1970 Langmuir 2009, 25, 1970-1979

Nanoprecipitation of Polymethylmethacrylate by Solvent Shifting:

1. Boundaries
Julien Aubry,† Francois Ganachaud,*,† Jean-Pierre Cohen Addad,‡ and Bernard Cabane*,§
Ingénierie et Architectures Macromoléculaires, Institut Charles Gerhardt CNRS UMR5253, ENSCM,
8 Rue de l‘Ecole Normale, 34296 Montpellier Cedex, France, Laboratoire de Spectrométrie Physique de
Grenoble, UMR CNRS/UJF 5588, UniVersité Joseph Fourier Grenoble I, BP 87, 38402 Saint-Martin
d’Hères Cedex, France, and Laboratoire PMMH, CNRS UMR 7636, ESPCI, 10 rue Vauquelin,
75231 Paris Cedex, France

ReceiVed September 12, 2008. ReVised Manuscript ReceiVed December 11, 2008

Nanoparticles of polymethylmethacrylate (PMMA) have been produced through the solvent shifting process (also
called the “Ouzo process”) in which water (nonsolvent) was added to a solution of PMMA in acetone or tetrahydrofuran
(solvent). At low concentrations of PMMA in the initial solution, and for large additions of water, the process yielded
PMMA nanoparticles with a narrow distribution of particle sizes. The mean particle diameter varied as a power law
of the initial PMMA concentration in the solvent, in agreement with the predictions from the Smoluchowski equation
for an aggregation process that has definite “start” and “stop” times. At higher PMMA concentrations, the mixing
process yielded microparticles coexisting with PMMA nanoparticles. The boundary that separates the Ouzo region
of compositions (PMMA nanoparticles only), from the “non-Ouzo” region (nano- and microparticles) has been determined.
This boundary does not appear to have any relation to the spinodal decomposition line of the ternary solutions: the
transition from Ouzo to non-Ouzo behavior must have another unknown origin.

Introduction process. The main drawback is the large dilution of thus-prepared

dispersions or emulsions, with mass fractions as low as 10-4.
“Solvent-shifting” is a very simple process by which hydro- The papers in this series address this issue, i.e., why nanopre-
phobic molecules are dispersed in water as tiny (at least cipitation is limited to such low concentrations and whether there
submicrometric) droplets or particles.1 The process appears are ways to overcome this limitation.
deceptively simple: the hydrophobic solute is first dissolved into Previous reports have already indicated that the production of
a polar solvent that is fully miscible with water, and then this metastable dispersions or emulsions is limited to a small region
solution is mixed with a large amount of water. The solution of the composition map, called the Ouzo region.9 When oils
becomes a nonsolvent for the hydrophobic molecules and it phase- were used as hydrophobic solutes, ethanol as the solvent, and
separates into small droplets or particles and a continuous phase. water as the nonsolvent, metastable emulsions were obtained
An everyday use of this process is the spontaneous emulsification only if the final mass fraction of solute was below 2 × 10-2. With
of alcoholic drinks that are diluted in water, such as Ouzo in organic dyes, steroids, carotenoids, or polymers, metastable
Greece and “Pastis” in France.2-5 Industrially, the “solvent dispersions of nanoparticles were likely obtained, again limited
shifting” route is used to manufacture submicrometric particles to similarly low solute concentrations.6,10-15 At present, the
of, for example, Vitamin A,6 dyes,7 and drugs.8 It has some physical cause of this limitation is not understood. According to
advantages over other processes such as high-pressure homog- one school of thought, the Ouzo region is the region where the
enization or grinding, in particular the very small sizes that can solution is sufficiently dilute to be metastable with respect to the
be achieved (often nanometric), the simple nature of equipment, concentration of hydrophobic solute; spontaneous fluctuations
the low energy costs, and the easy implementation as an online of the solute concentration produce very small nuclei of pure
solute, and only those nuclei that are larger than a critical size
* Corresponding author. E-mail: francois.ganachaud@enscm.fr. Tel: 33 grow by capturing other solute molecules. This nucleation-and-
4 67 14 72 96. Fax: 33 4 67 14 72 20 (F.G.). E-mail: bcabane@pmmh.espci.fr.
growth process would produce a dispersion of pure solute droplets
Tel: 33 1 40 79 47 15. Fax: 33 1 40 79 45 23 (B.C.).
†
Institut Charles Gerhardt CNRS UMR5253, ENSCM. or particles.9 Beyond the Ouzo boundary, the solution would be
‡
Université Joseph Fourier Grenoble I. unstable, and large-scale fluctuations of the solute concentration
§
CNRS UMR 7636, ESPCI. would grow spontaneously, as in spinodal decomposition
(1) For recent reviews on spontaneous emulsification by solvent shifting, see:
(a) Horn, D.; Rieger, J. Angew. Chem., Int. Ed. 2001, 40, 4330–4361. (b) Lopez- processes, yielding macroscopic aggregates. In this view, the
Montilla, J. C.; Herrera-Morales, P. E.; Pandey, S.; Shah, D. O. J. Dispersion Sci.
Technol. 2002, 23, 219–268. (c) Miller, C. A. Surf. Sci. Ser. 2006, 132, 107–126. (9) Vitale, S. A.; Katz, J. L. Langmuir 2003, 19, 4105–4110.
(2) Sitnikova, N. L.; Sprik, R.; Wegdam, G.; Eiser, E. Langmuir 2005, 21, (10) Brick, M. C.; Palmer, H. J.; Whitesides, T. H. Langmuir 2003, 19, 6367–
7083–7089. 6380.
(3) Grillo, I. Colloid Surf., A 2003, 225, 153–160. (11) (a) Van Keuren, E.; Georgieva, E.; Durst, M. J. Dispersion Sci. Technol.
(4) Carteau, D.; Pianet, I.; Brunerie, P.; Guillemat, B.; Bassani, D. M. Langmuir 2003, 24, 721–729. (b) Van Keuren, E. R. J. Dispersion Sci. Technol. 2004, 25,
2007, 23, 3561–3565. 547–553. (c) Van Keuren, E.; Bone, A.; Ma, C. Langmuir 2008, 24, 6079–6084.
(5) Scholten, E.; van der Linden, E.; This, H. Langmuir 2008, 24, 1701–1706. (12) Lannibois, H.; Hasmy, A.; Botet, R.; Chariol, O. A.; Cabane, B. J. Phys
(6) Horn, D.; Schmidt, H. W.; Ditter, W.; Hartmann, H.; Lueddecke, E.; II 1997, 7, 319–342.
Schmieder, K. U. S. Patent 4,522,743, 1985. (13) Stainmesse, S.; Orecchioni, A.-M.; Nakache, E.; Puisieux, F.; Fessi, H.
(7) (a) Kasai, H.; Nalwa, H. S.; Oikawa, H.; Okada, S.; Matsuda, H.; Minami, Colloid Polym. Sci. 1995, 273, 505–511.
N.; Kakuta, A.; Ono, K.; Mukoh, A.; Nakanishi, H. Jpn. J. Appl. Phys., Part 2: (14) Galindo-Rodriguez, S.; Allemann, E.; Fessi, H.; Doelker, E. Pharm. Res.
Lett. 1992, 31, L1132-L1134. (b) Texter, J.; Travis, W. B.; Flow, V. U.S. Patent 2004, 21, 1428–1439.
5,624,467, 1997. (15) Legrand, P.; Lesieur, S.; Bochot, A.; Gref, R.; Raatjes, W.; Barratt, G.;
(8) Violante, M. B.; Fischer, H. W. U.S. Patent 4,997,454, 1991. Vauthier, C. Int. J. Pharm. 2007, 344, 33–43.

