Chemical Engineering Science 63 (2008) 3366 -- 3378

Contents lists available at ScienceDirect

Chemical Engineering Science

journal homepage: w w w . e l s e v i e r . c o m / l o c a t e / c e s

Jet breakup and droplet formation in near-critical regime of carbon

dioxide--dichloromethane system
Lai Yeng Lee a,1 , Liang Kuang Lim a,1 , Jinsong Hua b , Chi-Hwa Wang a,∗
a Department of Chemical and Biomolecular Engineering, National University of Singapore, 4 Engineering Drive 4, Singapore 117576, Singapore
b Institute of High Performance Computing, 1 Science Park Road, #01-01 The Capricorn Singapore Science Park II, Singapore 117528, Singapore

A R T I C L E I N F O A B S T R A C T

Article history: The jet breakup and droplet formation mechanism of a liquid in the near-critical conditions of a
Received 8 December 2007 solvent--antisolvent system is examined with high-speed visualization experiments and simulated using
Received in revised form 7 April 2008 a front tracking/finite volume method. The size of droplets formed under varying system pressure at
Accepted 7 April 2008
various jet breakup regimes is measured with a Global Sizing Velocimetry, using the shadow sizing
Available online 11 April 2008
method. A stainless steel nozzle with 0.25 mm I.D and 1.6 mm O.D was used in this study. Experiments
Keywords:
were performed at fixed temperature of 35 ◦ C and system pressure in the range from 61 to 76 bar in
Computation the near-critical regime of the DCM--CO2 . At the near mixture critical regime for DCM--CO2 mixture, the
Supercritical fluid miscibility between the two fluid phases increases and the interfacial tension diminishes. This phase
Visualization behavior has important applications in particle formation using gas antisolvent (GAS) and supercritical
Droplet formation antisolvent (SAS) processes. The jet breakup and droplet formation in the near-critical regime is strongly
Particle formation dependent on the changes in interface tension and velocity of the liquid phase. An understanding of the
Multiphase flow droplet formation and jet breakup behavior of DCM--CO2 in this regime is useful in experimental design
for particle fabrication using SAS method.
© 2008 Elsevier Ltd. All rights reserved.

1. Introduction
Particle formation is an important application of supercritical
fluid technology (Jung and Perrut, 2001). Properties of supercritical
fluids that favor particle formation include liquid-like density and
gas-like viscosity (Richard and Dechamps, 2004). The well known
techniques for particle formation using supercritical fluids include
the rapid expansion of supercritical solutions (RESS) (Jung and
Perrut, 2001; Richard and Dechamps, 2004; Tom and Debenedetti,
1991; Debenedettiet al., 1993) and the gas/supercritical antisolvent
(GAS/SAS) processes (Jung and Perrut, 2001; Richard and Dechamps,
2004; Tom and Debenedetti, 1991; Randolph et al., 1993;
Subramaniam et al., 1997; Chattopadhyay and Gupta, 2001a,b,
2002a--c; Reverchon et al., 2003; Henczka et al., 2005).
Carbon dioxide (CO2 ) is used extensively as a supercritical fluid
in particle fabrication and pharmaceutical applications, which is
attributed to its desirable properties such as relatively accessible
critical point at temperature of 31.1 ◦ C and pressure at 73.8 bar,
∗ Corresponding author. Tel.: +65 6516 5079; fax: +65 6779 1936.
E-mail address: chewch@nus.edu.sg (C.-H. Wang).
1 The authors have equal contributions to this work.
abundance and its low toxicity (Jung and Perrut, 2001; Richard and
Dechamps, 2004; Tom and Debenedetti, 1991). Since most organic
solvents are miscible with CO2 at supercritical conditions, a low
residual solvent content can be easily achieved in the final product
without extensive downstream purification to remove excess organic
solvent (Ruchatz et al., 1997). In particular, the SAS process has been
used in numerous studies for pharmaceutical products. In the SAS
process, the substrate of interest is first dissolved in a suitable organic
solvent. The organic solution is then injected into supercritical fluid
which acts as an antisolvent by rapid mass transfer, resulting in
precipitation of the substrate.
Several factors may affect the particle size and properties
achieved from SAS process. This includes the phase behavior of the
ternary mixture and the hydrodynamics of the solution injected into
the supercritical phase. Considerable literature suggests that the
controlling parameter for particle size in the SAS process is the rate
of mass transfer (Reverchon et al., 2003; Henczka et al., 2005). This
is influenced by both the spray hydrodynamics of the organic solu-
tion and thermodynamic properties of the supercritical fluid phase.
Perez de Diego et al. (2005) reported the mechanism of particle
formation in subcritical and supercritical regimes for precipitation
from compressed antisolvent (PCA) process. The jet disintegration
mechanism for operation in subcritical and supercritical conditions

