You are on page 1of 10

J.

of Supercritical Fluids 59 (2011) 98107


Contents lists available at ScienceDirect
The Journal of Supercritical Fluids
j our nal home page: www. el sevi er . com/ l ocat e/ supf l u
Precipitation kinetics of PMMA sub-micrometric particles with a supercritical
assisted-atomization process
Hsien-Tsung Wu

, Ming-Wei Yang
Department of Chemical Engineering, MingChi University of Technology, 84 Gungjuan Rd., Taishan, New Taipei City 24301, Taiwan
a r t i c l e i n f o
Article history:
Received 8 April 2011
Received in revised form22 July 2011
Accepted 1 August 2011
Keywords:
Supercritical assisted-atomization
Polymer particles
Precipitation kinetics
a b s t r a c t
Sub-micrometric particles of PMMA were successfully prepared via a supercritical assisted-atomization
(SAA) process using acetone as a solvent and supercritical carbon dioxide as a spraying medium. The
effects of several key factors on the particle size were investigated. These factors included the concen-
tration of polymer solution, temperature in saturator and volumetric ow rate ratio of carbon dioxide
to polymer solution. The shape of the polymers primary particles is spherical with the arithmetic mean
size ranging from 82 nm to 176 nm and the mass-weighted mean size ranging from 127 nm to 300 nm.
As evidenced from the experimental results, the lower concentrations of polymer solution, optimized
volumetric ow rate ratios, and higher temperatures in saturator can effectively reduce the mean parti-
cle size. The precipitation kinetic parameters were determined from the particle size distributions with
the aid of the population balance theory. This study found the primary nucleation to be dominant in the
precipitation and diffusion may govern particle growth.
2011 Elsevier B.V. All rights reserved.
1. Introduction
Over the previous decades, several supercritical uid-assisted
methods have beendevelopedtoprepare ne particles withunique
characteristics in numerous industrial elds. These particulate pro-
cessing techniques can be roughly classied, based on the function
of supercritical uids as: (a) a solvent, e.g., the rapid expansion
of supercritical solution, RESS; (b) a solute, e.g., the particle from
gas saturated solution, PGSS; (c) an anti-solvent, e.g., supercrit-
ical anti-solvent, SAS; (d) a spraying medium, e.g., supercritical
assisted-atomization, SAA, and carbon dioxide assisted nebuliza-
tion with a bubble dryer, CAN-BD. Reverchon and Antonacci [1]
successfully developed a SAA process to produce micro- and
nano-metric particles for pharmaceutical compounds, salts, dyes,
catalyst, and superconductor precursors. The supercritical carbon
dioxide was dissolved rst in liquid solution containing water or
organic solvents with solute. The mixed solution was then sprayed
into a vessel via a nozzle at near atmospheric pressure under a
ux of hot N
2
. This process involves two steps of atomization.
First, the primary droplets are formed at the exit of the nozzle by
pneumatic atomization. Second, the secondary droplets are subse-
quently generated by rapidly releasing CO
2
inside of the primary
droplets (decompressive atomization). Then, the ultra-ne parti-

Corresponding author. Tel.: +886 2 2908 9899x4630; fax: +886 2 2908 3072.
E-mail address: stwu@mail.mcut.edu.tw(H.-T. Wu).
cles obtained by the evaporation of solvent and supersaturation of
the solute.
Reverchon et al. [2] investigated micronization of Pigment Red
60 with both the SAS and SAA methods, and discovered that nano-
metric particles could be formed using the SAS, with a low yield
for SAS micronization. However, micro-metric particles were pro-
duced using SAA, thus, the yield of the SAA process was larger
than 95%. The morphology of the resulting products using the
supercritical anti-solvent (SAS) process is closely dependent on the
phase regions where the precipitations are implemented. Micro-
metric aggregated ball-like particles were produced as the SAS
process passed by the vaporliquid coexistence region, whereas
nano-metric particles were obtained in the either supercritical
or superheated vapor region [3,4]. Reverchon and Antonacci [1]
micronized the PMMA particles using the SAA process. The exper-
imental results indicated that the morphology, particle size and
distribution of PMMA microparticles were dependent on the mass
ow ratio of carbon dioxide to liquid solution. While increasing
the mass ow ratio, the operating condition probably fell into the
vaporliquid coexistence region and resulted in the coalescence of
particles collected in the precipitator. Wu et al. [5,6] have inves-
tigated the dispersion of 1wt% of Pigment Red 177 and Violet 23
particles in propylene glycol monomethyl ether acetate (PGMEA)
using a supercritical uid-assisted dispersion process (SFAD). The
SFAD process maintained at the supercritical state is effective in
improving particle dispersion. As shown from these results, the
phase behavior of mixtures in the high-pressure chamber is one of
the crucial factors of the supercritical technology appliedtoparticle
0896-8446/$ see front matter 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.supu.2011.08.001
H.-T. Wu, M.-W. Yang / J. of Supercritical Fluids 59 (2011) 98107 99
Nomenclature
B
o
nucleation rate (no./s m
3
)
C mass concentration of polymer in the solution feed
stream(kg/m
3
)

d
no
arithmetic mean particle size (nm)

d
4,3
mass-weighted mean particle size (nm)
F
CO
2
volumetric ow rate of CO
2
at 276K and 6.5MPa
(cm
3
/min)
F
CO
2
(T
S
) volumetric ow rate of CO
2
at the temperature in
saturator and 6.5MPa (cm
3
/min)
F
l
volumetric owrate of polymer solution (cm
3
/min)
G linear particle growth rate (nm/s)
i order of growth rate in nucleation kinetics equation
(Eq. (9))
K
N
constant in nucleation kinetics equation (Eq. (9))
k
V
volumetric shape factor
L particle size (nm)