10.1021/la803000e CCC: $40.75  2009 American Chemical Society

Published on Web 01/26/2009
Phase Boundaries in Nanoprecipitation of PMMA Langmuir, Vol. 25, No. 4, 2009 1971

Ouzo boundary is the spinodal line of the ternary solution. Another the Ouzo boundary of the nanoprecipitation region, before finally
school of thought holds the opposite point of view that discussing the nature of the nanoprecipitation process. In the
interdiffusion of solvent and water always causes the solution second paper we shall discuss the influence of additives, such
to become unstable with respect to concentration fluctuations, as surfactants, on the control of nanoprecipitation and shelf
and that it is through a spinodal decomposition mechanism that stability. In the third paper we shall use the Flory-Stockmayer
the solution decomposes into solute particles and the solvent theory to calculate the location of the spinodal lines of the ternary
mixture.10 In this case, the Ouzo boundary could be related to systems, and compare this location with that of the Ouzo boundary.
a modification of the spinodal decomposition process according We will then propose an alternative explanation for the nature
to the solute concentration. of the Ouzo boundary, and a model for the switch from
Systematic experiments performed with a polymer as the nanoprecipitation to microprecipitation.
hydrophobic solute, can provide some critical information
regarding the formation of particles and the location of the Ouzo Experimental Section
region where only nanoparticles are obtained. Indeed, the
Materials and Methods. PMMA was purchased from Aldrich.
nanoprecipitation process has been largely used to generate Size exclusion chromatography in THF yielded a weight-average
nanoparticles or nanocapsules; for a recent review on the molar mass of 14700 g/mol with a polymolecularity of 1.54. Acetone
preparation of polymer-based colloids using the Ouzo effect, see and THF were respectively purchased from Analar Normapur and
ref 16. First, the choice of the polymer aims to prevent the growth Riedel de Haen with a purity of 99%. Brij56 or polyoxyethylene
of the nanodroplets through Ostwald ripening, or their loss through (10) cetyl ether, a nonionic surfactant of intermediate hydrophilic-
dissolution-recrystallization process, as observed when dispersing lipophilic balance (HLB) (12.9), was purchased from Aldrich and
oils and dyes, respectively: it is possible to keep the polymer systematically introduced in the water to ensure long-term colloidal
particles over extended periods of time. Second, the boundary stability of the dispersion. NaOH pellets were purchased from SDS
of the Ouzo region has been reported to follow a simple law in with a purity of 98%. High-performance liquid chromatography
(HPLC)-grade water was purchased from Panreac, and buffer
the case of polymer solutes. Indeed, while Stainmesse et al.13
solutions were obtained from Carlo Erba. Unless otherwise stated,
produced dispersions of polycaprolactone (PCL) nanoparticles all products were used as received.
through a solvent shifting process, they found that the maximum Particle size analysis was performed on filtered (polyester
polymer concentration (boundary of the Ouzo region) decreased membrane filters, pore size 1.2 µm) and undiluted dispersions using
exponentially with the S/W ratio. This trend was confirmed in back-scattered quasi-elastic light scattering (Nanotrac 250). This
numerous other solvent-shifting experiments.14,17,18 apparatus uses the Brownian motion of particles in a fluid to determine
In order to determine the nature of the processes that take particle size distributions (PSDs). The laser provides a beam of
place on either side of this Ouzo boundary (nucleation and growth? wavelength λ ) 780 nm, which is scattered by the particles so that
the signal frequency is shifted according to the Doppler effect caused
spinodal decomposition?), it is necessary to know the interaction
by their velocities. This scattered, frequency-shifted light is
free energies of the different components in the ternary system. transmitted to the photodetector together with a part of the initial
The system polymethylmethacrylate (PMMA)/acetone or tet- beam that is directly reflected to the photodetector without having
rahydrofuran (THF)/water has been extensively used in a related been scattered by particles. This reflected beam has the same
process, which is the manufacturing of porous membranes by frequency as the original laser and acts as a reference signal for
pouring water (nonsolvent) on a concentrated polymer solution. detection.22 The power spectrum of the interference signal is
In this case, the spinodal decomposition of the polymer solution calculated by high-speed fast Fourier transform digital signal
leads to the separation of polymer-rich and water-rich domains, processor hardware and inverted to give the PSD. The fact that the
which become the membrane and the pores, respectively. Thanks signal is taken at 180° from the original beam simplifies the
to previous work in this field, polymer-solvent and polymer- mathematics and makes it possible to analyze turbid, i.e., concentrated,
colloidal dispersions with sizes ranging from 0.8 to 6500 nm. The
nonsolvent interaction parameters are known,19-21 so that it should
Nanotrac software supplies volume and number PSDs using
be possible to make quantitative prediction for the phase separation Lorenz-Mie theory and intensity PSDs using Rayleigh-Debye
processes of the same ternary systems in the more dilute theory (see Supporting Information). For our measurements, we
composition regions. chose to indicate the intensity-average particle size dI because this
This paper (the first in a series of three) presents the results is generally the particle size provided by other dynamic light scattering
of the solvent shifting process performed in the PMMA/acetone/ instruments. The calculation of particle sizes requires the knowledge
water and PMMA/THF/water systems. We show that different of the viscosity and refractive index of the continuous phases, which
objects can be obtained, e.g., PMMA nanoparticles slightly were deduced from tables23,24 according to the water/solvent
swollen with solvent, or coexisting nanoparticles and large composition of the continuous phase. For each measurement we
also indicated the polydispersity in particle sizes, dV/dN, where dV
polymer lumps (referred to as “microparticles” in the rest of this is the volume-average particle size and dN is the number-average
paper), depending on the initial solution composition, or the particle size. The particle morphology was investigated using scanning
amount of added water. We first determine the characteristics of electron microscopy (SEM). SEM samples were prepared by
the nanoparticles that are produced by nanoprecipitation, and depositing a drop of the PMMA dispersion on an aluminum foil and
then for each system, we determine the equilibrium phase letting it evaporate at room temperature. The resulting film of PMMA
separation boundary of the ternary solutions (binodal line) and nanoparticles was then vacuum-coated with gold and analyzed.
Centrifugation of the dispersions was carried out at an acceleration
(16) Ganachaud, F.; Katz, J. L. ChemPhysChem 2005, 6, 209–216. of 60 × g for 30 min using a Sorvall GLC-2B centrifuge.
(17) Thioune, O.; Fessi, H.; Devissaguet, J. P.; Puisieux, F. Int. J. Pharm. Nanoprecipitation Process. PMMA was first dissolved in the
1997, 146, 233–238. organic solvent by stirring the powder in the solvent for one day.
(18) Murakami, H.; Kobayashi, M.; Takeuchi, H.; Kawashima, Y. Int. J. Pharm. The mass fraction of PMMA dissolved in the solvent ranged from
1999, 187, 143–152.
(19) Cheng, J.-M.; Wang, D.-M.; Lin, F.-C.; Lai, J.-Y. J. Membr. Sci. 1996, 10-5 to 10-2 and 2 × 10-2 for acetone and THF solutions, respectively.
109, 93–107.
(20) Lai, J.-Y.; Lin, S.-F.; Lin, F.-C.; Wang, D.-M. J. Polym. Sci., Part B 1998, (22) Plantz, P. E. ACS Symp. Ser. 1998, 103–129.
36, 607–615. (23) Noda, K.; Ohashi, M.; Ishida, K. J. Chem. Eng. Data 1982, 27, 326–328.
(21) Schuhmacher, E.; Soldi, V.; Nunes Pires, A. T. J. Membr. Sci. 2001, 184, (24) Nayak, J. N.; Aralaguppi, M. I.; Naidu, B.; Vijaya, K.; Aminabhavi, T. M.
187–196. J. Chem. Eng. Data 2004, 49, 468–474.
1972 Langmuir, Vol. 25, No. 4, 2009 Aubry et al.

