Colloidal Phase Separation of Concentrated PNIPAm Solutions

Caroline Balu, Michel Delsanti, and Patrick Guenoun*

LIONS, SCM-C.E.A. Saclay, F-91191 Gif sur YVette Cedex, France

Fabrice Monti and Michel Cloitre

Laboratoire Matiere Molle et Chimie (UMR 7167, ESPCI/CNRS), ESPCI, 10 rue Vauquelin,75321 Paris Cedex, France

ReceiVed September 22, 2006. In Final Form: NoVember 16, 2006

In the concentration range of 1-6 wt %, solutions of a thermosensitive polymer (poly-N-isopropylacrylamide(PNIPAm),Mw ) 1.4× 105 g‚mol-1) are shown to phase separate in the form of dense stable colloids of nearly purepolymer. Diffuse wave spectroscopy and small-angle neutron scattering both provide consistent measurements of thecolloidal size as a function of temperature. Results are in agreement with a Cahn regime of spinodal decompositionblocked at an early stage, prior to a growth that would lead to a macroscopic phase separation. [Early results of thiswork were presented at the 231st American Chemical Society National Meeting, Symposium on Amphiphilic Polymers,Atlanta, GA, 2006, March 26-30.]

Introduction

Phase separation of simple fluids, induced by an abrupt quenchin temperature, leads to a macroscopic segregation of phases.The laws describing the growth of the characteristic size of thedomains versus the time elapsed since the beginning of the quenchare well documented.1,2However, few fundamental studies havebeen carried out on polymer solutions to explain why the domainsize depends on polymer concentration and quench temperature.For such solutions, the large difference in viscoelastic propertiesbetween the two separating phases has led to original predictions3

that have been partly tested by experiment4 on polystyrene inorganic solvents. These predictions are significantly differentfrom growth laws observed for equal molecular weight mixtures.1

However, some experiments have been performed5 where noviscoelastic induced modification was reported.

The study of phase separation for polymers in aqueous solventshas been extensively triggered6-11by the increasing use of water-soluble thermosensitive polymers. On the applied side, thesegregation of phases may also lead to the synthesis of membranes,scaffolds,12 or microcapsules.13 It is then tempting to test ourunderstanding of phase separation mechanisms in situations wherehydrogen bonds or Coulomb interactions exist, going beyondthe simple case of van der Waals interactions present in organicmedia.

In this paper, we report on aqueous solutions of poly-N-isopropylacrylamide (PNIPAm), whose thermosensitive proper-ties are widely used in applications14 but for which nocomprehensive knowledge of phase separation mechanisms existsso far. Previous attempts have mostly concerned extremely dilutesolutions, before phase separation, to detect coil-globuletransitions on a single chain15 or the phase separation of dilutesolutions (<0.05 wt %) where stable colloids of PNIPAm havebeen detected.8,9 In this regime, PNIPAm colloids are stabilizedby charges born by the chain end-groups9 brought by the chargedinitiator of the polymerization reaction in water, but othermechanisms of stabilization have been suggested.11For instance,it has been proposed that hydrophilic groups may remain hydratedabove the phase separation temperature and sterically stabilizethe colloids,16but all predicted colloidal structures are anisotropic.Also, the ineffectiveness of colloid collision has been put forwardas a mechanism preventing colloid coalescence.17 This could beenhanced by the glassy nature of PNIPAm-rich spheres, providedthey are concentrated enough.18 However, none of the abovemechanisms have been able to account for the variation of thecolloidal size with temperature or concentration. Moreover,how phase separation proceeds for more concentrated solu-tions is still unknown, although a knowledge of the growthmechanisms in a wide range of concentrations is required to tunemore finely the structures obtained through phase separationprocesses.

