ions

eptized withofr of these

ormed atincreasingthtrend toinimum in

ment may

es from a

Journal of Colloid and Interface Science 260 (2003) 82–88www.elsevier.com/locate/jcis

In situ preparation of weakly flocculated aqueous anatase suspensby a hydrothermal technique

Juan Yang,∗ Sen Mei, and José M.F. Ferreira

Department of Ceramics and Glass Engineering, CICECO, University of Aveiro, 3810-193 Aveiro, Portugal

Received 11 July 2002; accepted 10 December 2002

Abstract

Weakly flocculated aqueous anatase suspensions were prepared in situ by hydrothermally treating amorphous titania particles pdifferent amounts of tetraethylammonium hydroxide (TENOH). The simultaneous formation of hydrous TiO2 polyanions in the presenceOH− and tetraethylammonium cations are two essential conditions for the peptization process to occur. The absence of eitheconditions will cause reprecipitation. Transmission electron microscopy (TEM) revealed that the morphology of the particles flow TENOH concentrations consisted of well-dispersed anatase crystals, changing to asterisk-like structured particles withconcentrations of TENOH. Because of the extremely high absolute zeta potential (over−70 mV in all the samples) and ionic strengvalues, nontouching particle networks may be formed in situ in the mother solution in all samples, as predicted by DLVO theory. Acoagulation was observed in the suspensions with increasing concentrations of TENOH due to a more pronounced secondary mthe particle pair potential curves. Assuming the particles remained in the secondary minimum throughout the hydrothermal treatlead to the formation of the asterisk-like hard agglomerates. This may arise from the condensation of the –OH-rich TiO2 particles or fromthe deposition of material in the interparticle gap during the particle growth process. The green packing density of slip-cast bodisuspension containing 20 wt% solids was around 46%. 2003 Elsevier Science (USA). All rights reserved.

Keywords: Anatase; Hydrothermal; Peptization; DLVO; Weakly flocculated

1. Introduction

Well-dispersed suspensions are promising in the process-ing of highly reliable ceramics by colloidal processing tech-nology [1]. Most colloidally stable systems can be character-ized by a more or less strong dependence of the magnitudeof the repulsion forces on the interparticle distance. The tra-ditional DLVO theory, which consists of the sum of electro-static interaction and the London–van der Waals dispersionforces, is only valid when the separation distance betweenthe particles is relatively large. Non-DLVO forces such assolvation forces, structural forces, or hydration forces inaqueous media come into play when two particle surfacesapproach closer than a few nanometers [2–4]. The long-range electrostatic repulsive interparticle potentials can leadto the formation of dispersed slurries, which show improvedflow behavior and high packing capacity during consoli-

* Corresponding author.E-mail address: yangjuan@cv.ua.pt (J. Yang).

dation. However, recent research studies have shown thatshort-range repulsive potentials could yield weakly attractiveparticle networks, and thus weakly flocculated suspensions,which could be consolidated by pressure casting to packingdensities comparable to those achieved from stable suspen-sions [5–9]. With the advantage of the consolidated bodiesshowing less fragile behavior and allowing the formation ofmore complex shapes, several approaches have been usedto develop short-range repulsive potentials. For example,the addition of indifferent electrolytes (such as LiCl [10],NH4Cl [5], or tribasic ammonium citrate [6]) to various ox-ide slurries (Al2O3 [6], ZrO2 [10], or SiO2 [5]) has beenreported to decrease the thickness of the counterion cloudaround each particle, giving rise to short-range repulsive po-tentials. Bergström [9], and Colic and Lange [11] used theadsorption of small-chain molecules at the surfaces of Al2O3

and Si3N4 powders, respectively, to create short-range repul-sive potentials.

Hydrothermal synthesis, in which chemical reactionscould take place in aqueous or organic–aqueous media un-

0021-9797/03/$ – see front matter 2003 Elsevier Science (USA). All rights reserved.doi:10.1016/S0021-9797(02)00190-X

J. Yang et al. / Journal of Colloid and Interface Science 260 (2003) 82–88 83

der the simultaneous generation of pressure upon heating,has been used to prepare highly crystalline TiO2 particlesat low temperatures. Recent researches revealed that highlycrystalline anatase or rutile particles could be obtained byhydrothermally treating TENOH or HNO3 peptized tita-nia sols, respectively [12,13]. Well-packed microstructurescould be obtained from the peptized samples due to the lowdegree of aggregation among the particles with respect to theunpeptized samples [14].

