COLLOIDS AND Colloids and Surfaces SURFACES A ELSEVIER A: Physicochemicaland Engineering AspectsJ73 (2000) 127-158 www.elsevier.nl{Jocate/colsurfa
A.S. Dukhin 8, P.I. Goetz a, T.H. Wines b, P. Somasundaran b,* a Dispersion TechnologyInc., 3 Hillside Ave~, Mt Kisco, NY 10549, USA b Columbia University, Rm. 911, .500 W. 12OthStreet, 1140Amsterdam Avenue, New York, NY 10027, USA
Received 28 September 1999; ~ 13 March 2000
Abstract
Two new ultrasound based techniques (acoustics and electroacoustics)otTer a unique opportunity to characterize concentrated dispersion, emulsions and microemulsions in their natural state, without dilution. Elimination of the dilution protocol is crucial for an adequate characterization of liquid dispersions, especially structured. Dilution changes the thermodynamic equilibrium in these systems and atTects their reological properties. Changes in equilibrium conditions can lead to variation of the particle size and can also affect surface chemistry. In this paper, a short review of the theoretical basis of the ultrasound techniques is given. Emphasis is placed on the theoretical models which are supposed to be valid in concentrated systems.These theories have been developed recently on the basis of 'cell model concept' for bOth acoustics and electroacoustics. This approach opens the way to implement particle-particle interaction into the theoretical model. Experiment proves that these theories are adequate in concentrated systems up to 45% vol. Second part of the paper is dedicated to the applications of acoustics and electroacoustics.The list of applications includes: ceramics, mixed dispersed systems, chemical-polishing materials, emulsions, food emulsions, microemulsions and latecies. @ 2000 Elsevier Science B.V. All rights reserved.
Keywordr: Acoustic spectroscopy; Electroacoustic spectroscopy; Ultrasound techniques
. Introduction capabilities for being successful. The first hard- ware for measuring acoustic properties of liquids The widespread acceptanceand commercializa- was developed more than 50 years ago at MIT [I] tion of acoustic spectroscopy has been slow to by Pellam and Galt. The first acoustic theory for develop. This technique has been overlooked by heterogeneoussystems was created by Sewell 90 many in academia and industry in the past, but years ago [2]. The general principles of the acous- has recently been showing increased levels of ac- tic theory were formulated 45 years ago by Ep- ceptance.This powerful method of characterizing Stein and Carhart [3]. There is a long list of concentrated heterogeneous systems has all the applications and experiments using acoustic spec- troscopy, see reviews [4,5]. Despite all of these . Corresponding author. Tel.: + 1-212-8542926;fax: -j developments, however, acoustic spectroscopy is 212-8548362. rarely mentioned in modem handbooks on colloid E-mail address:[email protected] (P. Somasundaran). science[6,7].
0927-7757/001$- see front matter 0 2000 Elsevier Science B. V. All rights reserved. PO: 80927-7757(00)00593-8 128 A.S. Dukhin et a/.jCol/oids and Surfaces A: Physicochem.Eng. Aspects 173 (2tXXJ)127~lSB
Acoustics is able to provide reliable particle size degree of complexity in fitting experimental re- information for concentrated dispersions without sults to theoretical models based on various any dilution. There are examples when acoustics acoustic loss mechanisms. The advent of high yields size information at volume fractions above speedcomputers and the refinement of these theo- 400/0.This in-situ characterization of concentrated retical models has made the inherent complexity systems makes the acoustic method very useful of this analysis of little consequence.In compari- and unique in this capability compared to alter- son, many other particle sizing techniquessuch as nate methods including light scattering where di- photon correlation spectroscopyalso rely on simi- " lution is required. Acoustics is also able to deal lar levels of complexity in analyzing experimental with low dispersed phase volume fractions and in rcsults. some systems can characterize down to below Acoustics has a related field that is usually 0.1% vol. This flexibility for concentration range referred to as 'electroacoustics' [8]. Electroacous- provides an overlap with classic methods for di- tics can provide particle size distribution as well lute systems. In the overlap range, acoustics size as ~-potential. This relatively new technique is characterization has been found to have excellent more complex than acoustics because an addi- agreement with these other techniques. tional electric field is involved. As a result, both Acoustics is not only a particle sizing technique, hardware and theory become more complicated. but also provides information about the mi- There are even two different versions of electroa- crostructure of the dispersedsystem. The acoustic coustics depending on what field is used as a spectrometer can be considered as a micro- driving force. Electrokinetic sonic amplitude rheometer. In acoustics, stressesare applied in the (ESA) involves the generation of sound energy caused by the driving force of an applied electric same way as regular rheometers, but over a very short distances on the micron scale. In this way, field. Colloid vibration current (CVI) is the phe- nomenon where sound energy is applied to a the microstructure of the dispersed system can be system and a resultant electric field or current is sensed.Currently, this feature of the acoustics is created by the vibration of the colloid electric only beginning to be exploited, but it is certainly double layers. very promising. Coming back to acoustics, it's lack of wide- Many people have perceivedacoustics to have a spread acceptancemay