Heteroaggregation, repeptization and stability in mixtures of oppositely charged colloids

d)tability up. The other

e mixed

Journal of Colloid and Interface Science 278 (2004) 115–125www.elsevier.com/locate/jcis

Heteroaggregation, repeptization and stability in mixturesof oppositely charged colloids

M. Rasaa,∗, A.P. Philipsea, J.D. Meeldijkb,c

a Van ’t Hoff Laboratory for Physical and Colloid Chemistry, Debye Institute, Utrecht University, Padualaan 8, 3584 CH Utrecht, The Netherlandsb Department of Cell Biology, Utrecht University, Padualaan 8, 3584 CH Utrecht, The Netherlands

c Debye Institute, Utrecht University, Padualaan 8, 3584 CH Utrecht, The Netherlands

Received 10 November 2003; accepted 18 May 2004

Available online 15 June 2004

Abstract

We report a study of mixtures of initially oppositely charged particles withsimilar size. Dispersions of silica spheres (negatively chargeand alumina-coated silica spheres (positively charged) at low ionic strength, mixed at various volume ratios, exhibited a surprising sto compositions of 50% negative colloids as well as spontaneous repeptization of particles from the early-stage formed aggregatesmixtures were found to contain large heteroaggregates, which were imaged using cryogenic electron microscopy. Electrophoretic mobility,electrical conductivity, static and dynamic light scattering and sedimentation were studied as a function of volume fraction of thdispersions to investigate particle interactions and elucidate the repeptization phenomenon. 2004 Elsevier Inc. All rights reserved.

Keywords: Heteroaggregation; Heterocoagulation; Repeptization; Colloidal dispersions; Silica; Electrophoresis; Cryogenic electron microscopy

if-amear-car

cless

sig-bedize,ag-ag--on

lus-ormnt.tailrly

yingag-esdal

lly

iallyvelye di-gthon iswtheim-

on-het-duc-en-0%

1. Introduction

Aggregation in colloidal dispersions containing two dferent types of particles may take place between the stype of particles (homoaggregation) or between unlike pticles (heteroaggregation). Several studies have beenried out on systems containing oppositely charged parti[1–4]. It was shown[1] that the relative size of the particledetermines the morphology of aggregation: if there is anificant difference in size, the small particles are adsoronto the large particles but if particles have comparable sgrowth of large fractal clusters may take place. Heterogregation, which occurs together with or without homogregation was studied in[2] where the morphology of clusters was observed after performing electron microscopydried samples. However, in the process of drying the cter morphology may change and clusters may even fdue to capillary forces induced by a receding liquid froEarly stage fractal cluster formation was analyzed in dein [3]. Kinetics of heteroaggregation (mainly in the ea

* Corresponding author. Fax: +31-30-2533870.E-mail address: [email protected] (M. Ra¸sa).

0021-9797/$ – see front matter 2004 Elsevier Inc. All rights reserved.doi:10.1016/j.jcis.2004.05.020

-

stage) also received much attention[5–7]. In [7] heteroag-gregation versus homoaggregation was studied by varthe ionic strength. Simulation of growth of large heterogregates, determination of fractal size of heteroaggregatand the effect of added electrolyte were recently presentein [4]. Papers[8,9] review the experimental and theoreticstudies on heteroaggregation, respectively.

We report in this paper the surprising stability of initiaoppositely charged particle mixtures with a (initially) nega-tive dispersion content between 0 and 50%. Commercavailable aqueous dispersions of positively and negaticharged silica particles were used. The dispersions weralyzed against ethanol, so that only the low ionic strenregime is considered. In such a case, homoaggregatiinitially inhibited. After the mixtures were studied at loionic strength, salt effects were investigated. The size ofmixed spherical particles was approximately the same. Silar types of particles as in[3] were used.

The repeptization observed in the stable mixtures, csisting of spontaneous redispersion of particles fromeroaggregates, was investigated by electrophoresis, contivity measurements, dynamic light scattering and sedimtation experiments. The mixtures containing more than 5

116 M. Rasa et al. / Journal of Colloid and Interface Science 278 (2004) 115–125

5

S15

e of

Table 1Samples obtained by mixing the negative and positive particle dispersions

Sample S1 Se Sd Sc Sb Sa S2 S3 S4 S

Positive Ludox (ml) 0 0.001 0.01 0.025 0.05 0.1 0.25 0.5 0.75 1Negative Ludox (ml) 5 5 4.99 4.975 4.95 4.9 4.75 4.5 4.25 4Φd (%) 100 99.9 99.8 99.5 99 98 95 90 85 80

Sample S6 S7 S8 S9 S10 S11 S12 S13 S14

Positive Ludox (ml) 1.5 2 2.5 3 3.5 4 4.25 4.5 4.75 5Negative Ludox (ml) 3.5 3 2.5 2 1.5 1 0.75 0.5 0.25 0Φd (%) 70 60 50 40 30 20 15 10 5 0

The mass fraction of both dispersions was 4% (corresponding to 2 vol%).Φd is the volume of the negative particle dispersion relative to the total volummixtures, which was 5 ml.

rmami-the

pli-the

e voeren

m-rgedicailldis-thiningrti-itive

-with

lesf dis, an

ilica.o aith

X8nedis-eredinstap-en

he

rtu-r thededp-ntra-nlynstely

lf a

e-s iniongerd to

ins.wasandredwithnolionsrpm,pre-ades in

tion

-SAsur-ere

o-redval-

50void

negative dispersion are strongly aggregated. Cluster fotion was studied by cryogenic transmission electroncroscopy (cryo-TEM) because this technique imagesclusters in situ and avoids drying effects which may comcate the conventional TEM pictures used so far. Even incase of aggregated samples, our observations at extremume ratio of positive and negative particle dispersions wdifferent from those presented in[3], where the aggregatiowas terminated at an early stage.

