Adsorption of cellulose derivatives on flat gold surfaces and on spherical gold particles

Journal of Colloid and Interface Science 328 (2008) 20–28

Contents lists available at ScienceDirect

Journal of Colloid and Interface Science www.elsevier.com/locate/jcis

Adsorption of cellulose derivatives on flat gold surfaces and on spherical gold particles Masoud Amirkhani a , Sondre Volden b , Kaizheng Zhu a , Wilhelm R. Glomm b , Bo Nyström a,∗ a b

Department of Chemistry, University of Oslo, P.O. Box 1033, N-0315 Oslo, Norway Ugelstad Laboratory, Department of Chemical Engineering, Norwegian University of Science and Technology (NTNU), N-7491 Trondheim, Norway

a r t i c l e

i n f o

a b s t r a c t

Article history: Received 10 June 2008 Accepted 5 September 2008 Available online 10 September 2008 Keywords: QCM-D Polymer adsorption Dynamic light scattering Gold particles Layer thickness

The adsorption of hydroxyethylcellulose (HEC), ethyl(hydroxyethyl)cellulose (EHEC), and their hydrophobically modified counterparts HM-HEC and HM-EHEC has been studied on planar gold and citratecovered gold surfaces by means of quartz crystal microbalance with dissipation monitoring (QCM-D), and on citrate-covered gold particles with the aid of dynamic light scattering (DLS). The QCM-D results indicate that larger amounts of polymer are adsorbed from aqueous solutions of HM-HEC and HM-EHEC on both substrates than from solutions of their unmodified analogues. The adsorption affinity for all the polymers, except EHEC, is higher on the citrate-covered surfaces than on the bare gold substrate. This indicates that more adsorption sites are activated in the presence of the citrate layer. The experimental adsorption data for all the polymers can be described fairly well by the Langmuir adsorption isotherm. However, at very low polymer concentrations significant deviations from the model are observed. The value of the hydrodynamic thickness of the adsorbed polymer layer (δh ), determined from DLS, rises with increasing polymer concentration for all the cellulose derivatives; a Langmuir type of isotherm can be used to roughly describe the adsorption behavior. Because of good solvent conditions for HEC the chains extend far out in the bulk at higher concentrations and the value of δh is much higher than that of HM-HEC. The adsorption of EHEC and HM-EHEC onto gold particles discloses that the values of δh are considerably higher for the hydrophobically modified cellulose derivative, and this finding is compatible with the trend in layer thickness estimated from the QCM-D measurements. © 2008 Elsevier Inc. All rights reserved.

1. Introduction The adsorption of polymers on solid surfaces is very important for practical purposes in different technological fields such as stabilization of colloidal dispersions, chromatography, development of biocompatible materials, and microelectronics. In materials sciences, it is used to control surface properties such as wetting, hardness, or resistance to aggressive environments. Depending on the application, two different approaches can be envisaged: either polymer grafting onto the surface by covalent bonds [1] or spontaneous physical adsorption [2]. The latter phenomenon is simpler to implement to implement, because no specific reaction sites are needed. However, the adsorbed polymer layer is easier to remove than a covalently bonded surface, and the adsorption approach puts constraints on the chemical nature and architecture of the polymers, which can be difficult to fulfill. Chemically dissimilar polymers usually have different affinities for a surface, which leads to preferential adsorption of the poly-

*

Corresponding author. E-mail address: [email protected] (B. Nyström).

0021-9797/$ – see front matter doi:10.1016/j.jcis.2008.09.013

© 2008

Elsevier Inc. All rights reserved.

mer with the highest segmental adsorption energy [3–5]. Polymer adsorption can be affected by many parameters such as the solvent power, charge density [6], molecular weight [7], conformation of the chains [3,4,8,9], and hydrophobicity of the polymer [10]. For highly charged polyelectrolytes the repulsion between the segments dominates the adsorption behavior [6]. Adsorption preference with respect to molecular weight occurs when macromolecules that differ only in chain length are adsorbed [7]. The conformation of chains at an interface can also have an impact on the adsorption characteristics of the surface [3,4,8,9]. In this work, we have focused on the effect of polymer hydrophobicity on the adsorption of polymer on flat gold or citrate-covered gold surfaces, and binding of polymer onto citrate-covered gold particles. To elucidate the effect of hydrophobicity on the adsorption features, non-ionic cellulose ethers and their hydrophobic modified analogues have been employed in this study. Non-ionic cellulose derivatives are of substantial industrial interest as they have found applications as stabilizers in colloidal dispersions [10], additives for improving rheological properties in aqueous solutions [11], as potential drug delivery devices [12–14], and in other areas [15]. An interesting category of amphiphilic

M. Amirkhani et al. / Journal of Colloid and Interface Science 328 (2008) 20–28

Fig. 1. Schematic illustration of the chemical structures of HEC, HM-HEC, EHEC, and HM-EHEC.

polymer systems are those that undergo phase transitions in response to external stimuli such as temperature and pH [16], wherein, for example, their solubility decreases as an effect of increased temperature, causing macroscopic phase separation at sufficiently high temperatures [17]. Thermoresponsive polymers such as ethyl(hydroxyethyl)cellulose (EHEC) have been widely investigated for drug delivery purposes, separations, and diagnostics applications [16]. Moreover, by modifying this type of polymer so that its hydrophobicity is enhanced, an increase in viscosity can be observed [18,19] as a result of more intense intermolecular associations. Hydroxyethylcellulose (HEC), on the other hand, is essentially a hydrophilic polymer and this cellulose derivative displays rather weak thermoresponsive properties. It has been established in previous papers that a large amount of EHEC is adsorbed onto polystyrene latex particles [20–22]. If we take into account that the EHEC polymer contains hydrophobic microdomains in addition to regions of hydrophilic nature, one would expect the adsorption behavior to be altered when surfaces of different hydrophobicity are employed and the polymer is exposed to further hydrophobic modification. The EHEC polymer involved in this study was modified by grafting branched nonylphenol groups onto the polymer backbone in a ratio of 1.7 mol%, thus producing a hydrophobically modified ethyl(hydroxyethyl)cellulose (HM-EHEC) that still is water-soluble [23]. In addition to these polymers, the hydrophilic hydroxyethylcellulose (HEC) and its hydrophobic analogue (HMHEC), modified with glycidyl hexadecyl ether, have been studied. The chemical structures of all studied polymers are displayed in Fig. 1.