be related to the fact that high degree of complexity. The operating princi- it yields too much, sometimes overwhelming in- ples are in fact quite straightforward. The acoustic formation. Instead of dealing with interpretation spectrometer generates sound pulses that pass of the acoustic spectra it is often easier to dilute through a sample system and are then measured the system of interest and apply light based tech- by a receiver. The passage through the sample niques. It was often naively assumed that the system causes the sound energy to change in dilution had not affected the dispersion character- intensity and phase.The acoustic instrument mea- istics. Lately, many researchersare coming to the sures the sound energy losses (attenuation) and realization that dispersed systemsneed to be ana- the sound speed.The sound attenuates due'to the lyzed in their natural concentrated form, and that interaction with the particles and liquid in the dilution destroys a lot of useful and important sample system, Acoustic spectrometers generally properties. operate with sound in the frequency range of The authors are optimistic about the future of , 1-100 MHz. This is a much higher sound fre- acoustics in colloid science. It is amazing what quency than the upper limit of the hearing which this technique can do especially in combination is only 0.02 MHz. with electroacoustics for characterizing eleCtric While the operating principles are relatively surface properties. It is hoped that this review will simple, the analysis of the attenuation data to allow one to taste the power and opportunities obtain particle size distributions does involve a related to these sound based techniques. A.S. Dukhin et al. / Colloid\' and Surfaces A: Physicochem.Eng. Aspects 173 (2O(XJ)127-158 129
2. Theoretical background field, and consequently to alternating electric cur- rent. As a result, a part of the acoustic energy is There are six known mechanismsof the ultra- transformed into electric energy and then irre- sound interaction with a dispersed system: (1) versibly to heat. viscous (av;J; (2) thermal (ath); (3) scattering (asc); Only the first four loss mechanisms (viscous, (4) intrinsic (ainJ; (5) structural (a.tr); and (6) thermal, scattering and intrinsic) make a signifi- electrokinetic (aeJJ. cant contribution to the overall attenuation spec- (1). The viscous losses of the acoustic energy tra in most cases.Structural lossesare significant occur due to the shear waves generated by the only in structured systems that require a quite particle oscillating in the acoustic pressure field. different theoretical framework. These four mech- These shear waves appear becauseof the differ- anisms form the basis for acoustic spectroscopy. ence in the densities of the particles and medium. Total attenuation measured with acoustic spec- This density contrast causes the particle motion trometer usually equals to the sum of these four with respect to the medium. As a result, the liquid partial attenuations: layers in the particle vicinity slide relative to each (X (Xvis + (Xlh + (xsc + (XiDI (1) other. This sliding non-stationary motion of the = liquid near the particle is referred to as the 'shear The contribution of electrokinetic lossesto the wave'. Viscous lossesare dominant for small rigid total sound attenuation is almost always negligi- particles with sizes below 3 ~m, such as oxides, bly small [9] and will be neglected.This opens an pigments, paints, ceramics, cement, graphite, etc. opportunity to separate acoustic spectroscopy (2). The reason for the thermal losses is the from electroacoustic spectroscopy beCauseacous- temperature gradients generated near the particle tic attenuation spectra is independent of the elec- surface. These temperature gradients are due to tric properties of the dispersed system. the thermodynamic coupling between pressure Following this distinction between acoustics and temperature. This mechanism is dominant for and electroacoustics, the corresponding theories soft particles, including emulsion droplets and will be considered separately. latex beads. (3). The mechanism of the scattering losses is 2.1. Theory oj acoustics quite different than the viscous and thermal losses.Acoustic scattering does not produce dissi- The most well known acoustic theory for het- pation of acoustic energy. This scattering mecha- erogeneoussystems was developedby Epstein and nism is similar to light scattering. Particles simply Carhart [3], Allegra and Hawley [10]. This theory redirect a part of the acoustic energy flow and as takes into account the four most important mech- a result this portion of the sound does not reach anisms (viscous, thennal, scattering and intrinsic) the sound transducer. This mechanism is impor- and is tenDed the 'ECAH theory'. This theory tant for larger particles (> 3 ~m) and high fre- describes attenuation for a monodisperse system quency ( > 10 MHz). of spherical particles and is valid only for dilute (4). The intrinsic losses of the acoustic energy systems. occur due to the interaction of the sound wave The tenD 'monodisperse' assumesthat all of the with the materials of the particles and medium as particles have the same diameter. Extensions of homogeneousphases on a molecular level. the ECAH theory to include polydispersity have (5). Structural lossesare caused by the oscilla- typically assumeda simple linear superposition of tion of a network of particles that are intercon- the attenuation for each size fraction. The tenD nected. Thus, this mechanism is specific for the 'spherical' is used to denote that all calculati
"'
~
tw t- 4~
:2' ~ ."j.- M i ~ ~- ~l . v . .. 12 .w.# ..,. ~ . I . J.1 . . 1.5,.j.-.i,~t.,'.,,'.j'lcc") 2.8 2.5 3.0 J.5 ~ ..w .~ u ... -, W--I ...
