Interaction between a Novel Gemini Surfactant and Cyclodextrin: NMR and Surface Tension Studies

Journal of Colloid and Interface Science 246, 191–202 (2002) doi:10.1006/jcis.2001.8058, available online at http://www.idealibrary.com on

Interaction between a Novel Gemini Surfactant and Cyclodextrin: NMR and Surface Tension Studies S. Abrahms´en-Alami,1 E. Alami,2 J. Eastoe, and T. Cosgrove School of Chemistry, University of Bristol, Bristol BS8 1TS, United Kingdom Received June 6, 2001; accepted October 22, 2001

INTRODUCTION The interaction between cyclodextrins, hydroxypropyl-βcyclodextrin (HPβCD), and hydroxypropyl-γ -cyclodextrin (HPγ CD) and a novel type of nonionic surfactant synthesized from a fatty acid has been investigated. The so-called nonionic heterogemini surfactant (NIHG750) contains two hydrophobic groups and two hydrophilic groups, composed of one monomethyl ethylene glycol and one secondary OH group, CH3 (CH2 )7 – CH[OH]–CH[O(CH2 CH2 O)16 CH3 ]–(CH2 )7 CN. Surface tension studies indicate that micelles form in NIHG750 systems in both the presence and the absence of small quantities (molar ratio (HPβCD : NIHG750) ≈ 2) of cyclodextrin (HPβCD or HPγ CD). This gives NIHG surfactants an advantage compared to singletailed nonionic surfactants, which generally lose their ability to micellize at much lower additions of cyclodextrins. However, the interaction between HPβCD and NIHG750 results in a disruption of the micellar aggregates at higher levels of cyclodextrin. In the dilute systems (CNIHG750 < 0.1% (w/w) aprox) prolate-shaped mixed aggregates (HPβCD and NIHG750) form, with a short ˚ respectively. and a long axis of the order of 8–9 and 17–20 A, These gradually aggregate into micellar-like structures at higher concentrations. In the aqueous bulk phase HPβCD interacts mainly with the hydrophobic part of NIHG750, but both NMR and surface tension measurements indicate that an interaction with the hydrophilic part of NIHG750, as well, may exist. This interaction results in a better packing of NIHG750 at air-water interfaces. However, at elevated temperatures results from turbidity measurements indicate that NIHG750 and HPβCD interact mainly through the hydrophilic part of the surfactant; a decrease in the cloud point temperature is observed. The interaction of the larger cavity molecule, HPγ CD, with NIHG750, on the other hand, seems to be relatively weak. The interaction, when present, most probably takes place through inclusion of the hydrophilic EO part of NIHG750. The results suggest that HPγ CD in combination with NIHG750 is a better solubilizing system than with HPβCD. °C 2002 Elsevier Science Key Words: NMR; self-diffusion; NOE; HPβCD; HPγ CD; cyclodextrin; nonionic heterogemini surfactant; interaction; complexes; surface tension.

1 To whom correspondence should be addressed at current address: AstraZeneca R&D M¨olndal, 431 83 M¨olndal, Sweden. 2 Current address: Applied Surface Chemistry, Chalmers University of Technology, 412 96 G¨oteborg, Sweden.

Cyclodextrins (CDs) and their derivatives have been used widely in pharmaceutical formulations where they function as stabilizing and solubilizing agents or as catalysts (1–3). They are composed of 6 to 8 (α-1,4)-linked α-D-glucopyranose units forming a rigid, torus-shaped conformation with a central cavity ˚ (Fig. 1). Approximately 20 current drug forof the size 5 to 8 A mulations employ cyclodextrins, more than half of which contain natural β-cyclodextrin (3). The lipophilic cavity of the CD molecule provides a microenvironment into which hydrophobic molecules or hydrophobic parts of molecules may be solubilized. When administered orally cyclodextrins have been shown to decrease blood cholesterol levels by effective complex formation with bile acid anions (4–6). Natural cyclodextrins are modified for various purposes, for example, to improve their low aqueous solubility or to decrease their toxicity. The hydroxyl groups are available as starting points for structural modifications, and various functional groups, such as alkyl and hydroxyalkyl groups, have been incorporated into the cyclodextrin structure (2). However, the binding constants between guests and alkylated and hydroxyalkylated CDs are usually lower than those of the parent cyclodextrins (1). Surface-active agents are an alternative means for solubilizing hydrophobic compounds that are frequently found in drug delivery systems (7). Nonionic ethylene oxide-based surfactants are particularly common ingredients in pharmaceutical formulations because of their low toxicity. Combinations of cyclodextrins and nonionic surfactants are increasingly used in pharmaceutical formulations. One reason for mixing two solubilizing agents is to decrease the concentration of each component so that toxic levels are not reached. Both the micellar core and the hydrophobic interior of the CD molecule provide a solubilization site for hydrophobic compounds in such a system. However, the solubilizing capacity may decrease if the surfactant itself forms an inclusion complex with CD and hence competes with the hydrophobic guest. Many drugs are surface active and are therefore often toxic since they interact with cell membranes. They may also have other disadvantages such as adsorption at surfaces of tubes and containers, for example. The addition of CD to a formulation of a surface-active drug is a way to overcome such a problem. Studies of the interaction between surface-active

191

0021-9797/02 $35.00

° C 2002 Elsevier Science

All rights reserved.

192

´ ABRAHMSEN-ALAMI ET AL.

FIG. 1. Structural formulae and 1 H NMR spectra (at 27◦ C) of a mixture of NIHG750 (1.5% (w/w)) and HPβCD (where the degree of substitution is denoted by n) at molar ratio 1 (HPβCD : NIHG750).