0009-2509/$ - see front matter © 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2008.04.015
 L.Y. Lee et al. / Chemical Engineering Science 63 (2008) 3366 -- 3378
was different and hence particles properties were also different. The
effect of flow rate on particle size in the subcritical regime was
reported. Carretier et al. (2003) investigated the hydrodynamics of
the SAS process by evaluating the macro and micromixing within
the precipitation vessel.
It was observed that most studies on SAS were carried out in
atomization of a gaseous plume mode in the turbulent jet regime
(Carretier et al., 2003; Shekunov et al., 2001). Varying parameters
such as nozzle diameter, solution flow rate, pressure and temper-
ature have no significant effect on the size and size distribution of
final particles obtained. Final particle size achieved is mainly depen-
dent upon the extent of mixing between the organic solvent and CO2
phases. Smaller and more uniform sized particles could be achieved
using coaxial (Henczka et al., 2005) and ultrasonic nozzles (Randolph
et al., 1993; Subramaniam et al., 1997) which provides better mixing
between the two phases.
The dichloromethane (DCM)--CO2 system is one of the most com-
monly used for fabrication of polymeric particles using SAS process
(Perez de Diego et al., 2005; Gokhale et al., 2007; Obrzut et al., 2007),
and the various flow regimes in the free jet of DCM in CO2 have been
reported (Kerst et al., 2000). DCM is a good solvent for Poly L lactide
(PLA) and several other pharmaceutical compounds. The mixture
critical pressure of DCM--CO2 at 35 ◦ C is 78 bar (Gokhale et al., 2007)
with a composition between 1.0 and 1.9 mol% DCM in CO2 (Reaves
et al., 1998). In this work, the droplet formation and breakup of DCM
jet in CO2 is studied. By operating in the regime close to the mixture
critical point (MCP) of DCM--CO2 system, and with careful manipu-
lation of the Reynolds number of the liquid phase, it is possible to
achieve dripping or Rayleigh disintegration of droplets where more
monodispersed droplets could be achieved. In the dripping mode,
the droplet size is a very strong function of the interface tension be-
tween the two phases. In the Rayleigh disintegration mode or jetting
mode, droplet size is approximately 1.5 times the jet diameter.
2. Materials and methods
2.1. Materials
Compressed CO2 (Air Liquide Paris, France) was purchased from
Soxal (Singapore Oxygen Air Liquide Pte Ltd.). DCM (DS1432, HPLC
grade) was purchased from Tedia (Fairfield, OH, USA). PLA was pur-
chased from Sigma Aldrich (St. Louis, MO, USA).
2.2. Experimental setup and visualization
The experimental setup used in this study is shown in Fig. 1.
Liquefied CO2 (C1, Polyscience refrigerating circulator) was intro-
duced into the high pressure vessel (HP, Jerguson 12-T-32, 70 cm3 )
by means of a high pressure pump (P2, Jasco HPLC pump) to the re-
quired pressure. The temperature in the vessel was controlled and
F2 BPR
V1
C2
P1
HP
C1 V2
CO2
P2
Fig. 1. Experimental setup. P1, P2: pumps for the polymer solution and liquified
CO2 , respectively. HP: high pressure vessel. C1, C2: temperature controllers. BPR:
back pressure regulator. F2: flowrate regulator. Temperature of the setup for droplet
and particle formation studies was fixed at 35 ◦ C.
3367
maintained at 35 ◦ C by use of a circulating heated water bath (C2,
Polyscience 712 circulator). The system pressure used lies in the
near-critical regime of DCM--CO2 from 61 to 76 bar. The composition
of the antisolvent phase in the vessel is maintained at 0--1% DCM in
CO2 at all times.
The high pressure vessel used in this study has two parallel
borosilicate glass windows which allow visualization along the en-
tire length of the vessel during the antisolvent process. High magnifi-
cation images were captured with a digital SLR camera (D2H, Nikon,
Japan) with a macro lens (105 mm Micro Nikkor, Nikon, Japan). High-
speed camera was (FastCam PCI, Photron Inc., USA) used with the
same macro lens to capture consecutive snapshots during the droplet
formation process. The high-speed camera can capture images up to
2000 frames per second. Shadow sizing system (Dantec Dynamics,
Denmark) is used for measurement of DCM droplet size during the
SAS process. This system is equipped with a high sensitivity cam-
era (Highsense MKII, Dantec Dynamics, Denmark) and is operated
based on backlighting with laser light and image analysis software
with an advanced edge detection algorithm (FlowManager PC, Dan-
tec Dynamics, Denmark).
2.3. Numerical simulation
In order to aid the current investigation, numerical simulation
was employed to study the formation process of the dripping mode
and the Rayleigh disintegration mode. The two-fluid system consist-
ing of a liquid jet emerging from a nozzle into a gaseous medium
is modeled in this study. A two-dimensional (2D) axis-symmetric
cylindrical-coordinate system was used, and the numerical simula-
tion was done using a front tracking/finite volume method.
2.3.1. Front tracking/finite volume method
The recent development of a front tracking/finite volume method
for multiphase flow simulation has provided a novel and robust
method to treat the moving interface between two fluids with large
density difference (Unverdi and Tryggvason, 1992; Hua and Lou,
2007). In this study, the front tracking/finite volume method pro-
posed by Hua and Lou (2007) is modified to take into account the
changes in the surface tension when the MCP of the DCM--CO2 sys-
tem is approached. In this method, a stationary fixed background
mesh is used throughout the whole computational domain, and a
set of adaptive front mesh is used to mark the moving interface.
Only one set of the momentum and continuity equations is solved in
the whole computational domain by treating the different phases as
one single fluid with variable material properties. The distributions
of physical properties such as density and viscosity are calculated
according to the position of the interface and the property differ-
ence between two fluid phases. The physical properties of the bulk
phases of DCM and CO2 (at 35 ◦ C) used in this study are summarized
in Table 1. The surface tension stress is computed on the front mesh
and then distributed to the fixed background mesh through the use
of a dirac-delta like distribution function as a body force term in
the momentum equation. The position of the interface is advected
explicitly with the velocity interpolated from the flow field on the
Table 1
Summary for physical parameters of CO2 and DCM at 35 ◦ C
DCM properties at 35 ◦ C
Density, L (kg m−3 ) 1289
Viscosity, L (Pa s) 3.7 × 10−4
61 bar 71 bar 76 bar
CO2 properties at 35 ◦ C
Density, G (kg m−3 ) 160 220 270
Viscosity, G (Pa s) 2.81×10−4 3.83×10−4 4.24×10−4
Interfacial tension coefficient,  (N m−1 ) 5.359×10−3 2.230×10−3 1.126×10−3
3368 L.Y. Lee et al. / Chemical Engineering Science 63 (2008) 3366 -- 3378

background gird. Such method has been applied successfully to var- indicator function can be written in the form of an integral over the
ious interfacial flow problems (Tryggvason et al., 2001). region where the liquid phase lies

2.3.2. Governing equations and boundary conditions I( 
x , t) = (x − x )(y − y ) dx dy (5)
The flow field simulation is formulated by solving the govern- 
ing equations of mass conservation and the momentum equations.
where (x − x ) is a delta function that has a value of one on the
Assuming the system to be isothermal, both liquid phase and gas
interface and zero everywhere else. x and y represent the two spatial
phases can be considered as incompressible. Hence, the mass con-
coordinates in the system. dx dy is the differential area that covers
servation of the whole domain for both the gas phase and the liquid
the region () where the inner liquid phase exists.
phase can be expressed as
It is well known that it is difficult to implement the dirac-delta

∇ · u ∗=0 (1) function numerically. Hence, a distribution function D(  x ) is used to
approximate the delta function and to estimate the fraction of the
 physical properties distributed across the artificial thickness of
where u ∗ is the velocity of the fluid.
To take into account the surface tension on liquid--gas inter- the interface. As a result, there is no sharp jump in the value of the
∗ ), an additional
face (FST interfacial force term is included in the physical properties crossing over the interface. In this study, the
Navier--Stokes equation. The governing equation for momentum distribution function given by Peskin (Peskin, 1977; Peskin and
Printz, 1993) is used:
        