L arithmetic average size ranging fromL


1
to L
2
(m)
L width of the size range (nm)
n population density at size L (no./nmm
3
)
n
o
population density of nuclei (no./nmm
3
)
m
3
thirdmoment of the populationdensity distribution
M
T
suspension density (kg/m
3
)
P pressure (MPa)
R volumetric ow rate ratio of CO
2
to polymer solu-
tion
R
2
square of correlation coefcient
T
S
temperature in saturator (K)
V
mixing
mixing volume in precipitator (cm
3
)
w mass of an individual particle
W weight fraction of particles in the size range L
1
to L
2
x
i
number of particles within a size range
mean residence time of the mixed suspension (s)

C
density of polymer particle (kg/m
3
)

CO
2
density of carbon dioxide (g/cm
3
)
Superscripts
calc calculated value
expt experimental value
formation and dispersion. Additionally, the investigators generally
focused on discussing the effect of process parameters for prepar-
ing nely particles with narrowparticle size distribution [1,3]. The
precipitation kinetics for the SAS process has been studied from
the particle size distribution with the aid of the population bal-
ance theory [79]. Reverchon and De Marco [10] organized the
numerous SAS literature results to determine the precipitation
mechanisms and particle morphologies. They proposed the pos-
sible particle formation mechanisms: liquid jet break-up and gas
mixing at supercritical conditions inSAS process are incompetition
to produce spherical microparticles or nanoparticles. The forma-
tion of crystal with various habits and dimensions depends on the
interactions with the liquid solvent used. Caputo et al. [11] used a
laser diffraction method to investigate the atomization mechanism
on the SAA process. Experimental data conrms that dissolved-
gas atomization allows for the formation of micrometric droplets
and produces ne-precipitates with controlled size and distribu-
tion. However, the modeling the precipitation kinetic for the SAA
process is limited.
In the present study, poly(methyl methacrylate) (PMMA) was
selected as a model polymer. Because it possesses excellent opti-
cal, electrical, and biocompatible properties, PMMA can be used
as spacers for the fabrication of liquid crystal displays and as
substrates for drug delivery [12]. The precipitation experiments
of PMMA particles in this study were conducted at various
concentrations of polymer solution, and at the volumetric ow
rate ratios of carbon dioxide to polymer solution over tempera-
tures in saturator ranging from 313.2K to 363.2K. The collected
samples were analyzed with a eld emission scanning electron
microscope (FESEM), and the particle size distributions and the
mean particle sizes were then determined using image process-
ing software. In addition to determining favorable conditions for
small particle formation, the precipitation kinetic parameters were
estimated by tting the experimental results to the MSMPR pop-
ulation balance equation. Based on the correlated results, the
mechanisms of nucleation and precipitation kinetics were then
considered.
2. Experimental
2.1. Materials
Poly(methyl methacrylate) (PMMA, average MW=15, 120, and
350kg/mol) was supplied by Aldrich, USA. Acetone (99.9% purity,
HPLCgrade) was purchasedfromTedia, USA. Carbondioxide (99.9%
purity) and nitrogen (99.9% purity) were purchased from Yung-
Ping Gas Co., Taiwan. These chemicals were used without further
purication.
2.2. Apparatus and operation
Fig. 1 is the schematic diagramof the SAA apparatus installed in
the present study. The apparatus consists of three main chambers
and feeding lines. The three chambers are: saturator (8), precipita-
tor (10), and separator (14). The saturator is a high-pressure vessel
(10cm
3
, AppliedSeparations, USA) loadedwithprotrudedstainless
steel packing (PRO-PAK

0.16 inch protruded packing, Scientic


Development Co., USA), assuring a large contact surface between
the liquid polymer solution and CO
2
. The solution obtained in the
saturator was sprayed through a spray nozzle (9) (I.D. 130m,
model: TP500004-TC, Spraying System Co., USA) into the precip-
itator. The saturator was circulated in a thermostatic water bath
(5), which was regulated to within 0.1K. The water bath tem-
perature was measured with a precision platinumRTD sensor (12)
(model: 1560, Hart Scientic Co., USA) to an accuracy of 0.02K.
A pressure transducer (13) (model: PDCR 4070, ranging 035MPa,
Druck, UK) with a digital display (model: DPI-280, Druck, UK) was
used to measure the operating pressure in the saturator, accurate
to 0.1%. The precipitator (10) (volume, 3.14dm
3
) was immersed
in a thermostatic air bath, which was regulated to within 0.5K.
A 0.5m metal frit (11) was mounted at the outlet (bottom) of
the precipitator to retain the polymer particles. A separator (14)
followed the precipitator to separate the solvent and gas mix-
ture at near atmospheric pressure. The ow rate and total volume
of carbon dioxide and nitrogen passing through the precipitator
were monitored using a wet test meter (15) (model: TG20, Ritter,
Germany) to an accuracy of 0.25%. Three feeding lines are a poly-
mer solution (4), carbon dioxide and nitrogen. A HPLC pump (3)
(model: PU-1580, JASCO, Japan) was used to deliver the polymer
solution and another high-pressure liquid pump (2) (model: NPL-
5000, Nihon Seimitsu Kagaku Co., Japan) for carbon dioxide. The
N
2
ow was controlled using a mass controller (6) (model: 251-
FKASBYAA, Porter, USA) froma cylinder, heated in an electric heat
exchanger (7) (series 93, Watlow, USA) and sent to the precipitator
to assist the liquid droplets evaporation.
The experimental procedure is described as follows. The satura-
tor temperature (T
S
) and the volumetric owrate of carbon dioxide
were at pre-set values. The precipitator temperature and N
2
were
specied to 333.2K and 0.8Nm
3
/h. As a steady-state was achieved,
100 H.-T. Wu, M.-W. Yang / J. of Supercritical Fluids 59 (2011) 98107
Fig. 1. The schematic diagramof the SAA apparatus.
the polymer solution was introduced into the saturator via the
pre-heating water bath (5) at a specied ow rate (the majority
of ow rate was 5ml/min). The carbon dioxide mixture that was
dissolved in the polymer solution and obtained in the saturator
was sprayed through an injection nozzle to produce atomization
of liquid droplets in the precipitator. Upon contact of the droplet
solution with the heated N
2
and the resulting evaporation of
acetone fromthe droplets, the polymer particles formed due to the
supersaturation of the solute. During the SAA process in this study,
the pressure in the saturator was almost constant (approximately
6.50.3MPa). The collected samples were removed from the
precipitator (10) and observed using a eld emission scanning
electron microscope (FESEM) (model 6500, JEOL, Japan). The
particle size distributions of samples and the mean values were
determined using Sigma Scan Pro5 software. Approximately 1000
particles on the FESEM pictures of every run were counted to
determinate the size distribution and the mean value. All precip-
itation runs were conducted at least twice at each experimental
condition. The obtained arithmetic mean particle size (

