Self-Assembling of Perfluorinated Polymeric Surfactants in Water. Electron-Spin Resonance Spectra of Nitroxide Spin Probes in Nafion Solutions and Swollen Membranes

2188 Langmuir 1994,10, 2188-2196

Self-Assembling of Perfluorinated Polymeric Surfactants in Nonaqueous Solvents. Electron Spin Resonance Spectra of Nitroxide Spin Probes in Nafion Solutions and Swollen

Membranes

Ewa Szajdzinska-Pietekt and Shulamith Schlick*

Department of Chemistry, University of Detroit Mercy, Detroit, Michigan 48219-0900

Andrzej Plonka

Znstitute of Applied Radiation Chemistry, Technical University of Lodz, 93-590 Lodz, Poland

Received January 10, 1994. Zn Final Form: April 18, 1994@

Structural details on the self-assembling of perfluorinated ionomer (Nafion) chains in solutions and in swollen membranes have been obtained from ESR studies in systems containing doxylstearic acid spin probes. Results previously obtained for aqueous systems are extended in this study to formamide (FA), ethanol (EtOH), andN-methylformamide (NMF) as solvents. The slow-motional ESR component detected in swollen membranes and solutions in FA has been assigned to spin probes bound to polymer aggregates. The additional, motionally averaged, component detected in FA solutions was assigned (at least in part) to spin probes associated with single chains. These assignments are similar to those in the aqueous systems. Comparison of the corresponding order parameters suggests that self-assembling of polymer amphiphiles occurs at higher polymer concentration and leads to less ordered aggregates in FA, compared to aqueous systems. The present ESR data do not indicate aggregation in NafioniEtOH and N a f i o m F solutions. The ESR spectra of the spin probes in membranes swollen by EtOH are consistent with a plasticizing effect of the solvent, rather than with a phase separated morphology. These conclusions are in agreement with two types of studies of Nafion membranes and solutions: by ESR in systems containing paramagnetic V02+ as the counterion and by 19F NMR. Based on the results for aqueous and FA systems, we propose a mechanism for the transition between the micellar structure in solution and the reverse micellar structure in the swollen membranes, which we call the fringed rod model. The model assumes that at high polymer concentrations some chains can be incorporated in more than one rod, thus effectively providing the cross-linking necessary for complete connectivity of the polymeric material.

I. Introduction

Nafion, a perfluorinated ionomer made by DuPont, has the formula given in Chart 1 and is widely used as an ion selective and separation membrane.'V2 Numerous studies have been directed toward elucidation of the structure and dynamics of Nafion membranes in the acid form or neutralized by various counterions, and swollen by ~o lven t s .~ - l~ Spectroscopic and scattering results for the membranes swollen by water suggest a phase separated

* Author to whom correspondence should be addressed. t On leave from the Institute of Applied Radiation Chemistry,

@ Abstract published in Advance ACS Abstracts, June 1, 1994. (1) Perfluorinated Ionomer Membranes; Eisenberg, A., Yeager, H.

(2) Structure and Properties of Ionomers; Pineri, M., Eisenberg, A.,

(3) Barklie, R. C.; Girard, 0.; Braddel, 0. J . Phys. Chem. 1988,92,

(4) Martini, G.; Ottaviani, M. F.; Pedocchi, L.; Ristori, S. Macro-

( 5 ) Martini, G.; Ottaviani, M. F.; Ristori, S.; Visca, M. J . Colloid

(6) Alonso-Amigo, M . G.; Schlick, S. Macromolecules 1989,22,2628. (7) Schlick, S.; Alonso-Amigo, M. G. Macromolecules 1989,22,2634. (8) Bednarek, J.; Schlick, S. J . Am. Chem. SOC. 1990,112, 5019; J .

(9) Schlick, S.; Gebel, G.; Pineri, M.; Volino, F. Macromolecules 1991,

(10) Lossia, S. A.; Flore, S. G.; Nimmala, S.; Li, H.; Schlick, S. J.

(11) Schlick, S.; Alonso-Amigo, M. G.; Bednarek, J. Colloid Surface

(12) Avalos, J.; Gebel, G.; Pineri, M.; Schlick, S.; Volino, F. Polym.

(13) Li, H.; Schlick, S. Polym. Prepr. (Am. Chem. SOC. Diu. Polym.

Technical University of Lodz, Lodz, Poland.

D., Eds.; American Chemical Society: Washington, DC, 1982.

Eds.; NATO AS1 Series, Reidel: Dordrecht, 1987.

1371.

molecules 1989,22, 1743.

Interface Sci. 1989, 128, 76.

Am. Chem. SOC. 1991,113, 3303.

24, 3517.

Phys. Chem. 1992,96, 6071.

A: Physicochem. Eng. Aspects 1993, 72, 1.

Prepr. (Am. Chem. SOC. Diu. Polym. Chem.) 1993, 34, 448.

Chem.) 1993,34, 446.

0743-7463/94/2410-2188$04.50/0

Chart 1. Nafion Ionomer, Acid Form

morphology, which consists of water pools surrounded by the hydrophobic fluorocarbon chains; the ionic head groups and the counterions are located a t the interface.' Simi- larities and differences between this structure and the reverse micellar solutions formed from the ternary mixture surfactant-water-oil have been proposed.1° In contrast to scattering studies, however, multifrequency electron spin resonance (MESR) and 19F NMR results do not support the concept of phase separation in membranes swollen by nonaqueous solvents such as methanol, ethanol, dimethylformamide, or t e t r a h y d r o f ~ r a n . ~ ~ ~ J ~ J ~

A soluble Nafion powder has been recently prepared14J5 and the self-assembling of the ionomer chains in water, formamide (FA), N-methylformamide (NMF), and ethanol (EtOH) solutions has been studied by small angle neutron and X-ray scattering (SANS and SAXS, respective1y).l6-l8 The results for all solvents are similar and have been interpreted in terms of aggregation of Nafion unimers into rodlike micelles arranged in a planar hexagonal array. In the proposed model the perfluoro backbone constitutes the solvophobic core of the rod, the pendant chains are

(14) Grot, W. G.; Chadds, F. U.S. Patent 0066369, 1982. (15) Martin, C. R.; Rhoades, T. A,; Ferguson, J . A.Ana1. Chem. 1982,

(16) Aldebert, P.; Dreyfus, B.; Pineri, M. Macromolecules 1986,19,

(17) Aldebert, P.; Dreyfus, B.; Nakamura, N.; Pineri, M.; Volino, F.