This lack of knowledge about phase separation mechanismspartly originates from the extreme difficulty in reaching twomacroscopic separated phases for PNIPAm solutions: a samplequenched in the diphasic state remains turbid and milky. In orderto quantitatively describe the phase separation in a wider rangeof concentrations (c> 1 wt %) and temperatures, we demonstratehere how the difficulties associated with such strongly turbid

* Corresponding author.(1) Nikolayev, V. S.; Guenoun, P.; Beysens, D.Phys. ReV. Lett. 1996, 76,

3144.(2) Solis, F.; Olvera de la Cruz, M.Phys. ReV. Lett. 2000, 84, 3350.(3) Onuki, H.; Taniguchi, J.J. Chem. Phys.1996, 106, 5761.(4) Tanaka, H.Phys. ReV. Lett. 1993, 71, 3158.(5) Xie, Y. L.; Ludwig, K. F.; Bansil, R.; Gallagher, P. D.; Konak, C.; Morales,

G. Macromolecules1996, 29, 6150.(6) Inomata, H.; Yagi, Y.; Otake, K.; Konno, M.; Saito, S.Macromolecules

1989, 22, 3494.(7) Tanaka, H.; Nishi, T.Jpn. J. Appl. Phys., Part 2: Lett.1988, 27, L1787.(8) Gorelov, A. V.; DuChesne, A.; Dawson, K. A.Physica A1997, 240, 443.(9) Chan, K.; Pelton, R.; Zhang, J.Langmuir1999, 15, 4018.(10) Laukkanen, A.; Valtola, L.; Winnik, F. M.; Tenhu, H.Macromolecules

2004, 37, 2268.(11) Aseyev, V.; Hietala, S.; Laukkanen, A.; Nuopponen, M.; Confortini, O.;

Du Prez, F. E.; Tenhu, H.Polymer2005, 46, 7118.(12) Guan, J. J.; Fujimoto, K. L.; Sacks, M. S.; Wagner, W. R.Biomaterials

2005, 26, 3961.(13) Yeo, Y.; Basaran, O. A.; Park, K.J. Controlled Release2003, 93, 161.

(14) Duracher, D.; Elaissari, A.; Mallet, F.; Pichot, C.Langmuir2000, 16,9002.

(15) Zhang, G. Z.; Wu, C.Phys. ReV. Lett. 2001, 86, 822.(16) Vasilevskaya, V. V.; Khalatur, P. G.; Khokhlov, A. R.Macromolecules

2003, 36, 10103-10111.(17) Wu, C.; Li, W.; Zhu, X. X.Macromolecules2004, 37, 4989.(18) Van Durme, K.; Van Assche, G.; Van Mele, B.Macromolecules2004,

37, 9596

2404 Langmuir2007,23, 2404-2407

10.1021/la0627821 CCC: $37.00 © 2007 American Chemical SocietyPublished on Web 02/02/2007

phase-separated systems can be overcome by combining diffusivewave spectroscopy (DWS) and small-angle neutron scattering(SANS).

Experimental Section

Materials and Methods.The neutral PNIPAm used, synthesizedby radical polymerization in organic solvent (dioxane) with a neutralinitiator (AIBN), was purchased from Polysciences, Inc. andcharacterized by aqueous gel permeation chromatography. Theweight-average molecular weight wasMw ) 1.4× 105 g‚mol-1 witha polydispersity index ofMw/Mn ) 3.0. For DWS and SANSexperiments, two solutions of PNIPAm of volume fractionsφ )0.98 and 6.14% were prepared with heavy water D2O at neutral pH.The temperature of the cloud point,TCP, which corresponds to thebeginning of the demixing process, was determined by measuringthe drop of transmission of the sample (λ ) 632.8 nm) upon slowheating (0.03°C/min.) and was found to be 30.6 and 29.3°C for0.98 and 6.14 vol %, respectively. TheTCPvalues measured in D2Oare higher by 0.6°C than those in H2O at equivalent volume fractions.These observations are in agreement with light scattering19 anddifferential scanning calorimetry20 experiments, which have shownthat D2O is a better solvent than H2O for PNIPAm. The decreasein TCPwith volume fraction in this concentration range is in agreementwith the results of Afroze et al.21 (solutions in H2O). For a givenpolymer concentration, rapid thermal quenches were performed fromambient temperatureTi to the quench temperatureTf above the cloudpoint. The sample cells of rectangular shape (Hellma, Germany)with an inner thickness of 1 or 2 mm were transferred from roomtemperatureTi into the thermostat at temperatureTf. The sampletemperature reachesTf exponentially with a characteristic time onthe order of 10-20 s. In this way, we study how the solution demixesat a definite temperature. When the characteristic time is larger thanthe intrinsic phase separation time scale, the final size should dependon the thermal pathway followed to reachTf.