Figure 1. Volume fraction of acetone in the particles as a function of Figure 2. Determination of the Ouzo boundary using filtration to remove
the volume fraction of acetone in the aqueous phase. Circles represent microparticles from the dispersion of nanoparticles. Ratio of scattered
swelling experiments on PMMA particles (first method): initial polymer intensity of filtered dispersion/scattered intensity of raw dispersions
fraction f fp ) 1.1 × 10-4 (black filled circles), f fp ) 3.1 × 10-4 (black (RIS) as a function of the mass fraction of PMMA in the final solution
open circles), f fp ) 1.2 × 10-3 (gray filled circles), f fp ) 1.5 × 10-3 (gray for two PMMA/acetone/water systems (Brij 56/PMMA ) 0.1 (b) and
open circles). Black squares are results obtained by weighing centrifuged 14 (O)).
gels after swelling (second method).
It was weighted again to calculate the original volume fraction of
Then, the aqueous phase (water, aqueous solution of Brij56, aqueous acetone in gel. Three dispersions with similar initial mass fractions
solution of NaOH or buffer) was poured quickly into the organic of PMMA in acetone (f pi ) 10-2) and mass fractions of acetone (fs
solution (process 1) under moderate magnetic stirring (200 rpm). ) 0.4) were tested. They differed only by the stabilizer added to the
The amounts of organic solution and aqueous phase and the mass aqueous phase (one was stabilized with Brij 56 and the other two
fraction of PMMA in the organic solution were chosen to reach the by NaOH). Two gels were extracted from each sample from which
desired final mass fractions of PMMA and solvent in the ternary the values of φSP and φSW are reported in Figure 1 (black squares).
system PMMA/solvent/water. These compositions are noted as f pi Equilibrium Phase Separation (Binodal Line). The binodal
(initial mass fraction of PMMA in the solvent), f fp (final mass fraction line is the equilibrium solubility limit of the polymer in the
of PMMA), and fs (mass fraction of solvent in the final solution). solvent-water mixture. It was determined in two ways. The first
A typical recipe was as follows: an organic solution was prepared method consisted of titrating with water a solution of PMMA in the
by dissolving 0.1 g of PMMA in 100 g of acetone. An aqueous solvent until the mixture turned milky. The composition at the binodal
solution was prepared by dissolving 0.1 g of Brij56 in 400 g of line was deduced from the mass of water added at the onset of
HPLC water. Then 9 g of aqueous solution were quickly poured into turbidity. The second method consisted of adding solvent to a
6 g of organic solution under moderate stirring (200 rpm). Stirring dispersion of PMMA nanoparticles (dI ) 100 nm) in water that was
was stopped right after mixing, and the resulting dispersion size was previously made by nanoprecipitation and from which the solvent
determined immediately after filtration through light scattering, had been removed by evaporation. The intensity of scattered light
without removing the solvent. was then followed until this intensity dropped sharply to nearly
Solvent Partition Coefficient. All solvents do partition between zero, indicating that polymer chains were dissolved in the
water and particles during the nanoprecipitation process. Two solvent-nonsolvent mixture: this composition was taken as a point
methods were used to measure the partition coefficient of acetone of the binodal line.
between PMMA particles and water. The first one was based on the Ouzo Boundary. This boundary separates a region (at low initial
controlled swelling of a primary aqueous dispersion, previously PMMA concentration) where solvent shifting produces nanoparticles
prepared through nanoprecipitation in the Ouzo domain and solvent- only, from another region where it produces nanoparticles and
evaporated. Small volumes of acetone were added step by step to microparticles. This boundary was determined using filtration (to
this dispersion, while the mixture was stirred at 200 rpm for 1 min, separate out the microparticles) and light scattering (to detect the
and then the intensity-average particle size was measured through reduction in the number of nanoparticles). Dispersions were made
light scattering. We found that the particle diameter increased linearly according to process 1 with different PMMA initial mass fractions
with acetone content, provided that the refractive index and viscosity but keeping the same mass fraction of solvent and using an aqueous
of the dispersion medium were corrected through the Nanotrac phase containing a small amount of dissolved Brij 56 (mass fraction
software (Figure S2). Since the aqueous PMMA dispersions for 2 × 10-4). Half of each dispersion was filtered through a polyester
these experiments were quite dilute (f pf < 2 × 10-3), we assumed membrane (pore size 1.2 µm), and the raw and filtered dispersions
that acetone added to the dispersion was mainly present in the were analyzed using the Nanotrac instrument, which provided the
continuous phase rather than in the particles so that φSW was PSD and a “loading index” (LI). The LI value was used to calculate
approximated as the volume fraction of acetone added to the the total scattered light intensity IS, since both parameters are strictly
dispersion. The fraction of solvent in the particles φSP was simply proportional (see Supporting Information and Figure S1). The ratio
calculated from swollen and nonswollen particle diameters (φSP ) of the intensity of filtered dispersions to the intensity of raw
1 - d03/d3; see Supporting Information for details). The partition dispersions, noted RIS, was then plotted as a function of final mass
coefficient K ) φSP/φSW was then determined by plotting φSP as a fraction of PMMA. An example is shown in Figure 2 for two Brij
function of φSW and measuring the slope of the fitted line. Figure 1 56/PMMA ratios (r ) 0.1 and r ) 14) and similar mass fractions
shows this plot for four dispersions with different initial weight of acetone (fs ) 0.4). As expected, RIS is constant and close to unity
fractions of PMMA in water and different particle diameters. at low final mass fractions of PMMA, indicating that such
The second method consisted of preparing a dispersion outside compositions were within the Ouzo range and that all particles passed
the Ouzo range at a sufficiently high mass fraction of acetone in through the filters. For dispersions containing microparticles that
order to get a gel-like sediment after centrifugation. This gel was were removed by filtration, the final polymer mass fraction was less
isolated, weighted, and then dried to remove acetone completely. than in the raw dispersions, and RIS decreased. The critical
Phase Boundaries in Nanoprecipitation of PMMA Langmuir, Vol. 25, No. 4, 2009 1973