In the present work, a method for in situ preparation ofweakly flocculated anatase suspension by hydrothermallytreating amorphous titania powder suspensions peptizedwith tetraethylammonium hydroxide (TENOH) was devel-oped. The addition of TENOH has a threefold role: (i) pep-tizing the amorphous titania, (ii) stabilizing the anatasesuspensions, and (iii) acting as an electrolyte to weakly floc-culate the suspensions.

2. Experimental

Titanium butoxide (Ti(OBu)4) (Aldrich, A.R. Grade) wasused as a Ti precursor without further purification. Amor-phous precipitates were obtained by adding dropwise a0.5-M isopropanol solution of the titanium butoxide intodeionized water ([H2O]/[Ti] = 150) followed by washingwith deionized water twice using a centrifuge. The whiteprecipitate was peptized at 70◦C by adding TENOH (Merck,A.R. Grade) at [TENOH]/[Ti] = 0.25, 0.50, and 1.00, cor-responding to the sample codes E1, E2, and E3, respectively.All the peptized sols were placed in Teflon containers afterfiltration and treated at 240◦C for 2 h using a heating rateof 3 ◦C/min. The resulting suspensions containing about1.5 wt% solids were washed with deionized water twice,followed by washing with absolute ethanol, using a cen-trifuge, and dried at 60◦C for further characterization. Thecentrifuged E3 sample was redispersed in the mother solu-tion in order to obtain a suspension containing 20 wt% solidsfor rheological characterization and consolidation of greensamples by slip casting.

The TiO2 powders were analyzed by X-ray diffraction(D/MAX-C, Rigaku), using CuKα radiation. The mor-phology of the particles from the mother suspensions wasexamined using a high-resolution transmission electron mi-croscope (Hitachi H9000-NA, Japan). The samples wereprepared by drying small amounts of the hydrothermallyderived suspensions on carbon-coated copper grids. Zetapotential measurements of the amorphous TiO2 precipi-tate under different concentrations of TENOH as electrolytewere performed using a zeta potentiometer (Coulter Delsa440 SX, USA). For comparison, a solution of 0.001 MKCl was also used as a background electrolyte. In orderto measure the zeta potential of the TiO2 after hydrother-mal treatment under the same ionic strength of the mothersuspensions, small amounts of centrifuged particles were re-dispersed into the respective decanted supernatants to form

dilute suspensions suitable for the measurements. Particlesize measurements of E2 at different pH values were alsocarried out with the same equipment (Coulter Delsa 440SX, USA), using a size mode. Infrared spectra of the pep-tized samples before and after hydrothermal treatment wererecorded in a FT-IR instrument (Mattson, 7000), using KBrpowder to dilute the solids and two crystalline silicon wafersto capture the sols.

Rheological behavior of the 20 wt% E3 suspension wasanalyzed with a controlled stress CSL rheometer (Carri50-MED, UK) at 20◦C, using a cone–plate system (angle0.02 rad, diameter 4 cm). The samples were presheared at arate of around 1000 s−1 for 1 min to break down any networkstructure, followed by 1 min at rest. The measurement wasperformed by stepping up to the highest stress within 1 minand stepping down to zero within 1 min. The microstructureof the fracture surface of the green body was observed undera scanning electron microscope (Hitachi S-4100, Japan).

3. Results and discussion

Colloidal sols were obtained by peptizing the suspensionof the amorphous TiO2 precipitate with TENOH at 70◦C.The change from a cloudy suspension to a colloidal sol oc-curred gradually, being faster with higher concentrations ofTENOH. This suggests that the previously formed amor-phous particles have been transformed into very small units,the size of which could approach the size of hydrated Ti ions.The transformation derived from the increased alkalinity ofthe media [15] with pH values of 11.0, 11.3, and 11.8 for thesamples E1, E2, and E3, respectively. However, under sim-ilar pH values, but in the presence of ammonia, it was notpossible to obtain clear sols from the same precipitate. Theonly difference between NH3·H2O and TENOH lies in thelarger size of TEN+ cations. Thus, this apparent transitionfrom amorphous TiO2 aggregates to colloidal sols seems tobe due mostly to the specific role played by the TEN+ ion inan alkaline environment, justifying the use of TENOH as apeptizer for TiO2 precipitates in previous works [12,16].