(4h)−2 1 + cos (x − xf ) · 1 + cos (y − yf ) , if |x − xf | < 2h; |y − yf | < 2h
D( 
x − xf ) = 2h 2h (6)
0, otherwise

where h is the grid size.

conservation can be expressed as
Taking the gradient of the indicator function I(x, t), a gradient
 function G( x , t), showing the changes cross the interface ( ), is given
j∗ u ∗    
+ ∇ · ∗ u ∗ u ∗ = − ∇p∗ + ∇[∗ (∇ u ∗ + ∇ u ∗ T )] as follows in an integral form:
jt ∗ 
 
+ (∗ − ∗g ) g ∗ + FST
∗ (2) G( 
x , t) = ∇I( 
x , t) = (x − xf )(y − yf ) n f dsf

 
where ∗ is the density of the fluid, p∗ is the pressure, ∗ is the = D( 
x −
x f , t) n f sf (7a)

viscosity of the fluid, g ∗ is the gravitational acceleration. Subscript f

g refers to the gas phase (CO2 ). 

The surface tension force on the twophase interface can be cal- where n f is the unit normal vector at the interface at an interfacial
culated as follows on the front mesh: element with an area of sf whose centroid is  x f . Thus, the following
 Poisson equation can be used to reconstruct the indicator function

∗
FST = ∗ ∗f n f (x∗ − xf∗ )(y∗ − yf∗ ) dsf∗ (3) I( 
x , t), which will be used in Eq. (4) to calculation fluid property
,f
f distribution on the background mesh for the fluid flow:

where (x∗ − xf∗ ) and (y∗ − yf∗ ) are the dirac-delta functions in the ∇ · ∇I( 
x , t) = ∇ · G( 
x , t) (7b)
x and y directions of the space coordinate system. ∗ is the surface
tension coefficient, ∗f is the curvature of the interface. In a similar manner, the interface tension force calculated using
Eq. (3) on the front mesh can be distributed to the background
2.3.3. Interface treatment and tracking mesh using the distribution function for the solution of momentum
To solve the momentum equation in the whole domain, the vary- equation (2):
ing physical properties of the fluid, which includes the gas phase and 
∗ () =
FST ∗ D(  −  )
FST,f (7c)
x x xf
the liquid phase, must be taken into account. Although the physical
properties are considered to be constant in each phase, there is a
With the calculated distribution of fluid property and also the inter-
change in the physical properties across the interface. The novelty of
facial force source in governing equations, the velocity field is ob-
the front tracking/finite difference method, proposed by Tryggvason,
tained on the fixed background mesh. The moving velocity of the
is that the interface, i.e. the moving front, is considered to have a
interface could be interpolated from the velocity field at the fixed
finite thickness of the order of the mesh size, instead of having zero
grid so that the interface is ensured to move at the same velocity as
thickness with sharp jump of fluid properties. In the region nearby
the surrounding fluids. The distribution function given by Peskin is
the interface, the physical property changes smoothly and continu-
used to interpolate the velocity of the interface from the background
ously from the value on one side of the interface to the value on the
field using
other side of the interface. The thickness of the interface thus de-
pends on the grid size and is kept constant during the computation.   
uf = D( 
xf −
x)u (8)
Thus, the governing equations could be solved in the whole domains
with variable property for both liquid and gas phases. In the front
Then, the interface is advected in a Lagrangian fashion
tracking method, the field distribution b(  x , t) of the physical fluid
properties throughout the whole domain can be calculated through  n+1 −  n = t 
x f x f uf (9)
b( 
x , t) = bg + (bl − bg ) · I( 
x , t) (4)
After the interface is advected, the mesh size and quality on the
where bl is the properties in the liquid phase (DCM), bg is the proper- interface may be deteriorated due to the deformation. Hence, front
ties in the gas phase (CO2 ) and I( 
x , t) is an indicator function which mesh adaptation has to be performed to improve the front mesh
has a value of one in the liquid phase and 0 in the gas phase. The quality.
 L.Y. Lee et al. / Chemical Engineering Science 63 (2008) 3366 -- 3378 3369

2.3.4. Numerical method

1. Calculate velocity of interface on the
To solve the above equation numerically, normalization is needed
front mesh (Eq. 8)
to improve the numerical calculation accuracy. In this study, the
following normalizations are applied to the characteristics variables:

x∗ u∗ t∗ ∗ 2. Advect interface (Eq. 9) to the new

x= ∗, u= , t= , =
R0 (g ∗ R0∗ )1/2 ∗1/2
R0 g ∗−1/2 l position and calculate the interface
tension (Eq. 3)
P∗ ∗  ∗
P= ∗ ∗ ∗, = ∗, = (10)
l g R0 l R0∗−1
3. Redistribute fluid physical properties
R0∗ is defined as the radius of the inner diameter of the nozzle,
(Eq. 4) and interfacial tension (Eq. 7c)
and subscript l represents properties of the liquid (organic solution)
to the background mesh for fluid flow
phase.
The re-formulated momentum equation can be re-expressed as

j( u ) 
+∇ · u u 4. Solve Navier-Stokes equation for new
jt velocity field (Eq. 2)
 1  
= −∇P + (1 − ) g + ∇ · [(∇ u +∇ T u )]
 Ar
Fig. 2. Flow chart for the iterative numerical procedure.
1 
+  n (x − xf )(y − yf ) dsf (11)
Bo f f f

where Ar is the Archimedes number, defined by (Bonometti and
Magnaudet, 2006)
∗3/2
∗ g ∗1/2 R0
Ar = l
∗l
Archimedes number is the ratio of gravitational force to viscous force
which characterizes the motion of liquids due to density differences.
This may be taken as the Reynolds number based on the gravitational
velocity (gR∗0 )1/2 . (Note: the relevant velocity scale is actually (1 −
g /l ) (gR∗0 )1/2 . Since the g /l value is in the order of 0.001, this
velocity scale is approximated by (gR∗0 )1/2 .)
Bo is the Bond number, defined by
∗ g ∗ R∗2
Bo = l ∗ 0