Diffusive Wave Spectroscopy.DWS experiments in transmissiongeometry were performed on an experimental set up described inref 22. The laser beam (λ ) 514.5 nm) was expanded to 6 mm atthe position of the sample of thicknessL ) 2 mm. The optical cellwas immersed in a large water thermostat bath. The autocorrelationfunctions of the scattered intensity were calculated using a BI9000ATcorrelator. The experimental electric field correlation functionsg(t)were fitted to the following expression:23

with γ ) 6k2D, wherek is the wave vector of the incident light, andD is a diffusion coefficient. The transport mean free pathl* in eq1 is the length over which the photon direction becomes uncorrelatedwith the incident direction. For any sample,l* was measured fromthe ratio of the scattered light between the solution and a referencelatex solution whosels* is known.24We made use of a latex solutionwhose measured diameter is 190( 10 nm, andls* was calculatedassuming a Mie scattering with an optical index of 1.595( 0.005.25

The applicability of formula 1 is restricted to isotropic multiplescattering and to the conditionL . l* (here,L > 10 × l*).

Small-Angle Neutron Scattering. SANS experiments wereperformed at the Laboratoire Le´on Brillouin at Saclay, on the PACEspectrometer.26 A wavelength of 1.2 nm and a sample-to-detectordistance of 4.57 m were used. The samples were contained in 1 or2 mm path quartz cells placed in a small furnace. The scatteredintensities were corrected for the intensity scattered by the emptycell and normalized by the incoherent scattering of a NIPAm monomersolution of 6.14 vol % in order to convert the data into differentialcross-sections per unit volume (I(cm-1)) following the standardprocedure.27

Electrical Mobility. Electrical mobility measurements(Delsa440SX, Coulter) were performed at a quench temperatureof 40°C in the two-phase region. Special care was taken to eliminatethe influence of thermal gradients (the main cause of errors) byrepeating the measurements at different electrical field values.

Results and Discussion

The evolution and the structure of the phase separation werefollowed by DWS and SANS experiments, and both measure-ments provided a stable signal after typically 10 min. We focushere on the interpretation of this steady state. To the naked eye,this steady state resembles a stable colloidal white turbid solution.Observation with confocal microscopy (LSM510, Zeiss) revealsthe existence of isolated PNIPAm domains adsorbed to the opticalwindow closest to the objective (Figure 1). Fits of correlationfunctions to formula 1 provide a series of diffusion coefficientsas a function of temperature for a concentration of 6.14 vol %(0.98 vol % is not concentrated enough to be an isotropic multiplescatterer). From theD values, we deduce the radiiRDWS usingthe Stokes-Einstein relation, assuming spherical shapes assuggested by confocal microscopy.RDWSandl* values, deducedfrom scattered intensities, decrease with temperature (Figure 2).The radii values are compatible with the sizes that can be deducedfrom the confocal pictures.

SANS measurements clearly obey a Porod behavior,27as shownin Figure 3. Neutron transmission measurements confirm that notemporal evolution occurs a few minutes after the quench. ThePorod regime supports the existence of sharp interfaces betweenthe two separated phases, namely, the colloids and the continuousmedium. Assuming a population of spheres, as suggested byconfocal microscopy and DWS analysis, enables one to extracta radius value from the Porod scattering by assuming that sharpinterfaces separate rich polymer spheres to a continuous phase

(19) Wu, C.; Wang, X.Macromolecules1999, 32, 4299.(20) Kujawa, P.; Winnik, F. M.Macromolecules2001, 34, 4130.(21) Afroze, F.; Nies, E.; Berghmans, H.J. Mol. Struct.2000, 554, 55.(22) Cloitre, M.; Borrega, R.; Monti, F.; Leibler, L.Phys. ReV. Lett.2003, 90,

068303.(23) Pine, D. J.; Weitz, D. A.; Zhu, J. X.; Herbolzheimer, E.J. Phys. (Paris)

1990, 51, 2101.(24) Rojas-Ochoa, L. F.; Romer, S.; Scheffold, F.; Schurtenberger, P.Phys.