(18) Gebel, G. Thesis, Universite Joseph Fourier, Grenoble, 1989.

54, 1639.

2651.

J . Phys. (Paris) 1988, 49, 2101.

0 1994 American Chemical Society

Self-Assembling Surfactants in Nonaqueous Solvents

Chart 2. Doxy1 Stearic Acid Spin Probes = 5,10

X = Li (DSA)

X = MEll-M. (DSE)

located at the periphery of the rods, and the ionic groups are at the rod-solvent interface. Recent 19F NMR results are in accord with this model for water and FA solutions; in EtOH, however, all 19F nuclei (from the backbone and the pendant chains, identified by their chemical shifts) are represented by motionally averaged signals, suggest- ing the formation of a true solution.9J2 An ESR study of Nafion neutralized by paramagnetic V02+ cations and dissolved in water and alcohols has provided additional support for the conclusions derived from the 19F NMR experiment^.'^

Additional details on the self-assembling of the polymer chains in solutions have been obtained from ESR studies, using doxylstearic acid spin probes, Chart 2.19920 In the latter paper (referred to as I), results have been reported for aqueous solutions of Nafion neutralized by Li+, doped with 5- and 10-doxylstearic acids (5DSA and lODSA, respectively), and with 10-doxylstearic methyl ester (10DSE).20 The dominant, slow motional, ESR signal detected in solutions (concentration 0.5-9% (w/w) iono- mer) and in membranes was assigned to spin probes bound to aggregated polymer chains. "he additional, motionally narrowed, signal detected in the solutions for 5DSA and lODSA has been assigned (at least in part) to spin probes associated with the unimers. The results in I have also suggested that the nitroxide group of lODSA and lODSE is located deeper inside the aggregates than that of 5DSA, but the 14N hypefine splittings for all probes reflect a polar environment. A more ordered structure at a given temperature in the membranes compared to solutions and an increase of the aggregate size a t higher ionomer concentrations have also been suggested in I.

The present work extends these studies to nonaqueous solvents: FA, NMF, and EtOH. The primary objective was to verify the structural model for the ionomer solutions and swollen membranes, in an attempt to reconcile the scattering and spectroscopic results. An additional objec- tive was to contribute to a better understanding of a fundamental problem: how self-assembling of amphiphilic molecules depends on the properties of the solvent. FA has been often used as a nonaqueous solvent for low molecular weight (usually priotated) amphiphiles; forma- tion of simple micelles, microemulsions, and liquid crystalline phases in this solvent has been demon- ~ t r a t e d . ~ l - ~ ~ The microemulsions formed in FA have the potential to become important reaction media, especially for reagents such as perfluorinated olefins, which do not dissolve in water.23 Comparison with the respective aqueous systems has indicated higher critical micelle concentrations, higher Kram points, and smaller sizes

(19) Lee, K. H.; Schlick, S. Polym. Prepr. (Am. Chem. SOC. Div. Polym.

(20) Szajdzinska-Pietek, E.; Schlick, S.; Plonka, A. Langmuir 1994, Chem.) 1989,30, 302.

10, 1101.

440.

(21) Rico, I.; Lattes, A. J. Phys. Chem. 1986, 90, 5870. (22) Binana-Limbele, W.; Zana, R. Colloid Polym. Sci. 1989, 267,

(23) Lattes, A.; Rico, I. Colloids Su$. 1989,35, 221. (24) Schubert. K. V.: Busse, G.: Strey, R.: Kahlweit, M. J. Phys. Chem. . . . . - .

1993, 97, 248. (25) Bergenstahl, B. A.; Stenius, P. J. Phys. Chem. 1987,91,5944. (26) McIntosh, T. J.; Magid, A. D.; Simon, S. A. Biochemistry 1989,

28. 7904.

. . . . - . 1993, 97, 248. (25) Bergenstahl, B. A.; Stenius, P. J. Phys. Chem. 1987,91,5944. (26) McIntosh, T. J.; Magid, A. D.; Simon, S. A. Biochemistry 1989,

28. 7904. ~

(27) Jonstromer, M.; Sjoberg, M.; Warnheim, T. J. Phys. Chem. 1990, 94,7549.

Langmuir, Vol. 10, No. 7, 1994 2189

Table 1. Physicochemical Parameters of Solvents parameters water FA EtOH NMF ref

ViscositYQ 0.89 3.30 1.08 1.65 32 dielectric constant6 78.39 111.00 24.55 182.40 32 surface tensionC 71.81 57.91 21.90 38.70 32 dipole momentd 1.85 3.73 1.69 3.83 42

acceptor number 54.8 39.8 37.9 32.1 33 H-bonding capability 1.17 0.71 0.83 33

and lower stability of the aggregates. It has been suggested that the difference between water and FA is not in the different interaction of the solvents with the head groups but with the hydrocarbon tails of the amphiphiles." "his aspect is important for our system, because of the higher solvophobicity of perfluorinated, compared to protiated, chains.28 Self-assembling pro- cesses have been also observed in NMF.25 To the best of our knowledge, however, there is no report in the literature on micelle formation in EtOH, except in the small-angle scattering studies of Nafion mentioned above.16-18