The Bond number is written as a ratio of the gravitational force to the
surface tension force. The relative importance of these three forces
dictates the mode where droplet formation occurs.
In this work, using a fixed, Cartesian, staggered grid, the momen-
tum equation is discretized with the finite volume method. As the
front is advected explicitly, the physical properties and surface ten-
sion are updated subsequently. Then the coupling of fluid velocity
and pressure is updated by solving the momentum equations and
the continuity equation using a SIMPLE scheme (Patankar, 1980).
The simulation process is thus robust even for large density ratio
because of the semi-implicit algorithm:
The procedures implemented are listed as shown in Fig. 2:
1. Based on the velocity field and the interface position at time step
t, the velocity of the moving interface on the front mesh can be
calculated according to Eq. (8).
2. The interface is moved to a new location for the next time step
t + t using Eq. (9). And the surface tension stress for each point
on the interface is calculated according to Eq. (3).
3. With the new interface location, the indication function I(x, t+t)
can be reconstructed. And the fluid physical properties such as
density and viscosity are redistributed in the simulation domain
onto the background mesh using Eq. (4) with appropriate values
for the liquid phase and the gas phase, respectively. The inter-
face tension force is also redistributed to the background mesh
using Eq. (7).
4. The equation for the conservation of mass and the momentum,
with the inclusion of the surface tension stress, are then solved
to get the new velocity field and pressure field for the next time
step (t + t).
(12a) 5. Steps 1--4 are repeated to continue with the calculation for the
next time step.
The radius of the domain where the simulation is performed has a
dimensionless length of 8, while the length of the domain is depen-
dent on the type of simulation performed so as to make the simu-
lation computationally efficient. Typically, the dripping mode has a
shorter domain, while the Rayleigh disintegration mode has a longer
domain. There are 20 grid points per dimensionless length. The ge-
ometry of the nozzle is set to mimic the actual physical geometry
of the setup and the liquid attachment point on the nozzle wall is
(12b) determined through experiments. Simulation was performed on the
IBM supercomputer p575 in the Institute of High Performance Com-
puting, Singapore.
3. Results and discussion
3.1. Interfacial tension of DCM--CO2 system
The interaction between high pressure CO2 with organic solvents
such as ethanol, acetone and DCM, provides favorable conditions for
particle formation using SAS process. The miscibility of organic sol-
vents in CO2 is strongly dependent on the pressure and composition
of the system. Sun and Shekunov (2003) investigated the surface
tension of ethanol in CO2 using laser interferometric microscopy,
and below MCP, stable droplets with surface tension decreasing with
increasing CO2 pressure is observed. Badens et al. (2005) also mea-
sured the pure dynamic interface tension of water, ethanol and DCM
in CO2 using the hanging drop method.
In this work, the interface tension of DCM in CO2 is determined
using image analysis of a pendant drop using the image analy-
sis method described by Hansen and Rodsrud (1991). Fig. 3 shows
the pendant drops of DCM in CO2 at varying CO2 pressures. The
size of pendant drop decreases with increasing system pressure,
which shows a decrease in interface tension at higher pressures.
The calculation of DCM--CO2 interface tension at varying pressures
from ambient to near MCP is shown in Fig. 4. The results obtained
are very similar to the interface tension measurements by Badens
et al. (2005). The interface tension of DCM in CO2 has important
3370 L.Y. Lee et al. / Chemical Engineering Science 63 (2008) 3366 -- 3378

80
51 Bars
70 61 Bars
60
50

L/d
40
30
20

Fig. 3. Pendant drops of DCM in CO2 taken using digital SLR at (a) 1 bar; (b) 21 bar; 10
(c) 41 bar and (d) 61 bar. The temperature of the system was fixed at 35 ◦ C.
0
0 500 1000 1500 2000 2500 3000
ReL
30
Present study Fig. 6. Jet stability diagram for CO2 --DCM system at 35 ◦ C and pressure of 51 and
25 Badens et al. (2005) 61 bar, respectively. L: jet length before disintegration, d: nozzle diameter, ReL =Ud/
,
Interfacial tension (mN/m)

where U is the characteristic velocity of the jet and

is the kinematic viscosity of
the supercritical CO2 . For each point on the graph, the sampling size is n = 3 and
20 the L/d value reported is the mean value of the measurements, and the error bar
represents the standard deviation.
15 y = -0.3129x + 24.446
R2 = 0.9914

10

0
0 20 40 60 80 100
CO2 Pressure (Bars)

Fig. 4. Comparison of measurements for DCM--CO2 interfacial tension at 35 ◦ C

between present study (DCM--CO2 interfacial tension at 308 K) and Badens et al.
(2005) (DCM--CO2 interfacial tension at 308 K). x: CO2 pressure (bar). y: interfacial
tension (mN/m).

Fig. 7. Axis-symmetric modes of jet disintegration in a DCM--CO2 system at 61 bar

and 35 ◦ C simulated using front tracking/finite volume simulations. (a) Dripping
mode at a flowrate of 0.5 ml/min; (b) Rayleigh disintegration mode at a flowrate of
0.8 ml/min.

Fig. 5. Different jet disintegration regimes observed in a high pressure CO2 --DCM
system. (a) Dripping mode; (b) Rayleigh disintegration mode; (c) wavy disintegra-
tion/sinuous wave mode; (d) atomization/turbulent jet mode. The temperature and
pressure of the system was 35 ◦ C and 61 bar, respectively. Flowrate of DCM is in-
creased from 0.5 to 5 ml/min to achieve the corresponding transitions from dripping
to turbulent jet modes.
implications on the various modes of droplet formation at varying
CO2 pressures. This in turn provides insights to the design of oper-
ating parameters to be used for particle fabrication experiments.
3.2. Jet break up regimes
The disintegration of a jet may be described by various regimes
in the stability curve (Grant and Middleman, 1966) Various flow
regimes of free jets at high ambient pressure have been reported in
the studies by Kerst et al. (2000). Fig. 5 shows the various modes of jet
disintegration that can be observed for the injection of DCM into CO2 .
At very low flow rates, dripping of DCM from the tube is observed
and no jet with a definable breakup length is formed (dripping mode,
Fig. 5a). As the flow rate of the liquid phase increases, a jet is formed
and the jet breakup is mainly influenced by surface tension forces.
The jet is predicted to be unstable only to symmetrical disturbances
(Rayleigh disintegration mode, Fig. 5b). Further increase in flow rate
shifts the jet disintegration mechanism from symmetric breakup to
transverse wave breakup (wavy disintegration mode, Fig. 5c). At very
high flow rate, the jet is turbulent at the nozzle exit and a turbulent
jet is formed (Fig. 5d).
The drop sizes obtained at various jet formation modes are very
different. In the dripping mode, the mass of droplet is proportional
to the surface tension of the liquid (Lefebvre, 1989). In the Rayleigh
 L.Y. Lee et al. / Chemical Engineering Science 63 (2008) 3366 -- 3378 3371