agents and CDs are hence a field of interest both from technical and fundamental points of view. Interactions between cyclodextrin and nonionic surfactants have been studied previously using methods such as fluorescence (8, 9), nuclear magnetic resonance (9), calorimetric titrations (10, 11), and surface tension (12–14). Most of these studies indicate that an interaction occurs with the hydrophobic part of the surfactants. Complexation ratios (CD:surfactant) from 1 to 2 have been reported for nonylphenol oligoethylene or Triton X nonionic surfactants and α-, β-, or γ -CD (8). Topchieva et al. reported studies which imply that an interaction may also take place between the hydrophilic ethylene oxide part of the surfactant and α- or γ -CD (12). A new class of surfactant referred to as a “heterogemini” has been used in the present study (15). This type of surfactant appears to be better than conventional compounds in certain important properties such as solubilization capacity. The surfactant used here, denoted NIHG750, is composed of two hydrophobic tails with nine carbons and two head groups, hydroxyl and oligoethylene oxide (EO)16 , respectively (Fig. 1). Studies of the interaction of this surfactant with cyclodextrins are particularly interesting due to its relatively bulky hydrophobic group. The aim of the present study is to investigate the interaction between a nonionic heterogemini surfactant, NIHG750, and cyclodextrin, i.e., hydroxypropyl-β-cyclodextrin (HPβCD) and hydroxypropyl-γ -cyclodextrin (HPγ CD) using NMR chemical shift, nuclear Overhauser enhancement (NOE), and selfdiffusion measurements. In addition, adsorption and aggregation have been further investigated by surface tension and turbidity

measurements. In the present study hydroxyalkylated CDs were used due to their higher solubility and the fact that they form soluble complexes with various guests that may be more easily characterized with the methods used here. EXPERIMENTAL SECTION

Materials Hydroxypropyl-β-cyclodextrin (Encapsin) and hydroxypropyl-γ -cyclodextrin were purchased from Jannsen and Fluka, respectively, and used as received. The HPβCD and HPγ CD contain 2.8 and 5.1% water and have degrees of modification of 0.64 and 0.6, respectively. The molecules have molecular weights, of 1300 and 1580 g/mol and are composed of 7 and 8 glucopyranose units. Their inner and outer diameters are 6 and 8 ˚ respectively (2). The synthesis of the nonand 15.4 and 17.5 A, ionic heterogemini surfactant, NIHG750 (Mw = 1030 g/mol), and the self-aggregation in aqueous solution have been described by Alami et al. (15). The molecular weight of the oxyethylene part of the surfactant is 750 g/mol which may explain the denotation NIHG750. Samples for surface tension measurements were prepared using double-distilled water as solvent. All solutions were clear and homogeneous at the measurement temperature, which was 20◦ C unless otherwise stated. NMR samples were prepared in D2 O (Isotec Inc, 99.9%) when not otherwise indicated. Some samples were also prepared by using a combination of 50% (w/w) D2 O and 50% (w/w) deionized H2 O. Since the viscosity of H2 O is different than that of D2 O, the self-diffusion

GEMINI-CYCLODEXTRIN INTERACTION

coefficients in the H2 O/D2 O mixture were corrected using Eq. [1] to enable direct comparison with the samples solubilized in D2 O. ¤ £ DD2 O = DH2 O/D2 O 0.5ηH2 O /ηD2 O + 0.5 .

[1]

Here ηH2 O and ηD2 O denote viscosities of nondeuterated and deuterated water, and DD2 O and DH2 O/D2 O are the self-diffusion coefficients in deuterated water and the mixture, respectively. NMR NMR chemical shift and nuclear Overhauser enhancement studies were carried out on a JEOL Alpha 500 MHz spectrometer. A NOE-DF pulse sequence was used and the resonance signal of interest was selectively irradiated during the experiment. Self-diffusion measurements, based on the Fourier transform pulsed-field gradient spin-echo (NMR PFGSE) technique were performed using JEOL FX 100 and JEOL Alpha 500 spectrometers equipped with a diffusion probe head capable of producing a maximum gradient strength of 0.6 and 1.2 T/m, respectively. A stimulated echo pulse sequence was employed for the determination of self-diffusion coefficients (16, 17). The attenuation of the spin-echo amplitude after Fourier transformation was sampled as a function of the duration of the applied gradient strength (G). The gradient pulse interval, 1, and the width of the gradient pulse, δ, were kept constant at 220 ms and in the range 3 to 6 ms, respectively, during separate experiments. The first rf pulse interval, during which spin-spin relaxation occurs was kept short (10 to 20 ms). Field gradient calibration was carried out by the use of known self-diffusion standards (H2 O/D2 O mixtures) (18, 19). The resonance due to residual solvents (H2 O, HDO) at 4.8 ppm (20◦ C) was used as an internal reference. The PGSE method measures molecular motion by probing the change in spin-echo attenuation in a pulsed-field gradient. For molecules undergoing unhindered Brownian motion, and for a single diffusing species, the attenuation of the signal intensities is given by Ii /I (0) = exp(−k Di ).

193

Tensiometry Equilibrium surface tensions were carried out using the dropvolume (Lauda TVT1) method. A detailed description of the instrument and measurement procedure can be found elsewhere (20). The calibration was checked using double-distilled water. All the tensiometry measurements were carried out with freshly made solutions. Turbidity The cloud point temperature (CPT) was determined by simple visual observation of clouding upon a temperature increase. The samples, in NMR tubes, were heated in a thermostat bath. RESULTS AND DISCUSSION

The 1 H NMR spectra of a mixture of surfactants and cyclodextrins are complex. This is exemplified by the assignment of 1 H NMR spectra for a mixture of NIHG750 and HPβCD in aqueous solution (D2 O) at a 1 : 1 molar composition presented in Fig. 1 together with the chemical structures (15, 21–24). NMR Self-Diffusion The pulsed-field gradient NMR method has not been used extensively to study interactions between cyclodextrins and guest molecules, and there are only a few reported studies (25–27). However, the method has been used frequently to study micellar systems and solubilization (28, 29). The pulsed-field gradient spin echo decays for proton resonances in an aqueous mixture of NIHG750 and HPβCD, at 1.5% NIHG750 and a molar ratio of cyclodextrin to surfactant of 0.5, are presented as a function of k = (γ Gδ)2 (1 − δ/3) in Fig. 2. The decay for the surfactant methyl (1) and methylene (2) protons proved to be single component and consequently Eq. [2] could be applied. In order to evaluate the experimental

[2]

Here k = (γ Gδ)2 (1 − δ/3), where γ represents the magnetogyric ratio of the nucleus under observation (in this case 1 H) and Di the self-diffusion coefficient of species i. In mixtures of two diffusing species, with the same chemical shift and wellseparated self-diffusion coefficients a double exponential decay function can be applied: Ii /I (0) = p exp(−k D A ) + (1 − p) exp(−k D B ).

[3]

Here p is the fraction of species “A”, weighted by its spin relaxation attenuation. Under the condition that the relaxation times are the same for the two species, p equals the molar ratio of of protons from species “A”.