II. Experimental Section Materials. The Nafion 117 membranes, with an equivalent

weight of 1100 g of polymer per mol of SOsH and a thickness of 0.178 mm, were obtained from DuPont. The soluble Naflon powder in the Li+ form was a gift from G. Gebel (Grenoble, France) and was prepared by solubilization of Nafion/Li+ in an autoclave at 250 "C in a 50/50 (v/v) water/EtOH mixture and evaporation of the solvent at %80 0C.16-18 The spin probes, 5DSA and lODSA from Aldrich and lODSE from Molecular Probes, Eugene, OR, were used as received. The solvents, FA (purified) from Baker & Adamson, EtOH (200 proof dehydrated alcohol) from US. Industrial Chemical Co., and NMF (99%) from Aldrich, were kept over molecular sieves and used without additional purifica- tion. The perfluoropolyether oil (MW = 800) was produced by Montefluos, Milan, Italy. Other chemicals were reagent grade. Relevant physicochemical properties of the solvents are sum- marized in Table 1; data for water are also given, for comparison.

Sample Preparation. All samples for ESR measurements were prepared in a glovebox, in an oxygen-free atmosphere. The NafionlLi powder was directly dissolved in FA, NMF, or EtOH (prebubbled with nitrogen) to the desired concentration in the range 1-25% (w/w). An appropriate amount ofLiOH was added to the solutions to ensure complete neutralization ofthe polymer. Ionomer solutions in EtOH and NMF are perfectly clear, while those in water and FA are opalescent.

ESR Measurements. ESR spectra were measured in the temperature range 125-360 K with a Bruker X-band spectrom- eter, Model ECS 106, equipped with the ESP 3220 data system for aquisition and manipulation and with the EM1 11 VT variable temperature unit.

Additional experimental details have been published.20

HI. Results ESR spectra of the spin probes were recorded in the

temperature range 125-360 K in Nafion solutions and swollen membranes, in the neat solvents, and in an EtOW HzO mixture (1:l by volume). The notation used is probe/ solvent for solutions in the neat solvent, pmWsolvent'S for ionomer solutions, and probdso1vent.M for swollen membranes.

Neat Solvents. The spin probes 5DSA and lODSA can be dissolved in sub-millimolar concentrations in all solvents. For lODSE such concentrations can be obtained in EtOH and NMF; in FA the solubility is about 1 order of magnitude lower.

At 300 K motionally narrowed ESR spectra are observed for all the solutions; the same hyperfine splittings Aim- (14N) were measured for all probes, within f0.03 G. The maximum anisotropic tensor component A,, determined

(28) Gebel, G.; Riatori, S.; Loppinet, B.; Martini, G. J. Phys. Chem. 1993, 97, 8664.

2190 Langmuir, Vol. 10, No. 7, 1994 Szajdzinska-Pietek et al.

16.2 H , S I lSB

c3 - lSa4 t /a I

33.0 34.0 35.0 36.0 37.0

Figure 1. Plot of Abo vs A,, (I4N) for the indicated solvents. The straight line is a least-squares fit to all experimental points, excluding data for the PFPE oil. The correlation betweenAho and the acceptor number (AN) for NMF, EtOH, FA, and HzO is shown in the inset.

Chart 3. Perfluoropolyether Oil (PFPE) CF,(OCF&F),OCF,

CF,

from the rigid limit spectra at 125 K, is identical for the three probes in NMF and EtOH solutions, within f0.2 G. In FA, phase separation occurs on freezing, as in water.20 A rigid limit spectrum at 110 K was obtained only for lODSA, and A,, was determined within f0 .5 G. The average Aiso vs A,, values for the three probes are plotted in Figure 1. Previous data for aqueous solutions20 are also included. The linear correlation is in agreement with the earlier results for 5-, 12-, and 16DSAprobes and their methyl esters in solvents of different p ~ l a r i t i e s . ~ ~ Higher values ofAiso and A,, in nitroxides are due to a higher spin density on the nitrogen nucleus in a more polar environ- ment. The data shown in Figure 1 indicate that there is no correlation between the hyperfine splittings and the dielectric constant E ofthe solvent; E increases in the order EtOH < H20 < FA < NMF, from 24.55 for EtOH to 182.40 for NMF (Table 1). Alinear dependence betweenAiso and the acceptor number (AN) of the solvent is shown in Figure 1 (inset). The AN parameter was introduced by Mayer et al. and is calculated from the 31P chemical shifts of triethylphosphine oxide in solvents that lower the electron density at the phosphorus atom due to the inductive effect.30 Other parameters commonly used as a measure of solvent acidity, such as the Dimroth-Reichardt pa- rameter, ET, or the hydrogen bond donating ability, a,31,32 do not parallel the observed variations of the hyperfine splitting constants.

The lODSE probe in a perfluoropolyether oil (PFPE, Chart 3)33 was also examined; the results (Aiso = 14.15 G and A, = 34.0 f 0.5 G ) do not fit the linear dependence shown in Figure 1 for the other solvents, suggesting that I4N splittings for doxy1 spin probes in perfluorinated and protiated media with different polarities cannot be directly compared.

NafiodEtOH Systems. ESR spectra of 5DSA/EtOWS and lODSE/EtOW-S have been examined. ESR spectra

(29) Griffith, 0. H.; Jost, P. In Spin Labeling. Theory and Applica-

(30) Mayer, U.; Gutman, V.; Gerger, W. Monatsh. Chem. 1976,106,

(31) Marcus, Y. Ion Soluation; J. Wiley: Chichester, England, 1985. (32) Fawcett, W. R. J. Phys. Chem. 1993, 97, 9540. (33) Sanguineti, A.; Chittofrati, A.; Lenti, D.; Visca, M. J. Colloid

tions; Berliner, L. J., Ed.; Academic Press: New York, 1976; p 501.