d
no
) could
be reproduced to within approximately 5%. The particle samples
were also determined the glass transition temperatures (T
g
) using
a differential scanning calorimeter (DSC, Q-20, TA, USA). Samples
of 310mg were placed in aluminumpans and sealed, and heated
from25 to 180

C at a rate of 10

C/min in nitrogen.
2.3. Experimental results
As noted by Reverchon and Antonacci [1], the efciency of
the SAA process depends on the solubilization of scCO
2
in the
liquid solution inside the saturator. The solubility of CO
2
in liq-
uids is related to the high-pressure vaporliquid equilibrium(VLE)
phase boundaries of CO
2
+acetone. The concentrations of PMMA
in the saturator are extremely dilute in our experimental runs. For
example, in the case of C=10kg/m
3
, R=0.8, the mass and molar
ratios of PMMA:acetone:carbon dioxide are 1g:98.5g:91.9g and
110
5
:1.7:2.1, respectively. As a consequence, the effect of the
presence of the PMMA on the phase behavior may be rather minor.
Even though a polymer-rich phase may be formed (VLLE) in the
H.-T. Wu, M.-W. Yang / J. of Supercritical Fluids 59 (2011) 98107 101
0 0.2 0.4 0.6 0.8 1
x
co2,
y
co2
0
40
80
120
P

(
b
a
r
)
T = 313.2 K (Chang et al. [13])
T = 313.2 K (Chiu et al. [14])
T = 333.2 K (Han et al. [15])
T = 353.2 K (Han et al. [15])
T = 373.2 K (Han et al. [15])
PR-Q2, T = 313.2 K
PR-Q2, T = 333.2 K
PR-Q2, T = 353.2 K
PR-Q2, T = 373.2 K
A B C D E F
Fig. 2. VLE phase diagramof the mixture of carbon dioxide +acetone. Points A, B, C,
D, E, and F are referred to R values of 0.8, 1.2, 1.6, 2.0, 2.4, and 2.8, respectively.
saturator, the fraction of this extraneous phase in the saturator
should be very small and may mix well with the acetone-rich
phase during the continuous injection process. We thus suggest
that the inuence of VLLE formation, if any, on the particles gener-
ation may be insignicant. The phase diagramof the binary system
(without PMMA-containing) can be used approximately to esti-
mate the phase behavior of the mixtures in the saturator and
represented the qualitative relation for selecting suitable operat-
ing conditions for the SAA processes. The VLE phase boundaries of
carbon dioxide +acetone were obtained using the PengRobinson
equationincorporating the two-parameter vander Waals one-uid
mixing rules. These equations have been used to t experimental
data at different temperatures from313.2Kto 373.2Kin literatures
[1315] andshowninFig. 2. The critical properties andacentric fac-
tors for pure components and used binary interaction parameters
have been presented in Tables 1 and 2. The precipitation experi-
ments were implemented at different spray angles of the nozzles
(illustrated in Fig. 1), concentrations of polymer solution (C), satu-
rator temperatures (T
S
), and volumetric ow rate ratios of carbon
dioxide to polymer solution (R = F
CO
2
/F
l
). The condition of the vol-
Table 1
Critical properties and acentric factors for pure components.
Components T
C
(K) P
C
(bar)
Carbon dioxide 304.1 73.8 0.239
Acetone 508.1 47.0 0.304
Table 2
Interaction parameters for carbon dioxide (1) +acetone (2) system.
T (K) k
12
l
12
P/P (%) y
2
(%)
313.2 0.0149 0.0179 2.6 0.64
333.2 0.0435 0.0220 4.0 0.80
353.2 0.0557 0.167 2.7 2.3
373.2 0.0869 0.217 3.9 3.8
Mixing rule: aM =

j
x
i
x
j
_
a
i
a
j
(1 k
ij
); bM =

j
x
i
x
j
b
i
+b
j
2
(1 l
ij
).
450 400 350 300 250 200 150 100 50 0
Particle size (nm)
0
0.1
0.2
0.3
F
r
a
c
t
i
o
n

o
f

P
a
r
t
i
c
l
e

N
u
m
b
e
r
Fig. 3. Theparticlesizedistributions of thesamples producedusingtwo sprayangles
of nozzles in SAA process (50

, solid line; 5

, dash line).
umetric ow rate of carbon dioxide is 276K and 6.5MPa. Table 3
lists the experimental conditions and results.
2.3.1. Effect of spray angle of nozzle
Two spray angles (5

, 50

) of nozzles were used in the SAA pro-


cess, the operating conditionof two runs (#1 and #4 inTable 3) was
C=10kg/m
3
, R=1.2, and T
S
=353.2K. Fig. 3 shows the particle size
distributions of the samples produced with different nozzles. The
sharp particle size distribution (PSD) was obtained as the larger
spray angle nozzle (50

) was used, and the mean particle size


increasedaccordingly while decreasing the spray angle of nozzle. In
small spray angle nozzle, the atomized liquid streamin nozzle out-
let approximated solid stream, the resulted PMMA therefore was
not favor the smaller particles sizes. The spray angle nozzle of 50

was thus utilized throughout all the rest of the experiments.