Fig. 8. Axis-symmetric modes of jet disintegration DCM--CO2 system obtained using front tracking/finite volume simulation. Transition from the dripping mode to the
Rayleigh disintegration mode at CO2 pressure of 71 bar and temperature of 35 ◦ C as the flowrate of DCM increases. The flowrate of DCM for a1--a3 is 0.500 ml/min, while
the corresponding one in b1--b3 is 0.525 ml/min. Panels a1 and b1 illustrate the outline of the flow from the simulation results. a2 and b2 show the contours for excess
pressure (relative to environmental pressure), while a3 and b3 show the velocity vector, plotted against the streamline of the flow in the system.

disintegration mode for viscous fluids, the diameter of droplet is in Section 2.2. As CO2 pressure increases, the transition between
approximately twice the diameter of the jet, and droplet sizes are various jetting modes shifts to the left of the curve.
fairly uniform. In the wavy disintegration mode, droplet sizes are The droplet formation of DCM in CO2 at elevated pressures in
less uniform and observed to be less than or about the same as the dripping and Rayleigh disintegration modes is investigated us-
the jet diameter. In the turbulent jet mode, typical droplet sizes ing both image analysis and front tracking/finite volume simula-
are much smaller than the jet diameter. The jet stability curve (L/d tion to understand the droplet formation mechanism in these two
vs. ReL ) for DCM in high pressure CO2 is shown in Fig. 6 (L: jet regimes and to assess their application to particle fabrication using
length before disintegration, d: nozzle diameter). ReL is the Reynolds SAS method. Figs. 7a and b show a snapshot of the axis-symmetric
number of the liquid leaving the nozzle. ReL = Ud/
, where U is the droplet formation process simulated using the front tracking method
characteristic velocity of the jet and
is the kinematic viscosity of for the dripping mode and Rayleigh disintegration mode, respec-
the supercritical CO2 . This figure is prepared by performing an image tively. At high pressure of 61 bar, it was observed from the simula-
analysis of the jet at 35 ◦ C at CO2 pressures of 51 and 61 bar, using tion that increasing the flowrate from 0.5 to 0.8 ml/min changes the
high magnification pictures obtained from the digital SLR mentioned droplet formation from dripping to Rayleigh disintegration modes.
3372 L.Y. Lee et al. / Chemical Engineering Science 63 (2008) 3366 -- 3378

Fig. 9. Axis-symmetric modes of jet disintegration DCM--CO2 system obtained using front tracking/finite volume simulation. Changing from the dripping mode to the Rayleigh
disintegration mode when the CO2 pressure is at 76 bar and temperature at 35 ◦ C. The flowrate of DCM in a1--a3 is 0.400 ml/min, in b1--b3 is 0.425 ml/min, and in c1--c3 is
0.450 ml/min. Panels a1, b1 and c1 show the outline of the flow from the simulation results. a2, b2 and c2 show the contours for excess pressure (relative to environmental
pressure). Panels a3, b3 and c3 show the velocity vector, plotted against the streamline of the flow in the system.
 L.Y. Lee et al. / Chemical Engineering Science 63 (2008) 3366 -- 3378
Fig. 10. Dripping mode of DCM in CO2 at different pressures obtained using front
tracking/finite volume simulation: (a) at 61 bar and 35 ◦ C,  = 5.36 × 10−3 N/m,
Deqv = 2.27 mm; (b) 71 bar and 35 ◦ C,  = 2.23 × 10−3 N/m, Deqv = 0.75 mm; (c) 76 bar
and 35 ◦ C,  = 1.13 × 10−3 N/m, Deqv = 0.50 mm.
Figs. 8 and 9 show the transition from dripping mode to Rayleigh
disintegration mode at CO2 pressures of 71 and 76 bar, respectively.
At higher pressures, the transition between dripping and Rayleigh
modes occur at lower DCM flowrates which is attributed to the cor-
responding decrease of interfacial tension close to the MCP. Also,
at the same flow rate but with decreasing surface tension, the sys-
tem that has a smaller surface tension will produce smaller droplets
due to the increasing relative strength of the gravitational effect as
compared to the surface tension effect, and the frequency of droplet
formation will increase.
3.3. Droplet formation in dripping mode
Fig. 10 shows the changes in the droplet size (at the point of pen-
dant drop breakage) in the dripping mode when the CO2 pressure
is increased from 61, 71 to 76 bar. An increase in CO2 pressure in-
creases the miscibility of CO2 and DCM phases and consequently, a
reduction in the interface tension between DCM--CO2 is observed.
Thus the Bond number of the system is increased in magnitude. The
size of the droplets in the dripping mode is a strong function of the
Bond number. This can be explained by the fact that at this mode,
the size of the pendant drop that is attached to the nozzle before
pinch-off is determined by the balance between the gravitational
force (it pulls the liquid downward in the direction of gravity) and
the surface tension force (it holds the droplet in place, resisting mo-
tion). From Fig. 10, the size of droplets obtained from simulation
decreased from 2.27, 0.75 to 0.50 mm with increasing pressure from
61, 71 to 76 bar. Accompanying the increase in pressure, the gravi-
tational effect also became increasingly important, as shown by an
increase in the Bond number from 0.039, 0.093 to 0.183.
Fig. 11 shows the pinch-off process when the weight of the pen-
dant drop is too large for the surface tension to hold it in place.
Comparing Panels a1--a3, b1--b3, c1--c3, d1--d3 and e1--e3, it was
observed that the numerical simulation result matches well with
the experimental observation, showing the pinch-off of the liquid
droplet. Panels a2, b2, c2, d2 and e2 show that pressure contours in
the system. The point of greatest pressure before droplet pinch-off
is at the neck (see Panels a2 and b2). After the pinch-off of droplet,
there is a gradual dissipation of the built-up pressure in the droplet,
while a large pressure is built-up in the retracting liquid column
that is attached to the nozzle (see Panels c2 and d2). The observed
pressure contour profile can be explained by analyzing the balance
of the pressure inside and outside the DCM--CO2 interface,
 