1235.

Interface Sci. 1993, 155, 402.

1 1 I 1 l ~ I 1 1 1 I

3320 3340 3360 3380 3400 31?0 111

V

3320 3340 3260 3310 ?400 3420 , 1 1 1 l , l l 1 1

Figure 2. X-band ESR spectra in the temperature range 180- 220 Kof lODSE/EtOH (neat) in (A) and of lODSE/EtOWS (23% (w/w)) in (B). Modulation amplitude was 0.5 G.

a t 300 K of both probes in the ionomer solution and in the neat solvent are similar, in terms of line widths and Aiso values. At lower temperatures the lines broaden, and below 200 K the spectra are typical of slow tumbling spin probes, reaching a rigid limit at 125 K and an extreme separation ca. 1 G higher in the ionomer solutions, compared to the solution in the neat solvent. Selected ESR spectra, of 10DSE/EtOH and lODSE/EtOWS (poly- mer concentration 23% (w/w)), are presented in Figure 2. The line widths in the temperature range 220-280 K are higher in the presence of Nafion. The differences with respect to the neat solvent do not exceed 5% for 5DSA but are larger for lODSE, up to 25% for the m = -1 (high field) signal a t 220 K. It is clear from Figure 2 that a t a given temperature lODSE spectra are more “rigid in the presence of Nafion than in neat EtOH. In both cases, however, the spectra indicate higher mobility of lODSE in comparison to 5DSA, which is opposite to the trend observed for the spin probes bound to the aggregated polymer in water and FA (see below) systems.

Figure 3 presents ESR spectra obtained at 300 K for 5DSA and lODSA spin probes in membranes swollen by ethanol; two spectral components are detected. The line widths of the narrow triplet are ~ 3 0 % larger than in neat ethanol. These widths, and the relative intensities of the two components, are very sensitive to the ethanol content in the membranes. Because of the high volatility of the solvent, it is difficult to reproduce the spectra quantita- tively. In drier samples the line widths of the motionally narrowed component are broader, and the slower com- ponent is more pronounced. Even for a clearly visible solvent excess, however, two components are detected and the line widths of the motionally narrowed component are broader compared with the neat solvent. The amount of lODSE retained by the membrane was less than l/10 that of BDSA and lODSA probes, suggesting that lODSE is more solvophilic with respect to EtOH, and is washed- out during sample preparation.20

N&on/N1VLF Systems. ESR spectra of spin probes 5DSA and lODSE have been examined in NafionNMF solutions and in swollen membranes. The probes were not retained in the membranes, suggesting even higher solvophilicity in NMF than in EtOH.

Motionally narrowed signals are obtained in 25% (w/w) Nafion solutions in NMF at 300 K, but the lines are broader compared to the neat solvent (especially for lODSE),

Self-Assembling Surfactants in Nonaqueous Solvents Langmuir, Vol. 10, No. 7, 1994 2191

1 1 ,

3320 3340 3360 3380 3400 3420 I61

Figure 3. X-band ESR spectra at 300 K for 5DSAIEtOH.M and 10DSAE3OH.M. Modulation amplitude was 2 G. Spectra are normalized to the same integral intensity. The low field slow motional (anisotropic) component is emphasized in the vertically expanded region.

l l l l l l l l l l l 1 1 l l l l l l I I l l l l W

3320 331 3x4 3380 3401 3424 3120 3110 339 3380 3Mo 31M $1 161

Figure4. X-band ESR spectra in the temperature range 230- 270 K of 10DSE/”F (neat) in (A) and of lODSE/NMF/S (25% (w/w)) in (B). Modulation amplitude was 0.5 G.

indicating lower mobility of the spin probes in the presence of Nafion. The effect is more pronounced at lower temperatures, as seen in Figure 4. For example, at 250 K a highly anisotropic spectrum is observed for the solution, while in the neat solvent the signal is isotropic. Spectral changes do not seem to be related to the melting point of the solvent (269.5 K).

Nafion/FA Systems. Figure 5 presents ESR spectra recorded at 300 K in two Nafion/FA solutions and in swollen membranes. For 5DSA/FA/S and 10DSA/FA/S (parts A and C of Figure 5 ) the motionally narrowed signal is dominant in the 1% (w/w) Nafion solution, while in the 22% (w/w) solution the slow motional signal is also detected; th is component is clearly seen in Figure 5C where a higher modulation amplitude was used, and in the vertically expanded portion of Figure 5A. Two spectral components were obtained for lODSE/FA/S in the entire range of concentrations. In membranes the contribution of the motionally narrowed signal for all three probes is negligible (<1%).

1310 3340 3360 3360 ?NO 3120

IL I

3320 3340 3360 3300 3400 3420 3320 3340 3360 3380 3400 3420

IGI IG1

Figure 5. X-band ESR spectra at 300 Kin 1% and 22% (w/w) Nafion/FA solutions and in the respective membrane, for 5DSA (A), lODSA (C), and lODSE (D). The slow motional component for the 22% solution is emphasized in the vertically expanded region. The modulation amplitude was 0.5 G in (A) and 2.0 G in (C) and (D). Spectra are normalized to the same integral intensity and then expanded vertically by the factor given on the right. The parameters used for the calculation of the order parameter are indicated in (A), lowest spectrum: 2A‘, = 2A,, between the downward arrows, and 2A- between the upward arrows. The relative intensities ofthe slow motional component in water and FA (in %) vs Nafion concentration are plotted in (B) for 5DSA.