FIG. 2. PGSE spin echo decays in a mixture at molar ratio 0.5 (HPβCD : NIHG750), for NIHG750 at 3.7 ppm (d), 1.3 ppm (j) and 0.9 ppm (.), and HPβCD at 5.1 ppm (s) and 1.13 ppm (h) at 1.5% NIHG750. The lines represent fitted single or double exponential decays (Eqs. [2–3]).

194

´ ABRAHMSEN-ALAMI ET AL.

the viscosity of cyclodextrin may become important at higher concentrations. The self-diffusion coefficient observed for NIHG750 in the absence of cyclodextrin indicates the formation of micelles. The solid line passing through the data for the self-diffusion coefficients of NIHG750 in Fig. 3 represents a fit to a two-state model defined by (15, 33) Dobs = p D1 + (1 − p)D2 .

FIG. 3. Self-diffusion coefficients of NIHG750 (s) and HPβCD (h) in mixed (molar ratio 2 (HPβCD : NIHG750)) (closed symbols) and pure systems (open symbols) of NIHG750 and HPβCD. The lines are guides to the eye.

data for the ethylene oxide peak (5), on the other hand, a double exponential function, Eq. [3], had to be used. The fast component is though to originate from monomethyl-PEG that was not converted to surfactant in the synthesis (15). The slowly diffusing species has within experimental error, the same self-diffusion coefficient as the one obtained for the end group of the surfactant, above, and corresponds to a mean value of the self-diffusion of aggregated and nonaggregated surfactant in the fast exchange limit. The self-diffusion coefficient of NIHG750 was determined from the slow component of the ethylene oxide protons (5) or from the self-diffusion of the hydrocarbon methyl protons (1). All spin echo decays for HPβCD and HPγ CD proton resonances proved to be single exponential. The self-diffusion coefficient was in most cases determined from the resonances at 1.1 ppm (HA) or at 5.0 ppm (H1), Fig. 1. The concentration dependence of the self-diffusion coefficients of pure NIHG750 and HPβCD in aqueous solutions and mixtures with a molar ratio 2 (HPβCD:NIHG750) is presented in Fig. 3. At surfactant concentrations, above the CMC (0.04% w/w or 0.4 mM (15)) cyclodextrin seems to dissolve the micelles. The self-diffusion coefficient of NIHG750 is higher than that observed in the binary NIHG750 system and very close to that of HPβCD. Mixed aggregates are most probably formed. The observed self-diffusion coefficients of HPβCD in this region are significantly lower than that in the absence of NIHG750. This indicates that HPβCD disrupts the micellar aggregates, and that small mixed aggregates are present, and probably composed of one or a few HPβCD molecules and a surfactant unimer. The self-diffusion coefficient of HPβCD decreases with increased concentration in both the presence and the absence of surfactant. In binary water-HPβCD solutions this decrease may be explained by the fact that cyclodextrin itself aggregates through formation of dimers linked by hydrogen bonding: these aggregates then cause obstruction effects (31, 32). Alternatively,

[4]

Here p is the fraction of micellized surfactant and Dobs , D1 and D2 are the observed self-diffusion coefficient and the selfdiffusion coefficient of micelles and monomers, respectively. The dimensions of HPβCD should be quite close to those reported for βCD, for which the external diameter, height of the ˚ respectively torus, and cavity diameter are 15.4, 7.9, and 6 A, (2). One alternative way to obtain an approximate dimension of HPβCD is to use Eq. [5] to calculate its hydrodynamic radius (R H ) from the self-diffusion coefficient at infinite dilution. The Stokes-Einstein equation, which is valid in the absence of obstruction at low concentrations, can then be used (34): D=

kT . 6π η R H

[5]

Here k is the Boltzmann constant and η the viscosity of the solvent (D2 O). The HPβCD molecule is approximated as a sphere in this case. The self-diffusion coefficient of HPβCD extrapolated to zero (DC0 D = 19.2 × 10−11 m2 /s) gives an estimated ˚ which is close to the dimensions hydrodynamic radius of 9 A, reported in the literature, above (2). When HPβCD binds to the surfactant molecule, possibly by threading onto either the hydrophobic or the hydrophilic part of the surfactant, the complex would be expected to have a shape close to a cylinder or a prolate ellipsoid. For spheroidal aggregates the self-diffusion 0 coefficient at infinite dilution (Daggregate ) may be estimated from the Stokes-Einstein or Perrin equation (34, 35), 0 = Daggregate

kT F(X ), 6π ηb

[6]

where b is the short axis of a spheroidal aggregate and F(X ) is a function depending only on the axial ratio (X ), which is the length of the long axis (a) divided by the short axis (b), of the aggregate. F(X ) equals unity for a spherical aggregate, simplifying to Eq. [5], and may for a prolate ellipsoid be written as (35) ¤± £ F(X ) = ln X + (X 2 − 1)1/2 (X 2 − 1)1/2 .

[7]

Setting the short axis of the aggregate equal to the hydrody˚ or the literature value for the namic radius of HPβCD (9 A) ˚ above, the long axis external radius of the βCD torus (7.7 A), ˚ usof the complex can be estimated to be between 17 and 20 A ing Eqs. [6] and [7]. The self-diffusion coefficients of HPβCD

GEMINI-CYCLODEXTRIN INTERACTION

195

and NIHG750 in the mixed system extrapolated to infinite dilution (DC0 D = 15.0 × 10−11 m2 /s) were used as an approximation for the self-diffusion coefficient for the complex. The length of ˚ and one extended hydrophobic alkyl chain of NIHG750 is 13 A ˚ and the hydrophilic Me-PEG750 has coil radius of about 10 A ˚ (36, 37). Hence, the results an extended chain length of 56 A obtained may very well suggest an aggregate where HPβCD is threaded onto a NIHG750 molecule with its Me-PEG group mainly in its coiled conformation. At higher concentrations the excluded volume of the complex extends its diffusion path and thus reduces the self-diffusion by so-called obstruction effects which for a prolate shaped aggregates may be described as (35) · 0 1 − 2φ Daggregate = Daggregate