2. Materials and methods

2.1. Sample description

The colloids in our experiments were originally comercial aqueous dispersions (DuPont) of negatively chasilica particles (Ludox HS-40) and positively charged silparticles (Ludox CL). In what follows these dispersions wbe referred to as the negative dispersion and positivepersion, respectively. The positive charge is due to alayer of alumina which covers the silica core. Accordto the suplier’s information, the average diameter of pacles is 12 nm. The counterions in the negative and posLudox dispersions are Na+ and Cl−, respectively, while theco-ions are OH− and H+, respectively. The study of mixtures of positive and negative particles was carried outsamples prepared as follows.

First series of mixtures (series I). It was previously ob-served[10] that ethanol addition to negative Ludox particflocculates the suspension because of the presence osolved silica which precipitates in ethanol. Consequentlyion exchange procedure is required to remove soluble sThe particles (contained in 200 ml of dispersion diluted tweight concentration of 20%) were successively mixed w100 g Dowex 50WX8 sulfonic acid resins and Dowex 1quaternary ammonium resin. The resins were conditioaccording to[11] and were stirred with the aqueous dpersion several hours. The resulting dispersion was filtto remove flocs, diluted with ethanol, and dialyzed agaethanol (technical grade) for one week. An attempt toply resins to Ludox CL failed: the dispersion gelated whit was mixed with the alkaline resins, probably because t

-

l-

-

positive Ludox particles are not stable at basic pH. Fonately, ion exchange turned out to be unnecessary fopositive Ludox particles; they are stable if ethanol is adin any amount, apparently because the alumina coating supresses the silica solubility. Consequently, the ion concetion in the positive particle dispersion could be reduced othrough dialysis. Mixtures of the two type of dispersiocontaining 4 wt% particles (corresponding to approxima2 vol%), were made according toTable 1. After injectingone dispersion into the other the mixture was stirred haminute.

A second series of mixtures (series II) was prepared tocheck the reproducibility of the observations made on sries I, and to obtain dust free samples for experimentwhich light scattering is involved. The positive disperswas subjected to dialysis as previously, but now for a lonperiod of 16 days. The negative dispersion was subjecteion exchange using AG 501-X8 (Bio-Rad) mixed bed resWe used 250 ml of the as-received dispersion, whichsubsequently diluted to a weight concentration of 20%,mixed with approximately 30 g of mixed resins and stirfor one hour. The procedure was repeated once morefresh resins. Then the dispersion was diluted with ethaand dialyzed against ethanol for one week. The dispersbased on ethanol were centrifuged three times (20001 h) to remove aggregates and dust. The mixtures werepared in a dust free environment and dilutions were musing filtered ethanol. The same composition scheme aTable 1and preparation method as in the previous secwere applied.

2.2. Experimental methods

Electrophoretic mobility of colloidal particles and conductivity of dispersions were measured with a DEL440SX (Coulter). Undiluted samples were used for meaing the conductivity while only the most turbid samples wdiluted in order to perform mobility measurements. The mbility was determined from a Lorentzian fit to the measupeaks at four scattering angles; the corresponding fourues of the mobility were averaged.

Static light scattering (SLS) was done using a FICAsetup. The aggregated samples were diluted in order to a

M. Rasa et al. / Journal of Colloid and Interface Science 278 (2004) 115–125 117

20

Table 2Dilutions with pure ethanol madefrom the original mixtures inTable 1, used for experiments in which light scattering was involved

Diluted sample Sb1 S31 S61 S62 S71

Proportion (ml mixture : ml ethanol) 0.4:10 0.2:10 0.2:10 1.5:10 0.25:10Used mixture Sb S3 S6 S61 S7

Diluted sample S72 S81 S82 S101 S131 S151

Proportion (ml mixture : ml ethanol) 1:10 0.5:12 1:4 1:20 1:20 1:Used mixture S71 S8 S81 S10 S13 S15

dedt filmi-in-

tupthe

am-

m-r (atof

mi-in-sedmi--pleaslterfiedntl vainis-h ad ationwassra-ple

par-

vo-icleforre-

el.

hesil-ses

dSa

3 areion

multiple scattering. The preparation of the mixtures avoidust as much as possible, so that the samples were notered prior to the measurements in order to keep thecrostructural properties unaltered. The wavelength of thecident linearly polarized light wasλ0 = 546 nm. Dynamiclight scattering (DLS) was done with a home made seusing a Malvern 7032 CE correlator (128 channels) onsame samples. Different degrees of dilutions of turbid sples were measured (seeTable 2). The wavelength of theincident linearly polarized light was 514.5 nm.

Sedimentation in gravitational field was studied on saples stored in a temperature controlled dark chambe20◦C). Cylindrical sedimentation tubes containing 5 mlsamples were used. Pictures were taken regularly.