Figure 6 presents ESR spectra in the temperature range 125-360 K for lODSE/FA/S (22% (w/w) Nafion) and lODSE/FA/M. Minor changes are detected up to 200 K, and above 220 K the extreme separation gradually decreases. Above the melting point of FA (275.7 K) the motionally averaged component appears in the solution and grows with increasing temperature. In the membrane the sharp three-line component (the high-field component is indicated by an arrow in Figure 6B) has a negligible contribution even at 360 K. We suggest that this signal is due to the free spin probe in the excess solvent added to ensure an equilibrium swelling of the membrane. Above 300 K the spectral range of the slow motional signal for lODSELFA/M decreases, indicating incipient averaging (Figure 6B). Comparison of the line widths for the three probes suggests that the probe mobilities in the membrane a t 360 K are in the order 5DSA > lODSA > 10DSE.

IV. Discussion This section consists offour parts. First, we will discuss

the results obtained for EtOH and NMF as solvents. Second, we will consider self-assembling of the ionomer in FA. Third, we will compare self-assembling in short-

2192 Langmuir, Vol. 10, No. 7, 1994 Szajdzinska-Pietek et al.

I I I I I V I I I 1 I

3320 3340 3360 3380 3400 3420 IGI

I

I I I I I

3320 3340 3360 3380 3400 3420 IGl

Figure 6. (A) X-band ESR spectra of 10DSE/J?A/S (22%) (w/w)) as a function of temperature. Modulation amplitude was 2 G in the range 125-290 K and 1 G in the range 300-360 K. (B) X-band ESR spectra of 10DSE/FA/M as a function of temperature. Modulation amplitude was 2 G in the range 125-290 K and 0.5 G in the range 300-360 K. The arrow indicates the low field isotropic signal due to the free probe in the excess solvent (see text).

chain perfluorinated surfactants and in perfluorinated ionomers. Finally, we will propose a model for the transition between the rodlike micelles in the ionomer solutions in water and FA, and the reverse micelles in membranes swollen by these solvents.

Structure of the Ionomer in EtOH and NMF. The spectra shown in Figure 2 do not reveal formation of large polymer aggregates in EtOH solutions. It could be argued that the absence of the slow-motional component at higher temperatures is due to higher solubilities of the spin probes in the neat solvent. We recall however that the 19F NMR studies, where no external probe is used, also indicate formation of a true Nafion solution in e t h a n ~ l . ~ J ~ Fur- thermore, no changes of the surface tension coefficient in going from neat EtOH to 23% (w/w) Nafion solution were detected (within f l dydcm), while in water a gradual decrease was observed from 0 to 9% (w/w) ofthe polymer.20 It seems that EtOH lacks the polarity and cohesive energy that favor molecular aggregation, even though it has high hydrogen bonding capability (Table 1). Our results are thus in agreement with the suggestion of Binana-Limbele and Zana,22 that micellization is governed by the cohesive energy, not by the hydrogen bonding.

Lower mobilities of the spin probes below 300 K (Figure 2) are detected in Nafion solutions, compared to the neat solvent. Since the effect is probe dependent, it is logical to assume that the mobilities reflect binding of the probe to unimers. Our results suggest that the binding is more efficient for the neutral probe lODSE, than for the charged BDSA probe.

If phase separation into polar and nonpolar domains existed in the membranes swollen by EtOH, the two spectral components (Figure 3) could be rationalized by spin probes located in the polymer phase and in the EtOH clusters, respectively. The line widths, however, indicate that the probes are significantly less mobile in the membranes than in the bulk solvent. We suggest therefore that EtOH, unlike water or FA (see next section), penetrates into and plasticizes the polymer chains; l9F NMRg and V02+ ESR13 experiments support this conclu- sion. The slow motional component may reflect probe molecules located further away from the ionic groups, in regions less accessible to the solvent.

The 2Az, values for the swollen membranes are inde- pendent of the probe; the extreme separation at 300 K, however, is higher for 5DSA than for lODSA (62.8 G as compared to 60.7 GI. This indicates lower mobility of the former probe and could be explained in terms of its lower solvophobicity and deeper penetration into more rigid regions of the membrane, less penetrated by EtOH. The probe solvophobicity in this relatively nonpolar solvent appears to be in the order BDSA > lODSA > lODSE, which is opposite to the case of FA and water. Similar experi- ments have been done earlier in this laboratory, for BDSA and 16DSA in NafionlNa membranes swollen by metha- n01.l~ In this study two-component spectra have been observed at lower temperatures, but at 300 K the motionally averaged signal only has been detected. The different result, compared to the EtOH system, may be related to the lower viscosity ofmethanol, and to its higher

Self-Assembling Surfactants in Nonaqueous Solvents

polarity, which leads to reduced penetration into the perfluorinated chains. This explanation is in accord with results obtained for V02+ in membranes swollen by the two alcohols: the rotational correlation time re at 300 K is significantly shorter in methanol than in ethan01.l~

The results for NMF solutions are qualitatively similar to those for EtOH solutions, except when comparing BDSA vs 10DSE data. In neat NMF the spectra at all tem- peratures are essentially the same for the two probes, but lODSE is less mobile than BDSA in the presence of Nafion. It appears that the spin probes in Nafion solutions associate with the polymer. It is possible, in spite of their solvophilic character, that a dynamic equilibrium exists between free probes and probes bound to unimers and even to aggregates. The presence ofthe latter is suggested by the lower mobility of lODSE/NMF/S in comparison to BDSA/NMF/S. However, the absence of the slow motional component at 300 K suggests that the aggregates are much smaller than those in water and FA (see below).

Self-Assembling in FA. As in I, we assign the slow motional and motionally narrowed spectral components, respectively, to spin probes bound to polymer aggregates and in the bulk solvent associated with Nafion unimers. In support of this assignment of the motionally narrowed component is the increased solubility of lODSE (by 1 order of magnitude) in 1% (w/w) Nafion solutions in FA, compared to neat FA. This behavior is similar to the increased solubility of hydrophobic probes in solutions of surfactants above the critical micelle concentration (cmcY4 and indicates formation of solvophobic regions from self- assembled ionomer chains even at low concentrations (1%).