¸ 1 , X (F(X ))3

[8]

where φ is the volume fraction of obstructing particles. The axial ratios (1.9 to 2.6) used in the calculation at infinite dilution are far too low to explain the experimental data at higher concentrations (using Eq. [8] and approximating the density of the complex to be close to 1 g/cm3 ). The decreased self-diffusion coefficients at higher NIHG750 concentrations are probably a result of aggregation of the prolate-shaped surfactant-cyclodextrin complexes into larger aggregates with micelle-like structures, as suggested by the surface tension measurements described below. The two highest surfactant concentrations investigated suggest a spher˚ ical aggregate with a radius estimated to be larger than 16 A, when taking into account that monomeric complexes as well as micellized complexes contribute to the observed self-diffusion coefficients in this region. This radius is significantly smaller than the radius estimated for pure NIHG750 surfactant micelles ˚ (15) due to the bulky shape envisaged for the complex. (40 A) This behavior is similar to that observed with surface-active cyclodextrins, studied by Auz´ely-Velty et al., and suggests small aggregation numbers (38). In Figs. 4a and 4b self-diffusion coefficients in pure systems and mixtures of NIHG750 and HPβCD are presented as a function of HPβCD concentration. The effect of HPβCD on NIHG750 aggregation/micellization is clearly displayed in these figures. At 1.5% (w/w) NIHG750, Fig. 4a, the self-diffusion of NIHG750 in the absence of HPβCD is slow and to a large extent determined by the diffusion of micelles. By addition of HPβCD to the system, an increase followed by a subsequent decrease of the NIHG750 self-diffusion coefficient is observed. At higher HPβCD concentration the self-diffusion coefficients of HPβCD and NIHG750 coincide. The self-diffusion coefficient of HPβCD is lower in the presence than in the absence of NIHG750. This reflects the formation of complexes with a size slightly larger than that of the HPβCD molecule itself. The amount of surfactant bound to cyclodextrin can be estimated assuming a two-state relation, Eq. [4], and assuming the amount of free surfactant molecules to be negligible. The two states are surfactant molecules in micellar aggregates (Dmic ) and com-

FIG. 4. Self-diffusion coefficients of NIHG750 (d) and HPβCD (j) as a function of HPβCD at (a) 1.5% and (b) 7% (w/w) NIHG750. The HPβCD selfdiffusion coefficients in water are also included (h). The data are normalized with the observed self-diffusion coefficient of the surfactant. The lines are guides to the eye.

plexed with cyclodextrin (Dcomplex ). Furthermore, the observed self-diffusion coefficient of HPβCD is assumed to equal the selfdiffusion coefficient of the complex (Dcomplex ). This is probably less valid at high HPβCD contents where the free HPβCD also contributes to the observed self-diffusion coefficient. The insert in Fig. 4a shows the fraction ( p) of surfactant complexed with HPβCD as a function of the amount HPβCD in the system. At a molar ratio of 2 to 3 (HPβCD : NIHG750) practically all NIHG750 is bound to HPβCD, i.e., the bound fraction approaches unity. At high surfactant concentration (7% (w/w)), Fig. 4b, the observed self-diffusion coefficient of NIHG750 does not vary much upon addition of HPβCD. This indicates that the surfactant micelles, at least to some extent, stay intact. The NIHG750 and the HPβCD self-diffusion coefficients in the mixture coincide at molar ratios higher than 2 to 3 (HPβCD : NIHG750), suggesting complex formation. Note that a similar observation was made in the system with 1.5% (w/w) NIHG750. Cyclodextrin may bind to surfactant unimers to an increasing amount upon HPβCD addition but also to micelles both resulting in decreased

196

´ ABRAHMSEN-ALAMI ET AL.

FIG. 5. Self-diffusion coefficients of NIHG750 (s) and HPγ CD (h) as a function of NIHG concentration in mixed (molar ratio 2 (HPγ CD:NIHG750)) (closed symbols) and pure systems (open symbols) of NIHG750 and HPβCD. The lines are guides to the eye.

self-diffusion coefficients of HPβCD. From geometrical considerations, cyclodextrin has the possibility of interacting with the hydrophobic tail as well as with the hydrophilic polyethylene oxide chain. Since micelles apparently also form in the presence of HPβCD one possibility is that the interaction to some extent occurs also with the hydrophilic part, at least at the high surfactant concentration investigated, Fig. 4b. A strong interaction with both hydrophobic chains of the surfactant would result in micellar breakup. NMR self-diffusion coefficients in mixtures of NIHG750 and a cyclodextrin molecule with a slightly larger hydrophobic cavity, hydroxypropyl-γ -cyclodextrin (HPγ CD), are presented in Fig. 5. The molar ratio was kept constant at 2 (HPγ CD : NIHG750). Only a slight increase of the NIHG750 self-diffusion is observed upon addition of HPγ CD. The effect of surfactant on the self-diffusion coefficient of HPγ CD is also quite small. This suggests that HPγ CD has little effect on the micellar structure of NIHG750 compared to HPβCD, see Fig. 3. Figures 6a and 6b show the effect of HPγ CD on self-diffusion coefficients at fixed surfactant concentrations of 0.9 and 3.5% (w/w), respectively. The results suggest a small micellar breakup at the low NIHG750 concentration as the amount of HPγ CD increases. At a molar ratio of about 2 (HPγ CD : NIHG750) the NIHG750 self-diffusion coefficient reaches a plateau. The selfdiffusion coefficient of HPγ CD is about the same in the presence and absence of NIHG750, which may indicate that cyclodextrin does not complex to any significant extent with the surfactant. It is possible that the interaction occurs over a short time scale, which is however sufficiently long to disrupt the micellar aggregates. The reason for micellar breakup, to a large extent, may also be due to a change in the surrounding media and that the surfactant micelles are smaller than in pure aqueous systems (39). The micelles are intact at high surfactant concentration,

however. It is clear from Fig. 6b that the self-diffusion coefficient of HPγ CD is to a small, but significant, extent affected by the presence of micelles at higher concentrations. The observed difference in self-diffusion coefficient of HPγ CD in the absence and presence of NIHG750 may not only be an effect of the obstruction caused by the micelles since it should reduce the HPγ CD self-diffusion coefficient only by about 2%, as calculated from the relation D/D0 = 1/(1 + φ/2) (30). Instead, a low extent of complexation is suggested. At HPγ CD concentrations above 10% (w/w), corresponding to a molar ratio of 2 (HPγ CD : NIHG750), the self-diffusion coefficients start to decrease and at even higher concentrations, at a molar ratio of 3.5, approach the observed HPγ CD self-diffusion coefficient in the mixture. These results suggest that HPγ CD may bind more tightly to the micelles. However, it is not possible to draw any more detailed conclusions. At this relatively high HPγ CD concentration the viscosity of the solution surrounding the micelles, and the obstruction this causes, may be so high that the NIHG750 self-diffusion may be affected (30, 32).