2.3.2. Effect of concentration of polymer solution
The effect of the solute concentration on particle size was
obtained at different PMMA concentrations (C) in acetone between
1 and 50kg/m
3
(runs #2 to #6 in Table 3), at the condition of the
saturator temperature (T
S
=353.2K) and R=1.2. The FESEMimages
of sample particles shown in Fig. 4 are always formed by discrete
spherical sub-micrometric particles and are collected at different
C. Fig. 5 shows that the lower concentration (C) of polymer solu-
tion can effectively reduce the particle size and sharply the size
distribution. Fig. 6 shows the mean particle size (

d
no
) varying with
the concentration of polymer solution. These results explained the
greater viscosity or surface tension of the higher concentration
solution might cause the larger liquid droplets and increased the
polymer particles size.
2.3.3. Effect of saturator temperature
The effect of the saturator temperature (T
S
) on the PMMA par-
ticles was also obtained at various temperatures from 313.2K to
363.2K (runs #4, and #7 through #9 in Table 3), at the condition
of the concentration of polymer solution (10kg/m
3
), and the vol-
umetric ow rate ratio (R) of 1.2. Fig. 7 shows the mean particle
size (

d
no
) of the PMMA varying with the saturator temperature,
indicating that the smaller PMMA particles were formed at higher
temperatures. Assuming that the solution in the saturator can form
102 H.-T. Wu, M.-W. Yang / J. of Supercritical Fluids 59 (2011) 98107
Fig. 4. FESEMimages of PMMA particles producedby SAAat different concentrations (C) of polymer solution: (a) C=1kg/m
3
, (b) C=10kg/m
3
, (c) C=30kg/m
3
, (d) C=50kg/m
3
.
Table 3
Experimental conditions and results.
Run C T
S
F
CO
2
R

dno

d
expt
4,3
PDI
(kg/m
3
) (K) (cm
3
/min) F
CO
2
/F
l
(nm)
a
(nm)
b
d
expt
4,3
/

dno
1
c
10 353.2 6 1.2 173 5 267 14 1.5
2
d
1 353.2 6 1.2 82 2 127 11 1.5
3
d
5 353.2 6 1.2 101 3 150 10 1.5
4
d
10 353.2 6 1.2 126 3 173 9 1.4
5
d
30 353.2 6 1.2 136 4 250 16 1.8
6
d
50 353.2 6 1.2 155 3 286 12 1.9
7
d
10 363.2 6 1.2 130 6 224 8 1.7
8
d
10 333.2 6 1.2 137 3 206 7 1.5
9
d
10 313.2 6 1.2 166 4 239 9 1.4
10
d
10 353.2 4 0.8 146 4 256 14 1.8
11 10 353.2 8 1.6 154 4 258 15 1.7
12
d
10 333.2 8 1.6 136 2 201 8 1.5
13
d
10 333.2 8 2.0 132 5 190 9 1.4
14 10 333.2 8.4 2.4 159 1 218 3 1.4
15 10 333.2 8.4 2.8 176 4 300 19 1.7
16
d
10 313.2 8 1.6 152 3 230 8 1.5
17
d
10 313.2 8 2.0 142 2 208 7 1.5
18
d
10 313.2 8.4 2.4 131 2 178 4 1.4
19 10 313.2 8.4 2.8 156 3 253 11 1.6
20
e
10 353.2 6 1.2 147 5 239 15 1.6
21
f
10 353.2 6 1.2 149 3 239 18 1.6
a
dno =

N
i=1
d
i
N
, arithmetic mean size.
b
d
expt
4,3
=

N
i=1
d
4
i

N
i=1
d
3
i
, mass-weighted mean size.
c
The spray angle nozzle of 5

.
d
Data used in the population balance model of the MSMPR precipitator.
e
MW=15kg/mol.
f
MW=350kg/mol.
H.-T. Wu, M.-W. Yang / J. of Supercritical Fluids 59 (2011) 98107 103
500 400 300 200 100 0
Particles size (nm)
0
0.1
0.2
F
r
a
c
t
i
o
n

o
f

p
a
r
t
i
c
l
e

n
u
m
b
e
r
C = 1 kg/m
3
C = 5 kg/m
3
C = 10 kg/m
3
C = 30 kg/m
3
C = 50 kg/m
3
0.3
Fig. 5. The particle size distributions of the samples produced by SAA at differ-
ent concentrations (C) of polymer solution from 1kg/m
3
to 50kg/m
3
(T
S
=353.2K,
R =1.2).
a homogenous liquid phase, the composition of carbon dioxide in
the solution depends on the volumetric owrate ratio (R) and lim-
its to the VLE phase envelopes. According the VLE diagram(Fig. 2),
at a constant pressure (P=6.5MPa), the point B refers to R=1.2 and
is near the compositionof carbondioxide insaturated compressed-
liquid(bubble point) at 353.2K. Andthen, the compositionof CO
2
at
the bubble point increases with decreasing temperatures, e.g., the
bubblepoint at 333.2Kis approximatelyR=1.62.0(points Cor Din
Fig. 2). Therefore, the volumetric owrate ratio (R=1.2) inthe satu-
rator temperature range of 313.2353.2K, the compositions of CO
2
in the liquid solution are nearly identical, smaller PMMA particles
formed at higher temperatures may be attributed to the viscosity
of solution decreasing with an increase of temperatures. However,
Fig. 7 presents a slightly reversed tendency at 363.2K. The larger
0 10 20 30 40 50
C (kg/m
3
)
70
90
110
130
150
M
e
a
n