Within experimental error (f0.3 G), the isotropic splitting Aiso for the three-line signal in Nafion solutions is the same as in neat FA. The line width is, however, affected by the presence of the polymer. For example, the peak-to-peak width of the m = 1 (low field) line of 5DSA increases from 1.11 Gin the neat solvent to 1.23 Gin 22% (w/w) Nafion solution. For this reason the two-component spectra could not be deconvoluted accurately by subtract- ing the signal observed in the neat solvent, as described in I for aqueous solutions. Only a lower limit of the slow motional signal contribution was estimated, by performing the spectral titration until the reversed-phase signal appeared. The results for BDSA are shown in Figure 5B; data deduced for 5DSA/water/S in I are also shown, for comparison. Identical results were obtained for the deconvolution for 10DSA/FA/S, within experimental error. The contribution of the slow motional signal for these two probes in 1% (w/w) Nafion solutions in FA is not higher than lo%, which is within the limits of the accuracy. The more solvophobic lODSE probe is less solube in neat FA, and the deconvolution procedure gives 99% and 70 f 10% of the slow signal in 22% and 1% (w/w) Nafion solutions, respectively. In all cases the fraction of the slow com- ponent in FA solution is significantly lower than in aqueous solutions of comparable Nafion concentration.20 This may be in part due to higher solubility of the spin probes in the former solvent; we note that lODSE is completely un- soluble in neat water. The fact that for BDSA in the 22% (w/w) Nafion/FA solution the contribution of the slow motional signal is lower than that in 0.5% (w/w) Nafiod HzO solution (70% as compared to 83%) cannot, however, be explained by the different solubilities in neat solvents alone. A more reasonable interpretation of these results is that in ionomer solutions in FA there are fewer micellar

Lmgmuir, Vol. 10, No. 7, 1994 2193

Table 2. Extreme Separation (G) of ESB Spectra for FA Systems at 300 K (W,) and at 126 K (M,)

5 DSA 10 DSA 10 DSE % (w/w) Nafion 2A:, 2Azz UZz 2Azz WZz 2Azz 1 69.8 71.2 61.9 70.3 4 70.2 12 69.8 22 54.0 70.0 60.8 71.2 62.2 70.7

63.3 71.2 ~ 6 5 ~ 60.0 69.9 63.0 71.0 av 69.9 f 0.1 71.1 =k 0.1 70.7 =k 0.4 a Estimated from the number of solvent molecules per so3- group

in the membrane.18

aggregates. In addition, in swollen membranes (which contain =35% (w/w) FA18 and where all polymeric material is aggregated) the contribution of the motionally narrowed signal is negligible, <l%.

Less efficient micellization in FA (compared to water) has been observed previously for short-chain surfactants, and explained in terms of a weaker repulsive interaction of the solvent with the solvophobic tails of the amphi- ~ h i l e s , ~ ~ - ~ ~ resulting in a higher cmc. A higher concen- tration of the unimers, in equilibrium with large micellar aggregates (rods), can therefore be expected for polymer solutions in FA, in agreement with the larger contribution of the isotropic signal observed experimentally. The increasing line width of this signal with polymer content indicates the decrease of the probe mobility, which we assign to the high polymer concentration. The effect was not detected in water solutions, probably because the maximum polymer concentration was only 9% (W/W).~O

Table 2 shows the extreme separation of the ESR signals a t 300 K(Wz2, shown by downward arrows in Figure 5A), and in the rigid limit at 125 K(2Azz). For BDSA and lODSA in solutions containing less than 22% (w/w) Nafion we have not been able to estimate WZz values with a reasonable accuracy, because the slow motional signals at 300 K are too weak. The relatively highAzJAzz ratios (from 0.77 for 5DSA/FA/S to 0.89 for 10DSA/FMM and 10DSE/FA/M) indicate considerable immobilization of the spin probes and are consistent with our assignment ofthe slow motional component to polymer aggregates. The 2Au values for all the probes in solutions and swollen mem- branes appear to be independent of polymer concentration. The average A,, values are the same as that determined for lODSA in neat FA (70 f 1 G), suggesting that the microenvironment of the probes in aggregated Nafion is fairly polar. It is possible that the probes “drag“ some solvent molecules into the aggregate, in analogy with the proposed penetration of water into protiated micellar systems.36 We cannot exclude, however, the effect of the perfluorinated polymer chains on the hyperfine param- eters, as discussed above for the probe solutions in neat solvents. As expected, the extreme separation at 300 K is lower

in solutions than in membranes. However, the data for lODSE do not reveal an increase of Nu with Nafion content. Such an increase was detected in aqueous solutions and was assigned to the formation of larger aggregates. I t appears that, due to weaker solvophobic interaction, the aggregates in FA do not grow much in size even at higher Nafion concentrations (up to 22% as compared to 9% in water), andor they have less compact (more disordered) structure, so that incorporation of additional unimers does not affect the mobility ofthe probe.

(34) Shinoda, K. In Colloidal Surfactants; Shinoda, K, Nakagawa, T., Tamamushi, B., Isemura,T., Eds.;Academichss: New York, 1963; Chapter 1.

(35) (a) Menger, F. M.; Jerkunica, J. M.; Johnston, J. C. J . h . Chen. Soc. 1978,100,4676. (b) Muller, N.; Simsohn, H. J. Phys. Chem. 1971, 75, 942.

