-NF2 and W-NF3 , respec-
Overall, Kd decreases with increasing PI. Based on experi-
mental observations, the mean drop diameter decreases with
increasing PI, leading to a longer drop residence time. How-
ever, as expressed by Eq. (3), Kd decreases with both effects Fig. 6 – Comparison of the experimental Dnf with present
(Fig. 5). and previous equations.

Please cite this article in press as: Bahmanyar, A., et al., Mass transfer from nanofluid drops in a pulsed liquid–liquid extraction column. Chem.
Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.01.024
CHERD-1483; No. of Pages 11
10 chemical engineering research and design x x x ( 2 0 1 4 ) xxx–xxx

Table 5 – The %AARE of the predicted Kd for previous and present correlations.
System’s name Eq. (28) Eq. (8) Eq. (9) Eq. (10) Eq. (11) Eq. (12) Eq. (17)
This New- Kronig Handlos Steiner Davis Temos
work man and Brink and Baron

W-NF1 5.7 48.4 33.2 24.4 104.7 205.6 109.3

W-NF2 4.0 50.4 38.0 24.4 113.4 263.6 137.2
W-NF3 6.3 44.6 33.8 24.6 115.7 263.4 129.7

The ability of the model presented in this work to pre-

dict mass transfer enhancements is shown in Fig. 7. This
figure depicts the variation of the mass transfer enhancement
at various PIs compared to experimental data for all chemi-
cal systems. According to the average of the residuals, very
good agreement between the experimental data and corre-
lated curves is achieved for all chemical systems.

6. Conclusions

The mass transfer characteristics of SiO2 –kerosene nanofluid

drops were investigated experimentally in a pulsed
liquid–liquid extraction column. Different pulsation intensi-
ties were maintained for fixed mass flow rates of dispersed
(Qd ) and continuous (Qc ) phases (with ratio Qc /Qd = 1.2). Mass
transfer occurred from the dispersed to the continuous
phase. The experimental results showed that the suspended
nanoparticles remarkably increase the mass transfer per-
formance of the base fluid. The addition of 0.1 vol% silica
particles to kerosene enhanced the mass transfer rate by 60%.
A new correlation was proposed to calculate the effective
mass diffusivity in terms of the nanoparticle volume fraction,
the Reynolds number, and the Schmidt number. It was shown
that the mass transfer coefficients calculated using this
correlation agreed well with the experimentally determined
values. Compared with other theoretical models that are
available in the literature, the model presented here displays
higher accuracy and precision.


Bird, R.B., Stewart, W.E., Lightfoot, E.N., 2002. Transport

Phenomena, second ed. Wiley, New York.
Chen, H., Yang, W., He, Y., Ding, Y., Zhang, L., Tan, C., Lapkin, A.A.,
Bavykin, D.V., 2008. Heat transfer and flow behaviour of
aqueous suspensions of titanate nanotubes (nanofluids).
Powder Technol. 183, 63–72.
Davies, J.T., 1966. Interfacial renewal. Chem. Eng. Prog. 62, 89–94.
Ding, Y., Chen, H., Musina, Z., Jin, Y., Zhang, T., Witharana, S.,
Yang, W., 2010. Relationship between the thermal conductivity
and shear viscosity of nanofluids. Phys. Scr. T139, 014078.
Eckert, E.R.G., Sakamoto, H., Simon, T.W., 2001. The heat/mass
transfer analogy factor, Nu/Sh, for boundary layers on turbine
blade profiles. Int. J. Heat Mass Transfer 44, 1223–1233.
Feng, W., Wen, J., Fan, J., Yuan, Q., Jia, X., Sun, Y., 2005. Local
hydrodynamics of gas–liquid nanoparticles three-phase
Fig. 7 – Experimental mass transfer enhancement at fluidization. Chem. Eng. Sci. 60, 6887–6898.
various PI’s for (a) W-NF1 , (b) W-NF2 and (c) W-NF3 . The Gao, L., Zhou, X., Ding, Y.L., 2007. Effective thermal and electrical
solid lines represent correlated curves for the mass transfer conductivity of carbon nanotube composites. Chem. Phys.
enhancement calculated using Eq. (28). Lett. 434, 297–300.
Garg, P., Alvarado, J.L., Marsh, C., Carlson, T.A., Kessler, D.A.,
Annamalai, K., 2009. An experimental study on the effect of
ultrasonication on viscosity and heat transfer performance of
Calculating the %AARE reveals that the prediction error for the
multi-wall carbon nanotube-based aqueous nanofluids. Int. J.
model presented in this work is 5.4%. The comparison of the
Heat Mass Transfer 52, 5090–5101.
experimental Dnf with those obtained from previous equations Godfrey, J.C., Slater, M.J., 1994. Liquid–Liquid Extraction
is also presented in Fig. 6. The average error of these models Equipment, second ed. John Wiley & Sons, New York, pp.
is −74%. 227–305.