FIG. 6. Self-diffusion coefficients of nonionic heterogemini surfactant (d) and HPγ CD (j) at (a) 0.9% (with 50 : 50 H2 O : D2 O as solvent) and (b) 3.5% (w/w) NIHG750 (with D2 O as solvent). The HPγ CD self-diffusion coefficients in water are also included (h). The data are normalized with the observed selfdiffusion coefficient of the surfactant. The lines are guides to the eye.

GEMINI-CYCLODEXTRIN INTERACTION

FIG. 7. The proton shift of resonances from (a) HPβCD H5 (d) and H6 (s) and (b) NIHG750 1(0.9 ppm) (d), 2a(1.3 ppm) (j), 2b(1.4 ppm) (m), 3(2.4 ppm) (s), and 6(1.6 ppm) (h) as a function of the content of HPβCD keeping NIHG750 content fixed at 1.5% w/w.

NMR Chemical Shift and Resonance Peak Shapes The formation of inclusion complexes of NIHG750 with HPβCD was evidenced by observing characteristic shifts of the HPβCD inner cavity 1 H resonance (H5) (Fig. 1) compared with those of pure macrocycle solutions (21, 24, 40). The chemical shift difference between cyclodextrin resonances (H5 and H6), see Fig. 1, in the absence of NIHG750 (δ0 ) and with NIHG750 (1.5% w/w) (δ) is presented as a function of the HPβCD concentration in Fig. 7a. In the present system the maximum chemical shift change is found at ratios of HPβCD:NIHG750 between 0.5 to 2. After this maximum the shift difference levels off. Similar observations, in particular upfield changes of β-CD cavity proton resonance in the presence of Triton X-100 surfactants, were reported by Smith et al. (9). The effect of HPβCD on the NIHG750 proton resonances was also investigated in aqueous solutions of 1.5% w/w NIHG750 with various concentrations of HPβCD. The changes in the NIHG750 proton resonances, δc , were generally quite small. Some of the changes observed, originating from the hydrocarbon

197

end group, are shown in Fig. 7b. The chemical shift is taken as the highest peak in the multiplets (when present). The chemical shift of the methyl protons (1, 0.9 ppm) at the end of one of the hydrocarbon chains changes only slightly (1δ = 0.02 upfield) upon addition of cyclodextrin. It should be noted that on micellization of NIHG750 in aqueous solution a larger chemical shift change in the opposite direction (1δ = 0.05 downfield) is observed. The upfield change of δc may therefore be taken as an indication of micellar breakup as cyclodextrin is added to the system, and possibly binding to HPβCD which is a less deshielding environment than the relatively more hydrophobic interior of the micelle. For other protons, such as the methylene protons (2b, 1.4 ppm), the resonance moves to lower fields upon addition of cyclodextrin, which is normally an indication of complex formation (41). Nevertheless, the changes in chemical shifts were too small, as is generally the case for aliphatic guests with small shielding effects, to be able to derive any information about the most probable average structure of the complexes in solution using the continuous variation method (24). The addition of HPγ CD to NIHG750 did not result in any measurable chemical shift changes indicating a weaker interaction. For both cyclodextrins no significant changes in chemical shifts were observed for resonances originating from the hydrophilic groups. This is expected since chemical shifts for the EO part of the surfactant in macromolecules as a result of complex formation are often difficult to observe. Not only do the chemical shifts of the NIHG750 hydrocarbon resonances change upon addition of cyclodextrin but a change in peak shape also occurs. In the absence of cyclodextrin the resonance peak (3) of the methylene proton next to the CN group is a clearly distinguished triplet (Fig. 8a). However, as HPβCD is added the triplet seems to separate into two triplets. Similar observations were made for the resonances of the methylene protons next to the previous protons (6) and the methyl protons (1) (Figs. 8b and 8c). The appearance of new peaks in the spectra may indicate that the exchange of NIGH750 is between two states with different chemical environments, a free state and as a complex bound to HPβCD, is sufficiently slow to give separate

FIG. 8. The effect of HPβCD on peaks 3 (a), 6 (b), and 1 (c), at a molar composition HPβCD : NIHG750 from bottom to top of 0, 0.24, 1, 3, 7, and 13. The NIHG750 content was 1.5% (w/w).

198

´ ABRAHMSEN-ALAMI ET AL.

FIG. 9. Nuclear Overhauser Enhancement results (at 27◦ C) for 1.5% NIHG750 and a molar composition of 7 (HPβCD : NIHG750). The 1 H NMR spectra of the sample at the bottom (a), and cross signals obtained after irradiating the H5 (b), 4 (c), and 3 (d) signals, respectively.

resonance frequencies. Similar observations have been reported for the NMR resonance lines of Triton X-100 in the presence of β-CD (9). In order to observe such a splitting of peaks the exchange rate between NIHG750 complexed with HPβCD to the noncomplexed state must be slow compared to the NMR time scale. The resonance frequency difference between the two peaks is of the order of 2 Hz and indicates an exchange time in the order of 0.05 s ((π 2 1ν 2 T2∗ /2)−1 ) (42). The exchange rate was sufficiently rapid, though, to observe a single exponential signal amplitude decay in the PGSE experiment, but this measurement is on a longer time scale (220 ms). Nuclear Overhauser Enhancement Intermolecular NOEs provide information about “through space” interactions in host-guest complexes (24). Direct information about proton distances may be obtained when they are closer to each other than a few angstroms. In Figs. 9a–d NOE measurements made on a mixture with 1.5% (w/w) NIHG750 and a molar ratio of 7 (HPβCD : NIHG750) (10.3% HPβCD (w/w)) are presented. Fig. 9b presents the NMR spectra obtained after selectively irradiating at a frequency corresponding to the H5 resonance at 3.75 ppm. After saturating this resonance, cross peaks are observed from both the hydrocarbon chain protons (0.8 to 1.5 ppm) and the ethylene oxide protons (3.7 ppm) of NIHG750, indicating these protons to be closer to the H5 protons than a few angstroms. One complicating factor that may arise when irradiating the H5, though, is that it has several close resonances that may also be