Particle sizes were determined using an atomic forcecroscope (AFM) (Nanoscope IIIa, Digital Instruments),tapping mode, according to[12]. In order to image the cluster formation as well as single particles in the ethanol basamples, we used the cryogenic transmission electroncroscopy (cryo-TEM). A “Quantifoil” carbon film was subjected to glow discharge to be made hydrophilic. The samwas prepared in a so called vitrification robot: the film wsuccessively immersed in the dispersion, blotted with fipaper (once for one second in our experiment), and vitriby shooting it into liquid ethane. In order to prohibit solveevaporation, the atmosphere was saturated with ethanopor. We imaged the particles in the thin ice film formedthe holes (2 µm in diameter) of the carbon film. A transmsion electron microscope (Tecnai 12, FEI) equipped witcryo-holder (626 cryotransfer system, Gatan) were uselow dose conditions to avoid melting the frozen dispersfilm. For example, at a magnification of 21,000 the doseof about 5 electrons per Å2. The exposure time was 500 mon a high sensitivity CCD camera (TIETZ). The acceletion voltage was 120 kV and the temperature of the sam−180◦C.

3. Results and discussion

AFM was used to determine the average diameter ofticles. For this purpose we used the height image (Fig. 1).Since the lateral diameter is strongly affected by tip conlution, height information was used to determine the partsize[12]. The average diameter of particles was 12.1 nmnegative particles (which confirms the nominal diameter

-

-

Fig. 1. Height AFM images of negative Ludox particles (above) and positivLudox particles (below), taken after theparticles were transfered to ethano

ported by DuPont) and 15.2 nm for positive particles. Tlatter is larger, due to the alumina layer which covers theica cores. The determined polydispersity was in both caof approximately 18%.

Direct observation of series I during the preparation ansedimentation (Fig. 2) showed that the samples Se up to(such subseries will be denoted by Se–Sa), S2 and Sunstable: even after a small quantity of positive dispers

118 M. Rasa et al. / Journal of Colloid and Interface Science 278 (2004) 115–125

eeks a

Fig. 2. First series (series I) of mixtures of negative and positive silica particles, 4 days after preparation. The sample composition is indicated in the sameorder inTable 1.

Fig. 3. Pictures of the second series of mixtures (series II) of negative and positive silica particles, taken one week after preparation (A), 5 wfterpreparation (B), and a zoomed in—picture obtained from B for samples Sa–S10 (C). The sample composition is indicated in the same order inTable 1.

sityeyen afandousag-were aner-lus-verySd

inmalls toonepar-n

, S2–6 themore

geslyini-

was mixed and stirred with the negative dispersion, denfluctuations and aggregate formation were observed (byand the aggregates sedimented fast. Surprisingly, eveter larger quantities of negative dispersion were mixedstirred with the positive dispersion (S8–S14), homogenesuspensions formed after stirring, in spite of the fact thatgregation was observed as soon as negative particlesinjected, as in all mixtures. Only samples S8 and S9 havincreased turbidity in comparison with the positive dispsion S15, indicating the presence of somewhat larger cters. Samples S2–S7 are very turbid and contain thuslarge clusters. Sedimentation was observed in samplesSa and S2–S7, but no sediment formed in S8–S14.

)-

e

–

For series II only minor differences were observedcomparison to series I, and this could be due to the sdifferences in the process of transferring the particleethanol. Several pictures of the mixtures were taken afterweek, five weeks, and three months, respectively. In comison with series I, inFig. 3A (taken after one week) one caobserve a similar behavior in the case of samples Sd–SaS7, and S10–S13, respectively. In the case of S5 and Ssupernatant was clear while S8 and even S10 scatterthan in the previous series.

Fig. 3B, taken five weeks later, shows interesting chanin some mixtures: samples S8and S10 became significantless turbid indicating a redispersion (repeptization) of

M. Rasa et al. / Journal of Colloid and Interface Science 278 (2004) 115–125 119

S72that

Wefterpar

oussednressam-upershowple

s. ThDLSively

andsamme-ro-

am-r to

d inex-percou

12.s nored

n-ini-lesuce

me

forthely an

he

mix-theof aandriesin

ationred

ons

epa-eslsost fors, thenly.

ionS15

s.

yiontheof

thegth

t,inr-

ince

ndtex-

tially clustered particles; the sedimentation velocity ofincreased in comparison toFig. 3A; the supernatants of Sand S3 showed different layers. We can conclude nowthe samples S8–S14 are stable suspensions.

The final observations can be summarized as follows.mention that the picture of series II, taken three months athe preparation, does not show notable changes in comison with Fig. 3B: (a) Samples S13–S10 are homogeneand no sediment is observed; (b) Sample S8 shows noiment; (c) Sample S7 shows sediment and a concentratioprofile of the supernatant. The sediment presents yield stwhich was also observed for the following discussed sples; (d) Samples S5 and S6 are sedimented but their snatant does not scatter light; (e) Samples S3, S2 and Sasediments and supernatants with several layers; (f) SamSb–Sd show sediments and homogeneous supernatantsupernatant of samples Sb and Sc was subjected tomeasurements which showed that it contains only negatcharged particles (seeSection 3.1).

After three months, all samples were shaken upstored to observe once more the process. Except forple S7, which showed a lower sediment volume and sowhat lower turbidity, the observations (a)–(f) were repduced for all the other samples.