In an attempt to understand the peptizing role of theTENOH and its influence on the interface charge propertiesof the amorphous TiO2 precipitate, a series of zeta potentialmeasurements were carried out in the presence of differentconcentrations of TENOH, as well as in the presenceof 0.001 M KCl. The results of decreasing pH runs arepresented in Fig. 1. It can be seen that there is a trend forthe electrophoretic curves to shift toward lower pH valueswith increasing concentrations of TENOH. This suggeststhat the OH− ions rather than TEN+ cations were firstadsorbed on the surface of the amorphous titania precipitateto form hydrous TiO2 polyanions. Then the TEN+ cationswere attracted to the polyanions, peptizing the system bydestroying the amorphous structure and leading to theformation of the colloidal sol. It was observed that atthe highest TENOH concentration used, the units of TiO2

84 J. Yang et al. / Journal of Colloid and Interface Science 260 (2003) 82–88

Fig. 1. Plots of zeta potential versus pH for the amorphous TiO2 precipitateunder various TENOH concentrations.

peptized by TENOH tended to disappear along the time ofmeasurement in the alkaline region due to a dissolution-likeprocess. Zeta potential measurements were even impossibleto perform in that pH region if higher concentrations wereused. However, the peptization process became graduallyless significant as the pH approached the isoelectric point(IEP). A reprecipitation process was clearly noticed in theacid region after the IEP, where the particles had highpositive values of zeta potential.

These results clearly indicate that for the peptizationprocess to occur in the presence of TENOH two essentialconditions are required. One of them is the formation ofhydrous TiO2 polyanions and the other is the charge and thesize of TEN+ cations. These two requirements would startto interact with each other immediately after the TENOHwas added to the suspension of TiO2 precipitate, leadingto a gradual breakdown of the amorphous structure intoextremely small units, which can possibly approach the sizeof hydrated Ti ions. Obviously, this interaction dictated bythe different-sign charges of hydrated Ti anions and TEN+cations cannot occur for pH values lower than the IEPdue to the absence of TiO2 polyanions in the strong acidambient. This explains why a precipitate appears under theseconditions.

Figure 2 shows the FT-IR spectra of the peptized soland of the powder obtained by hydrothermally treating thepeptized sol. Two absorption bands at 540 and 640 cm−1,which correspond to the typical vibration peaks of anatase,can be observed. This indicates that anatase nuclei have al-ready been formed in the peptized sol of E2. The XRDspectrum of the powder obtained after drying the peptizedsol at room temperature also indicates the formation of veryfine anatase crystals (Fig. 3). The powders obtained by dry-ing the peptized sol exhibits very broad peaks correspondingto the main anatase peaks (Fig. 3a), while the powdersobtained by hydrothermally treating the TiO2 amorphousprecipitates (Fig. 3b) show a lower degree of crystallinity of

Fig. 2. FT-IR spectra of E2 peptized samples, before (a) and after (b)hydrothermal treatment.

Fig. 3. XRD patterns of samples: (a) E2 sol before hydrothermal treatment,(b) unpeptized TiO2 precipitate after hydrothermal treatment, (c) E2 solafter hydrothermal treatment.�, anatase;�, brookite.

anatase compared with the sample obtained from peptizedsol (Fig. 3c). All these results strongly suggest that the hy-drated Ti ions, which have been formed and maintained inan ordered arrangement with the help of the TEN+ cations,tend to condense in an ordered structure. The condensationrate is expected to increase when thermally activated. Inany case, the arrangement of the spatial distribution of thesespecies in the sol seems to be a key point for the formation ofa thermodynamically more stable crystalline structure. Thecondensation process would involve the release of molecularwater and TEN+ cations, which would remain in the solu-tion. The controlled release of water and TEN+ ions allowsconcomitant delivery of the hydrous titania units from theamorphous precipitates to form crystalline anatase particles.