21

Earlier studies [24–28] on the adsorption of cellulose derivatives by means of ellipsometry and surface force experiments show that both the adsorbed amount and the conformation of the adsorbed polymers are sensitive to temperature changes and hydrophobicity of the surface. In both case it was found that hydrophobic interactions between the polymer and the surface constitute the main driving force for adsorption. One of the major drawbacks of ellipsometry and surface force techniques is that ellipsometry does not give detailed information about the conformation of the adsorbed polymers and the surface force method does not provide the adsorbed amount of polymer. One of the major drawbacks of ellipsometry and surface forcetechniques, is that ellipsometry does not give detailed information on the conformation of the adsorbed polymers and surface force does not give adsorb amount. In this investigation, the quartz crystal microbalance technique with dissipation monitoring (QCM-D) technique has been employed in order to obtain additional information about the adsorption properties of cellulose derivatives at hydrophobic surfaces and the effect of polymer concentration. In the studies mentioned above, changing the temperature altered the hydrophobicity of the systems, whereas in the present work a more straightforward approach is employed where the effect of hydrophobicity is disclosed by comparing the adsorption features of cellulose derivatives and their hydrophobically modified counterparts. The adsorption characteristics of HEC, HM-HEC, EHEC, and HMEHEC on flat gold and citrate-covered gold surfaces have been monitored with the aid of the quartz crystal microbalance with dissipation monitoring (QCM-D) technique. In addition, dynamic light scattering (DLS) was utilized to probe the hydrodynamic layer thickness of adsorbed cellulose derivatives on citrate-covered gold nanoparticles (GNP). Analyses performed with QCM-D provide an excellent tool for in situ studies of molecular adsorption at solid planar surfaces. Available information about systems subjected to this type of analyses includes reversibility of adsorption, adsorption kinetics and surface coverage, as well as layer structure/conformation information, all within the same simple experimental setup [29]. DLS is a well-established technique to monitor the hydrodynamic thickness of the adsorbed polymer layer onto nanoparticles in solution [30]. The spatial extension of an adsorbed layer can be determined by DLS via the hydrodynamic radius of particles with and without the adsorbed layer. DLS is routinely used to characterize layer thickness, with a noticeable sensitivity to long tails due to their influence on hydrodynamics. Adsorption protocols applied in this study involved aqueous solutions of HEC, HM-HEC, EHEC and HM-EHEC on gold-coated quartz crystals, citrate-buffered solutions of the above-mentioned polymers onto flat gold surfaces and on GNP. The aim of this work is to elucidate the effects of polymer structure and hydrophobicity on the adsorption features on planar and curved surfaces. 2. Materials and methods 2.1. Materials In the present work, a HEC sample with the commercial name Natrosol 250 GR (Lot. No. A-0382) acquired from Hercules, Aqualon Division, was used as the hydrophilic polymer and served as the precursor for the synthesis of the hydrophobically modified analogue (HM-HEC). The degree of substitution of hydroxyethyl groups per repeating anhydroglucose unit for HEC was 2.5 (given by the manufacturer), and the weight-average molecular weight was determined to be M w = 400,000 by intensity light scattering [31]. The hydrophobically modified analogue was synthesized according to a standard procedure [32], and the fine points of the synthesis as well as the details of the characterization of this sample

22

M. Amirkhani et al. / Journal of Colloid and Interface Science 328 (2008) 20–28

have been reported previously [31]. The chemical structure and purity of the HM-HEC were ascertained by 1 H NMR and the degree of hydrophobic modification (glycidyl hexadecyl ether groups, n-C16 H33 ) was determined from the peak ratios between the anisomeric protons and the methyl protons of the glycidyl hexadecyl chain (see reference [31]). The degree of substitution of the hydrophobic groups determined from NMR is 1 mol%. The EHEC and HM-EHEC samples were both supplied from Akzo Nobel Surface Chemistry AB, Stenungsund, Sweden. Before use the polymers were purified as described elsewhere [33]. Both the unmodified EHEC and the hydrophobically modified one (HM-EHEC) are ethyl(hydroxyethyl)cellulose ethers with the same molecular weight (M w ≈ 100,000), and the degree of substitution of ethyl and hydroxyethyl groups are DSethyl = 0.8 and MSEO = 1.8, respectively. The values of DS and MS correspond to the average number of ethyl and hydroxyethyl groups per anhydroglucose unit of the polymer. The manufacturer gave all the values stated here. The hydrophobically modified polymer is equivalent to the EHEC sample, but with branched nonylphenol chains grafted onto the polymer backbone. The degree of substitution was determined to be 1.7 mol% (ca. 6.5 groups per molecule) relative to the repeating units of the polymer by measuring the absorbance of the aromatic ring in nonylphenol at a wavelength of 275 nm. In order to remove salts and low molecular weight impurities, dilute solutions of all polymers were thoroughly dialyzed against Millipore water (ca. 7 days) before the cellulose was isolated by lyophilization. Regenerated cellulose with molecular weight cutoff of about 8000 (Spectrum Medical Industries) was utilized as dialyzing membrane. All the polymers have broad molecular weight distributions (M w / M n > 2), which is usually the case for most polysaccharides. When introducing a citrate solution to a gold surface, these molecules adsorb reversibly because of weak interaction with the support (data not shown). The pH of the citrate buffer was measured to be 8.55, and since citrate has pK a values of 3.14, 4.77 and 6.39 the molecules were assumed to be present in their anionic forms [34]. Contact angles were measured to be about 80◦ on quartz crystals with gold surfaces, and approximately 40◦ for citrate-coated quartz crystals with gold surfaces. When the materials are arranged from the most hydrophilic to the least hydrophilic surface, the coated quartz crystals can thus be ranked as Aucitrate > Au. Aqueous suspensions of GNP were purchased from Ted Pella, Inc., Redding, USA. The hydrodynamic radius of the gold particles measured by DLS is approximately 19 nm. The suspension of GNP was kept at 4 ◦ C and at this temperature the particles were stable with a virtually constant hydrodynamic radius for ca. 1 month. The GNP concentration in all final solutions is 3 × 1010 GNP/ml. The samples employed in the measurements were simply prepared by mixing polymer solutions of various concentrations with the suspension of nanoparticles at 25 ◦ C. 2.2. Quartz crystal microbalance with dissipation monitoring (QCM-D) Mass adsorption data were acquired at 25 ◦ C on a QCM-Z500 supplied by KSV Instruments Ltd., which monitors the reduction in vibration frequencies of a piezoelectric crystal when it is exposed to adsorbing species. The adsorption density (surface coverage) for each species was calculated using the Sauerbrey equation (Eq. (1)) [35] on data extracted from the 5th harmonic of the resonance:

m = −

C f n

,

(1)

where m is the change in mass, C is the sensitivity constant (17.8 ng cm−2 Hz−1 ) for the crystals used in this investigation,  f is the shift in resonance frequency for the respective harmonic

and n is the number of the overtone (n = 5). The surface coverage calculated by Eq. (1) is, however, to be considered as an approximation since there will be contribution from additional mass registered from water trapped within or associated with the adsorbed layer. Moreover, by applying values from all the harmonics in a measurement, layer thickness can be directly calculated from the software supplied by KSV [36], whereas the observed changes in dissipation ( D) provide information about the rigidity of the adsorbed layer. The dissipation factor D is defined by D=