P ∗ + ∗ (∇ u ∗ + ∇ u ∗T ) · n = ∗ ∗ (13)
3373
where the square bracket represents the difference between
"outside'' and "inside'' of the interface.
The pressure difference is strongly dependent on the surface ten-
sion and the curvature of the interface. In the axis-symmetric cylin-
drical coordinates, the surface curvature is a function of the radius of
the liquid column. Therefore, it can be seen that the point of highest
pressure is located at the region where the radius of a liquid col-
umn is the smallest. The velocity vectors are shown in Panels a3, b3,
c3, d3 and e3 of Fig. 11, plotted against the corresponding stream-
lines. As the droplets are formed and transported in the direction of
the gravity, the corresponding fluid phase velocity is in the down-
ward direction. Due to the confined nature of the chamber where
the droplets are flowing downwards, circulation in the surrounding
fluid can be found in the vicinity of the droplets. This phenomenon
is clearly visible with reference to the streamline plots illustrated on
the right-hand side of each of the figure panels.
To determine the size of the droplet, the shadow sizing system is
deployed. By analyzing the shadow image of the droplets (Figs. 12a
and b), the length of the major and minor axes of the droplets can
be obtained, and the size of the droplet can be calculated. Taking
into account the axis-symmetric nature of the system, the equiv-
alent diameter (Deqv ) of the droplets is then calculated (Fig. 12c).
The corresponding equivalent diameter at 66 and 71 bar is 1.31 and
1.07 mm, respectively. Fig. 12d illustrates the close match between
of the sizes of droplets achieved at different pressures during the
dripping mode.
3.4. Droplet formation and jet breakup in Rayleigh disintegration mode
As the inlet velocity is increased while the surface tension is
decreased by increasing the CO2 pressure, the dominant term in the
momentum equation (Eq. (11)) shifts from the fourth term (related
to the interface tension, involving 1/Bond number) to the third term
(related to the viscous stress, involving 1/Archimedes number). In the
Rayleigh disintegration mode, there is no formation of the pendent
drop. A thin jet is first formed as the liquid leaves the nozzle, and
then droplets are formed through the jet-breakup process.
One point to note is that due to the affinity of DCM to stainless
steel nozzle in the subcritical CO2 environment, the jet diameter is
weakly dependent on the surface tension (the point of contact of
the interface on the stainless steel nozzle and the contact angle is
determined by the balance of the stainless steel--DCM surface ten-
sion, stainless steel--CO2 surface tension, and the varying DCM--CO2
surface tension).
Results of the simulation show that the droplet size is very weakly
dependent upon the interfacial tension (Fig. 13). The above observa-
tion may be explained as the limit where the viscous effect is small,
the dominant jet-breakup wavelength is only dependent on the ra-
dius of the jet (Rayleigh, 1878, 1879). The surface tension has no
effect on the jet-breakup wavelength, and only plays a part in de-
termining the instability growth rate of the dominant jet-breakup
wavelength. Since the dominant wavelength is not affected by the
changes in the surface tension, it can be shown that the final droplet
∗
size (Ddroplet ) of an inviscid jet in such Rayleigh disintegration pro-
∗ ) through the following simple
cess is related to the jet diameter (Djet
function:
∗
Ddroplet ∗
= 1.48Djet (14)
The above expression was derived by noting that the dimensionless
wavenumber of Rayleigh disintegration of the inviscid jet is around
0.69 (non-dimensionalized by the jet circumference) (Rayleigh, 1878,
1879), and subsequently through the conservation of mass for the
section of the jet that is broken off to the formation of a spheri-
cal droplet. Further work to correlate the size (Lefebvre, 1989) has
3374 L.Y. Lee et al. / Chemical Engineering Science 63 (2008) 3366 -- 3378

Fig. 11. Comparison of experimental and simulated pendant drop formation in the dripping mode. Panels a1, b1, c1, d1 and e1: snapshots of the pinch-off point in dripping
mode at 71 bar and 35 ◦ C and DCM flowrate of 0.4 ml/min (time between each consecutive snapshots from top to bottom row is 0.001 s). Panels a2, b2, c2, d2 and e2:
outline of the pinch-off process obtained through the numerical simulation, and the contours for excess pressure (relative to environmental pressure) of the system. Panels
a3, b3, c3, d3 and e3: velocity vectors of the successive snapshots are plotted against the streamline of the flow in the system.

suggested that the ratio of diameter of droplet to the diameter of jet droplet radius = 2.63), for the simulated case of 71 bar is around 1.67
is around 1.89 for the case of a viscous jet. The ratio of the jet diam- (dimensionless jet radius=1.35, dimensionless droplet radius=2.25)
eter to the droplet diameter from Fig. 13 for the simulated case of and for the simulated case of 76 bar is around 1.70 (dimensionless
61 bar is around 1.75 (dimensionless jet radius = 1.50, dimensionless jet radius = 1.13, dimensionless droplet radius = 1.93). The surface
 L.Y. Lee et al. / Chemical Engineering Science 63 (2008) 3366 -- 3378
Fig. 11. (continued).
tension of the DCM in CO2 at 61 and 71 bar is 5.36 × 10−3 and
2.23 × 10−3 N/m, respectively. The surface tension of DCM in CO2
at 71 bar is less than half the value at 61 bar, however, there are no
significant differences in the jet and droplet diameter as shown in
Fig. 13. This shows that surface tension has negligible effect on the
dimensionless wavenumber of the Rayleigh disintegration of the jet,
and is close to the expected value. This, together with the slightly
changing affinity of DCM to the stainless steel nozzle, explains the
rather weak dependence of the changes of surface tension on the
droplet size, since only the jet diameter is weakly dependent on the
surface tension.
If a glass nozzle was used in place of the stainless steel nozzle,
different droplet size will be expected in the dripping mode as both
surface tension forces and liquid attachment area play important
roles on drop size in this regime. Using the stainless steel nozzle,
upward force balancing the weight of the drop is higher due to the
affinity of the drop to the nozzle wall and also the larger area where
the liquid is attached to the surface of the nozzle, resulting in larger
drop size. However, in the Rayleigh disintegration mode, drop size is
likely to be unaffected by change in the nozzle material as the drop
size is dependent on the jet diameter which is mainly similar to the
inner diameter of the nozzle.
3375
3.5. Particle formation
The SAS process is commonly used in pharmaceutical applications
for micronization of biodegradable polymers and drugs. In this work,
it was observed that the initial drop size of DCM in high pressure CO2
is strongly dependent upon the pressure in the dripping mode due to
a variation in the interface tension of the system. Particle fabrication
in SAS processes is affected by both the thermodynamic conditions
(pressure, temperature and composition, etc.) of the system as well
as the hydrodynamics of the organic phase into the high pressure
CO2 (flowrate, nozzle diameter, nozzle design, etc.).
Figs. 14a and b show the fabrication of PLA particles from a solu-
tion of 2% (w/v%) polymer in DCM into CO2 at 71 and 76 bar, respec-
tively. Dripping flow was observed at both pressures at a flowrate of
0.05 ml/min. From Fig. 14, it can be observed that at 71 and 76 bar,
operating at a dripping flow yields fairly uniform sized particles.
However, at lower pressure (71 bar), it was also observed that parti-
cles formed tend to agglomerate. At an even higher pressure closer
to the MCP of the system (76 bar), much higher miscibility between
DCM and CO2 is achieved as shown in the snapshot of the droplet
formation process at 76 bar. The particles formed at this condition
were found to be more discrete and spherical, and generally in the
3376 L.Y. Lee et al. / Chemical Engineering Science 63 (2008) 3366 -- 3378