Please cite this article in press as: Bahmanyar, A., et al., Mass transfer from nanofluid drops in a pulsed liquid–liquid extraction column. Chem.
Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.01.024
CHERD-1483; No. of Pages 11
chemical engineering research and design x x x ( 2 0 1 4 ) xxx–xxx 11

Jie, Y., Weiyang, F., Li., H.Z., 2005. Effect of packing on drop Prasher, R.S., Bhattacharya, P., Phelan, P.E., 2005. Thermal
swarms extraction of high viscosity solvents. conductivity of nanoscale colloidal solutions (nanofluids).
Hydrometallurgy 78 (1–2), 30–34. Phys. Rev. Lett. 94, 025901.
Khoobi, N., Bahmanyar, A., Molavi, H., Bastani, D., Mozdianfard, Prasher, R., Bhattacharya, P., Phelan, P.E., 2006.
M.R., Bahmanyar, H., 2013. Study of droplet behavior along a Brownian-motion-based convective–conductive model for the
pulsed liquid–liquid extraction column in the presence of effective thermal conductivity of nanofluids. J. Heat Transfer
nanoparticles. Can. J. Chem. Eng. 91, 506–515. 128, 588–595.
Koncsag, C.I., Barbulescu, A., 2008. Modelling the removal of Sahoo, B.C., Vajjha, R.S., Ganguli, R., Chukwu, G.A., Das, D.K.,
mercaptans from liquid hydrocarbon streams in structured 2009. Determination of rheological behavior of aluminum
packing columns. Chem. Eng. Process. 47, 1717–1725. oxide nanofluid and development of new viscosity
Krishnamurthy, S., Bhattacharya, P., Phelan, P.E., Prasher, R.S., correlations. Petrol. Sci. Technol. 27 (15), 1757–1770.
2006. Enhanced mass transport in nanofluids. Nano Lett. 6, Shenogin, S., Bodapati, A., Xue, L., Ozisik, R., Keblinski, P., 2004a.
419–423. Effect of chemical functionalization on thermal transport of
Kumar, A., Hartland, S., 1999. Correlations of mass transfer carbon nanotubes composites. Appl. Phys. Lett. 85,
coefficients in single drop systems and liquid–liquid 2229–2231.
extraction columns. Trans. Inst. Chem. Eng. 77, 372–384. Shenogin, S., Xue, L.P., Ozisik, R., Keblinski, P., Cahill, D.G., 2004b.
Lee, J.K., Koo, J., Hong, H., Kang, Y.T., 2010. The effect of Role of thermal boundary resistance on the heat flow in
nanoparticles on absorption heat and mass transfer carbon nanotube composites. J. Appl. Phys. 95, 8136–8144.
performance in NH3 /H2 O binary nanofluids. Int. J. Refrig. 33, Steiner, L., 1986. Mass transfer rates from single drops and drop
269–275. swarms. Chem. Eng. Sci. 41 (8), 1979–1986.
Li, C.H., Peterson, G.P., 2006. Experimental investigation of Temos, J., Pratt, H.R.C., Stevens, G.W.,1993. Comparison of tracer
temperature and volume fraction variations on the effective and bulk mass transfer coefficients for droplets. In:
thermal conductivity of nanoparticle suspensions Proceedings of the ISEC, vol. 93. Elsevier, Amsterdam, pp.
(nanofluids). J. Appl. Phys. 99, 084314. 1770–1777.
Nagy, E., Feczko, T., Koroknai, B., 2007. Enhancement of oxygen Treybal, R.E., 1990. Mass Transfer Operations, third ed. McGraw
mass transfer rate in the presence of nanosized particles. Hill, Japan, pp. 488–490.
Chem. Eng. Sci. 62, 7391–7398. Vajjha, R.S., Das, D.K., 2010. Measurements of Thermophysical
Namburu, P.K., Kulkarni, D.P., Dandekar, A., Das, D.K., 2007a. Properties of Nanofluids and Computation of Heat Transfer
Experimental investigation of viscosity and specific heat of Characteristics. LAP Lambert Academic Publishing,
silicon dioxide nanofluids. Micro Nano Lett. 2 (3), 67–71. Saarbrücken-Germany, ISBN: 978-3-8383-7214-3.
Namburu, P.K., Kulkarni, D.P., Misra, D., Das, D.K., 2007b. Viscosity Wen, J., Jia, X., Feng, W., 2005. Hydrodynamic and mass transfer of
of copper oxide nanoparticles dispersed in ethylene glycol gas–liquid–solid three phase internal loop airlift reactors with
and water mixture. Exp. Therm. Fluid Sci. 32, 67–71. nanometer solid particles. Chem. Eng. Technol. 28, 53–60.
Nan, C.W., Shi, Z., Lin, Y., 2003. A simple model for thermal Wen, D.S., Ding, Y.L., Williams, R.A., 2006. Pool boiling heat
conductivity of carbon nanotube-based composites. Chem. transfer of aqueous TiO2 -based nanofluids. J. Enhanced Heat
Phys. Lett. 375, 666–669. Transfer 13, 231–244.
Olle, B., Bucak, S., Holmes, T.C., Bromberg, L., Hatton, A., Wang, Xuan, Y., Li, Q., 2003. Investigation on convective heat transfer
D.I.C., 2006. Enhancement of oxygen mass transfer using and flow features of nanofluids. J. Heat Transfer 125, 151–155.
functionalized magnetic nanoparticles. Ind. Eng. Chem. Res. Xuan, Y., Li, Q., Hu, W., 2003. Aggregation structure and thermal
45, 4355–4363. conductivity. AIChE J. 49, 1038–1043.

Please cite this article in press as: Bahmanyar, A., et al., Mass transfer from nanofluid drops in a pulsed liquid–liquid extraction column. Chem.
Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.01.024