E diss 2π E stor

,

(2)

where E diss is the total dissipated energy during one oscillation cycle and E stor is the total energy stored during the oscillation. With the QCM-Z500, the dissipation changes following adsorption such that  D = D − D 0 , where D 0 is the dissipation factor of the pure quartz crystal immersed in solvent. The Au-QCM quartz crystal has a fundamental frequency ( f 0 ) of 5 MHz and an active sensor area of 20 mm2 , was acquired from KSV Instruments Ltd. The gold-coated quartz crystals were cleaned in a Piranha mixture consisting of H2 SO4 (98%) and H2 O2 (30%) in a 3:1 ratio for approximately 30 min, followed by rinsing with Milli-Q water and finally the crystals were dried under a stream of N2 gas. All crystals were used immediately after preparation. The QCM-D measurements were performed in a sequential deposition protocol [37], wherein successive aliquots of increasing concentrations (2 × 10−4 –2 × 10−3 wt%) of either an aqueous (ultra purified water, pH 7.0) or citrate buffered (sodium citrate, 10 mM) solution of the cellulose derivative were added to the crystal surface, followed by desorption in either ultra pure water or buffer solution after equilibrium was reached for the highest concentration added. 2.3. Analyses of the QCM-D data By plotting  D against  f (D– f plot) for a specific harmonic of the resonance, information about the state of the adsorbed material can be obtained [38–40]. If the plotted curve follows a firstorder linear relationship, the system adsorbs in a specific way, with the adsorbate remaining in the same adsorbed state throughout the experiment. However, if the curve reveals a break point, it is most likely that a change in the structure of the adsorbed layer has occurred, e.g., in the transition from a rigid film to a more viscoelastic layer, or through a spreading of the adsorbate on the surface. Despite the impact this may have on the calculated adsorbed mass, the Sauerbrey equation was deemed applicable since the changes in dissipated factor were within 10−6 per 5 Hz change in frequency for all the systems [38,41]. Many different isotherm models, such as the Langmuir [41], the bi-Langmuir [42], the Freundlich [43], Freundlich–Jovanovic [44] and the Freundlich–Langmuir isotherms [45], have been proposed for the adsorption of polymers in a liquid solution onto a solid surface. Most of these models are essentially empirical although theoretical derivations have been accomplished in some cases. Among all these models, the Langmuir model is probably the most popular one due to its simplicity and its good agreement with many experimental data. The Langmuir adsorption isotherm was used to fit the data in the present study. In spite of the simplicity of its theoretical foundation, the meaning of the parameters that can be derived is rather simple, and makes it possible to discuss the physical basis of the adsorption process. However, we have to notice that the Langmuir model of adsorption involves strongly restrictive hypotheses regarding the mechanism of adsorption that are obviously not fulfilled in the case of polymer adsorption. But since the Langmuir model is very simple, it is used as a reference and experimental data are often compared to this model when it can be fitted to them. So such a model is nevertheless useful since it allows

M. Amirkhani et al. / Journal of Colloid and Interface Science 328 (2008) 20–28

23

the quantitative comparison of adsorption data on similar systems [46–50]. The mass values acquired from the Sauerbrey equation were accordingly fitted to the Langmuir adsorption isotherm:

Γadsorbate =

Γmax K [ S ]free , 1 + K [ S ]free

(3)

and the values of the binding constants for the cellulose derivatives in the different systems were estimated. In Eq. (3), Γadsorbate denotes the surface coverage of the adsorbate (HEC, EHEC, and their hydrophobically modified analogues), Γmax is the maximum surface coverage, and K is a binding constant equal to the ratio of adsorption and desorption rate constants (K = ka /kd ). [ S ]free denotes the concentration of the adsorbate remaining in the bulk solution [51]. The Langmuir isotherm model was applied with the understanding that the conditions upon which it is based, inter alia, adsorption in monolayers, reversibility of adsorption, equivalence of adsorption sites, absence of lateral interactions between molecules on the surface, have not strictly been fulfilled. However, calculated values of K using Eq. (3) enables comparison between adsorption affinities affinity to each surface. This would be especially true when comparing values within equal systems, i.e., for the same adsorbent and in the same solvent.

Fig. 2. Adsorption isotherms (mass per area as function of concentration) of HEC and HM-HEC in aqueous and citrate-buffered solutions on gold-coated quartz crystals. The solid curves are fitted with the aid of the Langmuir isotherm.

A smaller β value corresponds to a broader distribution of relaxation times. The mean value of the relaxation time is given by

2.4. Dynamic light scattering (DLS)

∞

DLS measurements were performed using an ALV/CGS-8F multidetector version compact goniometer system, with eight off fiberoptical detection units, from ALV GmbH, Langen, Germany. The light source is a Uniphase cylindrical 22 mW HeNe laser, operating at a wavelength of 632.8 nm, with vertically polarized light. The beam was focused on the sample cell (10-mm NMR tubes, Wilmad Glass Co., of highest quality) through a temperature-controlled cylindrical quartz container (with two plane-parallel windows), which is filled with a refractive index matching liquid (cis-decalin). The temperature was kept at 25.0 ◦ C with an accuracy of ±0.1 ◦ C during the experiments. The scattering angle θ was varied from 39◦ to 141◦ with 17◦ increments, corresponding to a range of scattering wavevectors (q) between 8.8 × 106 and 2.5 × 107 m−1 . The parameter q is determined from [52,53] q=

4π n

λ

sin

θ 2

(4)

,

where n is the refractive index of the solvent (1.33 for water at 25 ◦ C), λ is the wavelength of the laser, and θ is the scattering angle. In the DLS experiments, the normalized intensity time autocorrelation function g 2 (q, t ) was measured [52,53], g 2 (q, t ) =

 I (q, 0) I (q, t ) ,  I (q, 0)2

(5)

where I (q, t ) is the scattered intensity at a given q and time t. The g 2 (q, t ) function is related to the normalized electrical field correlation function g 1 (q, t ) by the Siegert relation [54] assuming that the system is an ergodic media [52,53],

2  g (q, t ) = 1 + B  g 1 (q, t ) , 2

(6)

where B is the coherence factor of the equipment. All the correlation functions in this work were fitted by a single stretched exponential function:

  β  t . g (t ) = exp − 1

τ

(7)

The stretched exponential function describes the decay processes that have a distribution of relaxation times (τ ). The parameter β (0  β  1) measures the width of the distribution function.