Finally we added positive and negative dispersions (sples S15 and S1, respectively) to the mixture S6 in ordesee if mixtures with other composition can be reobtainethis way. The mixture S12 was very well reproduced: theisting aggregates in S6 redispersed after the positive dission was added and the obtained sample showed (after aple of days) a similar turbidity as the directly prepared SWhen we added negative dispersion, the mixture Sa waso well reproduced. In comparison with the directly prepaSa, a larger volume of sediment was observed.

Electrical conductivity of dispersions is determined maily by the concentration and the mobility of the free ionsthe solvent. The mobility of colloidal particles is negligble due to their much larger radius. However, the particmay influence the conductivity because they may prodsignificant electric fields in the suspension if their volufraction is high enough. A recent analytical study[13] ofconductivity and mobility of colloids led to closed-formequations for interacting particle suspensions but validlow zeta potential (smaller than 25 mV). As we will see,zeta potential is not low in our systems, so that here onsimple model will be used for conductivity. As it is showin Fig. 4, the conductivity decreases nonlinearly with tvolume fraction of negative particle dispersionΦd , whilea linear variation was expected for the nonaggregatedtures, since ion recombination is unlikely to occur. Inaggregated mixtures a release of ions (from the vicinitysingle particle) because of cluster formation may occurcould explain qualitatively the results. In the case of seII a slight initial increase could be observed. However,this case, the small increase can be a result of evaporof ethanol due to the small amounts of mixtures prepa

-

-

,

-

se

-

--

t

Fig. 4. Conductivity of the mixtures of negative and positive dispersiversus the relative volume fraction of negative particle dispersion.

(the measurements were performed five weeks after prration, period in which the vials were opened a few timfor light scattering measurements). This fact affected athe reproducibility of measurements on series II (the valuewere somewhat larger after a new measurement) excepthe stock sample, so that, in the case of the second serieerror bar is valid for the stocks samples (S15 and S1) oSeries I was measured immediately after the preparation.

Conductivity measurements may be used to estimateconcentration in the positive and negative dispersions (and S1, respectively), if the ion mobility is known,

(1)n0 = σ

e(µ+ + µ−),

whereσ is the conductivity of dispersion,e is the elementarycharge andµ+,− the mobilities of positive and negative ionAs discussed in the beginning of this paragraph,Eq. (1) isvalid for low volume fractions of colloids. In addition it mabe used for low ionic strength only (when ion–ion interactis negligible). In comparison with water based samples,conductivity of ethanol based dispersions is two ordersmagnitude smaller.

At this stage determination of the screening length incolloidal dispersions is useful. The Debye screening lenis given by

(2)λD = 1

κ=

(ε0εrkT

2e2n0

)1/2

,

whereε0 is the permittivity of vacuum,εr the relative per-mittivity of the solvent (ethanol),k the Boltzmann constanand T the absolute temperature. The mobility of ionsethanol (at infinite dilution) in the positive particle dispesion were taken from[14]. The results are presented inTa-ble 3. For negative particles, an estimation was done sthe mobilities were not known exactly. In both casesκR < 1,whereR is the radius of a particle. The particle radius aion concentration are both relatively small, in agreemenwith κR < 1, hence the Debye–Hückel approximation is

120 M. Rasa et al. / Journal of Colloid and Interface Science 278 (2004) 115–125

e

Sample RAFM (nm) λD (nm) Φ0 (mV) Uel(0) (kT ) Q

Positive particles 7.6 9.0 76.6 14.5 18|e|Negative particles 6.1 17.5 −76.6 11.6 10e

no

ralsignansnd

ol-

neg-sedinthe

andnm

thehav

par-t ofrgy

keron-gyclestheergyeas-tive14).r the

tionis-

eare

se.teds

dis-n.

ively.lues isility

ds,. Tak-

entpar-all

itive.unt

oth-Sa.gre-eticin a

ayernlysin-

sa ex-hemo-amehell

r-nsTheely

pected to be valid, even if the electrostatic potential islonger small[15].

Electrophoretic mobility measurements, done sevedays after the preparation of mixtures, showed that theof particle surface charge does not change after the trfer from water to ethanol. According to the discussion aresults in the previous paragraph, the conditionκR < 1 isachieved, so that the Hückel equation for mobility of cloids

(3)µ = 2ε0εrζ

3η

should provide reasonable results for both positive andative colloids. For the aggregated mixtures it cannot be uIn Eq. (3), ζ is the zeta electrostatic potential, whichthe case of organic media is usually approximated withsurface electrostatic potentialΦ0 [2]. By determining thesurface potential fromEq. (3) (seeTable 3), we can calcu-late the electrostatic attraction energy between a positivenegative particle. The diameter is similar (assumed 13.5in the following calculation) and the absolute values ofsurface potentials are the same because the mobilitiespractically the same absolute values. For the caseκR < 1,the DLVO electrical double-layer interaction energy[15]changes only its sign,

(4)Uel(x) = −4πε0εrR2Φ2

0

xexp

(−x − 2R

λD

),

wherex is the center-to-center distance between twoticles. For particles in contact, the result is independenthe Debye length and we found that the interaction eneis Uel(0) = −13kT . The attraction can be somewhat weabecause of neutralization of the charges which get in ctact. InTable 3, the maximum double-layer repulsive eneris presented separately for positive and negative partiThe distanceλi between the particle surfaces at whichelectrostatic attraction energy is equal to the thermal en(which we call the interaction length) can be estimatedily if we consider the case of a small amount of negadispersion introduced in the positive dispersion (e.g., SThen the Debye length is practically the same as that fopositive dispersion (Table 3) and one obtainsλi = 16 nm.These results clearly show that the electrostatic attracwill coagulate the positive and negative particles if the dtance between their surfacesis comparable with the Debylength. It remains to be explained why samples S8–S14so stable and why the aggregated particles can redisper

The mobilities for the two series of mixtures presenin Fig. 5give valuable information to account for stability a

-

.

e

.