p
a
r
t
i
c
l
e

s
i
z
e

(
n
m
)
170
Fig. 6. Mean particle size (

dno) varying with concentration of polymer solution at


T
S
=353.2K, and R=1.2.
310 320 330 340 350 360 370
T
S
(K)
120
140
160
180
M
e
a
n

p
a
r
t
i
c
l
e

s
i
z
e

(
n
m
)
Fig. 7. Mean particle size (

dno) varying with the saturator temperature (T


S
) at
C =10kg/m
3
, and R=1.2.
PMMA particles obtained at 363.2Kmight caused by being near the
glass transitiontemperature of PMMA (T
g
=372.2K), fusedtogether
as larger particles.
2.3.4. Effect of the volumetric ow rate ratio
The mean particle size (

d
no
) of micronized PMMA at differ-
ent volumetric ow rate ratios (R) and saturator temperatures are
shown in Fig. 8, indicating that the higher R values favor smaller
PMMA particles and have an optimized R value for different satura-
tor temperatures (e.g., R=1.2 at T
S
=353.2K; R=2.0 at T
S
=333.2K;
R=2.4 at T
S
=313.2K). The higher R values indicate that more CO
2
in the liquid solution and can enhance the secondary atomization
of the SAA process. Furthermore, the composition of CO
2
in the
liquid solution is limited to the VLE phase envelopes. Therefore,
the inuence of the volumetric ow rate ratio (R) on the parti-
cles size can accord with the phase behavior of the mixtures in the
saturator chamber [1]. Assuming the solution in the saturator can
forma homogenous liquid phase, Points AF in the phase diagram
(Fig. 2) refer to R-values of 0.8, 1.2, 1.6, 2.0, 2.4 and 2.8, respectively.
According to the VLE phase boundaries of different temperatures,
the composition of carbon dioxide in saturated compressed-liquids
(bubble point) were approximately R=1.2 (353.2K), R=1.62.0
(333.2K) and R=2.8 (313.2K), respectively. These results are con-
sistent with the optimized R values in Fig. 8. If the composition
of carbon dioxide in the solution is greater than the bubble point
(i.e., higher than the optimized R values), the operating conditions
probably into the vaporliquid coexistence region (CO
2
-rich vapor
and acetone-rich compressed liquid phases). The precipitation of
polymer in the saturator might be occurred, since CO
2
played as
anti-solvent (i.e., acetone soluble in CO
2
-rich vapor) and caused
higher concentration of polymer solution. And then, the larger par-
ticles were obtained [1]. However, the optimized R value (2.4)
of T
S
=313.2K was slightly less than the bubble point at 313.2K
(R=2.8), causebythemolefractionof CO
2
insaturatedcompressed-
liquid to be higher than 0.81, i.e., the anti-solvent effect of CO
2
.
2.3.5. Effect of the MWof PMMA
The effect of the molecular weight (MW) of PMMA on par-
ticles size was demonstrated in runs #4 (MW =120kg/mol),
#20 (MW =15kg/mol) and #21 (MW =350kg/mol), as shown in
Table 3. The lower MW (15kg/mol) of PMMA possessed a lower
104 H.-T. Wu, M.-W. Yang / J. of Supercritical Fluids 59 (2011) 98107
0.8 1.2 1.6 2 2.4 2.8
120
140
160
180
M
e
a
n

p
a
r
t
i
c
l
e

s
i
z
e

(
n
m
)
0.8 1.2 1.6 2 2.4 2.8
120
140
160
180
0.8 1.2 1.6 2 2.4 2.8
R = F
CO2
/F
l
120
140
160
180
(c) 313.2 K
(b) 333.2 K
(a) 353.2 K
Fig. 8. Mean particle size (

dno) varying with the volumetric ow rate ratio (R)


at C=10kg/m
3
and saturator temperatures: (a) T
S
=353.2K, (b) T
S
=333.2K, (c)
T
S
=313.2K.
glass transition temperature (T
g
) of 355.2K. The higher MW
(350kg/mol) of PMMA possesseda larger viscosity of polymer solu-
tionat C=10kg/m
3
(=0.41cp, 20

C, 120kg/mol; =0.49cp, 20

C,
350kg/mol). Both the lower and higher MWs of PMMA formed
larger particles, the particle size distributions were also broad
(PDI =1.37, 120kg/mol; PDI =1.6, 15 and 350kg/mol). DSC anal-
ysis implemented on untreated polymers is shown in Fig. 9, the
glass transition temperatures (T
g
) of PMMA are 355.2K, 372.2K,
and 378.2K (onset) for MWs of 15, 120, and 350kg/mol, respec-
tively. And then, the SAA processed polymer showed no obviously
different T
g
with untreated samples in three MWs of PMMA parti-
cles.
3. Precipitation kinetics
3.1. Population balance equation
The modeling of the SAA process is required for rationally
design, optimize andestimate the precipitationbehavior of the SAA
process. The population balance theory is widely used to deter-
minate the nucleation and growth parameters for a precipitation
process, and can be a valuable tool to discern primary or secondary
nucleation is dominant [16]. Jarmer et al. [7], Wu et al. [8] and
Boutin [9] successfully correlated their precipitation experiment
130 110 90 70
Temperature (
o
C)