Figure 3. SEM photographs of nanoparticles of PMMA synthesized in the Ouzo domain (a) and beyond the Ouzo boundary (b). The dispersions
were made by adding aqueous solutions (water and Brij 56) to solutions of PMMA in acetone. The insets show the respective PSDs of these dispersions,
obtained through light scattering.

concentration at which RIS decreases corresponds to the Ouzo Table 1. Initial Mass Fraction of PMMA, Mean Particle
Diameter of the Initial Aqueous Dispersions of PMMA in
boundary. This boundary is a function of the final mass fraction of
Water, and Obtained Partition Coefficient Using the Swelling
solvent, but it is independent of the surfactant concentration: RIS Experiments (First Method)
started to decrease at f pf ) 8.10-5 and at f pf ) 10-4 for Brij 56/
PMMA ratios of 0.1 (black filled circles) and 14 (hollow circles), initial mass mean particle
respectively. The full Ouzo boundary was determined for solvent fraction of PMMA diameter of partition
mass fractions ranging from fs ) 0.05 to fs ) 0.5. in the aqueous the particles in coefficient
Mixing Processes. In order to test the effects of mixing order and dispersion aqueous solution (nm) (slope)
kinetics, four processes were compared. As reported before, process 1.1 × 10-4 150 1.19
1 consisted of adding in one shot the aqueous phase, of largest 3.1 × 10-4 95 0.49
volume, into the organic phase. This produced the fastest mixing of 1.2 × 10-3 130 0.62
the two phases. Process 2 consisted in dropwise addition of the 1.5 × 10-3 110 1.00
aqueous solution to the organic solution. This technique produced
at a larger PMMA mass fraction, beyond the Ouzo boundary.
a regular decrease in the solubility of PMMA, resulting in near
equilibrium phase separation. Process 3 consisted in dropwise addition In this case, the micrographs show both nanoparticles and much
of the organic solution to the aqueous solution. This has been the larger objects with smooth surfaces and sizes on the order of a
preferred way to synthesize nanoparticles by nanoprecipitation few micrometers (see PSD obtained by light scattering in the
according to the literature.10,13,25,26 Process 4 was similar to 1, but inset of Figure 3b; note that objects above 6.4 µm are not detected
part of the solvent (10% of the full amount) was introduced in the by our instrument).
aqueous solution instead of being used in the initial PMMA solution. As acetone is fully miscible with PMMA and with water,
particles made through solvent shifting should contain a fraction
Results of solvent in equilibrium with solvent in continuous phase. The
Most solvent shifting experiments presented here were partition coefficient (K) of acetone in the PMMA/acetone/water
performed by adding in one shot a large volume of aqueous system was determined by two methods, as indicated in the
phase into the organic phase (process 1, see the Experimental Experimental Section. As this partition coefficient is a thermo-
Section). This is a robust process that yielded metastable dynamic constant, its value should be influenced neither by the
dispersions (i.e., dispersions with colloidal stability) provided initial mass fraction of PMMA nor by the initial mean particle
that the aqueous phase was made with extremely pure water and diameter. This is verified since all K values determined through
that it contained a surfactant, or sodium hydroxide, or both. the first method are in the range of 0.49 to 1.19 (Table 1), a rather
Indeed, we found that the quality of the aqueous phase (pH and narrow range for a partition coefficient estimate. The mean value
ionic strength) had a strong effect on particle sizes. Using HPLC- of the partition coefficient with this method was estimated at K
grade water, we obtained populations of nanoparticles with mean ) 0.82 with a standard deviation of 0.32. The second method
diameters that were reproducible to better than 3%. We first was used to verify that K values did not depend on the nature
describe the sizes, morphologies, and compositions of the obtained of the stabilizer. With this method, the mean value of the partition
particles (nano and microsized ones) and then specifically explain coefficient was K ) 1.08 with a standard deviation of 0.09 (note
how the phase diagram was constructed. that the value of K is overestimated, because of the remaining
Particle Shapes and Compositions. The particles produced cosolvent weighed in the gel-like sediment before being dried).
in different regions of the composition map were imaged by It is obvious from these results that particles made by nanope-
SEM. Figure 3a presents the micrographs of dispersions made cipitation are systematically swollen by solvent (Vide infra). The
with PMMA dissolved in acetone at a low polymer mass fraction, total amount of solvent in particles depends on K and on the total
below the Ouzo boundary. In this case, the micrographs show mass fraction of solvent in the dispersion. Still, it is important
nanoparticles only, with regular shapes and a narrow distribution to note that the relative increase in particle diameters caused by
of sizes. Figure 3b presents the micrographs of dispersions made this swelling is less than 15%, therefore solvent extraction would
not affect the particle size significantly.
(25) Yu, W.; Tabosa do Egito, E. S.; Barratt, G.; Fessi, H.; Devissaguet, J. P.; Composition Boundaries. The maps of compositions for the
Puisieux, F. Int. J. Pharm. 1993, 89, 139–146.
(26) Bouchemal, K.; Briancon, S.; Perrier, E.; Fessi, H. Int. J. Pharm. 2004, ternary systems are given in the representation proposed by Vitale
280, 241–251. and Katz,9 where compositions are plotted according to the final
1974 Langmuir, Vol. 25, No. 4, 2009 Aubry et al.

Figure 4. Composition boundaries for nanoprecipitation in the PMMA/ Figure 5. Ouzo boundary in the alternative representation. Vertical scale:
acetone/water and PMMA/THF/water ternary systems. Filled symbols: initial mass fraction of PMMA in the solvent, log scale. Horizontal
points of the equilibrium phase separation boundary (binodal curve, scale: solvent/water ratio. Symbols: (9) PMMA/acetone/water and (b)
bold line) determined through titration and precipitation (black: from PMMA/THF/water. The data show an exponential decay of the initial
acetone; gray: from THF) and through dissolution (black triangle: mass fraction of PMMA at the Ouzo boundary with the solvent/water
acetone). Open symbols: points of the Ouzo boundary (dashed-dotted ratio. The dashed line is the limit obtained by Stainmesse et al. for PCL
line) determined by filtration followed by measurements of the intensity in acetone.13
of scattered light (black: acetone; gray: THF). Compositions for which
different mixing processes were compared are displayed with “×” for
PMMA/acetone/water and with “|” for PMMA/THF/water.