Fig. 2. Themlal properties of various liquids.
shows that particle size must be below 10 J.1Infor the frequency range from 1 to 100 MHz. This {~ restriction is helpful for characterizing small particles. coPC By restricting frequency and particle size with z=(l +j)a j the longwave requirement one can use the simpler 2r" explicit ex6ression for the thermal losses (1.thob- where j is imagine number, At the same time the long wave requirement provides a sufficient simplification of the theory PM- for implementing particle-particle hydrodynamic -;-r- interaction into the theory of the viscous losses.It has been done in the work [18] on the basis of the 'coupled phase model' [19,20]. This new theory [21] works up to 400/0volume. where This theory yields expression for the complex 9'1CPn wavenumber I assuming viscous lossesas the only t- 2a2 one mechanism of the particles interaction with the sound wave: Ff= 61C'1aQ(Up- um) Sn..ICA ~ % ~ ~ . . O'BIioa-I " ...... ~O"".~ . . . I . . . . . . !-a.eor,. RUTILE vol...,.. t\-8dIo8 Ia % . 10 20 J4I ~ SO .~ ; , I '. --" I I , "" f 'c" . "- -- ...... ~ -1°1 . . - I . . b.""" 0"""'. ty .. -: . . . O'Brlta-Leota.d.-:Y § . 0 - d.-:Y.dilpene c Q, . -20- . - ty. poI,.IBpene i " 88 i ...... *: . Fig. 3. Electrokinetic /;-potential calculated from the measured colloid vibration current (CVI) at various volume fractions using different electroacoustic theories for silica Ludox and for rutile R-746 from DupOnt. 134 A.S. Dukhin el aI. / Colloids and Surfaces A: Physicochem.EnK. Aspects 173 (2lXXJ) 127-158 / w . " is dynamic viscosity, up and Umare velocities m =-+jam of the particles and liquid in the laboratory frame Cm of references, n is drag coefficient specified in The simplest formula expressing the scattering Appendix A, M* is stress modulus which can be losses in terms of densities and sound speeds can expressedin terms of densities and sound speeds be derived from Eq. (9) for a single scattering: as following: a = ~w4a3 K 1-- pmC;' )2 + (Prp - Pmrm )2] 2 2 sc 2c~ 3 Ppc; 2pp + Pm M* = PpPmCpCm lpPmC,2,.+ (I -lp )ppc; (10) Expression (6) specifies the value of viscous It is seen that scattering lossesdepend on fre- losses: quency very strongly. According to the authors' experience scattering is important only for large IXvis = - 1m! (7) particle (> 3 ~) and at high frequencies (> 10 This theory can be used also for calculating MHz). sound speed of the dispersions where viscous There is another approach to acoustics which lossesare dominant. employs a 'short wave requirement'. It was intro- duced by Riebel [21]. This approach works only (J) c~, (8) for large particles above 10 ~ and requires lim- Rei ited input data about the sample. This theory may Experiment described in papers [22,23] confirms provide an important advantage in the case of validity of this theory for sound speed. emulsions and latex systems when the thermal Expressions for calculating intrinsic (Xintand expansion is not known. scattering (xsclosses for long wave limit are given There is opportunity in the future to create a in the papers of McClements [4,11,12,24].He uses mixed theory that could use a polynomial fit term 'Iossless scatterers' for describing sound merging together 'short' and 'long' wave ranges propagation through the system when dissipative theories. Such combined theory will be able to mechanisms of viscous and thermal losses are cover a complete particle size range from nanome- negligible. Intrinsic attenuation in such a system ters to millimeters for concentrated systems. can be expressedas following [15]: There are two recent developments in the the- ory of acoustics which deservedto be mentioned (1 ~ Pm(Xp - dispersion consisting of spherical particles in a Now let one consider the measurementof ESA Newtonian liquid, he suggeststhat the thermody- which occurs when an alternating electric field is namic approach be explored as far as possible. applied to the disperse system [7]. If the ~-poten- This new theory operates with notions of particle tial of the particle is greater than zero, then the velocities and temperature fluctuations. This very oscillating electrophoretic motion of the charged promising theory yields some unusual results dispersed particles generatesa sound wave. [26,27J. It has not been yet used, as far as is Both electroacoustic parameters CVI and ESA known, in commercially available instruments. can be experimentally measured.The CVI or ESA spectrum is the experimental output from elec- 2.2. Theory of electroacoustics troacoustic spectroscopy. Both of these spectra contain information about ~-potential and PSD, Whereas acoustic spectroscopy describes the however, only one of the electroacousticspectra is combined effect of the six separate loss mecha- required becauseboth of them contain essentially nisms, electroacoustic spectroscopy, as it is the same information about the dispersedsystem. presently formulated, emphasizes only one of The conversion electroacoustic spectra into the these interaction mechanisms, the electrokinetic PSD requires theoretical model of the electroa- losses. coustic phenomena. This conversion procedure is In acoustic spectroscopy sound is utilized as much more complicated for electroacousticscom- both the excitation and the measured variable, paring to the acoustics. The reason of the addi- and therefore there is but one basic implementa- tional problems relates to the additional field tion. In contrast, electroacoustic spectroscopy involved in the characterization: electric field. The deals with the interaction of electric and acoustic theory becomes much more complicated because fields and therefore there are two possible imple- of this additional field. mentations. One can apply a sound field and For some time O'Brien's theory [28,29] has measure the resultant electric field which is re- been considered as a basis for electroacoustics ferred to as the colloid vibration potential (CVP), including concentrated systems.This theory intro~ or conversely one can apply an electric field and duces a notion of dynamic