τ  ≡

  β    t τ 1 exp − dt = Γ , τ β β

(8)

0

where Γ is the gamma function. The mean relaxation times for all the samples are q2 -dependent, which is the sign of a diffusive process. For spherical particles, the hydrodynamic radius can be calculated via the Stokes–Einstein equation D=

kT 6π R h η

,

(9)

where k is the Boltzmann constant, T is the temperature, R h is the hydrodynamic radius of the particle, and η is the viscosity of the fluid. 3. Results and discussion 3.1. QCM-D In this section the adsorption isotherms and the change in dissipation on flat surfaces of different materials will be presented and discussed for the investigated polymers. The QCM-D results are discussed in the framework of the simple Langmuir model. The adsorption data have been collected at conditions where the adsorption equilibrium for all the polymer solutions has been achieved. 3.2. Adsorption of HEC and HM-HEC Adsorption isotherms from QCM-D for HEC and HM-HEC on flat surfaces of gold and citrate-covered gold are depicted in Fig. 2. The general trend is that the amount of adsorbed polymer initially increases with increasing polymer concentration and progressively the adsorption isotherms flatten out. It is obvious that HM-HEC has a higher propensity for adsorption on both substrates than HEC. It is shown that the adsorption data, especially at higher polymer concentrations, can for all systems be fairly well described by the Langmuir model (Eq. (3)), and the solid curves represent best fits to the data. However, a close inspection of the data discloses significant deviations from the Langmuir equation at low polymer concentrations. The model predicts a faster adsorption of the polymer than the weak concentration dependence observed at

24

M. Amirkhani et al. / Journal of Colloid and Interface Science 328 (2008) 20–28

Table 1 Normalized frequency shift of the 5th harmonic of the resonance and calculated values for the mass adsorbed and the dissipation-to-frequency ratio at 0.002 wt%, together with values for binding constants, surface coverage, and layer thickness at 0.002 wt% onto gold and citrate-covered gold surfaces calculated by the Langmuir model for HEC and HM-HEC Surface

Polymer

 f /n (Hz)

Mass (ng/cm2 )

| D /( f /n)| (×10−6 Hz−1 )

K (1/wt%)

Γmax (ng/cm2 )

Layer thickness (nm)

Au

HEC HM-HEC

21.2 ± 6.4 76.1 ± 7.4

377 ± 114 1355 ± 132

0.10 ± 0.03 0.13 ± 0.01

4043 1087

429 2228

4.2 16.5

Au + citrate

HEC HM-HEC

34.6 ± 2.0 80.0 ± 4.8

616 ± 36 1424 ± 85

0.16 ± 0.01 0.14 ± 0.01

3551 923

715 2763

7.9 17.2

low concentrations. The reason for this divergence may be that at these very low concentrations, the change of the bulk concentration upon polymer adsorption can be significant and this effect has not been accounted for in the use of the model. Another complication may be that at these low concentrations it can take a very long time before equilibrium has been reached. A conspicuous feature is the high amount of adsorbed HM-HEC, on both surfaces, in comparison with that of HEC. Previous adsorption studies [20–22] of aqueous solutions of EHEC and HM-EHEC have revealed that the more hydrophobic counterpart has usually a higher affinity to adsorb onto different substrates. An interesting feature is the fact that a higher surface coverage is observed for gold surfaces with citrate-layer, and this effect is more pronounced for HEC than HM-HEC (see Fig. 2). Citrate ions have a strong affinity for gold surfaces and are frequently employed to stabilize colloidal materials [55,56] by conferring a significant negative charge and thus an electrostatic repulsion between surfaces or particles. In addition to long-range Coulumbic interactions, short-range repulsive forces were found between a gold colloid and gold surface with periodic steps that appeared to be of the approximate dimensions of the citrate molecule. This indicates a secondary barrier to aggregation of colloids due to multiple layers of citrate ions. It is possible that this layer of citrate ions activate more adsorption sites on the gold surface for bonding of more HEC chains to the surface. In the case of HM-HEC, this enhanced surface affinity has less impact on the amount of adsorbed polymer because the hydrophobic groups occupy a lot of sites on the gold surface with high affinity for adsorption on gold, and only a low fraction of additional sites will become accessible on the citrate-covered surfaces. In Fig. 3, plots of the change in dissipation  D as a function of the variation of frequency for the 5th harmonic of the resonance − f /n are displayed for HEC and HM-HEC on gold and citrate-capped gold surfaces. The solid lines represent best fits to the data. The typical feature for HEC adsorbed on both substrates is the break points in the plotted curves, which may indicate that the nature of the adsorption process is changed, i.e., there are several kinetic regimes. For HEC adsorbed on the gold-citrate substrate, a stronger increase in  D with increasing polymer concentration is observed after the break point in the plot, whereas on gold  D falls off at high HEC concentrations after the break point. The adsorption feature for HEC on gold with a citrate-layer is interpreted as a sign of enhanced viscosity, i.e., the polymer first tend to flatten on the surface, before at higher concentrations the chains are forced to reorganize to accommodate more of the polymer on the surface. This is achieved when the chains are protruding out from the surface into the solution and thereby increases its viscous drag and the dampening of the system. When it comes to HEC on gold, the decrease in  D after the break point, together with the fact that this system has the highest affinity (highest value of the binding constant K ; see Table 1) between polymer and surface of the studied HEC and HM-HEC systems, indicates that the polymer chains rearrange and adsorb in a more flattened out conformation at higher concentrations. The lower values of Γmax and layer thickness (cf. Table 1) for HEC on gold surface announce that the