Fig. 5. Electrophoretic mobility of the mixtures of negative and positivepersions versus the relative volume fraction of negative particle dispersio

well as for structures in the aggregated samples, respectThe trend for both series is similar. A higher absolute vaof the mobility of negative particles in the second serieobserved, which can explain the shift of the zero mobpoint to the left (for this series).

Samples S8–S15 exhibit positive mobilities of colloiwith close values, but a small decrease can be observeding also into account that they are stable and no sedimwas observed, we may conclude that either the negativeticles are well screened by the positive particles in a smcluster or the negative particles somehow become posThe second hypothesis may be true only for a limited amoof negative particles dispersed in the positive systemerwise we cannot explain the instability of samples Se–Surprisingly, samples S7 and S6, which are strongly aggated, show similar positive values of the electrophormobilities. We may assume in this case that the chargescluster cancel out and most of the particles in the outer lare positively charged. Hence, the mobility, determined oby the surface potential of aggregates, is close to that ofgle positive particles.

At higher values ofΦd , the mobility sharply decreasebecoming negative. The unstable samples S2 and Sd–Shibit negative values of mobilities close to the mobility of tsingle negative particles in the sample S1. The negativebility for aggregated samples can be explained in the sway as above, but now most of the particles in the outer sof the clusters are negatively charged.

In the vicinity of the volume fraction of negative dispesion ofΦd = 0.78, the mobility is close to zero. That meathat the net charge of particle clusters is very small.outer shell is composed of both positively and negativ

M. Rasa et al. / Journal of Colloid and Interface Science 278 (2004) 115–125 121

Sing a

gre-., atoox 1-n ofuted

to

The

d,m-eenFM.yus

DLSasdueerved

Thethethean-

m-

tive,. ItentsF)lita-hat1,

arti-

Fig. 6. Diffusion coefficients versusscattering vector obtained from DLmeasurements for negatively charged particles and mixtures containsmall quantity of positive dispersion.

charged particles. This may explain why the largest aggates formed in samples withΦd close to the value of 0.78

Light scattering was performed on samples of series IIfew days after the preparation. When the turbidity washigh, we measured diluted samples, marked with indeor 2 according toTable 2, in order to avoid multiple scattering. This fact may have an influence on the dimensioaggregates but we will make comparisons between dilsamples.

SLS performed on samples Sb1, S31, S61–S81 shows nominima in their angular intensity profile, which points

a large polydispersity of aggregates in size and shape.DLS measurements are discussed in detail below.

In Fig. 6 the diffusion coefficients for samples S1, SSb1 and S31 are presented. The radius of particles in saple S1 varied (with the increasing scattering angle) betw22 and 16 nm, which exceeded the size measured with AThe same result was obtained after the sample was highldiluted and after a high dilution of the original aqueodispersion was measured. The size determined from(which is always larger than the AFM or TEM one) whowever beyond our expectancy. This is most probablyto some aggregates present in these samples, also obsin the cryo-TEM image (Fig. 7D), in which we found in-deed aggregates with radii of about 16 nm or larger.polydispersity of single particles, of aggregates, and alsopossible presence of some residual impurities explaindependence of the diffusion coefficient on the scatteringgle.

Much lower diffusion coefficients than in the case of saple S1 were obtained for Sd, Sb1, and S31 as a result ofsignificant aggregation. These results are rather qualitataking into account the poor stability of these samplesis thus clear why the dynamical contrast of measuremand the fit with the intensity autocorrelation function (IAC[16] for monodisperse spheres were not so good. Quatively, the above mentioned diffusion coefficients show teven small volumes of positive dispersion mixed with Slead to large aggregate formation. Thus, the positive p

Fig. 7. (A, B) Cryo-TEM pictures of sample S2, which is a mixture of negative and positive particles as indicated inTable 1. (C) TEM picture of S2 (thespecimen was dried in air). (D) Cryo-TEM picture of sample S1, which contains only negatively charged particles.

122 M. Rasa et al. / Journal of Colloid and Interface Science 278 (2004) 115–125

-and

luste

lesnglewellthe

es.ofsizetheega

n

e-meFoweults

sta-ning) orened

seeS13

ents.

redtivearyder-ing.uld

plesicles,

diusggre-mea-plesega-veryremplessize.derstiga-rma-ue.rmallnec-ir.lowthee

sta-onsuesi-

Fig. 8. Diffusion coefficientsD and particle radiusR versus scattering vector obtained from DLS measurements for positively charged particlesmixtures containing positive dispersion down to 50%.

cles are not screened by the negative ones and the cgrowth does not stop at an early stage.