1mm

z
r 80
75

Pressure (bar)
70
65
60
B
55 Simulation
Experimental
50
0 0.5 1 1.5 2 2.5
A Equivalent diameter (Deqv)

Fig. 12. Measurement of droplet size of DCM in CO2 at 35 ◦ C and at different pressures using shadow sizing system: (a) 66 bar, Deqv = 1.31 ± 0.02 mm; (b) 71 bar,
Deqv = 1.07 ± 0.06 mm; (c) calculation of equivalent diameter (Deqv ) of droplet using an axis-symmetric model. Taking into account the axis-symmetric nature of the droplet
√
3
and the system, Deqv is calculated based on Deqv = A2 · B and (d) comparison of experimentally determined Deqv with numerical simulation in the dripping mode.

range of 2--5 m. Fig. 14c illustrates particles fabricated at conditions

of 76 bar and flowrate of 0.2 ml/min (left panel) and 3 ml/min (right
panel). At 0.2 ml/min, the particle size distribution achieved is similar
as observed in Fig. 14b. However, when operating in the turbulent
jet mode at higher flowrate of 3 ml/min, the particle size become
significantly more polydispersed.
Figs. 15a and b show PLA particles obtained from solution of 2%
(w/v%) polymer in DCM at a flowrate of 0.05 ml/min into CO2 at 81
and 91 bar, respectively. At these conditions above the MCP of the
system, it was observed that particle size distribution is more non-
uniform than the particles shown in Figs. 14a and b. At this operating
regime, the interfacial tension between DCM and CO2 diminished
and axis-symmetric dripping mode is not observed. The solution
enters the vessel as a gaseous plume. Figs. 15c and d show particles
obtained at flowrate of 4 ml/min at 81 and 91 bar, respectively. On
comparison with particles from Figs. 15a and b, it was observed that
at the conditions above MCP, solution flowrate has no obvious effect
on the final particle size and size distribution.

4. Conclusions

The droplet formation process of DCM in high pressure CO2 is

investigated in this work. It was observed that the interface ten-
sion of the DCM--CO2 system is significantly affected by the pres-
sure of the surrounding CO2 gas phase, especially at pressures close
to the mixture critical point. Dripping, Rayleigh disintegration, wavy
disintegration and turbulent jet were clearly observed using high-
speed visualization of the DCM--CO2 system at elevated pressures.
Fig. 13. Simulation of Rayleigh disintegration of DCM jet in CO2 at 35 ◦ C and
high pressure. (a) Typical Rayleigh disintegration droplet formation process, 76 bar, The dripping and Rayleigh disintegration modes yield more mono-
∗
flowrate of DCM = 0.5 ml/ min, Ddroplet ∗
/Djet ≈ 1.75; (b) simulated case, 61 bar, disperse droplets in contrast to the wavy disintegration and tur-
∗ ∗
flowrate of DCM = 0.8 ml/ min, Ddroplet /Djet ≈ 1.75; (c) simulated case, 71 bar, bulent jet modes. The axis-symmetric modes of droplet formation
∗
flowrate of DCM = 0.55 ml/ min, Ddroplet ∗ ≈ 1.67; (d) simulated case, 76 bar,
/Djet were simulated using a front tracking/finite volume simulation code
∗
flowrate of DCM = 0.45 ml/ min, Ddroplet ∗ ≈ 1.70.
/Djet and compared with experimental observations. Very close match of
 L.Y. Lee et al. / Chemical Engineering Science 63 (2008) 3366 -- 3378 3377

Fig. 14. PLA particles fabricated from the SAS process at 35 ◦ C at (a) 71 bar with DCM flowrate of 0.05 ml/min into CO2 ; (b) 76 bar with DCM flowrate of 0.05 ml/min into
CO2 ; (c) 76 bar with DCM flow of 0.20 ml/min into CO2 (left panel) and 3.00 ml/min into CO2 (right panel).

Fig. 15. PLA particles fabricated from the SAS process at 35 ◦ C with a DCM flowrate of 0.05 ml/min into CO2 at (a) 81 bar; (b) 91 bar; and with a DCM flowrate of 4.00 ml/min
into CO2 at (c) 81 bar; (d) 91 bar.
3378 L.Y. Lee et al. / Chemical Engineering Science 63 (2008) 3366 -- 3378