Fig. 3. Change in dissipation ( D) as a function of the change of frequency for the 5th harmonic of the resonance (− f /n) for HEC and HM-HEC.

adsorbed polymer layer is more compact on gold than on the goldcitrate surface. These results indicate that pre-coating of the gold surfaces with citrate leads to an overall higher surface-coverage for HEC than on the pure gold substrate. It should be mentioned that the concentration of citrate in the bulk solution was kept constant throughout the experiments, counteracting desorption of the passivating ligand emanating from changes in the equilibrium between surface-confined and bulk citrate. A complete displacement of citrate from the surface as a consequence of the introduction of a cellulose derivative is deemed unlikely, an assumption supported by the observation that HEC adsorbs in larger amounts on gold when added in citrate buffered solution (Table 1). These findings seem to indicate that the citrate coating of the surface activate more adsorption sites, and thereby facilitates the binding of HEC to the substrate. A different pattern of behavior is detected in the  D– f plots for adsorption of HM-HEC (Fig. 3b). In this case, no breaking points in the  D– f graphs are revealed and  D increases almost lin-

M. Amirkhani et al. / Journal of Colloid and Interface Science 328 (2008) 20–28

25

Table 2 Normalized frequency shift of the 5th harmonic of the resonance and calculated values for the mass adsorbed and the dissipation-to-frequency ratio at 0.002 wt%, together with values for binding constants, surface coverage, and layer thickness at 0.002 wt% onto gold and citrate-covered gold surfaces calculated by the Langmuir model for EHEC and HM-EHEC Surface

Polymer

 f /n (Hz)

Mass (ng/cm2 )

| D /( f /n)| (×10−6 Hz−1 )

K (1/wt%)

Γmax (ng/cm2 )

Layer thickness (nm)

Au

EHEC HM-EHEC

16.8 ± 1.1 29.9 ± 2.0

299 ± 20 532 ± 36

0.18 ± 0.04 0.19 ± 0.06

3968 863

470 1006

5 .9 5 .3

Au + citrate

EHEC HM-EHEC

20.0 ± 2.8 75.4 ± 7.5

356 ± 50 1342 ± 134

0.15 ± 0.03 0.21 ± 0.01

7001 5518

441 1837

3 .1 18.7

Fig. 4. Adsorption isotherms (mass per area as function of concentration) of EHEC and HM-EHEC in aqueous and citrate-buffered solutions on gold-coated quartz crystals. The data are fitted with the aid of the Langmuir isotherm.

early with increasing frequency. This feature suggests that the HMHEC chains do not undergo a significant conformational alteration in the course of the adsorption process. This indicates a situation where the polymer chains are progressively added to the surfaces without conformational changes of the adsorbed layer. Another scenario that cannot be excluded on the basis of these results is the formation of multilayers, where the hydrophobic moieties in the fist layer interact with hydrophobic groups of neighboring chains and thereby build-up a multiple-layer of polymer chains. Although the values of the layer thickness (Table 1) are large, it cannot, based on these results, be established that multilayers of HM-HEC are formed on these surfaces. By directly observing a systematic spreading of the harmonics in the frequency measurements, both HEC and HM-HEC were found to form viscoelastic rather than rigid films on the gold surfaces. On both types of surfaces, the adsorption results show that the values of the layer thickness as well as the amounts of adsorbed polymer are appreciably higher for the hydrophobically modified analogue. 3.3. Adsorption of EHEC and HM-EHEC EHEC is an amphiphilic polymer with hydrophilic and hydrophobic microdomains distributed randomly along the polymer backbone. The hydrophobically modified analogue is a polymer to which low amounts of bulky hydrophobic side chains have been grafted onto the backbone. Fig. 4 shows the adsorption isotherms for aqueous solutions of EHEC and HM-EHEC determined from QCM-D on gold and citratecovered gold. The adsorption isotherms for all the systems, except HM-EHEC on citrate-capped gold, approach a plateau-like region at higher polymer concentrations, corresponding to monolayer formation. The results for both EHEC and HM-EHEC show that the amount of polymer adsorbed onto the surfaces is larger for the citrate-covered gold, which again indicates that the citrate layer

stimulate the surface to accommodate more polymer. This finding is common for all the investigated cellulose derivatives. The enhanced affinity for polymer adsorption is much more pronounced for HM-EHEC. The fact that the amount of adsorbed polymer on both substrates is much higher for HM-EHEC emphasizes the importance of the attached hydrophobic tails for the adsorption process (see Table 2). It is clear from Table 2 that in spite of that both the amount of adsorbed polymer and surface coverage are higher for HM-EHEC than for EHEC on gold, a higher value of the layer thickness is observed for EHEC. This may be a sign of that a more compressed adsorption layer is formed for HM-EHEC on gold. A different picture emerges for the adsorption of EHEC and HMEHEC on citrate-covered gold. In this case the value of the layer thickness is much higher for HM-EHEC (Table 2), which may indicate that the HM-EHEC chains assume an extended conformation out into the bulk. This is attributed to a better solvent power in the citrate buffer and these amended thermodynamic conditions strengthen the excluded volume effect and promote extension of the polymer chains from the surface into the bulk solution. In Fig. 5, plots of  D versus − f /n have been constructed for the adsorption of EHEC and HM-EHEC on gold and citratecovered gold surfaces. The data for EHEC on both surfaces disclose break points in the graphs, with a stronger increase of  D at higher polymer concentrations. This suggests that a conformational change of the chains on the surfaces from a more compact layer to extended chains on the surface occurs at higher polymer concentrations. For HM-EHEC on gold, the value of  D initially falls off until the break point of the curve is reached. Since HM-EHEC is the most hydrophobic polymer and the gold surface is hydrophobic, this observation may announce a dehydration of the surface upon adsorption. Beyond the break point,  D increases and this also suggests a higher value of the layer thickness. As was the case for HEC on gold, the  D versus − f /n plot of HM-EHEC on citrate-covered gold reveals a lowered rate of dissipation change after the break point. It should be noted that the change in dissipation for HM-EHEC on gold covered with citrate is 2–4 times larger than for the other systems. Moreover the calculated layer thickness of HM-EHEC on citrate is 3–6 times higher (Table 2) than for EHEC or the other HM-EHEC system on the considered types of surfaces. This may advocate a type of adsorption where the chains are stacked tightly together so that their lateral movement on the quartz crystal is restricted, hence reducing the apparent damping of the system. The general picture that emerges is that HM-EHEC displays substantially higher values of Γmax on both surfaces as compared with the unmodified counterpart (Table 2). This difference in monolayer coverage (Γmax ) is reflected in the binding affinities of the cellulose derivatives to the different surfaces. All the studied cellulose derivatives are expected to possess randomly distributed adsorption domains, which causes the polymer to interact with the surface on multiple sites. This suggests that the binding affinity of the polymer is higher compared to species with fewer anchor points, and they are expected to adsorb in a more outspread fashion, limiting the number of chains able to bind to the surface. However, the observed features may be caused by the enhanced hydropho-