In Fig. 8 the results for samples S151, S131, S101, andS81 are presented. The radius of positive Ludox partic(between 65 and 45 nm, depending on the scattering awas determined using S15, several dilutions of S15, asas a very diluted aqueous sample (obtained directly fromoriginal Ludox dispersion) without noticeable differencThe DLS radius of positive particles is larger than thatnegative particles. The same explanations for the largerof positive particles compared to the AFM size and fordependence on scattering angle, mentioned above for ntive particles, are assumed.

Samples S131 and S101 show negligible differences idiameter of particles in comparison with S151 and the samepolydispersity. Only S81 contains significant heteroaggrgates. A higher dilution of S8 led to approximately the saresults as S81. All these samples led to a good fit with IACfor monodisperse spheres, and the measurements sha very good dynamical contrast which make these resquantitatively more accurate.

From mobility measurements we concluded that thebility of these samples is either due to a proper screeof negative particles (which stops the growth of clustersdue to their recharging. If the negative particles are screand are surrounded by positive particles, then we shoulda larger radius in the DLS measurements of samples

r

)

-

d

Fig. 9. Radius of the negative particles determined from DLS measuremusing the supernatant of samples Sb and Sc as well as the sample S1

and S10 for example, in comparison with that measufor S15. Referring to the sample S10, the ratio of posito negative particles, without taking into account primaggregates, is approximately 2.3 and it is not easy to unstand how such a ratio can result in a sufficient screenThe hypothesis of recharging of negative particles shothus receive more attention (seeSection 3.1).

It is also interesting to mention that in the case of samSc and Sb, the supernatant contains only negative partaccording to the DLS determination of radius (Fig. 9), donethree months after the preparation. The larger value of raat the smallest angle, for sample S1, is due to the homoagation process: this sample was shaken up before thesurement while the upper part of the supernatant of samSc and Sb was extracted with a syringe. The homoaggrtion process is thus measurable at small angles but isweak. FromFig. 9we can infer that all positive particles aclustered and sedimented, so that, indeed, in these sathe growth of aggregates does not stop at a small cluster

TEM and cryo-TEM. Since the diameter of positive annegative particles is similar, we expect fractal like clustin samples Sd–Sa and S2–S7. Optical microscopy investions suggested that this is indeed the case, but the confition we give is obtained by using the cryo-TEM techniqThe images of sample S2 (Fig. 7A) show large fractal clusteformation, although, some areas show the presence of sfractals or chains (Fig. 7B). Fig. 7Cpresents for comparisoa conventional TEM picture of S2, i.e., the image of a spimen prepared after the colloid thin film was dried in aThe contrast is better in this case since the constraint ofdose is not required anymore. The cryo-TEM picture ofsample S1 (Fig. 7D) shows single particles, except for somhomoclusters also observed in DLS experiments.

3.1. Repeptization and stability of mixtures

We deal here in more detail with the unexpectedbility of samples S8–S14. According to the discussiin the previous subsection, the stability is very likely dto a “charge-inversion” of negative particles to a po

M. Rasa et al. / Journal of Colloid and Interface Science 278 (2004) 115–125 123

-s-atan

x-tion

aftea-

s.cen-re-nt.o aes.

asag

sureof

ity

Hent

on-uesg-ngeCl

usre-

-ies,ngetizares

-aarg-

umeionsag-

reddel,

tials,ed inosi-for. As

ergy

an

taticfor

-itz

nols

cy

s

re-the

-qual

iteosi-le

e, the

Fig. 10. Electrophoresis measurementson positive particles (S15), negative particles (S1), supernatant of the sedimented positive particle dipersion (SS15), and negative particles dispersed in that supern(SS15+ S1).

tive value. To verify this hypothesis, the following eperiments were done three months after the preparaof samples. No noticeable changes were observedthis time in the mobility of positive particles (the remesured value was+1.06 µm cm/(V s)), while an increaseto −0.45 µm cm/(V s) was noticed for negative particleThe positive dispersion S15 was sedimented by ultratrifugation (25000 rpm, 4 days). The supernatant wasmoved and subjected to an electrophoresis measuremeshowed a peak corresponding to zero mobility, but alssmall one corresponding to the mobility of positive particlThe negative particle dispersion was added (Φd = 50%) tothe supernatant. Initially an increase in the turbidity wobserved (due to aggregates) but after a few hours thegregated particles redispersed. The electrophoretic meaments showed now a peak corresponding to a mobility+0.22 µm cm/(V s) and no peaks in the negative mobilregion. These results are presented inFig. 10. Recharging ofinitially negative particles is not due to the slightly acid pof the supernatant used in the above mentioned experimWe added HCl to the negative dispersion in different ccentrations, measured in terms of conductivity. At the valof conductivity similar to or larger (up to one order of manitude) than the conductivity of the supernatant, no chain the mobility was observed. Only when much more Hwas added (conductivity of order of mS/cm) the mobilitywas close to zero. This behavior of silica mobility verspH, in ethanol, is actually known and we reobtained thesults presented in[2]. It is known that the solubility of silicain ethanol is very small[17], but alumina may dissolve significantly, resulting in positively charged aluminate specwhich adsorb onto the negative silica particles and chatheir charge. It was observed that the process of repeption depends on concentration and composition of mixtubeing much faster at small values ofΦd . Probably, after positive charges are adsorbed, more alumina dissolve, untilfinal equilibrium is reached. For samples S2–S7, the rech

t

r

It

--

.