2194

L u w

Y P h

b 4

0 s

0.80

em

0.60

0.50

Langmuir, Vol. 10, No. 7, 1994

-- -- -- -- --

1 \ * "I 0.50

O S

a80

o.m

0.60

0.50

-*

- - -- -*

* - n --

* I

i 0.40 s

230 250 270 290 3x) 330 350 Temperature [KI

Figiwe 7. The order parameter S as a function of temperature for the three probes in membranes swollen by FA and in 22% Nafion solutions. The straight lines for each system are least- squares fits to the experimental data.

A more quantitative comparison of the anisotropic spectra in solutions and membranes at different tem- peratures can be obtained by calculating the order parameter S according to eq 1

where A,, is half the extreme separation (A',,) and A ~ n is half the distance between the first minimum and the last maximum in the spectra, as indicated by arrows in Figure 5A. The hyperfine tensor components areArr,Ayr, and A,,, and Aiso = (A, + hY + Az,)/3.

Taking into account that A, values are independent of polymer content and close to the value for neat FA, we have assumed Aiso values equal to those measured in the neat solvent, as reported for aqueous systems in I. The results are shown in Figure 7, for the three probes in the most concentrated Nafion solutions and in the swollen membranes.

The variation of the order parameter in FA and water systems is qualitatively the same. At a given temperature S values increase from solutions to membranes, and at a

Szajdzinska-Pietek et al.

given polymer content they change in the order S(5DSA) S(1ODSA) 5 S(lODSE), the difference between the three

probes being higher in solutions than in membranes. The slopes of S vs T plots are higher for 5DSA than for lODSA and lODSE and, for the latter two probes, slightly higher in solutions than in the membranes. These findings indicate that, in both aqueous and FA systems, BDSA is most mobile and is located in the less ordered peripheral region of the aggregate, while the other two (more solvophobic) probes penetrate deeper inside.20

The most important difference between aqueous and FA systems is that the S values in FA are lower; data for the 22% Nafion/FA solution are close to those for 0.5% Nafion/HzO solution. In membranes swollen by FA the S values at 300 K are lower by 8%, 5%, and 4% for BDSA, lODSA, and lODSE, respectively. Furthermore, the slopes of the straight lines in Figure 7 are higher than those obtained in aqueous systems, for the solutions and the swollen membranes. This result supports the suggestion given above, that in FA the molecular packing in the aggregates is more loose than in water. Consequently, the mobility of the probes may be higher, even in the environment of a more viscous solvent. The less ordered structure of NafiordFA aggregates is in agreement with the higher specific area occupied by one charge on the surface of the rod in FA, deduced from small angle scattering experiments.l'

Self-Assembling in Perfluorinated Short-Chain vs Polymeric Surfactants. Formation of micelles in aque- ous solutions of ammonium pentadecafluorooctanoate (AmPFO) has been studied by Ristori and Martini, using BDSA and 16DSA, and two cationic spin probes differing in size.36 For the large cationic spin probes micellization of the probe was detected, based on the appearance of a very broad ESR signal, whose intensity increases with probe concentration. Probe molecules associated with surfactant micelles were also observed; the ESR spectrum consists of an isotropic triplet, with broader components than for sub-cmc surfactant concentrations. The ESR spectrum was simulated with little or no anisotropy of the dynamic parameters. Similar spectra were also detected for the DSA probes associated with the surfactant micelles. Aggregation of perfluoropolyethers surfactants (MW 621 to 768) in aqueous solutions has been studied by Martini et al. using spin probes.37 The ESR spectrum at ambient temperature of a small cationic probe associ- ated with the micelles consists of an isotropic triplet and the lines are broader than in neat water.

The results presented for aqueous and FA solutions of Nafion differ significantly from those discussed ab0ve,3~J~ primarily in the detection of a slow-motional spectral component, which was assigned to spin probes associated with polymer aggregates. Preliminary simulations of this spectral component for lODSE/water/S (Nafion concen- tration 9% (w/w)) a t 300 K indicate that typical values of the rotational diffusion tensor components RII and RI are, respectively, 50 x lo8 and 0.1 x lo8 rod/s, leading to an anisotropy factor N (=R,JRJ of this value can be compared with N = 2 for BDSA in A ~ I P F O . ~ ~ The anisotropy used t o simulate the spectrum of lODSE in the membranes at 300 K is even higher, N - 1000. To the best of our knowledge such high values ofN have not been reported previously for perfluorinated surfactants. We suggest that the formation of large aggregates, and the

(36) Ristori, S.; Martini, G. Langmuir 1992, 8, 1937. (37) Martini, G.; Ottaviani, M. F.; Ristori, S.; Lenti, D.; Sanguineti,

(38) Szajdzinska-Pietek, E.; Pilar, J.; Schlick, S. To be submitted for A. Colloids Surf. 1990,45, 177.

publication.

Self-Assembling Surfactants in Nonaqueous Solvents

ensuing immobilization of the spin probes, is due to the high M w of Nafion, a 2 x lo6.

Finally we note that the results obtained in this report and in I seem to validate the initial assumption that the protiated probes associate with the perfluorinated short- chain and with polymeric surfactants and their ESR spectra can provide structural and dynamical details about the host. This conclusion is in agreement with a recent study of fluorinated and protiated mixed micelles by SANS: scattering curves have suggested formation of mixed micelles having the same composition, and a very narrow size d i s t r ib~ t ion .~~

Transition between Micelles (in Solutions) and Reverse Micelles (in Membranes). The ESR results obtained for membranes neutralized by paramagnetic counterions (Cu2+, V02+), or doped with nitroxide spin probes, are in agreement with the 19F NMR studies. These spectroscopic studies suggest that aqueous and FA systems are unique in producing extensive aggregation in solutions and a phase separated morphology in swollen membranes. Results obtained for other solvents such as ethanol, methanol, THF, DMF, and NMF suggest that membranes are plasticized in the solvents and that large aggregates are not formed in the solutions. The difference between the solutions and the membrane swollen by these solvents can be viewed as a decrease in the interaggregate distance with increasing polymer concentration.