26

M. Amirkhani et al. / Journal of Colloid and Interface Science 328 (2008) 20–28

Fig. 5. Change in dissipation ( D) as a function of the change of frequency for the 5th harmonic of the resonance (− f /n) for EHEC and HM-EHEC.

bicity of the polymer, leading to a more rapid adsorption. It is reasonable to assume that these polymers display a loop-train-tail conformation on the surfaces in agreement with the conformation of most flexible polymers adsorbed on surfaces [57]. The loops and trains play an important role for the adsorbed layer in the QCM-D measurements, whereas the low density tails protruding out into the bulk play a less prominent role. 3.4. Dynamic light scattering In the DLS experiments, only gold particles covered with a citrate layer were investigated because bare gold particles without charges have a tendency to form aggregates and we tried to avoid this complication by using stabilized particles. DLS is a powerful technique with which to estimate the hydrodynamic thickness of the adsorbed polymer layer δh , which can be calculated by subtracting the bare particle radius from that of the polymer-covered particle. The value of δh is sensitive to the contribution from tails, which extend far into the solution, in contrast to other techniques, e.g., ellipsometry where trains and loops play an important role. Usually δh is much larger than the corresponding layer thickness obtained from other methods. The QCM technique provides directly the amount of polymer adsorbed to a planar surface, whereas the layer thickness is obtained from the model employed in the fitting of the QCM-D data. The layer thickness from QCM-D is essentially determined by the denser portions of the adsorbed layer near the interface, i.e., the train segments and loops, but the method is insensitive to the presence of far extended tails, which are at low segment concentrations.

Fig. 6. Plot of the first-order field correlation function versus time for citrate-covered gold particles in the presence of various concentrations of EHEC or HM-EHEC at a scattering angle of 90◦ . The inset plot shows the polymer concentration dependence of the mean relaxation time.

3.5. Hydrodynamic thickness of the adsorbed layer for HEC, HM-HEC, EHEC, and HM-EHEC The decay of the time correlation function can be well described by Eq. (7) for all the samples at the considered conditions. Fig. 6 shows time correlation function data (at a scattering angle of 90◦ ) for citrate-covered gold particles in the presence of various concentrations of EHEC or HM-EHEC. The general trend is that the tail of the correlation function is shifted toward longer times as the polymer concentration is raised. A very similar behavior is also observed for the GNP in the presence of HEC or HM-HEC (data not shown here). It is obvious from the inset that the mean relaxation time τ  increases monotonously with increasing polymer concentration for both systems, and the values of τ  are higher (larger particles) for the adsorption of the hydrophobically modified analogue. There are two possibilities for higher values of the relaxation time in suspensions of particles with added polymer, namely aggregation of particles or polymer adsorption on the particles. Gold nanoparticles are extremely sensitive to changes in aggregation state, and the onset of aggregation or flocculation is accompanied by a marked color change of the suspension. Such changes were not observed, either with the naked eye or with UV–vis. At the low polymer concentrations considered here, the bulk viscosity is only increased by 20% at the highest polymer concentration (0.1 wt%). In view of this, the viscosity of the solvent medium governs the diffusion of the colloidal particles. This suggests that the particles grow as the polymer concentration rises, i.e., the thickness of the adsorbed polymer layer increases. The fittings (by means of Eq. (7)) of the correlation functions for the GNP in solutions of different

M. Amirkhani et al. / Journal of Colloid and Interface Science 328 (2008) 20–28

27

concentrations the value of δh is higher for HEC than for HM-HEC, and this behavior can be rationalized in the following way. The solubility of HEC in water is much better than that of HM-HEC. At a high surface coverage of a polymer in a good solvent, the long tails are forced to extend due to crowding so the values of δh are larger than at lower surface coverage. For HM-HEC with much lower solubility in water, a more compact adsorbed layer is formed. It has previously been recognized from experimental studies [10,60,61] that on worsening the solubility, there is a concomitant tendency to increase the adsorbed amount and decrease the adsorbed layer thickness. For the adsorption of EHEC and HM-EHEC on the particles, a different picture emerges because both polymers are hydrophobic and water cannot be considered as a good solvent for either of the polymers. EHEC is characterized by randomly distributed hydrophobic microdomains. The results from QCM-D revealed that the adsorbed amount is higher for HM-EHEC than for EHEC and this finding is consistent with the higher values of δh observed for HM-EHEC as compared to the corresponding values of δh for EHEC (Fig. 7b). This finding indicates a higher affinity for HM-EHEC to adsorb onto the colloidal particles. In a previous study [21,22] on the adsorption of EHEC and HM-EHEC on different substrates, a similar trend was reported. The solid curves in Fig. 7 have been fitted to the data with the aid of a Langmuir inspired expression (δh = P K [ S ]free /(1 + K [ S free ]), where P is a constant). The overall picture is that the polymer concentration dependence of the hydrodynamic layer thickness can for all systems be fairly well described by the Langmuir type of relationship. 4. Conclusions

Fig. 7. The hydrodynamic thickness of the adsorbed layer as a function of polymer concentration for the cellulose derivatives indicated. The solid curves have been fitted with the aid of a modified version of the Langmuir equation (see text for details).