electrophoretic mobil- measure the resultant acoustic field which is re- ity Jid' It declares that CVI and/or ESA are ferred to as the electronic sonic amplitude (ESA). proportional to this parameter with coefficient First let one consider the measurementof CVP. which is independent on frequency and particle When the density of the particles Pp differs from size: that of the medium Pm.the particles move relative ESA(CVI) = Ccal Pp - Pm I. It has been determined that acoustic spectra is ogy. All of them claim to be able to characterize affected by bubbles. An acoustic theory de- emulsions in the wide droplet size range. There scribing sound propagation through bubbly are some major differences between them. For liquid has been created by Foldy in 1944 [32], instance Opus was designed initially for large and confirmed experimentally in 1940-1950 particles only becauseit employs the 'short wave- [33- 35]. length requirement' [25]. 2. Contribution of bubbles to sound speed and There are also three electroacoustic spectrome- attenuation depends on the bubble size and ters on the market: Matec ESA 9800, the Acous- sound frequency. For instance, a 100 ~ bub- tosizer of Colloidal Dynamics and the DT -200 of ble has a resonance frequency of about 60 Dispersion Technology. KHz. This frequency is reciprocally propor- There is only one instrument which provides tional to the bubble diameter. A bubble of 10 both features, acoustics and electroacoustics to- ~m diameter will have a resonance frequency gether, and this is the DT-1200 Acoustic and of about 0.6 MHz. Electroacoustic Spectrometer of Dispersion 3. Acoustic spectroscopy of dispersed systems Technology. operates with frequencies above 1 MHz and Comparison of the different instruments lies usually up to 100 MHz. The size of the bub- beyond the scope of this review. The DT -1200 bles must be well below 10 ~m in order to was used for all experiments described in this affect the complete frequency range of acoustic work. A description of this instrument is given in the papers [23,36- 38] including accuracy and pre- spectrometer. cision tests. . 4. Bubbles with sizes below 10 ~ are very un- stable as is known from general colloid chem- There is one more recent development made by istry and the theory of flotation. 'Colloid-sized Dispersion Technology, which might be of the big gas bubbles have astonishingly short lifetimes, interest, especially for on-line measurementsand normally between 1 ~s and 1 ms'. [35]. They routine quality control. It is electroacoustic probe simply dissolve in liquid because of high described in details in the pending patent [38]. curvature. This probe implements a completely new method Bubbles can only affect the low frequency part of characterizing particle size and ~-potential of the acoustic spectra below 10 MHz. The fre- which is based on the new electroacoustic theory quency Tangefrom 10 to 100 MHz is available for [23] for concentrated colloids. This probe mea- particle characterization even in the bubbly liq- sures CVI as well as DT-I200. The difference is uids. Acoustic spectrometer can do both, sense that sensing electric electrode is placed on the bubbles and characterize particle size. This con- surface of the ultrasound transducer. This design clusion can be confirmed with thousands of mea- eliminates contribution of attenuation. Sample surements performed with hundreds of different volume is much smaller, probe diameter is about systems.Sensitivity to bubbles, in fact, is an im- 3 cm only. It can be easily immersed into the portant advantage of acoustics over electroacous- small sample container. In addition, it makes trics. The presence of bubbles may affect the possible to characterize colloids on-line. properties of the solid dispersed phase. For in- stance, bubbles can be centers of aggregation which makes them an important stability factor. 4. ,\pplications 4. J. Ceramics 3. Measuring technique Determination of both the particle size distriBu- Currently, there are three acoustic spectrome- tion and the ~-potential of ceramic slurries is of ters on the market: Ultrasizer of Malvern, Opus key importance in optimizing performance. The of Sympatec and DT-IOO of Dispersion Technol- particle size of the slip is closely related to inho- 138 A.S. Dukhin et at. / o,ltoids and Surfaces A: Physicochem.Eng. Aspects 173 (2lXXJ)127-158 mogeneities, which in turn relate to fracture size distributions for a variety of commonly used origins as well as shape distortion/cracking during ceramic materials. drying, pyrolysis and sintering. Furthermore, the For many applications it is important to recog- ~-potential of the slurry particulates can be used nize particle size sub-populations in the final as a tool for optimizing chemical dosage to slurry. Such bimodal distributions might result achieve the desired colloid stability and size from agglomeration of primary particles caused distribution. by non-optimum dispersant addition, or in the Traditional measurementsof particle size and following example, from an intentional addition ~-potential usually involve light scattering or sedi- of a secondsize fraction. Fig. 6 shows the acoustic mentation techniques and require extreme dilu- attenuation spectra for three 10 vol.