affected. Therefore, spectra were also recorded after selectively irradiating more well-separated signals originating from both the hydrophilic and the hydrophobic part of NIHG750, (4) and (3), respectively (Figs. 9c and 9d). In both cases cross peaks originating from HPβCD appear. The results indicate that the surfactant interacts both with the hydrophobic interior of HPβCD and with its exterior. Studies were also carried out on samples composed of NIHG750 and HPγ CD. This cyclodextrin has a slightly larger cavity size than that of HPβCD and the interaction with hydrocarbon chains is therefore expected to be weaker. In Figs. 10a–c NOE spectra for mixtures containing 1.7% (w/w) NIHG750 and HPγ CD at a molar ratio of 1.5 (HPγ CD:NIHG750) are presented. The results clearly show that cross peaks appear which would indicate closeness in space between protons of the hydrophilic part (peak 4) of the surfactant and HPγ CD (H6 at the rim of the hydrophobic cavity). Cross peaks observed with the hydrophobic part of the surfactant and HPγ CD are weaker than those observed for HPβCD as might be expected based on the NMR diffusion experiments above. Equilibrium Surface Tensions In Fig. 11, the variation in surface tension with concentration of NIHG750 is presented for the pure surfactant system, as well as for the system containing HPβCD in a molar ratio 2 (HPβCD : NIHG750). At NIHG750 concentrations lower than the CMC of NIHG750 the surface tension is about the same in the system with and without HPβCD. A decrease in surface

GEMINI-CYCLODEXTRIN INTERACTION

199

FIG. 10. Nuclear Overhauser enhancement studies (at 27◦ C) for 1.7% NIHG750 and a molar composition of 1 : 15 (HPγ CD : NIHG750). The 1 H NMR spectra of the sample at the bottom (a), and cross signals obtained after irradiating the H6 (b) and 4 (c) signals of the cyclodextrin and surfactant respectively.

tension is observed above the CMC of the pure NIHG750 system, however. This indicates that a complex forms which is more surface active than NIHG750 alone. At the same time, HPβCD seems to increase the CMC (to about 5 times higher concentration). This may be directly related to complexation with the strongly hydrated HPβCD molecule. The fact that a CMC exists in the presence of HPβCD indicates that interaction with the hydrophobic part of the surfactant is not so strong that it completely screens the surface activity of NIHG750 and that micelles may still form in the system. Therefore, it is possible that a complex

FIG. 11. The surface tension at molar ratios of HPβCD : NIHG750 equal to 0 (s) and 2 (d), and HPγ CD : NIHG750 equal to 2 (j). The lines are guides to the eye.

is partly formed with the hydrophilic part of NIHG750 or that some hydrophobic tails are uncomplexed and instead able to be incorporated in a micellar core. This results in a larger hydrophilic group for the complex, which promotes adsorption to the air-water interface compared to micellization (39). Nevertheless, the decrease in surface tension and the increase in CMC might also be explained as a solvent effect and is not necessarily due to an interaction between NIHG750 and HPβCD, since the surface tension of aqueous solutions of HPβCD is lower than that of pure water (39). However, this does not explain the fact that the surface tension does not decrease below the CMC, and that an increase in surface tension above the CMC is observed at higher additions of HPβCD, below. In Fig. 12a the surface tension measurements at two concentrations of NIHG750 (1% and 0.09% w/w (w/w)) are presented as a function of added HPβCD. Initially, a decrease in surface tension is observed upon addition of HPβCD. A maximum surface tension depression is observed at a molar ratio 4 and 2 (HPβCD : NIHG750), respectively, for the high and the low NIHG concentrations. Thereafter, the surface tension increases with added HPβCD. These changes in surface tension indicate interactions between the two components. A decrease in surface tension is probably associated with the formation of complexes that are more surface active than the surfactant itself, which give rise to a better packing at the air-water interface. The surface activity of the HPβCD itself (Fig. 12a) is probably a necessary condition for this interaction to take place (43). The increase in the surface tension at higher levels of HPβCD could be a consequence of complete solubilization of the hydrophobic part of the surfactant, or an increase in surfactant-complex volume

200

´ ABRAHMSEN-ALAMI ET AL.

FIG. 12. (Left) Surface tension as a function of HPβCD content at fixed NIHG750 contents 0.96 (s), 0.082 (d) and 0 (h)% (w/w) NIHG750. (Right) Surface tension as a function of HPγ CD content at fixed NIHG750 contents of 1.0 (s), and 0 (h)% (w/w) NIHG750. The lines are guides to the eye.

meaning a lower packing efficiency. Hodul et al. and Cserhati et al. proposed that the minimum in surface tension was due to an interaction with the hydrophobic part of nonionic surfactants that results in a more hydrophilic and bulkier complex (13, 14). This interpretation is possible also in the present system keeping in mind that a very strong and complete interaction with the hydrophobic part of the surfactant would screen the hydrophobic groups and result in a complex with much lower surface activity. In Fig. 11 the variation of the surface tension with concentration of NIHG750 is presented for the system containing HPγ CD in a molar ratio 2 (HPβCD : NIHG750). In contrast to the system with HPβCD, the surface tension is lower in the whole concentration range with HPγ CD than for the pure surfactant. A decrease in surface tension is also observed above the CMC (at 1% (w/w)) on addition of HPγ CD, Fig. 12b. The fact that the surface tension does not increase further after increasing HPγ CD indicates that a solubilization of the hydrophobic part of NIHG750 is not possible or not strong in this case. It is instead possible that the interaction to some extent occurs with the hydrophilic part of the surfactant. However, the decrease in surface tension and CMC may not necessarily be an effect of complex formation since NMR studies suggest a relatively weak interaction. It may instead be an effect of HPγ CD acting purely as an organic additive (44). It is known that addition of waterstructuring promoters such as sugars (fructose and xylose) to aqueous solutions of ethylene oxide-based nonionic surfactants markedly decreases surface tension and CMC (39).