polymers and various concentrations disclose that the size distribution of the particles is rather narrow (β = 0.8–0.9). The value of β for the bare gold particles is 0.9, which demonstrates that the polydispersity of the GNP is practically not affected by the adsorption of polymer. As long as the polymers do not fully cover the particle surfaces (equilibrium adsorption is not reached), the probability of bridging to another particle can be high. However, since the value of β is virtually not affected by the addition of polymer the bridging flocculation does not constitute a serious problem. The relaxation mode is always diffusive (q2 -dependent), and via the Stokes–Einstein relationship (Eq. (9)) values of δh can be calculated. The general trend depicted in Fig. 7 for the adsorption of HEC and EHEC and their modified analogues onto citrate-covered gold surfaces is that δh initially increases strongly with polymer concentration and levels out at high concentrations. This is a typical behavior that has been reported [10,21,22,57–59] for the adsorption of polymers of various natures onto colloidal particles. For the HM-HEC system a plateau-like region (δh ≈ 50 nm) is reached already at a low polymer concentration (ca. 0.02 wt%), which suggests that the adsorption isotherm is of the high-affinity type. The hydrophobic moieties of HM-HEC promote enhanced adsorption and this explains why the adsorbed layer is thicker for HM-HEC than for HEC at low polymer concentrations. This is compatible with the finding from QCM-D, namely that the adsorbed amount of polymer on citrate-covered gold is larger for HM-HEC than for HEC. It should be noted that only low polymer concentrations are covered by the QCM-D measurements. However, at higher polymer

Quartz crystal microbalance with dissipation monitoring (QCMD) and dynamic light scattering (DLS) were employed to study the adsorption HEC, EHEC, and their hydrophobically modified analogous (HM-HEC and HM-EHEC) onto planar gold and citratecovered gold surfaces, and on citrate-covered gold particles. The QCM-D experiments disclosed that HM-HEC as well as HM-EHEC had higher propensity for adsorption on both substrates than their unmodified counterparts. This is reflected in the values of the surface coverage (ng/cm2 ) on gold surface (Γmax,HEC = 429, Γmax,HM-HEC = 2228, Γmax,EHEC = 470, Γmax,HM-EHEC = 1006) and on citrate-covered gold (Γmax,HEC = 715, Γmax,HM-HEC = 2763, Γmax,EHEC = 441, Γmax,HM-EHEC = 1837). This demonstrates that the hydrophobic segments of the polymer have a higher ability for adsorption, and a larger amount of polymer is adsorbed onto the surfaces. The trend for all the polymers, apart from EHEC, is that the adsorption affinity is higher on the citrate-covered gold surface than on the bare gold substrate. This suggests that more adsorption sites are activated in the presence of the citrate layer. The experimental adsorption isotherms for all systems can be fairly well portrayed by the Langmuir equation at higher polymer concentrations. However, at very low polymer concentrations significant deviations between the experimental QCM-D data and the Langmuir isotherm are observed. Graphical representations of  D versus − f /n reveal breaking points in the plots for the adsorption of HEC, EHEC and HM-EHEC on both types of substrates. This change in behavior can probably be attributed to conformational change of the polymer chains in the adsorbed layer as the polymer concentration increases. For solutions of HM-HEC, no break point in the plot is detected for either of the surfaces. This feature suggests that the mode of adsorption is not altered during the course of the adsorption process. From DLS measurements, the hydrodynamic thickness of the adsorbed polymer layer δh was determined for all the polymers on citrate-covered gold particles. For all polymers, the adsorption

28

M. Amirkhani et al. / Journal of Colloid and Interface Science 328 (2008) 20–28

isotherms could be reasonably well portrayed by a Langmuir type of expression. The results for the HEC and HM-HEC derivatives showed a thicker layer for HEC (δh ≈ 90 nm) than for HM-HEC (δh ≈ 50 nm) at high polymer concentrations. Water is a good solvent for HEC and because of the crowding effect at higher polymer concentrations the adsorbed polymer chains are extended and protrude out in the bulk. The solubility of HM-HEC in water is poorer, and in this case we expect a more compact layer around the particle. At very low polymer concentrations, the values of δh are higher for HM-HEC than the corresponding values for HEC, and this trend is consistent with the QCM-D results (δHEC = 7.9 nm and δHM-HEC = 17.2 nm). In this study, the QCM-D data cover much lower concentrations than the DLS experiments. For the adsorption of EHEC and HM-EHEC onto the gold particles, the values of δh are larger for HM-EHEC (δh ≈ 150 nm) than for EHEC (δh ≈ 90 nm) over the considered concentration interval. In this case, water is not a good solvent for neither of the polymers and the bulky hydrophobic groups play an important role for the layer thickness. The trend of δh for these systems is compatible with the layer thickness estimated from the QCM-D experiments (δEHEC = 3.1 nm and δHM-EHEC = 18.7 nm). Although the numerical values of the parameters differ noticeably because of the different surfaces and experimental techniques, the trends are the same. Acknowledgments The authors gratefully acknowledge the Research Council of Norway (NFR) for financial support within the NANOMAT program, project No. 163529/S10 and under the FRINAT program, project No. 177556/V30. References [1] R.S. Farinato, P.L. Dubin (Eds.), Colloid–Polymer Interactions: From Fundamentals to Practice, John Wiley & Sons, New York, 1999. [2] F. Caruso, R.A. Caruso, H. Möhwald, Science 282 (1998) 1111. [3] G.J. Fleer, M.A. Cohen Stuart, J.M.H.M. Scheutjens, T. Cosgrove, B. Vincent, Polymers at Interfaces, Chapman and Hall, London, 1993. [4] G.P. Van der Beek, M.A. Cohen Stuart, G.J. Fleer, Langmuir 5 (1989) 1180. [5] A. Silberberg, J. Chem. Phys. 48 (1968) 2835. [6] D.I. Gittins, F. Caruso, J. Phys. Chem. B 105 (2001) 6846. [7] D.F. Siqueira, J. Reiter, U. Breiner, R. Stadler, M. Stamm, Langmuir 12 (1996) 972. [8] A. Takahashi, M. Kawaguchi, Adv. Polym. Sci. 46 (1982) 1. [9] Yu.S. Lipatov, T.T. Todosijchuk, V.N. Chornaya, in: K. Esumi (Ed.), Polymer Interfaces and Emulsions, Marcel Dekker, New York, 1999. [10] M. Malmsten, F. Tiberg, Langmuir 9 (1993) 1098. [11] D.N. Schulz, J.E. Glass (Eds.), Polymers as Rheology Modifiers, American Chemical Society, 1991. [12] M. Scherlund, A. Brodin, M. Malmsten, J. Colloid Interface Sci. 229 (2000) 365. [13] K. Lindell, S. Engström, Int. J. Pharm. 95 (1993) 219. [14] L. Ryden, P. Edman, Int. J. Pharm. 83 (1992) 1. [15] R. Dönges, Br. Polym. J. 23 (1990) 315. [16] M. Malmsten, Surfactants and Polymers in Drug Delivery, Marcel Dekker, New York, 2002. [17] G. Karlström, A. Carlsson, B. Lindman, J. Phys. Chem. 94 (1990) 5005. [18] A.-L. Kjøniksen, S. Nilsson, K. Thuresson, B. Lindman, B. Nyström, Macromolecules 33 (2000) 877.