Evidence for the release of the TEN+ ions during thecondensation process was given before through TG analysisof the E3 sample after hydrothermal treatment [17]. The totalweight loss of about 3 wt% up to 1200◦C was mostly dueto the evaporation of physically and chemically adsorbedwater, and only a very small part of the weight loss could

J. Yang et al. / Journal of Colloid and Interface Science 260 (2003) 82–88 85

Fig. 4. TEM morphology of anatase particles from the samples with increasing concentrations of TENOH (E3> E2> E1).

derive from the removal of the organic TEN+ cations. Thisindicates that most of the TEN+ cations would remainin solution. In this sense, the function of TENOH duringpeptization and hydrothermal treatment might be thought ascatalytic, accelerating the formation of crystalline anatase.

The morphology of the particles observed by TEM(Fig. 4) reveals that well-dispersed anatase nanocrystalswere obtained at the lowest TENOH concentration (E1).With increasing concentrations of TENOH, samples E2 andE3, even though some discrete crystals are observed in thesesamples, there is a trend to agglomerate formation, whichis more significant in E3 as asterisk-like particles form.A sharp increase in crystal size is observed when the con-centration of TENOH passes from E1 to E2, whereas aless noticeable increase of crystal size occurs with a furtheraddition to E3. This effect of increasing concentrations ofTENOH on particle size can be understood by consideringthat the adsorption density of the OH− species will in-crease and the peptized units will more resemble the anatase

structure as observed by IR (Fig. 2), while the concentra-tion of counterions also increases. As a consequence, deepersecondary minima are gradually formed at decreasing in-terparticle distances, as suggested in Fig. 5. This may leadto coalescence of neighboring primary particles during thegrowth process and to the formation of asterisk-like agglom-erates as Fig. 4 shows in the case of E3 concentration.

Particle sizes estimated from specific surface area (BET)are significantly larger than thed101 values estimated bythe Scherrer formula, as seen in Table 1. This differencearose from the formation of the agglomerates, whichmightbe formed through the condensation of –OH groups amongdifferent growing particles or to the deposition of material inthe interparticle gap during the particle growth process.

These results suggest that colloidal interactions amonggrowing particles would be determinants of both particlesize and shape. Table 1 shows the zeta potential values ofthe peptized particles of samples E1, E2, and E3, dispersedin the mother solutions with measured pH values of 11.0,

86 J. Yang et al. / Journal of Colloid and Interface Science 260 (2003) 82–88

Fig. 5. Pair potential curves for the hydrothermally prepared TiO2 particlesin their mother solutions under different TENOH concentrations.

Table 1Particle size and zeta potential data of the various samples

Sample E0 E1 E2 E3

TENOH/Ti 0.00 0.25 0.50 1.00Particle size (nm)d101

a 8.0 26.7 36.4 52.3dBET 10.8 51.8 93.5 134.1

Zeta potential (mV) –b −73.2 −81.4 −82.51/κ (nm) –b 1.4 1.0 0.7

a d101 is used as particle size for calculating the interaction energy.b This measurement was difficult to perform in the absence of an

electrolyte.

11.3, and 11.8, respectively. The absolute values of zetapotential are relatively high for all samples and increase firstfrom −73.1 mV for E1 to−81.4 mV for E2, becomingalmost unchanged with a further TENOH addition to E3.The first increase of zeta potential can be attributed tothe enhancement of the negative charge at the surfaces ofparticles due to adsorption of OH− groups, as well as to therelatively large size of the TEN+ counterions that promotethe formation of a thick electrical double layer. However,with a further addition of TENOH, the outer part of theelectrical double layer will become too crowded with TEN+ions and, as a consequence, its thickness will decrease due tothe screening effect. Accordingly, a possible increase in thesurface charge will not be accompanied by a concomitantincrease in zeta potential.

Therefore, the observed morphology of particles could beunderstood in terms of colloidal stability based on the classicDLVO theory, which accounts for the long-range repulsiveelectrostatic potential energy (VE) and the attractive vander Waals potential (VA) for the total potential energy. Theinteractions between two particles of equal radiusa with a

surface-to-surface separationh can be expressed as [18]

(1)VA = −A6

(2

s2 − 4+ 2

s2 + lns2 − 4

s2

), s = 2a + h

a,

(2)VE = 2πεrε0aψ20 ln

(1+ exp(−κh)),

where A is the nonretarded Hamaker constant,εrε0 isthe product of the dielectric constant of the solvent andthe permittivity of free space,Ψ0 is the surface potential(here, zeta potential was used instead, assuming approximatevalues between both), and 1/κ is the Debye–Hückel length,which can be estimated from the ionic strength for a 1:1electrolyte by the expression

(3)1/κ =(εrε0kBT

2ne2z2

)1/2

,

where kB is the Boltzmann constant,T is the absolutetemperature,n is the ionic concentration (which was approx-imately replaced by the TENOH concentration here),e is theelectron charge, andz is the valence of the ion pair.