m
W

/

m
g
120 k, untreated
120 k, SAA processed
350 k, untreated
350 k, SAA processed
15 k, SAA processed
15 k, untreated
Fig. 9. DSC thermal histograms of untreated and SAA processed PMMA for three
MWs of PMMA.
results for the SAS process with the population balance equation
for a mixed-suspension, mixed-product-removal (MSMPR) pre-
cipitator. The following assumptions were made for an MSMPR
crystallizer: no crystals were in the feed stream, all crystals were
the same shape and characterized by a chosen linear dimension L,
no break-down of crystals occurred, a well-mixed crystallizer with
mixed product removal (no classication at withdrawal) were at
steady-state. Thus, for an MSMPR crystallizer, the population bal-
ance equation is simplied as well-known exponential population
distribution
n = n
o
exp
_
L
G
_
(1)
where n
o
is the population density of nuclei (i.e., the size is close
to zero), G is the particle growth rate, and is the mean residence
time of the mixed suspension. If the particle size distribution of
a sample follows Eq. (1), the logarithmic population density (lnn)
should vary linearly with particle size (L) and the kinetic parame-
ters, ln n
o
and 1/G, which can be determined fromthe intercept
and the slope, respectively. By denition, the nucleation rate B
o
is
then calculated from
B
o
= n
o
G (2)
The original particle size distribution (e.g., Fig. 5) must be con-
vertedintothe populationdensityn(L) prior toproceeding withthe
analysis of precipitation kinetics. The population density in each
size interval can be determined [16] by
n

L
=
M
T
W
k
V

L
3
L
(3)
where
C
is the particles density (1188kg/m
3
for PMMA), L is the
arithmetic average size ranging from L
1
to L
2
, and L =L
2
L
1
is
the width of the size range. The volumetric shape factor k
V
for a
sphere equals /6. Wis the weight fraction of particles in the size
range L
1
L
2
. To calculate W, the mass of each individual particle
H.-T. Wu, M.-W. Yang / J. of Supercritical Fluids 59 (2011) 98107 105
500 400 300 200 100 0
L (nm)
30
32
34
36
38
l
n
n

(
n
o
.
/
m
3

n
m
)
Fig. 10. Population density plots for the experimental run #2 (, MT =0.098kg/m
3
,
long dash line, R
2
=0.92), run #4 (, MT =0.982kg/m
3
, short dash line, R
2
=0.88), run
#6 (, MT =4.908kg/m
3
, solid line, R
2
=0.94).
wis estimated by assuming that the PMMA particle is a sphere with
diameter d, i.e.,
w =
d
3

C
6
(4)
The suspension density M
T
in Eq. (3) was calculated on a clear
liquid volume basis [7] by
M
T
=
_
F
l
F
l
+F
CO
2
(T
S
)
_
C (5)
where C (kg/m
3
) is the concentration of PMMA in the solution feed
stream and F
l
is the volumetric ow rate of the polymer solution.
F
CO
2
(T
S
) is the volumetric owrate of carbon dioxide, which is cal-
culated from the density (
CO
2
) of carbon dioxide at the saturator
temperature [17].
3.2. Correlation results and discussions
Since the operation conditions of the higher volumetric ow
rate ratio (R) might into the vaporliquid coexistence region and
produced the obvious separation of vapor and liquid phases. The
assumption of the well-mixed crystallizer was not available. Thus,
this study correlated experimental results of 14 runs of less than
the optimized R with the MSMPR population balance equation, and
investigated the precipitation kinetics of the SAA process. The cor-
related results are shown in Table 4. Fig. 10 shows the correlation
of the population density distribution with Eq. (1) for runs #2,
#4, and #6. The correlation coefcients (R
2
) of tting were 0.92,
0.88, and 0.94, respectively. A linear relationship was observed for
each individual run. The SAA precipitation approximately follows
the MSMPR population balance equation, Eq. (1). The precipitation
kinetic parameters, n
o
and G, are calculated from the intercept
and the slope of the straight line, respectively, as listed in Table 4.
The overall averaged correlation coefcient (R
2
) is 0.92. The values
of the determined kinetic parameters need to meet the following
constraint for an MSMPR precipitator:
M
T
= k
V

C
m
3
= 6k
V

C
n
o
(G)
4
(6)
where m
3
is the third moment of the population density distribu-
tion and given by
m
3
=
_

0
L
3
n(L)dL =
_

0
L
3
n
o
exp
_

L
G
_
dL = 6n
o
(G)
4
(7)
The data points used in the correlation are adjusted to obey the
constraint. Typically, the data points of smaller sizes are discarded
in the data tting.
While the value of G is obtained from the above tting, the
mean residence time should further be specied to calculate the
growth rate G. It could be estimated from
=
V
mixing
F
l
+F
CO
2
(T
S
)
(8)
where V
mixing
is the mixing volume, the volume of injected poly-
mer solution within one second was assumed as a reference value
of V
mixing
in each run and the relative growth rate G was thus calcu-
lated, as shown in Table 4. Uchida et al. [18] measured the growth
rates of a naphthalene single crystal insupercritical carbondioxide.
According to their experimental results, the growth rates were in
the range of (10
7
to 10
8
m/s), falling between those in the liquid
phase (10
5
to 10
7
m/s) and in the vapor phase (<10
8
m/s) at the
samesupersaturation. Thecalculatedrelativegrowthrates, as listed
inTable4, areintheorder of magnitudeof 10
7
m/s, whicharecom-
parable to the literature values [18]. Furthermore, the calculated
nucleation rates B
o
(=n
o
G), also listed in Table 4, are in the range
of (0.80 to 3.52) 10
19
no./s m
3
. Compared, with the results from
other studies, such as Muhrer et al. [19] (the gas anti-solvent pro-
cess, GAS, 10
14
no./s m
3
), Weber et al. [20] and Turk [21] (the rapid
expansion of supercritical solutions, RESS, 10
20
to 10
32
no./s m
3
),
Jarmer et al. [7] (SAS, 10
18
no./s m
3
), Wu et al. [8] (SAS, 10
16
to
10
17
no./s m
3
) and Boutin [9] (SAS, 10
14
to 10
19
no./s m
3
), the cal-
culated nucleation rates of this study are reasonable.
Table 4 shows that the growth rate G and the dominant particle
sizes (L
d
=3G) increased with increasing suspension density M
T
over run #2 to run #6, in which M
T
varies 50-fold, from 0.098 to
4.91kg/m
3
, under the same mean residence time. The growth rate
(G) of particles was determined fromthe slope (1/G) in the pop-
ulation density plots. The PDIs (