[19] K. Thuresson, S. Nilsson, B. Lindman, in: J.F. Kennedy, G.O. Phllips, P.A. Williams, L. Piculell (Eds.), Cellulose and Cellulose Derivatives: Physico-Chemical Aspects and Industrial Applications, Woodhead Publishing, 1995, pp. 323–329. [20] R.A. Lauten, A.-L. Kjøniksen, B. Nyström, Langmuir 17 (2001) 924. [21] R.A. Lauten, A.-L. Kjøniksen, B. Nyström, Langmuir 16 (2000) 4478. [22] A.-L. Kjøniksen, F. Joabsson, K. Thuresson, B. Nyström, J. Phys. Chem. B 103 (1999) 9818. [23] B. Nyström, K. Thuresson, B. Lindman, Langmuir 11 (1995) 1994. [24] M. Malmsten, B. Lindman, Langmuir 6 (1990) 357. [25] M. Malmsten, P.M. Claesson, E. Pezron, I. Pezrone, Langmuir 6 (1990) 1572. [26] M. Malmsten, P.M. Claesson, Langmuir 7 (1991) 988. [27] P.M. Claesson, M. Malmsten, B. Lindmant, Langmuir 7 (1991) 1441. [28] I. Pezron, E. Pezron, P.M. Claesson, M. Malmsten, Langmuir 7 (1991) 2248. [29] K.A. Marx, Biomacromolecules 4 (2003) 1099. [30] C. Mangeney, F. Ferrage, I. Aujard, V. Marchi-Artzner, L. Jullien, O. Ouari, E.D. Rekai, A. Laschewsky, I. Vikholm, J.W. Sadowski, J. Am. Chem. Soc. 124 (2002) 5811. [31] N. Beheshti, H. Bu, K. Zhu, A.-L. Kjøniksen, K.D. Knudsen, R. Pamies, J.G. Hernández Cifre, J.G. de la Torre, B. Nyström, J. Phys. Chem. B 110 (2006) 6601. [32] T. Miyajima, T. Kitsuki, K. Kita, H. Kamitani, K. Yamaki, US Patent 5,891,450, April 6, 1999. [33] K. Thuresson, G. Karlström, B. Lindman, J. Phys. Chem. 99 (1995) 3823. [34] R.C. Weast (Ed.), Handbook of Chemistry and Physics, CRC Press, Boca Raton, 1980. [35] G. Sauerbrey, Z. Phys. 155 (1959) 206. [36] T. Viitala, Modeling the Mechanical Properties of Adsorbed/Deposited Layers at Air–Solid and Liquid–Solid Interfaces with the Dissipative KSV QCM-Z500, KSV Instruments Ltd., Helsinki, 2005, pp. 1–33. [37] S.H. Brewer, W.R. Glomm, M.C. Johnson, M.K. Knag, S. Franzen, Langmuir 21 (2005) 9303. [38] F. Hook, M. Rodahl, P. Brzezinski, B. Kasemo, Langmuir 14 (1998) 729. [39] D.E. Otzen, M. Oliveberg, F. Hook, Colloids Surf. B Biointerfaces 29 (2003) 67. [40] X.D. Su, Y. Zong, R. Richter, W. Knoll, J. Colloid Interface Sci. 287 (2005) 35. [41] S. Farrokhpay, G.E. Morris, D. Fornasiero, P. Self, J. Colloid Interface Sci. 274 (2004) 33. [42] P. Sajonz, M. Kele, G. Zhong, B. Sellergren, G. Guiochon, J. Chromatogr. A 810 (1998) 1. [43] H. Li, L. Nie, S. Yao, Chromatographia 6 (2004) 425. [44] R.J. Umpleby II, S.J. Baxter, Y. Chen, R.N. Shah, K.D. Shimizu, Anal. Chem. 73 (2001) 4584. [45] C. Baggiani, G. Giraudi, C. Giovannoli, C. Tozzi, L. Anfossi, Anal. Chim. Acta 504 (2004) 43. [46] M.J. McGuire, J. Addai-Mensah, K.E. Bremmell, J. Colloid Interface Sci. 299 (2006) 547. [47] H. Kim, G. Guiochon, Anal. Chem. 77 (2005) 6415. [48] K. Eskilsson, B.W. Ninham, F. Tiberg, V.V. Yaminsky, Langmuir 15 (1999) 3242. [49] M. Raposo, L.H.C. Mattoso, O.N. Oliveira Jr., Thin Solid Films 327–329 (1998) 739. [50] T.C. Clancy, S.E. Webber, Macromolecules 28 (1995) 2561. [51] P.C. Hiemenz, R. Rajagopalan, Principles of Colloid and Surface Chemistry, Marcel Dekker, New York, 1997. [52] W. Brown (Ed.), Dynamic Light Scattering: The Method and Some Applications, Clarendon Press, Oxford, 1993. [53] B.J. Berne, R. Pecora, Dynamic Light Scattering, Wiley, New York, 1976. [54] A.J.F. Siegert, Rad. Lab. Rep. No. 465, Massachusetts Institute of Technology, 1943. [55] J. Turkevich, P.C. Stevenson, J. Hillier, Discuss. Faraday Soc. 11 (1951) 55. [56] J. Turkevich, P.C. Stevenson, J. Hillier, J. Phys. Chem. 57 (1953) 670. [57] K. Matsuura, A. Tsuchida, Y. Okahata, T. Akaike, K. Kobayashi, Bull. Chem. Soc. Jpn. 71 (1998) 2973. [58] F. Csempesz, K.F. Csáki, Langmuir 16 (2000) 5917. [59] S. Kapsabelis, C.A. Prestidge, J. Colloid Interface Sci. 228 (2000) 297. [60] T. van den Boomgaard, T.A. King, T.F. Tadros, H. Tang, B. Vincent, J. Colloid Interface Sci. 66 (1978) 68. [61] T. Cosgrove, T.L. Crowley, K. Ryan, J.R.P. Webster, Colloids Surf. 51 (1990) 255.