ReV. E 2002, 65, 051403.(25) Brandrup, J.; Immergut, E. H.Polymer Handbook; John Wiley: New

York, 1989.

(26) Website of the Laboratoire Le´on Brillouin (C.E.A. Saclay, France).http://www-llb.cea.fr (accessed Jan 2007).

(27) Higgins, J. S.; Benoit, H. C.Polymer and Neutron Scattering; ClarendonPress: Oxford, 1994.

g(t) )[ 3L5l*

+ 45] × [sinh(xγt) + 2

3xγt cosh(xγt)]

(1 + 49

γt) sinh(Ll*

xγt) + 43

xγt cosh(Ll*

xγt)(1)

Figure 1. View by confocal microscopy of adsorbed PNIPAm beadsin D2O (black domains) on the optical window (T ) 34°C,φ ) 0.98vol %). The bar is 2µm.

Colloidal Phase Separation of PNIPAm Solutions Langmuir, Vol. 23, No. 5, 20072405

poor in polymer. We first assume two phases of pure PNIPAmand pure D2O with density contrast lengthsb(i). One then gets

whereS is the spheres area in the volumeV. The backgroundvalueB takes into account both the incoherent background andthe scattered intensity by the collapsed chains in the dilutecontinuous phase. TheB values (on the order of 0.3 cm-1 for6.14 vol %) show that the polymer volume fraction in thecontinuous phase is negligible (on the order of 0.1 vol %).

For spheres of radiusRN one getsS/V ) 3φ/RN. As seen inFigure 4, radii values deduced from the SANS analysis coincidewith values measured by DWS. The slope in Figure 4 is foundto be 1.0( 0.1. If the spheres are not made of pure PNIPAmbut contain water, the contrast is reduced by a factor ofx2, wherex is the volume fraction of PNIPAm inside the spheres. So far,we have neglected the decrease in the self-diffusion coefficientD measured by DWS with the bead volume fractionφ.28 Thisdecrease overestimates the DWS radius by about 10% for a

volume fraction of 6.14 vol %. These corrections lead to adetermination of the PNIPAm volume fractionxbetween 90 and100 vol %. It is interesting to note that this is a much higher valuethan the one measured between 30 and 40% (refs 11 and 29) fora more dilute concentration range (c e 0.03 wt %).

Two main features are now to be discussed: Why does theseparation of phases stop at some mesoscopic stage, and whatdetermines the measured radius value as a function of quenchdepth? The first aspect means that coalescences are not effectiveto reduce the interfacial area of the polymer phase, which behavesas an assembly of stable colloids. Colloids can be stericallystabilized versus van der Waals attraction by a swollen layer ofsolvated segments,11but this picture seems to contradict the Porodregime observed by SANS and with the poor solvent state of thepolymer. Another hypothesis is the presence of charges at thebead surface. These charges cannot structurally belong to thepolymer, which is a neutral one, synthesized with a non-ionicinitiator in an organic solvent. However, ions were detected fromcapillary electrophoretic measurements (Waters capillary ionanalyzer) in PNIPAm solutions in H2O at a level on the orderof 10-5 M, although standard dialysis procedures were used topurify the polymer. This is still an order of magnitude above theimpurity level that we measured in ultrapure deionized H2O andD2O by the same technique. From studies on specific ion effectson the water solubility of PNIPAm,30 it has been shown thatanions bind strongly to the amide dipole. In our case, a surfaceinteraction involving a small ion quantity and able to adsorb andcharge the PNIPAm spheres is fully possible. We then performedelectric mobility measurements on a solution in deionized H2O(0.1 wt %, neutral pH), which provided negative values on theorder of -5.9 ( 0.2 × 10-8 m2 V-1 s-1 and enabled us toestimate that, at 40°C, PNIPAm beads bear typically one chargeper 270 nm2. A DLVO calculation, which balances van der Waalsattraction and electrostatic repulsion, shows that such a chargedensity is able to stabilize beads with a radius of 200 nm up toan ionic strength of 10-3 M. For this calculation, we used aHamaker constant on the order of 1-2 kBT, relevant for apolymer-water-polymer interface.31 To check this hypothesis,LiCl (a salt known to induce no shift in the phase diagram of