-

ing phenomenon should occur too, but because the volof positive dispersion is smaller, there were not enoughfor the negative particles to be recharged. Thus fractalgregation persisted in time together with instability.

The charge of particles before the mixture were prepacan be determined in the frame of the Debye–Hückel modiscussed in the previous section:

(5)Q = 4πε0εrΦ0R

(1+ R

λD

).

Using the determined Debye lengths and surface potenwe obtained for the particle charges the results presentTable 3. Consequently, there should be more than 10 ptive monovalent ions in the solvent of positive dispersioneach negative particle for the redispersion to be possiblewill be discussed below, the van der Waals attraction enin these systems is comparable withkT .

The total DLVO potential energyUtot [15] for two sin-gle initially negative particles is obtained by adding the vder Waals energyUW to Eq. (4) taken now with the signchanged,

(6)UW(x) = −A

6

[2R2

x2 − 4R2+ 2R2

x2+ ln

(1− 4R2

x2

)],

whereA is the Hamaker constant andx is the center-to-center distance between the two particles. The electrosenergy is calculated using the parameters determinednegative particles (presented inTable 3). The Hamaker constant for silica in ethanol was calculated by using the Lifshmodel[18],

(7)A = 3

4kT

(ε1 − ε2

ε1 + ε2

)2

+ 3hνe

16√

2

(n21 − n2

2)2

(n21 + n2

2)3/2

,

whereε1 andε2 are the static dielectric constants of ethaand silica, respectively,n1 andn2 are the refractive indexeof ethanol and silica (in the visible spectrum),h is the Planckconstant, andνe is the main electronic absorption frequenin the UV (of the order of 3× 1015 Hz). A barrier of en-ergy of the order of 10kT resulted from this calculation, aexpected (not presented).

It is interesting to see if the theory supports thedispersion of heteroaggregated particles. According toelectrophoretic mobility measurements, the initially negatively charged particles recharge to a surface potential eto Φ ′

0 = 16.9 mV (obtained by usingEq. (3)). This sur-face potential corresponds to a charge ofQ = +4|e|. Inthis caseEq. (4) is not valid anymore because of the qularge difference between the surface potential of the ptive particle (Φ0 = 76.6 mV) and of the recharged partic(Φ ′

0 = 16.9 mV). For the case of two particles with the samradius and Debye length, but different surface potentialselectrostatic potential energy is given by[19]

(8)U ′el(x) = 2πε0εrRΦ0Φ

′0 exp

(−x − 2R

).

λD

124 M. Rasa et al. / Journal of Colloid and Interface Science 278 (2004) 115–125

ace

Ac-as ad balle theitive

sur-sur-tha

aals

pa-s,hre-is-butativeitheturen

forsaltatic-

al isinally

lesOS1

d atag-

they the

n

atestione ex-ur-alu-

saltfor-

terss.s.

swasar-

re-f the

m-be

a re-wasthe

mo-ore-roup

Fig. 11. Total DLVO potential energy perkT of a positive particle andrecharged particle versus center-to-center distance. The minimum distanbetween the particle surfaces was 0.23 nm.

Equation (8), which is correct forκR > 1, is assumed togive good qualitative results for the caseκR < 1. In a clus-ter, multiple interactions should be taken into account.cordingly, the results discussed below can be regardedestimation for the case of a heteroaggregate composetwo particles. We may think to a mixture with only a smamount of negative dispersion, such as S14. In this casDebye length is practically the same as that one for posparticles (seeTable 3). By usingEqs. (6), (7) and (8)the totalpotential energy, presented inFig. 11, shows only repulsiveinteraction if there is a minimum distance between thefaces of the particles (inFig. 11the minimum distance wa0.23 nm). This minimum distance may come from the sface roughness of the particles. It cannot be excludedalso the alumina layer plays a role in reducing van der Winteraction, such thin layers being usually porous.

We observed that the van der Waals energy is comrable tokT in the initial positive and negative dispersionrespectively: after salt (LiNO3) was added in high enougconcentration (of order of 0.1 M), no flocculation or agggation was observed by eye. The turbidity of positive dpersion only slightly increased after a period of 10 daysthe system remained stable. When the positive and negdispersions were mixed, aggregates were not observed e(however, they appeared one week later making the mixturbid). Thus the particle charge was screened well and if vader Waals had been significantly larger thankT the systemswould have flocculated. The DLVO total potential energythe positive and negative dispersions, respectively, afterwas added, showed indeed a very thin residual electrosbarrier of the order ofkT , while for surface-to-surface distances larger than 0.1 nm, the total interaction potentilarger than−kT . This shows that in the systems studiedthis paper the particle association/dissociation is practiccontrolled by the electrostatic interactions.