the depression of the CPT is related to the hydrophilic lipophilic balance (HLB) of the surfactant, decreasing with decreased HLB (45, 46). In Fig. 13 the variation of the cloud point temperature of aqueous solutions containing NIHG750 and HPβCD in a molar ratio 2 and the CPT obtained for the surfactant only, are presented; the latter has been given previously (15). A clear decrease in the CPT is observed with added HPβCD. An interaction with the hydrophobic part of the surfactant would render the complex more hydrophilic and hence increase the CPT. The opposite behavior is observed, however, which strongly indicates that the interaction with the hydrophilic part of the surfactant is dominant in the whole concentration range at temperatures near CPT. This somewhat contradicts the results obtained by NMR, above, which indicate interactions with the hydrophobic part as well. However, this could be explained taking the model suggested by Karlstr¨om et al. into account (47, 48). This considers a change in the conformation of the ethylene oxide chain upon heating from a more hydrophilic conformation (all gauche) at low temperatures to a hydrophobic one (all trans) at higher temperatures. It is therefore possible that the interaction with the aliphatic chain is present at low temperatures, and as the polyethylene oxide chain gradually takes a more hydrophobic conformation, at higher temperatures, its interaction with cyclodextrin becomes more important: HPβCD screens this hydrophobicity. The hydrophobic part of the surfactant, on the other hand, becomes more soluble upon increases in temperature, which may make it less prone to interact with cyclodextrin. In the inset of Fig. 13 the variation of the CPT in systems containing 1% (w/w) of NIHG750 is presented as a function of added HPβCD, indicating an optimum in CPT depression at a molar ratio of about 2 (HPβCD : NIHG750). At the optimum the complex would hence be composed of an average of two HPβCD molecules per surfactant. At higher relative amounts of HPβCD the CPT increases probably as a result of increased interaction with the hydrophobic alkyl chains of the surfactant.

Temperature-Induced Phase Separation (Clouding) One important, and generally observed, property of nonionic surfactant solutions is separation into two phases, one surfactantrich and one surfactant-poor, upon heating. The so-called cloud point temperature characterizes this phenomenon. In general,

FIG. 13. The cloud point temperatures of mixtures of NIHG750 and HPβCD, at molar ratios NIHG750 : HPβCD equal to 1 : 0 (s) and 1 : 2 (d). Insert : 1% (w/w) NIHG750. The lines are guides to the eye.

201

GEMINI-CYCLODEXTRIN INTERACTION

Addition of HPγ CD to aqueous solutions of NIHG750 did not significantly affect the cloud point temperature, which indicates a weaker interaction in accordance with the NMR results. SUMMARY

From geometric considerations both HPγ CD and HPβCD cyclodextrins can interact with the hydrophobic tail and with the hydrophilic polyethylene oxide chain of NIHG750. The cross˚2 sectional areas of HPβCD and HPγ CD are about 30 and 50 A (2). Comparing this with the cross section of a hydrocarbon chain ˚ 2 )(36) and the EO chain (<25 A ˚ 2 )(49), one may conclude (20 A that it is possible to include only one hydrocarbon tail or EO chain into HPβCD, but two into HPγ CD. The NMR studies indicate that HPβCD interacts with NIHG750 and forms inclusion complexes. At low NIHG750 concentration the interaction eventually results in micellar breakup. However, micelles are still formed at a molar ratio of 2 (HPβCD : NIHG750) as shown by surface tension measurements. This is an important observation and gives NIHG surfactants an advantage compared to singletailed nonionic surfactants, which generally lose their ability to micellize at much lower molar ratios (≈1 for βCD/Triton X-100)(12) as a consequence of interaction with βCDs. This is probably related to the larger size of the hydrophobic group of NIHG750. NMR chemical shift and NOE studies indicate that interaction occurs mainly with the hydrophobic but also to some extent with the hydrophilic parts of NIHG750. NMR self-diffusion data indicate that a NIHG750 molecule complexes with at least 2–3 HPβCD molecules. This value is slightly lower than that reported in the literature, 2 cyclodextrin molecules per single-chain hydrocarbon surfactant (C h > 8) (12), and is probably due to the steric hindrance present for the hydrocarbon chains of NIHG750. The aggregates formed have a prolate shape with a short and a long axis in the order of 8–9 and 17– ˚ respectively, which at higher concentrations aggregate 20 A, into micelle-like structures. Similar rod-like complexes composed of β-CD and Pluronics have been suggested in the literature (43, 50). Further information about the structure of the aggregates may be obtained by computer modeling or experimental methods, such as small-angle neutron scattering, and form a good basis for further studies. At high NIHG750 concentrations, micelles are able to form in the presence of HPβCD, and it is possible that the interactions, to some extent, occur also with the hydrophilic parts of the surfactant. The interaction of native β-CD with oligo or polyethylene oxides has in most cases been shown to be very weak or absent. However, current research in our laboratory indicates an interaction of the polyethylene oxide part of Pluronic F127 with methylated β-CD (50). Fujita et al. (51) have also shown that β-CD interacts with ethylene oxide monomers in an end-blocked polyroxatane based on a copolymer of poly(ethylene glycol) and poly(propylene glycol) at room temperature. It is possible that the interaction with the adjacent hydrophobic groups of the surfactant promotes an interaction with the hydrophilic group.

Cloud point temperature measurements indicate that an interaction of HPβCD with the hydrophilic EO part of NIHG750 may exist. This is probably related to the change of the EO chain to a more hydrophobic conformation at elevated temperatures. The interaction of HPγ CD with NIHG750 seems to be relatively weak. NMR self-diffusion data indicate that the NIHG750 micellar structure is distrupted to a lesser extent upon addition of HPγ CD than for HPβCD. The interaction, when present, most probably takes place through inclusion of the hydrophilic EO part of NIHG750. NMR self-diffusion data at high concentrations of both NIHG750 and HPγ CD show an increase in aggregate size, which is a clear indication of such interaction (Fig. 6b). It has been suggested that HPγ CD is able to host two ethylene oxide chains in the same cyclodextrin cavity (49). Topchieva et al. (12) suggest that novel architectures, containing two hydrocarbon tails, may form based on nonionic single-tailed ethylene oxide surfactants and γ -CD. This type of structure is less probable for the HPγ CD/NIHG750 complexes. For HPβCD the interaction with the hydrophobic part of the surfactant may reduce the solubilization capacity of a system containing both surfactants and cyclodextrins. However, this is probably a weak interaction compared to the interaction envisioned for a solubilized sparingly soluble compound. HPγ CD, on the other hand, interacts only weakly with the hydrophilic EO part of the surfactant. This suggests that HPγ CD in combination with NIHG750 is a better alternative as a solubilizing system than a combination with HPβCD. ACKNOWLEDGMENTS We acknowledge the Swedish Research Council and the SNAP for financial support, and Dr Martin Murray for skillful assistance with the JEOL Alpha 500.