% alumina tion of the ceramic slip. This dilution step often slurries: a 0.36 ~m sample, a 2 ~ sample, and a changes both the size distribution and the ~-po- 1:1 mix of the two. The theoretical spectra fit tential of the sample, thereby distorting the very quite precisely the experimental data giving high information being sought. Characterizing the con- confidence in the results. Fig. 6 shows the result- centrated sample directly would allow one to real- ing size distributions for the two single compo- istically judge the true agglomeration status of the nent slurries (blue and black curves) as well as the slip and to optimize the dosage of various chemi- bimodal distribution for the mixed slurry (red cal additives in situ. In contrast, measurementsof curve). the diluted sampleswith traditional methods often In many ceramic applications the ceramic slip is reveal only the primary size of the raw materials actually a mixture of more than one solid compo- since the usual sample preparation steps of dilu- nent. Traditional optical or sedimentation tech- tion, chemical modification, stirring and perhaps niques can not provide correct interpretation of even sonication have destroyed most of the useful such mixtures and typically assumethat all parti- information about the original slurry. cles have a common set of physical properties. In Acoustic spectroscopy can provide accurate contrast, commercially available software for particle size data even in concentrated slurries. acoustic spectroscopy has evolved to the point Fig. 4 shows the measured attenuation spectra that allows the specification of at least two classes and corresponding particle size distribution for of disperse particles. For example, Fig. 7 shows four different alumina slurries. Table 19ives a an example of a mixed system of alumina and comparison of the measuredparticle size with the zirconia particles. This Figure shows the attenua- manufacturer's data derived from traditional tion spectra for three 5 vol.% slurries: a 2 ~ methods. single component alumina, a 0.3 ~m single com- The agreementbetween acoustic and traditional ponent zirconia, and a I: 1 mix of the two ingredi- methods is quite good becausecare was taken in ents. Again, the theoretical spectra fit the formulating the concentrated dispersion with an experimental data quite precisely giving high optimum level of surfactant to insure good disper- confidence in the results. Fig. 7 shows the result- sion of the primary particles. In many real-world ing single mode particle size distribution for each cases, the final dispersant dose may have been oxide measuredseparately, as well as two separate simply extrapolated from very dilute measure- single mode distributions measuredfor each com- ments and one may obtain a larger particle size in ponent in the mixed slurry system. the actual slurry than predicted from these dilute It is not always appreciated that the particle measurementsof the raw materials. It is almost size distribution of a slurry is not simply a func- always better to characterize the particle size of tion of the primary size of the constituent ingredi- the slurry. When this does not compare with ents, but instead is a result of many complex dilute measurements one needs to examine chemical and mechanical operations on the sys- whether the chemical formulation is adequate. tem. The ~-potential of the system is one parame- Acoustic spectroscopyis applicable to virtually ter that can be used to investigate this complex all ceramic materials. Fig. 5 shows typical particle relationship. Fig. 8 compares ~-potential data for A.S. Dukhin et aJ./ Colloids and SurfacesA: Physicochem.Eng. Aspects 173 (2()(X) 127-158 J.6 w. .~. 1.2 ~ ;C.. .; Q~ 08 ~ 04 0.0 10-1 10') 10° 10' 102 Diameter (pm) i E"- = ~ = .~ ~ = = .. = .( 10. I{} I 10J Frequency (MHzl Fig. 4. Measured particle siu distributions for four Sumitomo aluminas demonstrates the wide particle size range capability of acoustic spectroscopyand the ability to provide quality control of a wide range of raw materials. Acoustic Spectra for same four aluminas fit precisely with theoretical curve based on output particle siu distribution. a typical rutile and alumina sample using electroa- coustic data. The pH at which the ~-potential goes to zero is referred to as the isoelectric pH. Differ- Table I Median particle size of Aluminas Sumitomo ent materials may have quite different isoelectric points as is evident from. this figure. If good Acoustics, DT - t 200 Manufacturer stability is desired, then one needs to operate far enough from the isoelectric point to achieve a ~-potential in excessof say 20-30 mY, either plus or minus. For the alumina shown this would suggest that to obtain optimum stability for this 140 A.S. Dukhin et oJ.jCoIloids and surfaces A: Physicochem.Eng. Aspects 173 (2lXXJ)127-158 readjusted to a point either significantly below or above this new isoelectric point. But how can this dramatic shift in the isoelectric point be ex- plained? The explanation is actually quite simple. The initial slurry had a very small level of con- tamination, which was insignificant in terms of the overall stability of the system.However, under acid conditions this minor component dissolved. Upon subsequentchange to more alkaline condi- tions, this dissolved material re-precipitated on the surface of the major silicon nitride compo- nent. Now this minor component, although present only in seemingly insignificant quantity, neverthelessdominated the surface chemistry of the silicon nitride material. By realizing that the final state is dependent on the history of the Fig. 5. Versatility of acoustic method is illustrated by particle sample the process could be modified to accom- size distributions for variety of ceramic materials. modate this change. alumina, one should adjust the pH to avoid the 4.2. Chemical-polishingmaterials pH range between 8 and 11 where the ~-potential is less than 20 mV. This complex relationship Modem chemical-mechanicalpolishing materi- between ~-potential and particle size distribution als (CMP) present a new challenge for measuring can be easily understood using acoustic spec- techniques.Three aspectsof this application cause troscopy [36]. difficulties when using instruments basedon tradi- In the real world, the situation is sometimes tional techniques. First, the particle size of a even more complex. The particle size and ~-poten- typical CMP slurry is too small for sedimentation tial is not just a function of the final chemical based instruments or electric zone instruments. state of the system, but may depend also on the Typically, the mean size of CMP materials is history of how the system reached this state. In approximately 100 om with no particles, or hope- other words, the complete history of the sample fully just a few, larger than 500 nm. Second, the may be important