mass fraction of PMMA (horizontal axis) and the mass fraction

of solvent (vertical axis) (Figure 4). The mass fraction of PMMA
was so low (10-5 to 10-3) that a log scale was adopted for the
x-axis.
The equilibrium phase separation boundary of the ternary
solutions (the binodal line) was determined either through slow
water addition to a PMMA solution, or through dissolution of
a previously prepared aqueous dispersion of PMMA (see the
Experimental Section for procedures). In the representation using
a log scale for the polymer composition, this boundary is a nearly
horizontal straight line at a constant mass fraction of solvent
(Figure 4). The boundary is at a lower mass fraction of solvent
in the case of THF, because THF/water mixtures are better solvents
than acetone/water mixtures (see paper 3 of this series).
The Ouzo boundary was determined according to the result
of the precipitation process. Dispersions with a PMMA mass Figure 6. Composition map of the ternary system PMMA/acetone/
fraction below this boundary contained nanoparticles only. water according to the state of dispersions after centrifugation. Horizontal
scale: final PMMA mass fraction; vertical scale: final acetone mass
Beyond the Ouzo boundary, they contained both nanoparticles fraction. Black squares indicate dispersions that yield powder-like
and microparticles, which were stopped by the filter (pore size sediments, open circles are for dispersions that yield flaky sediments,
1.2 µm; see SEM images Figure 3). This made it possible to and open triangles are for dispersions that yield gel-like sediment. The
determine the boundary through a measurement of the scattered dashed line is the binodal line, the solid line is the centrifugation boundary,
and the dotted line is the Ouzo boundary as determined by light scattering.
light intensity (see the Experimental Section). For instance, the
measurements presented in Figure 2 indicate that microparticles
were produced and retained by the filter when the initial mass PMMA decreases exponentially with the solvent/water ratio
fraction of PMMA was 2.5 × 10-4 (corresponding to a final (Figure 5). According to this simple law, it is the initial
mass fraction of f fp ) 10-4 because the final acetone mass fraction concentration of PMMA in the organic solvent that is relevant
was fs ) 0.4). This limit was roughly the same, regardless of the for the Ouzo boundary, rather than the final concentration in the
Brij 56/PMMA ratio, within the range studied here. ternary mixture.
An odd feature of the Ouzo boundaries (Figure 4) is their belly We also determined the Ouzo boundary through centrifugation
like curvature, which turns back to lower final mass fractions of experiments (Figure 6). When centrifugation was applied to a
PMMA at low fractions of solvent. However, low fractions of dispersion of nanoparticles made in the Ouzo domain, a white
solvent are obtained by adding large amounts of water, so that sediment, fully redispersable by a vortex, was collected. On the
this feature may be an effect of the dilution by water. A different other hand, when microparticles were present besides nanopar-
representation was used by Stainmesse et al.,13 who plotted the ticles (dispersion made outside the Ouzo domain), the sediment
initial rather than the final concentration of polymer (in their could not be redispersed. We attributed this limit to the fact that,
case PCL) at the Ouzo boundary. When the data of Figure 4 are when microparticles are present in the samples, they coalesce
replotted in this way, it appears that the initial mass fraction of more easily under the pressure of the centrifugation than do
Phase Boundaries in Nanoprecipitation of PMMA Langmuir, Vol. 25, No. 4, 2009 1975

Figure 7. Mean particle diameter (a) and polydispersity index (b) as functions of the initial mass fraction of PMMA in acetone. Each set of data
is for a different final mass fraction of solvent (b 0.05, 1 0.17, 3 0.25, 9 0.3, 0 0.35, 2 0.4). Black dots represent compositions within the Ouzo
domain, whereas red dots correspond to compositions that are beyond the Ouzo domain.

nanoparticles (see Discussion section). Outside of the Ouzo Table 2. Intercept, Slope, and Best Fit Coefficient (r2) Obtained
boundary, we also observed that, at large solvent fractions (fs > from Fitting Particle Size vs Concentration Plots (in Log-Log
0.35), a gel was recovered, whereas, at low solvent fractions (fs Scale) Starting from Acetone (Runs 1-6, Figure 7a) and THF
(Runs 7-14, Figure 8)
< 0.35), a white solid-like film was observed (noted as flakes
in Figure 6). This difference surely originates from the glass run solvent fraction intercept slope r2
transition, Tg, of the solvent/PMMA mixtures, which crosses the 1 0.05 625 0.23 0.98
room temperature at this acetone fraction: if the room temperature 2 0.17 1350 0.31 0.97
is below Tg, particles aggregate but do not coalesce, whereas, 3 0.25 775 0.25 0.95
above Tg, coalescence produces a gel-like sediment at the bottom 4 0.3 1650 0.34 0.96
5 0.35 1915 0.36 0.99
of the tube. The minimum content of acetone to introduce in the
6 0.4 2225 0.35 0.94
formulation to soften the nanoparticles was estimated at 0.3 to 7 0.05 1000 0.30 0.91
0.4 in mass fraction, a value that is close to the one theoretically 8 0.10 975 0.28 0.91
calculated, i.e., 0.22 (see Supporting Information for estimated 9 0.15 1410 0.32 0.91
theoretical Tg’s of acetone/PMMA mixtures).27 In conclusion, 10 0.20 895 0.25 0.97
centrifugation and filtration techniques give Ouzo boundaries 11 0.25 970 0.23 0.99
that are quite close, meaning that the same phenomenon is 12 0.30 780 0.16 0.98
13 0.40 6530 0.48 0.97
observed, i.e., the generation of increasing numbers of micro- 14 0.50 11360 0.46 0.91
particles while crossing the Ouzo boundary.
Particle Sizes. The sizes of nanoparticles in the final dispersions whether the dispersions were prepared in the Ouzo and non-
were measured systematically using the light scattering instru- Ouzo ranges. The exponents are listed in Table 2, runs 1-6. All
ment. Series of samples were prepared through the same process, of them but the first one are close to 1/3, indicating that the
with the same addition of water and surfactant (Brij 56/PMMA volume per particle was proportional to the concentration of
ratio ) 2:1) to PMMA solutions that had increasing concentrations PMMA in the initial solution. There was also a slow (logarithmic)
of PMMA. For dispersions prepared in the Ouzo range of rise of the polydispersity index, indicating broader size distribu-
compositions, there was a single population of nanoparticles, tions at higher initial mass fractions of PMMA (Figure 7b).
with a narrow size distribution (see the inset of Figure 3a). In
this case, we systematically measured the mean diameter of this Figure 8a presents the variation of the mean nanoparticle
population as a function of PMMA concentration. For dispersions diameter for the PMMA/THF/water system. In this case, there
prepared in the non-Ouzo range, there were two populations were two distinct behaviors. At large additions of water (THF
corresponding respectively to nanoparticles and to microparti- mass fractions below 0.25), the behavior was the same as in the
cles (inset in Figure 3b). In that case, the measurements were acetone system: the variation of the mean particle diameter was
carried out on filtered samples in order to evaluate the mean still a power law of the PMMA mass fraction (exponent 1/3),
diameter of the nanoparticle population only. Figure 7a presents and there was no effect of solvent content, nor was there any
the variation of the mean nanoparticle diameter with the mass effect of the Ouzo boundary. However, at small additions of
fraction of PMMA in the initial solution for the PMMA/acetone/ water (THF mass fractions g 0.25), the diameters were much
water system. In each series, the mean diameter varied as a power larger, and they followed a very different behavior. At low PMMA
law of PMMA mass fraction (i.e., a straight line in the log-log concentrations (Ouzo range), the mean diameter was independent
representation of Figure 7a). This trend was true regardless of of the initial mass fraction of PMMA, but it increased with the
mass fraction of THF. At high PMMA concentrations (non-
(27) Besides calculation, we also tried to experimentally determine the Tg of Ouzo range), the diameters increased both with the initial mass
PMMA/solvent particles as a function of the mass fraction of solvent in particles
p
fraction of PMA and with THF content. The polydispersity index
(f s ) using thermogravimetric analysis (TGA) and differential scanning calorimetry
(DSC) devices. As solvent partitioning is a thermodynamic phenomenon, analyses
also became quite large, especially so at high THF content (Figure
were carried out on simple PMMA/acetone/water mixtures, rather than on Ouzo 8b).
particles, to prevent edge effects due to their small sizes. Using five PMMA/ Mixing Mode. All experiments presented so far were
acetone/water mixtures, we found values in a narrow range (due to the volatility
of the solvent), however in fair agreement with theoretical curves (see details in performed using process 1. As mentioned above, the results were
Supporting Information). reproducible and followed clear trends, provided that the quality
1976 Langmuir, Vol. 25, No. 4, 2009 Aubry et al.