REFERENCES 1. Stella, V. J., and Rajewski, A., Pharm. Res. 14, 556 (1997). 2. Bekers, O., Uijtendaal, E. V., Beijnen, J. H., Bult, A., and Underberg, W. J. M., Drug Dev. Ind. Pharm. 17, 1503 (1991). 3. Loftsteinn, T., Pharm. Tech. 23, 40 (1999). 4. Cabrer, P., Alvares-Parilla, E., Meijide, F., Seijas, J. A., Nunez, E., and Tato, J., Langmuir 15, 5489 (1999). 5. Tan, Z. J., Zhu, X. X., and Brown, G. R., Langmuir 10, 1034 (1994). 6. Bender, M. L., and Komiyama, M., “Cyclodextrin Chemistry.” SpringerVerlag, New York, 1978. 7. Florence, A. T., and Attwood, D., “Pharmaceutical Physical Chemistry.” The Macmillan, Hong Kong, 1994. 8. Buschmann, H.-J., Cleve, E., and Schollmeyer, E., J. Incl. Phenom. Macrocyclic Chem. 33, 233 (1999). 9. Smith, V. K., Thilivhali, N. N., Munoz de la Pena, A., and Warner, I. M., J. Incl. Phenom. Mol. Rec. Chem. 10, 471 (1991). 10. Buschmann, H.-J., Jansen, K., and Schollmeyer, E., J. Incl. Phenom, Macrocyclic Chem. 37, 231 (2000). 11. Eli, W., Chen, W., and Xue, Q., J. Incl. Phenom. Macrocyclic Chem. 36, 439 (2000). 12. Topchieva, I., and Karesin, K., J. Colloid Interface Sci. 213, 29 (1999). 13. Hodul, P., Talaba, P., Srokova, I., Marcincin, A., and Peterova, M., Tenside Surf. Det. 34, 169 (1997). 14. Cserhati, T., and Szejtli, J., Carbohydr. Res. 224, 165 (1992).

202 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32.

´ ABRAHMSEN-ALAMI ET AL.

Alami, E., and Holmberg, K., J. Colloid Interface Sci. 233, 230 (2001). Stilbs, P., Prog. Nucl. Magn. Reson. Spectrosc. 19, 1 (1987). Tanner, J. E., J. Chem. Phys. 52, 2523 (1970). Callaghan, P. T., LeGros, M. A., and Pinder, D. N., J. Chem. Phys. 79, 6376 (1983). Holz, M., and Weingartner, H., J. Magn. Res. 92, 115 (1991). Dukhin, S. S., Kretzschamar, G., and Miller, R., “Dynamics of adsorption at liquid interfaces.” Elsevier, Netherlands, 1995. Ganza-Gonzalez, A., Vila-Jato, J. L., Anguiano-Igea, S., Otero-Esinar, F. J., and Blanco-M´endez, J., Int. J. Pharm. 106, 179 (1994). Demarco, P. V., and Thakkar, A. L., J. Chem. Soc., Chem. Commun. 2 (1970). Gonzalez-Gaitano, G., Sanz-Garcia, T., and Tardajos, G., Langmuir 15, 7963 (1999). Schneider, H. J., Hacket, F., and Rudiger, V., Chem. Rev. 98, 1755 (1998). Auz´ely-Velty, R., P´ean, C., Djeda¨ıni-Pilard, F., Zemb, Th., and Perly, B., Langmuir 17, 504 (2001). Lin, M., Jayawickrama, D. A., Rose, R. A., DelVisco, J. A., and Larive, C. K., Analytica. Chim. Acta 307, 449 (1995). Rymden, R., Carlfors, J., and Stilbs, P., J. Incl. Phenom. 1(2), 159 (1983). Lindman, B., Soderman, O., and Wennerstrom, H., in “Surfactant Solutions” (R. Zana, Ed.), Vol. 22, p. 295. Dekker, New York, 1987. Stilbs, P., in “Solubilisation in Surfactant Aggregates” (S. D. Christian and J. F. Scamehorn, Eds.), p. 367. Dekker, New York, 1995. J¨onsson, B., Wennerstr¨om, P. G., Nilsson, P. G., and Linse, P., Colloid Polym. Sci. 264, 77 (1986). Tsoucaris, G., and Rysanek, N., NATO Sci. Ser. Ser. C 519, 385 (1999). Paduano, L., Sartorio, R., Vitagliano, V., and Costantino, L., J. Solution Chem. 19, 31 (1990).

33. Lindman, B., Puyal, M.C., Kamenka, N., Rymden, R., and Stilbs, P., J. Phys. Chem. 88, 5048 (1984). 34. Hiemenz, P. C., “Principles of Colloid and Surface Chemistry,” Dekker, New York, 1986. 35. Jonstr¨omer, M., J¨onsson, B., and Lindman, B., J. Phys. Chem. 95, 3293 (1991). 36. Tanford, C., J. Phys. Chem. 76(21), 3020 (1972). 37. Devenand, S., and Selser, J. C., Macromolecules 24, 5943 (1991). 38. Auz´ely-Velty, R., Djeda¨ıni-Pilard, F., D´esert, S., Zemb, T., and Perly, B., Langmuir 16, 3727 (2000). 39. Rosen, J. M., “Surfactants and Interfacial Phenomena,” Wiley, New York, 1988. 40. Connors, K. A., “Binding Constants.” Wiley New York, 1987. 41. Matsui, Y., and Tokunaga, S., Bull, Chem. Soc. Jpn. 69, 2477 (1996). 42. Harris, R. K., “Nuclear Magnetic Resonance spectroscopy,” p. 123. Longman, Hong Kong, 1986. 43. Klamykin, A. A., Topichieva, I., and Zubov, N., Colloid Polym. Sci. 273, 520 (1995). 44. J¨onsson, B., Lindman, B., Holmberg, K., and Kronberg, B., “Surfactants and Polymers in Aqueous Solutions.” John Wiley, Chichester, UK, 1999. 45. Corti, M., Minero, C., and Degiorgio, V., J. Phys. Chem. 88, 309 (1984). 46. Gantz, G. M., in “Nonionic Surfactants” (J. J. Schick, Ed.), p. 736. Dekker, New York, 1967. 47. Karlstr¨om, G., J. Phys. Chem. 89, 4962 (1985). 48. Karlstr¨om, G., and Lindman, B., in “Organized Solutions” (S. E. Friberg and B. Lindman, Eds.), Vol. 44, p. 49. Dekker, New York, 1992. 49. Harada, A., Li, J., and Kamachi, M., Nature 370, 126 (1994). 50. Joseph, J., Ph.D. thesis, Bristol Univ., Bristol, 2001. 51. Fujita, H., Ooya, T., and Yui, N., Macromolecules 32, 2534 (1999).