as is illustrated in Fig. 8, a case range in the size of the particles may be greater history of a plant manufacturing silicon nitride. than 1000:I which also eliminates many classical The red curve shows the ~-potential as the pH is techniques. Third, CMP systems are typically decreased to an acid condition. From just this shear sensitive. Shear caused by the polishing data alone one would think that a pH of 7-8 process itself or the delivery system may cause would provide adequate ~-potential for stability. unpredictable assembly of the smaller particles However, operating experience in the plant indi- into larger aggregates.However, these aggregates cated otherwise. A clue to the problem was found may be weakly formed and easily destroyed by by reversing the titration towards more alkaline subsequent sonication, high shear, or dilution. conditions. The reverse titration, shown by the Therefore, any technique which requires dilution blue curve, revealed that the isoelectric point de- or other sample preparation steps may in fact creasedby 2 pH units after the slurry was exposed destroy the very aggregatesthat one is attempting to this acid condition. Investigation of plant oper- to quantify by measurement.It was suggestedthat ations showed that the slurry processingnormally CMP systemsmust be characterizedas is, without included such an acid wash. In order to obtain any dilution or sample preparation. adequate stability following this acid condition, Acoustic spectroscopy provides an exciting al- the data suggests that the process pH must be ternative to more classical methods. However, A.S. Dukhinet al. / Colloidsand Surfaces A: Physicochem.BIIg. Aspects 173 (2fXXJ)127-158 141 there is one feature of acoustic spectroscopy with another dominant mode has not yet been which thus far has not been described sufficiently investigated. Yet, it is just this feature, the ability in the literature: namely the ability to characterize to recognize a small sub-population, which is a bimodal PSD. Although Takeda et al. [39] most critical for CMP studies. The paper [40] demonstrated that acoustic spectroscopyis able to addressesthis important issue. characterize bimodal distributions of mixed alu- Unfortunately, there is no agreement in the mina particles, the ultimate sensitivity in detecting literature as to the number of larger particles one very small sub-population in combination which might be allowed in a CMP slurry. For the '" .~ = .Q ~ DI O.. ~ Q ~ Fig. 6. The high sensitivity of acoustic attenuation spectra is illustrated by the large difference in the attenuation spectra for two different size alumina slurries as well as a 1:1 mixture of each. The validity of the theory is established by the good fit between the theoretical prediction and the experimental data. Each peak in the bimodal distribution for the mixed slurry agreesprecisely with a corresponding peak in the size distribution for the single component slurry. 142 A.S. Dukhin et aJ./ Colloids and Surfaces A: Physicochem.Eng. Aspects 173 (2tXXJ)127-158 . 50:50 ~ ;"" IJ..ury -, 2.0 .f\- . ., 16 ';; ."-in. iatmi.-. = . ,Q .. e = 1,2 Q .. ~ (,f) ~ 0.8 ~~, I " 0.4 O.o~ ~-: ~~:::;, .!,:'-:~c.~~~~"-..:--., 10-1 10-1 100 101 102 Di8Defer IfBni Fig. 7. The attenuation spectra for alumina, zirconia, and 1:1 mixture shows that three caseshave quite different spectra and that best-fit theoretical distribution find good solution for each case. Particle size distribution for each component in mixed slurry can be measured. The size distribution for each component in the mixed slurry agrees well with the particle size for each component measured separately. moment it will be assumed that only one large size assuming a certain model for describing the particle of 1 J!In size might be allowed per 100000 sound attenuation in terms of the physical proper- small 100 nm particles. This target sensitivity ties of the system. It follows therefore that this corresponds to large particles amounting to I % of target sub-population sensitivity needs to be the total weight of all particulates. translated into a corresponding precision and ac- Of course, an acoustic spectrometer does not curacy specification for the attenuation measure- directly measure particle size. In fact, it measures ment. It was shown in the paper [40] that, from a an attenuation spectra and calculates the particle theoretical standpoint, the required precision is ~ A.S. Dukhin et aI./ColJoids and Surfaces A: Physicochem.Eng. Aspects 173 (2(XJO)127-1S8 143 roughly 0.01 dB cm - I MHz - '. The set of exper- particles was added to a single component slurry iments were perfonned with a single component of smaller particles. Two slurries were used for the system of Dupont silica Ludox- TM to confinned small particles: Ludox- TM and Cabot SS25.Two that DT-1200 acoustic spectrometer indeed meets Geltech silica with nominal sizes 0.5 and 1.5 J1m this target requirement. were used as the model large particles. It was A second set of experiments was then made to shown that the change in the attenuation spectra test whether the attenuation spectra changed re- was statistically significant when the large parti- producibly when a small amount of the larger cles amounted to at least 2% of the total weight of "'---'-' ~ ~ alumina ~ ~ ~ t ,'~ ' II . . u 'j'~"I""7""-~- - .B . i ~ . ~ ...... rutile .. " silicon nitride / /--~ \ Decreasing pH ./ \ :J ;/ \ ""(ij ~ c \ G) \ "[ tV - 8 G) N '" \. , """""' " '-" Fig. 8. Slurry isoeiectric point suggestsoptimum pH for achieving stability. Curve for alumina slurry suggestsavoiding the range pH 9-10, whereas titania curve suggestsavoiding the range pH 3-4. Slurry stability depends not only on chemical state but how one reachesthis state. Titration of production silicon nitride to low pH indicates high ~-potential and good stability at pH 7-8. However, processengineers found that this pH did not provide good operating performance. Actual processincluded an acid wash of slurry. Back titration to alkaline conditions shows dramatic shift in isoelectric point requiring shift in plant operating conditions. ~ 144 A.S. DukhiII et aI.j CoUoidr and Surfaces A: Physicochem.EIrg. Aspects 173 (2IX») 127-158 .4 .l e:-:"" .. .. .0 f-- ~I c 0.8 -.