Figure 8. Mean particle diameter (a) and polydispersity index (b) as functions of initial mass fraction of PMMA in THF. Each set of data is for
a different final mass fraction of solvent (9 0.05, b 0.1, 2 0.15, 10.2, 0 0.25, O 0.3, 4 0.4, 3 0.5). Black dots represent compositions within the
Ouzo domain, whereas red dots correspond to compositions that are beyond the Ouzo domain.

Table 3. Compositions Used for the Comparison of Mixing Processes: PMMA/Acetone/Water System (Runs 15-18) and PMMA/THF/
Water System (Runs 19-22)a
dI (dV/dN)
run solvent f × 10
i
p
3
fs f f
p × 10 3
process 1 process 2 process 3 process 4
15b acetone 0.02 0.5 0.01 74 (1.2) 116 (1.3) 129 (1.5) 81 (1.2)
16c acetone 1 0.2 0.2 84 (1.3) 207 (1.5) 124 (1.6) 78 (1.3)
17b acetone 10 0.2 2 201 (1.7) 648 (1.3) 215 (1.4) 208 (1.4)
18b acetone 10 0.5 5 237 (1.7) 916 (1.2) 214 (1.3) 238 (1.4)
19b THF 0.1 0.5 0.05 131(1.2) 183 (1.2) 142 (1.2) 185 (1.2)
20c THF 1 0.2 0.2 76 (1.3) 219 (1.2) 79 (1.3) 112 (1.3)
21b THF 10 0.2 2 186 (1.6) 633 (2.0) 188 (1.6) 211 (1.6)
22b THF 10 0.5 5 272 (1.8) 1090 (2.2) 239 (1.6) 300 (1.9)
a i
f p: initial mass fraction of polymer; fs: final fraction in solvent; f fp: final fraction in polymer; dI: mean particle diameter in intensity (in nm); dV/dN:
polydispersity index. b Concentration of hydroxide sodium in the aqueous phase: c ) 10-4 mol · L-1. c Concentration of hydroxide sodium in the aqueous
phase: c ) 10-5 mol · L-1.
of water was strictly controlled. However, since solvent shifting shifting. According to the experimental results presented above,
is a nonequilibrium process that starts with a nonhomogenous there are different ranges of composition and water addition that
situation (the initial mixing), there is the possibility that variations lead to the formation of different types of objects, e.g.
in the mixing protocol could lead to different final results. In nanoparticles (swollen with solvent) or microparticles (i.e.,
order to evaluate this possibility, the results of other mixing micrometric polymer lumps, also swollen with solvent). Ac-
procedures (processes 1, 2, 3, and 4 described in the Experimental cordingly, for each range of initial conditions, we examine what
Section) were compared. The comparison was performed quite decomposition mechanisms are possible.
systematically, using two polymer concentrations in the Ouzo Nanoparticle Formation at Very High Supersaturation.
region and two in the non-Ouzo region, two solvent/nonsolvent When large amounts of water are added to very dilute PMMA
ratios, and two different solvents (see Table 3 and Figure 4). In solutions, the result of solvent shifting is quite simple: the mixture
all cases, the same HPLC-grade water was used, and NaOH in decomposes to yield a single population of nanoparticles with
two different concentrations (Table 3) was added to control the a narrow distribution of diameters (Figure 3a). The original
colloidal metastability (see also the next paper of this series). hypothesis put forward by Vitale and Katz9 was that the
In both systems (acetone and THF), the processes that produced decomposition occurred through a nucleation and growth
fast mixing at the final composition (processes 1, 3, and 4) gave mechanism. In this mechanism, a few critical nuclei of pure
populations of particles with similar mean diameters and similar solute are formed when the solution is sufficiently supersaturated,
polydispersities. Process 2 was different, as it gave much larger and then these nuclei grow by capturing solute molecules from
particles, especially so at the highest PMMA concentrations. the surrounding solution. The signature of the nucleation and
This difference was expected, since process 2 (dropwise addition growth mechanism is that the number of particles remains equal
of water to the PMMA solution) causes a slow rise in the to the number of nuclei. At the end of growth, the mass per
supersaturation, which must lead to heterogeneous nucleation at particle equals the solute mass concentration divided by the
low supersaturation levels. In both systems (acetone and THF), number concentration of particles. As the number of nuclei varies
the most dilute PMMA concentrations were different, since the exponentially with the supersaturation, one should expect higher
sizes obtained through process 2 were closer to those obtained solute concentrations to yield a much higher number of nuclei
with the other processes. This was also expected, since precipita- and therefore smaller particles. Indeed, it is observed quite
tion must be much slower at these low PMMA concentrations, generally for nucleation and growth processes that higher solute
and therefore the initial mode of mixing may not matter. concentrations yield particles that are smaller and much more
numerous.28 This is contrary to the behavior observed here in
Discussion the conditions of very high supersaturation. Indeed, the mean
The aim of this discussion is to determine the mechanisms by
which the polymer-solvent mixture decomposes after solvent (28) Dirksen, J. A.; Ring, T. A. Chem. Eng. Sci. 1991, 46, 2389.
Phase Boundaries in Nanoprecipitation of PMMA Langmuir, Vol. 25, No. 4, 2009 1977