Salt effect. Electrostatic interactions between particcan be adjusted by adding certain amounts of salt. LiN3was added to samples S1 (the negative dispersion) and(the positive dispersion) which were subsequently mixevolume ratios of 3:1.3 (the composition corresponds to

ny

t

r

t

5

Fig. 12. Effects of salt addition on the stable sample S10 (A) and onaggregated sample S6 (B). The molar salt concentration is indicated bnumbers in the two pictures.

gregated sample S6,Fig. 12B) and 1.3:3 (the compositiocorresponds to stable sample S10,Fig. 12A). Fig. 12Ashowsthat salt addition prohibits the repeptization. The aggregpersist in time and sediment, the profile of sedimentabeing dependent on the salt concentration. This could bplained by the fact that a higher density of positive ions sround the negative particles preventing the absorption ofminate positive ions and the repeptization.Fig. 12Bshowsthe opposite effect of salt in sample S6. The increasingconcentration up to 0.3 M reduces the heteroaggregatemation. Beyond this limit, somewhat large heteroclusform because the salt dissociatesless at high concentrationHowever, all sediments observed still present yield stres

4. Conclusions

Mixtures of initially oppositelycharged silica particlewere studied at low ionic strength. Heteroaggregationobserved in the group of mixtures with smaller positive pticle content while stability and only relatively small agggates, which decayed in time were observed in the rest omixtures.

The particle repeptization and the stability of the saples containing up to 50% initial negative dispersion canexplained by the recharging of the negative particles assult of adsorption of aluminate species. This explanationconfirmed after the negative particles were dispersed insolvent of the positive dispersion and the electrophoreticbility was measured, and is consistent with the electrophsis and light scattering measurements done on these g

M. Rasa et al. / Journal of Colloid and Interface Science 278 (2004) 115–125 125

edis

n-ountox-osi-ounxi-was

on-dis-s

stersial ispargednot

tialparar-

aceionpar

thesre-

lp-by

ap-

ol-

hem.

304

e

165

68)

50

65,

bic

-

e

ress,

hem.

of mixtures. The time-dependent process of aggregate rpersion scales with the volume ratio of the mixtures.

Mixtures with a higher negatively charged particle cotent exhibited large aggregates. When only a small amof positive particle dispersion is present (down to apprimately 0.2%), the samples were unstable and all ptive particles aggregated and sedimented. When the amof positive particle dispersion was higher (up to appromately 15%) the sediment, which presented yield stress,accompanied by various supernatant profiles.

The conductivity of the series of mixtures showed a nlinear behavior versus the volume fraction of negativepersionΦd while the electrophoreticmobility measurementshowed a zero point mobility aroundΦd = 78%. In this casewe assumed that the particles in the outer shell of the cluare both negative and positive so that the surface potentclose to zero. In the case of unstable samples most of theticles in the outer shell of the clusters are negatively char(since a negative mobility was measured) but that wasenough however for achieving stability.

The Debye length, particle charge, and DLVO potenenergy were determined and discussed for the unmixedticles. The calculated total potential energy for a positive pticle and a recharged particle (which have different surfpotentials) supports quantitatively the particle redispersform heteroaggregates. In the investigated systems, theticle association/dissociation is practically controlled byelectrostatic interaction. Salt addition had opposite effecton the stability of initially unstable and stable mixtures,spectively.

Acknowledgments

Dr. B. Erné and N. Zuiverloon are thanked for heful discussions. This work was financially supported

-

t

-

-

-

NWO–CW (Nederlandse Organisatie voor Wetenschpelijk Onderzoek–Chemische Wetenschappen).

References

[1] K. Furusawa, C. Anzai, Colloids Surf. 63 (1992) 103.[2] G. Wang, P. Nicholson, J. Am. Chem. Soc. 84 (2001) 1250.[3] A.Y. Kim, J.C. Berg, J. Colloid Interface Sci. 229 (2000) 607.[4] A.Y. Kim, K.D. Hauch, J.C. Berg, J.E. Martin, R.A. Anderson, J. C

loid Interface Sci. 260 (2003) 149.[5] A.M. Puertas, A. Fernandez-Barbero, F.J. de las Nieves, J. C

Phys. 114 (2001) 591.[6] A.M. Puertas, A. Fernandez-Barbero, F.J. de las Nieves, Physica A

(2002) 340.[7] W.L. Yu, M. Borkovec, J. Chem. Phys. B 106 (2002) 13106.[8] A.M. Islam, B.Z. Chowdhry, M.J. Snowden, Adv. Colloid Interfac

Sci. 62 (1995) 109.[9] S. Usui, Prog. Membrane Surf. Sci. 3 (1972) 223.

[10] C. Pathmamanoharan, A.P. Philipse, J. Colloid Interface Sci.(1994) 519.

[11] H.J. van den Hul, J.W. Vanderhoff, J. Colloid Interface Sci. 28 (19336.

[12] M. Rasa, B.W.M. Kuipers, A.P. Philipse, J. Colloid Interface Sci. 2(2002) 303.

[13] H.J. Keh, J.M. Ding, Langmuir 18 (2002) 4572.[14] G. Kortüm, Treatise on Electrochemistry, Elsevier, Amsterdam, 19

p. 252.[15] E.J.W. Verwey, J.Th. Overbeek, Theory of the Stability of Lyopho

Colloids, Dover, New York, 1999.[16] J.K.G. Dhont, An Introduction to Dynamic of Colloids, Elsevier, Am

sterdam, 1996, p. 144.[17] P.A. Buining, L.M. Liz-Marzan, A.P. Philipse, J. Colloid Interfac

Sci. 179 (1996) 318.[18] J. Israelachvili, Intermolecular and Surface Forces, Academic P

London, 1992, p. 183.[19] A.M. Puertas, A. Fernandez-Barbero, F.J. de las Nieves, J. C

Phys. 115 (2001) 5662.