- ~~ 0 . (".b8I 12 .~ . ~ Ludes-TM ~ 0.6 - . . . ~I"~ ~ := . ".- (~tI.~ < * ~-.-o.-- . A -- ~---""'" 0.0 10' 10 I 10. r~uency (MHz) I 7 6 '" "~. ~ .5 .=.. u "4j .. ~ g ) Q. 0 10 10 10. 10' Diameter(ilml Fig. 9. Attenuation spectra measuredfor Ludox TM silica, Gdtecb 0.5 and 1.5 silica, Cabot SS25and SSI2silica. Total solid content is 12 wt.% except for SS25 which is 25 wt.%. Particle size distributions corresponding to the measured attenuation spectra. all particulates.Expressed another way, the detec- pare the particlesize detennined by acousticspec- tion limit for this 12 wt.% slurry correspondedto troscopy for the five 12 wt.% test slurries with a sub-populationwhich was only 0.24 wt' % in independentdata from the manufacturers.The tenDSof the total sampleweight, or 0.24 g of valuesof the mediansize in eachcase is shownin large particlesper 100g of the slurry. Table2. It is interestingto note that thereis some The attenuation spectra and corresponding differencebetween the acousticallymeasured data PSD for all five singlecomponent silica slurriesis and that provided by the manufacturer.In large shownin Fig. 9. Thesetests allowed one to com- part this is relatedto differencesin the characteri- ~ A.S. Dukhin el al. / Colloid\' and Surfaces A: Physicochem.Eng. Aspects 173 (2(XKJ)127-158 145 zation technique. For instance, the size of the The final question is to determine whether the Ludox- TM slurry is determined by Dupont using calculated PSD calculates a correct bimodal dis- a titration method. This method yields an average tribution for these mixed model systems. The size on an area basis. Acoustic spectroscopy gives DT-1200 always calculates a lognormal and a a size on a weight basis, which for a polydisperse bimodal distribution which best fits the experi- system will always be somewhat larger than an mental data. These two PSD are best in the sense area based size. In addition acoustic spectroscopy that the fitting error betweenthe theoretical atten- implies some assumption about real dispersedsys- uation calculated for the best PSD and the exper- tem when particle size is being calculated from the imental attenuation is minimized. These fitting attenuation spectra. Any measuring technique errors are important criteria for deciding whether does the same. These assumptions and variation the lognormal or bimodal PSD is more appropri- in physical properties which are involved into the ate for describing a particular sample. For in- calculation can cause some variation in size as stance, the PSD is judged to be bimodal only if well. the bimodal fit yields substantiaUy smaller fitting Successfulreproducibility and reasonableagree- error than a lognormal PSD. It was shown that ment with other techniques then encouraged one fitting errors for the bimodal PSD is better than to test the ability to correctly determine bimodal the lognormal for aU of the mixed systems over PSD. Ludox TM or CMP SS12 small particles the whole concentration range and for both the were used as the major component of a slurry. The Geltech 0.5 or Geltech 1.5 were used as large and larger sized particles. According to the 'large' and 'larger' particles in the minor compo- fitting errors, aU PSD in the mixed systems are nent of the mixed slurry. In each case the minor bimodal, which of course is correct for these fraction was added to the Ludox- TM or the CMP known mixed systems. SS12systems in steps. Each addition increasedthe Details about error analysis implemented into relative amount of the larger particles by 2%. The the DT instruments software are described in the attenuation spectra was measured twice for each papers [36,40]. mixed system in order to demonstrate Acoustics yields information only about parti- cle size. It turned out that it is not sufficient for reproducibility. Fig. 10 illustrate the results of these mixed characterizing CMP slurries. Information about system tests. It is seen that attenuation increases ~-potential is very valuable for proper characteri- with increasing amounts of the 'large' or 'larger' zation of these materials. Electroacoustics opens particles. The increase in the attenuation with the way to perform this characterization with very increasing doses of the Geltech content is in all high quality of reproducibility and speed.Fig. 11 cases significantly larger than precision of the shows results of pH titration of silica Ludox and instrument. This demonstrates that the DT-1200 CMP slurry produced by ECC. It is seen that data contains significant information about the electroacousticsis able to characterize ~-potential small amount of large particles. below 1 m V with precision about 0.1 m V. This degreeof accuracy and precision exceedsanything Table 2 imaginable with microelectrophoresis Median particle size for various silica samples 4.3. Emulsion.S' There are many instances of successfulcharac- terization of the particle size distribution and ~-potential of emulsion droplets. There are two quite representative reviews of these experiments published by McOements [3] (acoustics) and Hunter [8] (electroacoustics)(Fig. 12). 146 A.S. Dukhin et a/.j Colloids and surfacesA: Physicochem.ERg. Aspects 173 (2lXXJ)127-158 ~ ~ '"E = ~ c 0 O'; ~ = c ~ <: 2.0 I. ~ 1.6 '" .~ 14 C +- SS25diluted + 3.S-/-Geltech 0.5 .Q - SS2Sdiluted + 2-/. Gel~h 0.5 ~ 1.2 ~ . SS2S,NVI diluled .~ ~ 1.0 Q C/) 0.8 A. 0.6. 0.4 :71 0.2 O.O~ 10.J 10" 10 I 10" 10' Diameter I,m) Fig. 10.Attenuation spectra and corresponding particlc sizedistribution for CabotSS25 silica diluted down to 12n% with various additions of Geltech 0.5 silica. Total solid content is 12 wt.