diameter increases regularly with the initial concentration of

PMMA (Figures 6 and 7). Moreover, the nucleation and growth
model predicts that the number of nuclei should vary rapidly
with solvent quality, and therefore with the concentration of
nonsolvent. Consequently, the mean particle diameter ought to
become smaller at higher water additions. This is also in conflict
with the behavior observed here at high supersaturation. Indeed,
in such conditions (water content > 0.4), the mean diameter is
independent of water content. For this high supersaturation regime,
an alternative is the nucleation-aggregation mechanism.28 In
this mechanism, the number of nuclei is so high that that they
have frequent encounters. The probability of encounters is
proportional to the square of the number of particles. It is then
assumed that each encounter causes aggregation of the two
particles, so that the number of aggregates per unit volume, n,
varies as
dn k Figure 9. Mean particle diameter as a function of the final mass fraction
) - n2 (1) of PMMA. Symbols are experimental data, and lines are theoretical fits
dt 2 according to a nucleation-aggregation particle formation mechanism.
where the rate constant k has the value calculated by Smolu- Each set of data is for a different final mass fraction of acetone (b 0.05,
1 0.17, 3 0.25, 9 0.3, 0 0.35, 2 0.4). The contour of the symbols and
chowski:29-31 the color of the lines correspond to acetone mass fraction (pink for fs
) 0.05, green for fs ) 0.3, and blue for fs ) 0.4). The color inside the
8kBT symbols is black if the compositions are within the Ouzo domain and
k) (2)
3η red if they are outside.
with kB ) 1.3806 × 10-23 J · K-1 being the the Boltzman constant,
PMMA/acetone/water system, R(fs) is well determined on the
T being the temperature in Kelvin, and η being the dynamic
range studied (0.05 < fs < 0.4) (Table 4).
viscosity of the dispersive medium. In a first approximation, the
Thus, if the recombination goes on for a set time t, the volume
rate constant k can be taken as being independent of particle
per particle will be proportional to the final polymer mass fraction
size.29 Equation 1 can then be integrated to yield the variation
f pf, and the particle radius will be proportional to the cube root
with time of the number of particles:
of this mass fraction. This is the case for our experiments (Figures
1 1 k k 7 and 8). Moreover, plotting diameter as a function of f pf or fpi
) + t≈ t (3) produces the same slope. This analysis assumes that all particles
n n0 2 2
recombine at the same rate and grow to the same mass, as in a
The number of aggregates can be written as a function of the hierarchical process where the PSD is perfectly monodisperse;
mass fraction of solvent (fs), the initial mass fraction of PMMA in the more general case, the diffusion limited cluster aggregation
in solvent (fpi), the densities of the dispersive medium (Fsol) and (DLCA) model30 shows that the smaller particles recombine
of the particles (Fp), and the mean particle diameter (d). This more frequently than the larger ones, so that the final size
yields eq 4 (see detailed calculations in Supporting Information). distribution is still monopopulated, as observed in the experiments
(see Figure 3).
6fsf piFsol The time t during which aggregation takes place will be constant
n) (4)
Fpπd3 if the aggregation process has a definite start time and a definite
stop time. The start time is the time at which the primary particles
Combining eqs 3 and 4 gives the variation of d3 as a function are formed. At very low PMMA concentrations, the mixing of
of aggregation time t: solvent and nonsolvent is faster than the encounter time of
macromolecules. With large additions of water, the net interaction
8kBTFsol fs f pi free energy of macromolecules in this solvent mixture is quite
d3 ) ×t (5)
π Fpη high, much higher than thermal energies. Hence aggregation
starts right after mixing, and the start time is the mixing time,
f pf, f pi, and fs are linked by the following equation:
provided that mixing is performed through a one-shot process.
f pf ≈ f pi fs (6) The stop time is the time at which the particles stop aggregating.
This will occur through the adsorption of species that cause the
3
Introducing eq 6 in eq 5 leads to the final expression of d : particles to repel each other. Earlier work by Lannibois et al.12
demonstrated that surfactants can be used for that purpose. In
8kBTFsol f
d3 ) f t ) 1.037 × 10-20R(fs)fpft (7) the present work, we have also used the addition of OH- ions
πFpη p to achieve colloidal stability. Thus the stop time is the time at
The rate coefficient R(fs) contains parameters that depend solely which adsorption of surfactant or potential determining ions
on fs: variations of Fsol and η with fs are deduced from literature occurs.
tables (represented in Figure S2),23,24 whereas Fp is deduced A quantitative comparison of predictions of the theoretical
from φSP, which is also a function of fs (see Figure S3). For the model provided by eq 7 with experimental results is presented
in Figure 9, which shows plots of particle diameter d versus f pf.
(29) Evans, D. F., Wennerström, H. The Colloidal Domain; Wiley-VCH: New For the sake of clarity, only three linear fits are represented, with
York, 1994. a slope of 1/3. The comparison of experimental data with the
(30) Broide, M. L.; Cohen, R. J. Phys. ReV. Lett. 1990, 64, 2026.
(31) Debenedetti, P. Metastable Liquids; Princeton University Press: Princeton, predictions makes it possible to determine for each mass fraction
NJ, 1996. of solvent the aggregation stop time t, as shown in Table 4. We
1978 Langmuir, Vol. 25, No. 4, 2009 Aubry et al.

Table 4. Values of the Rate Constant r(fs) and of the

Aggregation Stop Time t Determined from the Fits of
Experimental Data (Figure 9) to Eq 7
mass fraction of acetone fs R(fs) aggregation stop time t (sec)
0.05 868.0 2.75
0.17 724.7 2.93
0.25 674.0 1.34
0.3 659.0 1.76
0.35 657.1 1.58
0.4 670.5 2.49

find that the aggregation stop time is nearly independent of the

mass fraction of acetone, since its values fluctuate around an
average value of 2.1 s with a standard deviation of 0.6 s. The
discrepancy between theoretical predictions and experimental
data at low solvent fraction (fs ) 0.05) may originate from the
diluting effect of water, which supposedly affects both the rate Figure 10. Thermodynamic boundaries for the system PMMA/acetone/
and frequency of particle encounters. water. The bold black and green lines are calculated binodal and spinodal
The comparison with the predictions of the nucleation- lines, respectively, using the Flory-Stockmayer theory of polymer
solutions with the following interaction parameters: χPMMA/water ) 2.34,
aggregation model was also carried out for the PMMA/THF/ χPMMA/acetone ) 0.13, and χacetone/water ) 0.95. The black symbols are the
water system with fs < 0.2. In this case also, good correlations points of the equilibrium phase separation boundary, determined in the
were found when fitting the mean particle diameter dI by a power present work for (9) the PMMA used throughout the paper (Mn )
law of f pi with exponent 1/3. However the partition coefficient 14 700 g · mol-1, Mw/Mn ) 1.54); (2) a very monodisperse PMMA of
similar molar mass (Mn ) 14 920 g · mol-1, Mw/Mn ) 1.03); ([) a higher
of THF in water and PMMA was not determined, so the value molar mass PMMA (Mn ) 56 500 g · mol-1, Mw/Mn ) 1.78), and (b)
of φSP as a function of fs is not known. Consequently, it was not a large molar mass monodisperse PMMA (Mn ) 140 000 g · mol-1 and
possible to calculate the aggregation stop time. Mw/Mn )1.13), the corresponding points of which were determined by
Growth of Larger Particles at Low Supersaturation. When Lai et al.20 The thin lines are the dilution pathways followed by the
smaller amounts of water were added to a PMMA/THF solution, ternary solutions during solvent shifting experiments. The red squares
are the points of the Ouzo boundary determined for the PMMA described
the particles were much larger and more polydisperse in size in this study and best fitted by the bold red line. Solvent shifting processes
(Figure 8, THF mass fraction > 0.25). Moreover, the variations that end within this boundary (i.e., at lower polymer concentrations)
of their mean diameter were opposite to those described above. yield dispersions of nanoparticles. Solvent shifting processes that end
Indeed, the mean diameter became independent of the PMMA above this boundary (i.e., at higher polymer concentrations) yield mixtures
of nanoparticles and microparticles.
concentration fpi, but it increased with the THF content. In terms
of numbers of particles, this indicated that there were fewer
particles and that their number was determined by the super- centration fluctuations of very small scales. Thus, large-scale
saturation of the final solution. This behavior is no longer fluctuations in PMMA concentrations would grow spontaneously
compatible with the nucleation-aggregation model described and give rise to the microparticles: indeed, the volume of polymer
above. On the other hand, it matches the predictions of the solution that yields a microparticle is rather large (on the order
nucleation-and-growth model mentioned previously.31 Indeed, of a few micrometers according to SEM images; see Figure 3b).
when the THF content is high, the PMMA/THF/water solution However, the Ouzo boundary shown in Figure 4 has the wrong
has a high metastability, and it can evolve only through nucleation slope for a spinodal line: it goes to higher polymer concentrations
of PMMA particles that are larger than the critical nucleus size. when the solvent quality gets worse. Still, in order to conclude
At high THF contents, nucleation events are rare, and therefore on this point, it is necessary to determine the spinodal line of the
these particles are far apart; consequently, they do not encounter ternary solutions. We have calculated the spinodal line of
each other. Instead, they evolve by collecting isolated PMMA the PMMA/acetone/water system, and Karlström calculated the
macromolecules. Their growth must stop when they have collected binodal line, using the Flory-Stockmayer theory of polymer
all the available macromolecules (i.e., all the dissolved PMMA solutions.32 The interaction parameters for PMMA were taken
in excess of the solubility limit). Hence, the number of critical from literature values, χPMMA/water ) 2.3420 and χPMMA/acetone )
nuclei determines the final particle size. If the THF content is 0.13, and the best fitting of literature experimental data20 led to
higher, the supersaturation is less, nucleation is more difficult, χacetone/water ) 0.95 (detailed calculations will be presented in the
a smaller number of nuclei will be formed, and the resulting third paper of this series). Figure 10 presents the calculated binodal
particles will be larger, as observed (Figure 7). The final size was and spinodal lines, together with the dilution pathways that are
also found to be approximately independent of the initial PMMA followed by the ternary solution during mixing. The binodal line
concentration. This is because of a compensation of two effects: matches quite satisfactorily the experimental points determined
when there is more PMMA, the supersaturation is higher, which in the present work, whatever the molar mass and/or polydispersity
yields more nuclei, but there is also more PMMA to feed the of the PMMA chains (see Figure 10 caption for details), and also
growth of these nuclei. the points determined by Lai et al.,20 hence the choice of
FormationofMicroparticlesatHighPMMAConcentrations. interaction parameters is correct. The spinodal line is quite close
At high polymer concentrations, beyond the Ouzo boundary, to the binodal at high polymer concentrations, but not at the very
solvent shifting produced microparticles in addition to nano- low polymer concentrations explored in the present work. The
particles. Vitale and Katz proposed that this change to a larger dilution lines cross this large region that is between the binodal
length scale could result from a crossover to a regime in which line and the spinodal line, where the dilute solutions are
it decomposes through spinodal decomposition. Indeed, spinodal
(32) (a) Thuresson, K.; Karlström, G.; Lindman, B. J. Phys. Chem. 1995, 99,
decomposition involves small concentration fluctuations over 3823. (b) Piculell, L.; Lindman, B.; Karlström, G. Surf. Sci. Ser. 1998, 77, 65–
large scales, whereas classical nucleation involves large con- 141.
Phase Boundaries in Nanoprecipitation of PMMA Langmuir, Vol. 25, No. 4, 2009 1979