d
4,3
/d
no
) of runs #2, #4, and #6 in
Table 3 are also related to the slope of the population density plots
in Fig. 10. A greater PDI value indicates a broad particle size dis-
tribution and represents a smaller slope in the population density
plot or a greater growth rate (G) of polymer particles. For exam-
ple, the SAA prepared sample with the greater suspension density
M
T
(run #6) has a board particle size distribution (PDI =1.9) with a
smaller slope value or greater growth rate (G=717nm/s) of poly-
mer particles. On the contrary, the sample with smaller suspension
density (run #2) has a narrow particle size distribution (PDI =1.5)
withasmaller growthrate(G=323nm/s) of polymer particles inthe
SAA process. Additionally, non-parallel lines (the different slopes)
exhibit on the population density plots for runs #2, #4, and #6, as
shown in Fig. 10. This suggests that the different size distribution
can be achieved by changing the suspension density of a system,
i.e., the primary nucleation may be dominant in the SAA precip-
itation. The primary nucleation rates are often correlated with a
power-lawkinetic expression [16]:
B
o
= n
o
G = K
N
G
i
(9)
Through a linear regression, the best-t to Eq. (9) for this study
was obtained as
B
o
= n
o
G = exp(32.44)G
1.90
(10)
The correlation coefcient (R
2
) of this tting is 0.92. Accord-
ingly, if the primary nucleation is predominant, the kinetic order
i would be greater than 2 [7]. Therefore, the correlated result, Eq.
106 H.-T. Wu, M.-W. Yang / J. of Supercritical Fluids 59 (2011) 98107
Table 4
Correlated results fromthe population balance model of the MSMPR precipitator.
Run
CO
2
MT 10
16
n
o
G G 10
18
B
o
d
expt
4,3
(kg/m
3
) (kg/m
3
) (no./m
3
nm) (nm) (s) (nm/s) (no./m
3
s) (nm)
2 122 0.098 2.47 31.7 0.098 323 7.99 127 11
3 122 0.491 3.26 44.9 0.098 458 14.9 150 10
4 122 0.982 3.54 55.0 0.098 561 19.9 173 9
5 122 2.945 4.74 62.8 0.098 640 30.3 250 16
6 122 4.908 4.91 70.4 0.098 717 35.2 286 12
7 115 0.931 2.99 55.0 0.093 591 17.7 224 8
8 140 1.110 3.48 55.7 0.111 502 17.5 206 7
9 171 1.324 3.36 61.6 0.132 465 15.6 239 9
10 122 1.404 2.48 63.1 0.140 450 11.1 256 14
12 140 0.857 3.71 52.1 0.086 609 22.6 201 8
13 140 0.697 4.79 45.9 0.070 658 31.6 190 9
16 171 1.027 3.80 55.7 0.103 543 20.6 230 8
17 171 0.839 3.88 49.6 0.084 591 22.9 208 7
18 171 0.588 4.24 42.9 0.071 605 25.7 178 4
(10), shows that the kinetic order i is close to 2, which is in accor-
dance with the primary nucleation. The result is consistent with
that fromthe population density plots shown in Fig. 10. Addition-
ally, the mass-weightedmeansize (

d
4,3
) for anMSMPRprecipitator
is given by [22]

d
4,3
= 4G (11)
Introducing Eqs. (10) and Eq. (6) into Eq. (11) yields

d
calc
4,3
= 4G =
_
4
i+3

i1
M
T
6
C
k
V
K
N
_
1/(i+3)
(12)
The value of

d
4,3
can also be calculated from the experimental
PSD by

d
expt
4,3
=

i
x
i

L
4
i

i
x
i

L
3
i
(13)
where x
i
is the number of particles within the size range and

L
i
is
the mean size of each size range. The experimental d
4,3
of each run
is listedinTable 4. The calculatedvalues of

d
4,3
, fromEq. (12), agree
with the experimental

d
4,3
to an average absolute deviation of 6.5%
over the entire experimental conditions of Table 4.
0.0032 0.003 0.0028 0.0026
1/T
S
(K)
6.1
6.2
6.3
6.4
l
n