We would like to comment now on the transition between the rodlike micelles in solutions of the ionomer in water and FA, and the reverse micelles in the swollen mem- branes, and to propose a mechanism for this transition, which we call thefringed rod model. The basic assumption is that at low ionomer concentrations only part of the chain segments are incorporated in a rod, while other segments are outside the rods, as more disordered material in contact with the solvent (Figure 8A): the rods are fringed. As the concentration of the polymer increases, parts of the chains "dangling" outside the rods become incorporated in other rods. Because the average chain contains about 200 repeat units and the maximum extended chain length is %4000 A, it is reasonable to imagine that one chain can belong to more than one micellar rod. The process continues and the disordered chain segments self-assemble and provide cross-links between the existing rods, so that the entire mass of polymer becomes connected; this process is shown in Figure 8B. In support of this mechanism is the high degree of order detected for the spin probes in water and FA solutions, and the even greater degree of order in the swollen membranes.

For an aqueous solution containing 10% (w/w) ionomer, the distance between nearest rods is %140 A, and the rod diameter is e60 A.18 The distance between the rods is much smaller for a higher ionomer concentration; we note that in membranes the ionomer concentration is %80%. As the connectivity of the polymer material increases, solvent becomes trapped between the connected rods. Some disorder is still expected, due in part to the rigidity of the perfluorinated chains, and leads to only partial enclosing of the solvent pools, i.e. to solvent connectivity between the pools. This conclusion is very important, because it helps explain the enhanced diffusion of cations and oxygen in the membranes+l compared to Teflon, the perfluorinated material that does not contain polar groups.

(39) Caponetti,E.;Martino,D.C.;Floriano,M.A.;Triolo,R.Langmuir ms,9, 1193. (40) Sperling, L. H. Physical Polymer Science; Wiley-Interscience:

(41) Maiti, B.; Schlick, S. Chem. Mater. 1992,4,458. (42) Handbook of Chemistry and Physics, 71th ed.; CRC Press: Boca

New York, 1986; p 162.

Raton, FL, 1990-1991.

Langmuir, Vol. 10, No. 7, 1994 2195

Figure 8. Fringed rod model for the transition between the rods in aqueous or FA solutions and the reverse micelles in the membranes: (A) isolated rod; (B) assembling of "dangling" polymer segments from two parallel rods i n t ~ two additional rods at high polymer concentrations.

The structure shown in Figure 8B, which contains four rods formed at high polymer concentration from two parallel rods initially present in a hexagonal arrangement, would therefore leave a solvent pool with a diameter significantly smaller than 80 A, perhaps approaching the literature value of %50 A.1 In addition to formation of more rods by incorporating the "dangling" chain segments, we can also expect that the size of the aggregates will increase with increasing polymer concentration, especially for water as solvent. This suggestion is based on the increase in the extreme separation of the slow-motional component in Nafion solutions measured in I as the polymer concentration increases. At this time we propose that more and larger aggregates are formed in the membranes, compared to solutions.

The proposed mechanism for the micelle - reverse micelle transition is similar in some respects to the fringed micelle model for the formation of single crystals of polymers, especially in the description of the connectivity and the presence of order in the system.q0

The results obtained in this study will now be compared with the conclusions of the scattering experiments. The spectroscopic evidence (from ESR and NMR studies) supports the formation of large polymer aggregates in water and FA, in agreement with scattering data. These data, however, do not support a similar conclusion in EtOH and NMF solutions. It is possible that in these solvents the aggregates are highly permeated by the solvents, still leading to appreciable scattering from the swollen objects. The calculation of the rod dimensions17J8 is based on dense material, with a density identical to that in the bulk (e = 2); if the material is highly permeated by the solvent, the density of the "object" and the object radius will be smaller. On the more local scale of the spectroscopic

2196 Langmuir, Vol. 10, No. 7, 1994

measurements (ESR and NMR), these solvent-permeated objects reflect chains with a high degree of dynamic reorientation.

V. Conclusions ESR spectra of doxy1 stearic acid probes do not indicate

formation oflarge aggregates in Nafion/EtOH and Nafiod NMF solutions, in agreement with 19F NMR, but contrary to SANS and SAXS conclusions. The behavior of the spin probes in membranes swollen by EtOH can be reconciled with a plasticizing effect of the solvent rather than with a phase separated morphology. This conclusion is in agreement with ESR studies of Nafion membranes and solutions containingV02+ as the counterion, and with 19F NMR results.

The self-assembling of ionomer chains in FA solutions and swollen membranes is qualitatively similar to that in the corresponding aqueous systems. Quantitatively, self- assembling of perfluorinated polymer amphiphiles in FA solutions is less efficient and leads to less ordered aggregates, as observed earlier for low molecular weight protiated surfactants.

Szajdzinska-Pietek et al.

The results reported in this paper for FA systems, together with results obtained for aqueous systems, suggest a mechanism for the transition between the micellar structure in solution and the reverse micellar structure in the swollen membranes, which we call the fringed rod model. The model assumes that at high polymer concentrations one chain can become part of more than one rod, thus providing the cross-linking necessary for complete connectivity of the polymeric material.

Acknowledgment. This research was supported by the National Science Foundation Grants DMR-8718947 and DMR-9224972 (Polymers Program) and IN"-8922643 (US-Poland Collaborative Research), and by NATO Grant CRG 901093 (US-France Collaborative Research Pro- gramme). We thank G. Gebel for preparing for us the soluble Nafion powder, S. Ristori (Florence) for the perfluoropolyether oil, DuPont Company for donating the perfluorinated membranes, and B. Quinn of BASF Labo- ratories for the surface tension measurements.