The energy of interaction between two particles was esti-mated using a computer program developed by Linhart andAdair [19]. The calculated pair potentials for the particlesdispersed in the mother solutions are plotted in Fig. 5 usinga Hamaker constant of 5.65×10−20 J [20] and the other datalisted in Table 1. It can be seen that increasing concentrationsof TENOH cause the growth of the maximum electrostaticbarrier, the decrease of the range of repulsive forces, andthe appearance and accentuation of a secondary minimum ofseveralkT thermal energy. The strong repulsive barriers ob-served, especially in E2 and E3, which would increase withparticle size as reported by Look and Zukoski [3], preventsparticles from agglomerating into the primary van der Waalsminimum. The simultaneous increase of the repulsive barrierand of the depth of the secondary minimum with increasingconcentrations of TENOH lead to the in situ formation ofweakly flocculated suspensions without adding other surfaceactive agents such as dispersants or polyelectrolytes.

For sample E1, the combination of a high surface poten-tial and a low ionic strength will allow particles to approacheach other in a very shallow secondary minimum wherethe adhesion is too weak and easily reversible, as sug-gested by Israelachvili [2]. Under these conditions particlescan grow separately, as shown in Fig. 4. The secondaryminimum gradually deepens with increasing concentrationsof TENOH, allowing the primary particles to get closerduring the growth process. This closer approach and the ther-mal fluctuations occurring along the hydrothermal treatmentcould promote the condensation of –OH groups exiting atthe surfaces of different particles or the deposition of mater-ial in the interparticle gap during the particle growth process.The condensation might be favored when the surfaces of theparticles are rich in OH− groups. This explains why hard-agglomerated asterisk-like structures have been formed inthe samples E2 and E3.

This interpretation of the electrostatic effect of TENOHon particle size is also supported by the evolution of average

J. Yang et al. / Journal of Colloid and Interface Science 260 (2003) 82–88 87

particle/agglomerate size and/or zeta potential with pH. Infact, the measured average particle size and/or zeta potentialfor the sample E2, under the same concentration of TEN+(0.09 M), were 288 nm/−43.4 mV, 379 nm/−38.4 mV,509 nm/−32.1 mV, and 533 nm/−20.0 mV, for the pH val-ues 12.0, 9.0, 7.4, and 5.4, respectively. All of these resultssuggest that the electrostatic interactions would play a pre-dominant role in the stabilization of the suspension, althoughthe steric stabilization effect of TEN+ ions cannot be dis-carded. Future work needs to be done to clarify this point.

According to the discussion above, weakly flocculated ordispersed anatase suspensions can be prepared by carefullyadjusting the required amounts of TENOH added as pep-tizer to the amorphous titania powder suspensions. Besidescolloidal stability, the concentration of peptizer also affectsthe size and morphology of the anatase particles.

The influence of the dispersing conditions on the flowbehavior of the sample E3 is reported in Fig. 6 for suspen-

Fig. 6. Flow behavior of the suspensions containing 20 wt% of E3 particles.

sions containing 20 wt% solids. The slurry prepared from thedried powder without addition of TENOH exhibits the high-est viscosity at any shear rate, while the slurry obtained byredispersing the same dried E3 particles into the centrifugedmother solution shows flow behavior comparable to that ofthe slurry obtained by concentrating the mother suspensionof the E3 sample. This reduction of viscosity in the presenceof TENOH suggests that weakly flocculated suspensionscan be prepared by adding TENOH to the suspensions ofthe dried anatase powders or could be formed in situ fromthe hydrothermally treated mother suspensions which hadbeen peptized with TENOH. An electrostatic model canbe proposed for the stabilization since all the samples ex-hibit relatively high values of zeta potential. Figure 7 showsthat a homogeneous microstructure could be obtained byslip-casting from the suspension containing 20 wt% solidsdispersed into the mother solution. The achieved green pack-ing density was around 46%, which is fairly high consideringthe high specific surface area of the powder and the asterisk-like morphology of the particles.