G

(
n
m
/
s
)
Fig. 11. Arrhenius plot of particles growth rate (G) at C=10kg/m
3
and R=1.2.
AnArrhenius plot of the growthrate is showninFig. 11, inwhich
the experimental runs of #4and#7to#9are included. According to
the correlated results, the activation energy of the particle growth
is approximately 4.5kJ/mol, suggesting that diffusion (activation
energy less than 20kJ/mol), rather than surface integration (acti-
vation energy approximately 4060kJ/mol), controls the particle
growth rate [22].
4. Conclusions
Sub-micrometric particles of PMMA were produced via a super-
critical assisted-atomization (SAA) process using acetone as a
solvent and supercritical carbon dioxide as a spraying medium. The
lower concentrations of polymer solution and higher temperatures
at the saturator were determined to be favorable for reducing the
PMMA particle size. At the optimized volumetric ow rate ratios,
the composition of CO
2
in the feed streams was near the bubble
points of the saturator temperatures. The MSMPR population bal-
ance equation is capable of representing the precipitation kinetics.
Reasonable values of the nucleation rates and growth rates were
obtained fromthe data correlation. As evidenced fromthe experi-
mental data analysis, it was suggested that the primary nucleation
was dominant in the SAA precipitation, and that the growth of the
polymer particles was governed by diffusion, rather than surface
integration.
Acknowledgement
The authors gratefully acknowledge the nancial support of the
National Science Council, Taiwan, through Grant No. NSC97-2221-
E131-013.
References
[1] E. Reverchon, A. Antonacci, Polymer microparticles production by supercritical
assisted atomization, J. Supercritical Fluids 39 (2007) 444452.
[2] E. Reverchon, R. Adami, I. De Macro, C.G. Laudani, A. Spada, Pigment red 60
micronizationusing supercritical uids basedtechniques, J. Supercritical Fluids
35 (2005) 7682.
[3] H.T. Wu, H.M. Lin, M.J. Lee, Ultra-ne particles formation of pigment green 36
in different phase regions via a supercritical anti-solvent process, Dyes and
Pigments 75 (2007) 328334.
[4] E. Reverchon, G. Caputo, I. De Marco, Role of phase behavior and atomiza-
tion in the supercritical anti-solvent precipitation, Industrial and Engineering
Chemistry Research 42 (2003) 64066414.
[5] H.T. Wu, M.J. Lee, H.M. Lin, Supercritical uid-assisted dispersion of ultra-ne
pigment red 177 particles with blended dispersants, J. Supercritical Fluids 39
(2006) 127134.
[6] H.T. Wu, K.Y. Huang, K.T. Lee, Supercritical uid-assisted dispersion of C.I. Pig-
ment Violet 23inanorganic medium, Powder Technology 206(2011) 322326.
[7] D.J. Jarmer, C.S. Lengsfeld, T.W. Randolph, Nucleation and growth rates of
poly(l-lactic acid) microparticles during precipitation with a compressed-uid
antisolvent, Langmuir 20 (2004) 72547264.
H.-T. Wu, M.-W. Yang / J. of Supercritical Fluids 59 (2011) 98107 107
[8] H.T. Wu, M.J. Lee, H.M. Lin, Precipitation kinetics of pigment blue 15:6 sub-
micro particles with a supercritical anti-solvent process, J. Supercritical Fluids
37 (2006) 220228.
[9] O. Boutin, Modeling of griseofulvin recrystallization conducted in a supercriti-
cal antisolvent process, Crystal Growth and Design 9 (2009) 44384444.
[10] E. Reverchon, I. De Marco, Mechanisms controlling supercritical antisolvent
precipitate morphology, Chemical Engineering J. 169 (2011) 358370.
[11] G. Caputo, R. Adami, E. Reverchon, Analysis of dissolved-gas atomization:
supercritical CO
2
dissolved in water, Industrial and Engineering Chemistry
Research 49 (2010) 94549461.
[12] C. Elviraa, A. Fanovichb, M. Fernndeza, J. Fraileb, J.S. Romna, C.
Domingo, Evaluation of drug delivery characteristics of microspheres of
PMMAPCLcholesterol obtained by supercritical-CO
2
impregnation and
by dissolution evaporation techniques, J. Controlled Release 99 (2004)
231240.
[13] C.J. Chang, C.Y. Day, C.M. Ko, K.L. Chiu, Densities and Pxy diagrams for carbon
dioxide dissolution in methanol, ethanol, and acetone mixtures, Fluid Phase
Equilibria 131 (1997) 243258.
[14] H.Y. Chiu, M.J. Lee, H.M. Lin, Vaporliquid phase boundaries of binary mixtures
of carbon dioxide with ethanol and acetone, J. Chemical and Engineering Data
53 (2008) 23932402.
[15] F. Han, Y. Xue, Y. Tian, X. Zhao, L. Chen, Vaporliquid equilibria of the
carbon-dioxide +acetone system at pressures from (2.36 to 11.77) MPa and
temperatures from(333.15 to 393.15) K, J. Chemical and Engineering Data 50
(2005) 3639.
[16] A.D. Randolph, M.A. Larson, Theory of Particulate Processes, Academic Press,
NewYork, 1971.
[17] Isothermal properties for carbon dioxide, NIST Chemistry WebBook, NIST Stan-
dard Reference Database No. 69 March 2003, Release, National Institute of
Standard and Technology, USA, http://webbook.nist.gov/chemistry/.
[18] H. Uchida, A. Manaka, M. Matsuoka, H. Takiyama, Growth phenomena of sin-
gle crystal of naphthalene in supercritical carbon dioxide, Crystal Growth and
Design 4 (2004) 937942.
[19] G. Muhrer, C. Lin, M. Mazzotti, Modeling the gas anti-solvent recrystallization
process, Industrial and Engineering Chemistry Research 41 (2002) 35663579.
[20] M. Weber, L.M. Russel, P.G. Debenedetti, Mathematical modeling of nucleation
and growth of particles formed by the rapid expansion of a supercritical solu-
tion under subsonic conditions, J. Supercritical Fluids 23 (2002) 6580.
[21] M. Turk, Inuence of thermodynamic behaviour and solute properties on
homogeneous nucleation in supercritical solutions, J. Supercritical Fluids 18
(2000) 169184.
[22] J.W. Mullin, Crystallization, 4th ed., Butterworth-Heinemann, Oxford, 2001.