(28) Qiu, X.; Wu, X. L.; Xue, J. Z.; Pine, D. J.; Weitz, D. A.; Chaikin, P. M.Phys. ReV. Lett. 1990, 65, 516.

(29) Kujawa, P.; Aseyev, V.; Tenhu, H.; Winnik, F. M.Macromolecules2006,39, 7686.

(30) Zhang, Y. J.; Furyk, S.; Bergbreiter, D. E.; Cremer, P. S.J. Am. Chem.Soc.2005, 127, 14505.

(31) Hunter, R. J.Foundations of Colloid Science; Clarendon Press: Oxford,1987.

Figure 2. Evolution of the bead sizeRDWS (circles) measured byDWS and of the measuredl* (squares) with quench temperature (φ) 6.14 vol %).

Figure 3. Neutron scattered intensities, whose backgroundB hasbeen subtracted, by a solution of PNIPAm at 6.14 vol %:T ) 45°C (squares) andT) 34°C (triangles). Aq-4 dependence is observed(Porod scattering).

I(cm-1) )2π(b(PNIPAm)- b(D2O))2S

V × q4+ B (2)

Figure 4. Radii measured by DWS as a function of the radii measuredby SANS at 6.14 vol % for different temperatures. The line in theFigure is of slope 1.

2406 Langmuir, Vol. 23, No. 5, 2007 Balu et al.

PNIPAm up to 0.1 M32) was added to a solution in H2O priorto phase separation. This addition induces the destabilization ofthe colloidal dispersion for a salt concentration greater than 10-2

M, and phase separation proceeds to a macroscopic scale. Thistends to confirm that charges are essential to explain the colloidalPNIPAm phase separation.

We now discuss the dependence of radii on temperature. Theradii of the beads decrease upon increasing the quench temperature(Figure 5), as found in the dilute range by Kujawa et al.29 whena true thermal quench is followed. To explain this variation, wepropose that, after an initial stage of molecular diffusion ofpolymer molecules, electrostatic repulsions prevent any collisionof the polymer beads. For a phase separation following a spinodaldecomposition mechanism, the initial diffusion stage is knownas the Cahn regime.33 In this linear regime, concentrationinstabilities grow exponentially with time, and a time-independentfastest growing wavelengthΛ emerges. This regime occurs beforenonlinear mechanisms take over and makesΛ grow with time34

by coalescence or hydrodynamics. The wavelengthΛ sets thespatial scale of separation, and the domain sizeR is naturallyassumed to be proportional toΛ. The growth of polymer domainsis stopped before the end of Cahn’s regime by surface charges(adsorbed ions), which prevent the nonlinear growth of thedomains. The time for an ion to diffuse overR is smaller by 2orders of magnitude than the time needed for two polymer beadsto diffuse and collide. Following van Aartsen35 and Binder,36

one knows that, assuming a Flory-type free energy with aninteraction parameterø independent of concentration,Λ varieswith temperature according to

where, for the concentration under study,TandTSare the quenchtemperature and the spinodal temperature, respectively. FollowingBinder, the prefactorΛ0(Φ) is, up to a constant on the order of

1, equal to (a/xφ), wherea is a monomer length. For describingthe PNIPAm phase diagram, a concentration dependentø isneeded:21 ø ) A(φ) + B(φ)T, but this preserves the temperaturedependence in the form

whereB′ andB′′ are the first and second derivatives ofB versusΦ, respectively.