the simulation with high-speed camera results were obtained in the Chattopadhyay, P., Gupta, R.B., 2001b. Production of griseofulvin nanoparticles using
dripping mode. supercritical CO2 antisolvent with enhanced mass transfer. International Journal
of Pharmaceutics 228, 19--31.
In the dripping mode, the droplet size decrease with increase Chattopadhyay, P., Gupta, R.B., 2002a. Methods of forming nanoparticles and
in CO2 pressure due to a corresponding decrease in the interface microparticles of controllable size using supercritical fluids and ultrasound. US
tension. This was observed in the droplet size measurements using Patent Pub. No. 2002/0000681.
Chattopadhyay, P., Gupta, R.B., 2002b. Protein nanoparticles formation by
the shadow sizing system as well as in the simulation results. In the supercritical antisolvent with enhanced mass transfer. A.I.Ch.E. Journal 48,
Rayleigh disintegration mode, the droplet size is weakly dependent 235--244.
on the interface tension of the system and is proportional to the Chattopadhyay, P., Gupta, R.B., 2002c. Supercritical CO2 based production of
magnetically responsive micro- and nanoparticles for drug targeting. Industrial
diameter of the jet. The operation of SAS process in the window
and Engineering Chemistry Research 41, 6049--6058.
close to the mixture critical point at low flowrate provides a means Debenedetti, P.G., et al., 1993. Rapid expansion of supercritical solutions (RESS):
of particle fabrication with possible size control. Operating in the fundamentals and application. Fluid Phase Equilibria 82, 311--321.
Gokhale, A., et al., 2007. Effect of solvent strength and operating pressure on the
dripping mode near the mixture critical point, the initial droplet size
formation of submicrometer polymer particles in supercritical microjets. Journal
can be manipulated by small changes in the pressure of CO2 . of Supercritical Fluids 43, 341--356.
Grant, R.P., Middleman, S., 1966. Newtonian jet stability. A.I.Ch.E. Journal 12 (4),
Notation 669--678.
Hansen, F.K., Rodsrud, G., 1991. Surface tension by pendant drop. I. A fast standard
instrument using computer image analysis. Journal of Colloid and Interface
Ar Archimedes number Science 141, 1--9.
Henczka, M., et al., 2005. Particle formation by turbulent mixing with supercritical
Bo Bond number antisolvent. Chemical Engineering Science 60, 2193--2201.
dSf∗ interface length element Hua, J.S., Lou, J., 2007. Numerical simulation of bubble rising in viscous liquid. Journal
of Computational Physics 222, 769--795.
f variables in the fluid phase Jung, J., Perrut, M., 2001. Particle design using supercritical fluids: literature and
g variables in the outer gas phase patent survey. Journal of Supercritical Fluids 20, 179--219.
 Kerst, W., et al., 2000. Flow regimes of free jets and falling films at high ambient
g acceleration due to gravity, m s−2 pressure. Chemical Engineering Science 55, 4189--4208.
l variables in the inner liquid phase Lefebvre, A.H., 1989. Basic processes in atomization. In: Atomization and
P pressure Sprays.Hemisphere Publishing Corporation, Bristol, PA, USA.
Obrzut, D.L., et al., 2007. Effect of process conditions on the spray characteristics of
ReL Reynolds number a PLA + methylene chloride solution in the supercritical antisolvent precipitation
R0 inner radius of the capillary tube process. Journal of Supercritical Fluids 42, 299--309.
 Patankar, S.V., 1980. Numerical Heat Transfer and Fluid Flow. Hemisphere,
u velocity vector
Washington, DC, USA.
∗ dimensional quantity Perez de Diego, Y., et al., 2005. Operating regimes and mechanism of particle
formation during the precipitation of polymers using the PCA process. Journal
Greek letters of Supercritical Fluids 35, 147--156.
Peskin, C.S., 1977. Numerical analysis of blood flow in the heart. Journal of
∗f curvature of the interface Computational Physics 25, 220--252.
Peskin, C.S., Printz, B.F., 1993. Improved volume conservation in the computation
 dynamic viscosity of flows with immersed boundaries. Journal of Computational Physics 105,
 density 33--46.
∗ surface tension coefficient Randolph, T.W., et al., 1993. Sub-micrometer-sized biodegradable particles of poly
L-lactic acid via the gas antisolvent spray precipitation process. Biotechnology
Abbreviation Progress 9, 429--435.
Rayleigh, L., 1878. On the instability of jets. Proceedings of the London Mathematical
CO2 carbon dioxide Society 10, 4--13.
Rayleigh, L., 1879. On the capillary phenomenon of jets. Proceedings of the Royal
Ddroplet droplet diameter, mm Society of London 29, 71--97.
Deqv equivalent diameter, mm Reaves, J.T., et al., 1998. Critical properties of dilute carbon dioxide + entrainer
Djet jet diameter, mm and ethane + entrainer mixtures. Journal of Chemical and Engineering Data 43,
683--686.
DCM dichloromethane Reverchon, E., et al., 2003. Role of phase behavior and atomization in the supercritical
antisolvent precipitation. Industrial and Engineering Chemistry Research 42,
Acknowledgments 6406--6414.
Richard, J., Dechamps, F.S., 2004. Supercritical fluid processes for polymer particle
engineering---applications in the therapeutic area. In: Elaissari, A. (Ed.), Colloidal
This research was supported by Agency of Science, Technology Biomolecules, Biomaterials, and Biomedical Applications. Marcel Dekker, New
and Research (A*STAR) and National University of Singapore (NUS) York, pp. 429--475.
Ruchatz, F., et al., 1997. Residual solvent in biodegradable microparticles: influence
under the Grant number R279-000-208-305. The authors acknowl- of process parameters on the residual solvent in microparticles produced by
edge the scientific discussion and inputs from Professors Kenneth A. the aerosol solvent extraction system (ASES) process. Journal of Pharmaceutical
Smith and Yen Wah Tong during various phases of this project. Science 86 (1), 101--105.
Shekunov, B.V., et al., 2001. Particle formation by mixing with supercritical
antisolvent at high Reynolds number. Chemical Engineering Science 56,
References 1433--2421.
Subramaniam, B., et al., 1997. Methods for a particle precipitation and coating using
Badens, E., et al., 2005. Laminar jet dispersion and jet atomization in pressurized near-critical and supercritical antisolvents. US Patent No. 5,833, 891.
carbon dioxide. Journal of Supercritical Fluids 36, 81--90. Sun, Y., Shekunov, B.Y., 2003. Surface tension of ethanol in supercritical carbon
Bonometti, T., Magnaudet, J., 2006. Transition from spherical cap to toroidal bubbles. dioxide. Journal of Supercritical Fluids 27, 73--83.
Physics of Fluids 18, 052102. Tom, J.W., Debenedetti, P.G., 1991. Particle formation with supercritical fluids---a
Carretier, E., et al., 2003. Hydrodynamics of supercritical antisolvent precipitation: review. Journal of Aerosol Science 22, 555--584.
characterization and influence on particle morphology. Industrial and Tryggvason, G., et al., 2001. A front tracking method for computations of multiphase
Engineering Chemistry Research 42, 331--338. flow. Journal of Computational Physics 169, 708--759.
Chattopadhyay, P., Gupta, R.B., 2001a. Production of antibiotic nanoparticles using Unverdi, S.O., Tryggvason, G., 1992. A front-tracking method for viscous,
supercritical CO2 as antisolvent with enhanced mass transfer. Industrial and incompressible multi-fluid flow. Journal of Computational Physics 100, 25--37.
Engineering Chemistry Research 40, 3530--3539.