4. Conclusions

Weakly flocculated aqueous anatase suspensions wereprepared in situ by hydrothermally treating titania amor-phous particles peptized with TENOH. We have shown thatthe addition of TENOH could perform a threefold func-tion, namely, peptizing the amorphous titania, stabilizing theanatase suspensions, and acting as an electrolyte to weaklyflocculate the suspensions. The peptization process in thepresence of TENOH implies the simultaneous formation ofhydrous TiO2 polyanions and the presence of large TEN+cations, which promote a dissolution-like process of the

Fig. 7. Fracture surface, as observed by SEM, of the green slip-cast body obtained from the 20 wt% E3-TENOH dispersed suspension.

88 J. Yang et al. / Journal of Colloid and Interface Science 260 (2003) 82–88

hy-t ofore

derin

y.hertionlesn ofth

m th

the000d to

ress,

m.

Am.

m.

7)

75

m.

. 81

60.

94)

ed.,

96.on,

of

amorphous precipitate. The as-formed extremely smalldrous TiO2 units, which inherit the structural arrangemenanatase, facilitate the formation of thermodynamically mstable crystalline structure at low temperature.

The size and morphology of the particles formed undifferent TENOH concentrations could be explainedterms of the classical DLVO theory of colloidal stabilitThe formation of asterisk-like agglomerates at the higTENOH concentrations could be due to the condensaof –OH groups existing at the surfaces of different particcaught in a deeper secondary minimum or the depositiomaterial in the interparticle gap during the particle growprocess. Homogeneous green bodies were obtained froweakly flocculated aqueous anatase suspensions.

Acknowledgments

The first and second authors gratefully acknowledgeFCT, Portugal, for grants under Contracts SFRH/3553/2and SFRH/6648/2001. The authors are also indebteDr. Augusto L.B. Lopes for his help with the TEM.

References

[1] F.F. Lange, J. Am. Ceram. Soc. 72 (1989) 3.

e

[2] J. Israelachvili, Intermolecular and Surface Forces, Academic PLondon, 1991.

[3] J.-L. Look, C.F. Zukoski, J. Am. Ceram. Soc. 78 (1995) 21.[4] H. Kamiya, M. Mitsui, H. Takano, S. Miyazawa, J. Am. Cera

Soc. 83 (2000) 287.[5] J.J. Guo, J.A. Lewis, J. Am. Ceram. Soc. 83 (2000) 266.[6] E.P. Luther, J.A. Yanez, G.V. Franks, F.F. Lange, D.S. Pearson, J.

Ceram. Soc. 78 (1995) 1495.[7] W.A. Ducker, Z. Xu, J.N. Israelachvili, D.R. Clarke, J. Am. Cera

Soc. 77 (1994) 437.[8] M. Colic, G.V. Franks, M.L. Fisher, F.F. Lange, Langmuir 13 (199

3129.[9] L. Bergström, C.H. Schilling, I.A. Aksay, J. Am. Ceram. Soc.

(1992) 3305.[10] S. Klein, M. Fisher, G. Franks, M. Colic, F.F. Lange, J. Am. Cera

Soc. 83 (2000) 513.[11] M. Colic, G. Franks, M. Fisher, F.F. Lange, J. Am. Ceram. Soc

(1998) 2157.[12] J. Yang, S. Mei, J.M.F. Ferreira, J. Am. Ceram. Soc. 83 (2000) 13[13] R.R. Bucsa, M. Grätzel, J. Am. Ceram. Soc. 79 (1996) 2185.[14] K.-N.P. Kumar, J. Kumar, K. Keizer, J. Am. Ceram. Soc. 77 (19

1396.[15] R.H. Perry, C.H. Chilton, Chemical Engineers’ Handbook, 5th

McGraw–Hill, New York, 1973.[16] A. Chemseddine, T. Moritz, Eur. J. Inorg. Chem. (1999) 235.[17] J. Yang, S. Mei, J.M.F. Ferreira, J. Am. Ceram. Soc. 84 (2001) 16[18] R.J. Hunter, Foundations of Colloidal Science, Vol. 1, Clarend

Oxford, 1995.[19] R.V. Linhart, J.H. Adair, Stabil 4.5 for Windows, Department

Materials and Engineering, University of Florida, 1996.[20] L. Bergström, Adv. Colloid Interface Sci. 70 (1997) 125.