In Figure 5, temperature variations of the measured radii aresuccessfully fitted to a lawR) R0 x(TS/T-TS). The fit providesvalues of the spinodal temperatures,TS(0.98 vol %)) 31.5(0.2°C andTS(6.14 vol %)) 32.2( 0.1°C, which are consistentwith the cloud point temperature measurements. This is inagreement with the kinetic study by Inomata et al. where theCahn regime was already suspected.6 The prefactorR0 is on theorder of 30 nm and is larger than the theoretical expressions (forinstance, witha ) 1 nm atφ ) 6.14 vol %, one hasΛ0 ) 4 nm).It is worth mentioning that such high values of the prefactor havealready been reported in the few experimental studies of theCahn regime of polymer solutions,35,37,38 all for hydrophobicpolymers. The key factors underlying the prefactor estimationare the range of interactions and the description of theasymmetrical dynamics between the polymer and the solvent.The range of interactions could be enlarged in our case becauseof the importance of hydrogen bonds. The dynamics probablyneeds a detailed description as in Onuki and Taniguchi,3 beyondBinder’s original approach. These two reasons may add up toexplain the observed discrepancy. A last remark is that thespherical shape of the domains supports an absence of domaincoalescence, which should lead to nonspherical shapes since theconcentrated PNIPAm phase at 90 vol % is likely to be glassy.18

Conclusion

The present study shows that rather concentrated PNIPAmsolutions, close to the overlap concentration, undergo a frozenphase separation. The polymer rich phase is nearly made of purepolymer. A probable mechanism for the grain stabilization is anadsorption of residual ions at the grain’s surface. This isreminiscent of the blocked poisoned coalescence of hydrophobicmolecules due to the adsorption of charges or surfactantimpurities.39,40This raises the question of whether a true neutralsurface can exist in water. This makes the polymer-rich phasea colloidal phase whose size can be controlled by quench depth.The size variation with temperature is compatible with a Cahnregime of spinodal decomposition.

Acknowledgment. The authors thank the C.N.E.S. forfinancial support, A. Bourdette and V. Burckbuchler for help inthe mobility measurements, and L. Auvray, L. Belloni, A.Halperin, D. Hourdet, I. Iliopoulos, and F. Winnik for fruitfuldiscussions.

LA0627821

(32) Freitag, R.; Garret-Flaudy, F.Langmuir2002, 18, 3434.(33) Cahn, J. W.Acta Metall.1961, 9, 795.(34) Langer, J. S.; Bar-on, M.; Miller, H. D.Phys. ReV. A 1975, 11, 1417.(35) van Aarsten, J.Eur. Polym. J.1970, 6, 919.(36) Binder, K.J. Chem. Phys.1983, 79, 6387.

(37) Kuwahara, N.; Kubota, K.Phys. ReV. A 1992, 45, 7385.(38) Kojima, J.; Takenaka, M.; Nakayama, Y.; Hashimoto, T.Macromolecules

1999, 32, 1809.(39) Marinova, K. G.; Alargova, R. G.; Denkov, N. D.; Velev, O. D.; Petsev,

D. N.; Ivanov, I. B.; Borwankar, R. P.Langmuir1996, 12, 2045.(40) Lannibois, H.; Hasmy, A.; Botet, R.; Chariol, O. A.; Cabane, B.J. Phys.

II (France) 1997, 7, 319.

Figure 5. Variation of radii measured by SANS with temperature(circles, 6.14 vol %; squares, 0.98 vol %). Lines are fits to Cahn’sspinodal law explained in the text.

Λ ) Λ0(Φ)xTS

xT - TS

(3)

Λ ) Λ0(Φ)2

xT - TSx2B - 2B′(1 - 2Φ) - B′′Φ(1 - Φ)(4)

Colloidal Phase Separation of PNIPAm Solutions Langmuir, Vol. 23, No. 5, 20072407