metastable, but they never enter the region below the spinodal to the polymer concentration beyond which such “larger” particles
line, where the solutions would become thermodynamically are produced and become susceptible to coalescence.
unstable. Hence, a spinodal decomposition process is ruled out. Conclusion
The transition from Ouzo to non-Ouzo behavior must therefore
have another origin. Nanoparticle formation through solvent shifting results from
a dilution of the solute molecules in a bad solvent (water), causing
Since the polymer concentration in the ternary solution remains
the nucleation of very small aggregates of solute molecules,
much lower than the spinodal decomposition boundary, the early followed by aggregation of these nuclei (nucleation-aggregation
steps of decomposition must still go through a nucleation step, mechanism). This aggregation process has a start time (the dilution
followed by aggregation of the nuclei if the supersaturation is time) and a stop time (the onset of colloidal stability, due to the
high enough. Therefore the system must still produce a population adsorption of ions or surfactant molecules). Consequently, the
of nanoparticles during the early stages of decomposition. The final size reached by the nanoparticles at the end of aggregation
final state still contains numerous nanoparticles, coexisting with varies as a power law of the solute concentration, as predicted
a few microparticles. This could be caused by a mechanism in by the Smoluchowski kinetic model. The population of nano-
which the larger particles would be able to coalesce, while the particles has a narrow size distribution, as in all aggregation
smaller ones would escape coalescence. Such behavior is not processes that follow the DLCA mechanism. This mechanism
that predicted by the Derjaguin-Landau-Verwey-Overbeek requires large additions of nonsolvent (water) to the initial polymer
(DLVO) theory for solid particles,33 although it has been observed solution. At smaller additions of water, the nucleation barrier is
in the case of emulsions.34 Indeed, when two droplets stabilized high, the nuclei are larger and less numerous, and they grow
by surfactant collide, they deform and for a while remain separated through the capture of nonaggregated solute molecules (nucleation-
by a flat surfactant/water/surfactant film. Then the probability of and-growth mechanism). Both mechanisms yield dispersions of
coalescence is the probability that a localized event such as a particles that are swollen with solvent, and have colloidal stability
loss of surfactant35 or a curvature inversion36 will take place through adsorption of surfactants or potential determining ions.
somewhere in the surfactant film that separates the droplets. This Both mechanisms are limited to very small solute concentrations,
probability is proportional to the area of the film, and therefore otherwise the solvent-swollen particles become large enough to
to the square of the droplet radius. It has been shown that, as a coalesce and yield a secondary population of much larger
consequence of this mechanism, emulsion droplets that are larger microparticles.
than a critical size coalesce very rapidly.34 In common emulsions,
this critical size depends on the nature of the surfactant film; in Acknowledgment. It is a pleasure for us to thank Gunnar
the present case, it would depend on the solvent content of the Karlström for the application of the Flory-Stockmayer theory
ternary solutions, since that determines the adsorption or surfactant to the calculation of the binodal line of the PMMA/acetone/
molecules or potential determining ions. Thus, the formation of water solutions. J.A. thanks the Languedoc Roussillon region
microparticles in a population of nanoparticles that are swollen and the CNRS for his Ph.D. grant, Dr. Abdellatif Manseri for
with solvent could result from coalescence events that take place DSC measurements, and Kevin Roger for providing some of the
when the largest particles of the population encounter each other, data for Figure 10.
but do not take place during collisions of the smaller particles Supporting Information Available: Detailed information on
of the population. The Ouzo boundary would then correspond the Ouzo boundary as determined by the Nanotrac apparatus, particle
swelling, and derived relevant parameters; the experimental and
(33) Vervey, E. J. W.; Overbeek, J. T. G. Theory of the Stability of Lyophilic
theoretical evolution of acetone-swollen PMMA glass transition as a
Colloids; Elsevier: New York, 1948. function of the mass fraction of acetone in PMMA; and the nucleation
(34) Sonneville, O. Biliquid foams Ph.D. Thesis, University Paris VI, 1997. and aggregation detailed theory. This material is available free of charge
(35) Exerowa, D.; Kashchiev, D.; Platikanov, D. AdV. Colloid Interface Sci. via the Internet at http://pubs.acs.org.
1992, 40, 201.
(36) Kabalnov, A.; Wennerstrom, H. Langmuir 1996, 12, 276. LA803000E