%. Legend shows the fraction of the total solid content corresponding to the silica Geltech. Some results of a recent investigation are that affects surface chemistry. This factor was presented that were not published before. tested for reverse water-in-oil emulsion. The oil Various factors that affected stability, size phase was simply commercially available car lu- and ~-potential of the emulsion droplets were bricating oil diluted twice with paint thinner in investigated. order to reduce the viscosity of the final sample. One of the most important parameter affect- Fig. 13 illustrates results for emulsions prepared ing emulsions is the surfactant concentration with 6% by weight of water. A.S. Dukhin et al. / Colloids and Surfaces A: Physicochem.£IIg. Aspects 173 (2lXXJ)127-158 147 This Figure shows the attenuation spectra for size distribution of relatively stable emulsions. In three samples. The first sample was a pure oil many instances emulsions are found that are not phase and exhibited the lowest attenuation. It is stable at the dispersed volume concentration re- important to measurethe attenuation of the pure quired to obtain sufficient attenuation. signals dispersion medium when a new liquid is evalu- (usually above 0.5%). Hazy water in fuel emul- ated. In this particular case, the intrinsic attenua- sions (diesel, jet fuel, gasoline) may exist at low tion of the oil phasewas almost 150dB cm-1 at water concentrations of only a few 100 ppmv 100 MHz which is more than seven times higher (0.01%) of dispersedwater. Attempts at character- than for water. This intrinsic attenuation is a very izing these systems without added surfactant re- important contribution to the attenuation of ul- sulted in unstable attenuation spectra and water trasound in emulsions. It is the background for droplets were discovered to separate from the characterizing emulsion system. bulk emulsion and settle out on the chamber The emulsion without added surfactant was walls. This problem is less important for thermo- measured twice with two different sample loads. dynamically stable microemulsions. As the water content was increased the attenua- tion became greater in magnitude. For this sys- 4.4. Food emulsions tem, the attenuation was found to be quite stable with time. Addition of 1% by weight sodium bis Food emulsions is a very important and 2-ethylhexyl sulfosuccinate (AOT) changed the prospective application for acoustics and electroa- attenuation spectra dramatically. This new emul- coustics. These emulsions are quite concentrated sion with modified surface chemistry was mea- which eliminates traditional light based tech- sured two times in order to show reproducibility. niques for characterization purposes. Dilution The corresponding particle size distribution is protocol is not suitable for emulsions becauseit shown in Fig. 13 and indicates that the AOT destroys the original distribution of droplets. converted the regular emulsion into a microemul- Fig. 14 illustrate that acoustic spectrometer is sion as one could expect. able to provide reproducible and distinguishable These experiments proved that the acoustic attenuation spectra for a wide spectrum of dairy technique is capable of characterizing the particle products. One can see that attenuation spectra pH .' 7 . 9 I ...~. tc '~ ~ -181 " ;;.. \ ~c, backtitration .! -20 s '\ i .. , ECC, titration -38 Ludox .. . ! " 'f. -40 Fig. II. Titration of silica Ludox TM at 10 wt.% and chemical-mechanicalpolishing silica ECC ~ 148 A.S. Dukhin et oJ./ ColloidS'and Surfaces A: Physicochem.Eng. Aspects 173 (2000) 127-1. 2.4 N = 2-0 '" ~ e ~ 16 :2.= c ~ ~ ,---, 8Il~ 9:03:44 AM .2 12 '.\ ~ ~..- -.- ;: .41.'-' --- . 11/161'15K:S~:33 AM = - . . = - - --- 8/1~ 8:4.~:OJAM ~ 08 8/1619511:35:20AM < --- - 8/16I9~8:26:05 AM 0.4 00 10' 10' Frequency (MHz) reflects a fat content. There is a general trend of Nevertheless, other oils with close properties increasing attenuation with increasing fat content. can be used. In this particular case sound speed Attenuation spectra contains information about wasassumed equal to 1440m s- I, thermalexpan- droplets size. This PSD can be calculated using sion to 7.2 x 10-41 K -I, intrinsic attenuation to certain input parameters. In the caseof emulsions 5 dB cm I MHz-I at 100 MHz which is 25 sound speed,thermal expansionand intrinsic atten- times higher than water. Content of fat is 25 wt. % uation of the fat oil need to be known. In principle for ranch and 17 wt. % for maonese.Droplets'size these parameters are easily measurable if one has distribution for ranch salad dressing and may- fat oil as a one phase liquid. Unfortunately, dairy onesecorresponding to these input parametersare fat oil is not available at least here as a pure liquid. given on Fig. 14. A.S. Dukhin et a/.jColloids and surfaces A: Physicochem.Eng. Aspects 173 (2tXXJ)127-158 149 It is important to mention that different input verse microemulsions (water in oil) without the parameterswould lead to the different droplet size complication introduced by additional co-surfac- distributions. However, even this PSD are tant. Such a co-surfactant (usually alcohol) is somewhat helpful for understanding difference required by many other reverse microemulsion betweensize of the different dairy products. It can systems.This simplification makes the alkane/wa- be used for this purpose especially keeping in ter/AOT system a model for studying reverse mind high reproducibility illustrated on Fig. 15. microemulsions. There have been many studies devoted to char- 4.5. Microemu/sions acterization of these practically important sys- tems. Reverse emulsion droplets have been used The miJc.tureof heptane with water and AOT is as chemical micro reactors to produce nano size a classic three component system. It has been inorganic and polymer particles with special prop- widely studied due to a number of interesting erties that are not found in the bulk form [41-45]. features it exhibits. This system forms stable re- These microemulsion systems have also been a 2.0 .-G _AOT 'O", \(n i ~ I%AOT ~ 16 -'","00 E --. .. ,\ 0,0 J c~ :':" , 10' 10 I 10' FnqlleDCy (MHzI ~ , 1% \(r!' .~ ~ -+- l%AOT -= ... AOT i .t ~ Q C/) Q, Fig. 13. Attenuation and corresponding particle size distribution of 6 wt.% water-in-car oil emulsion and microemulsion created by adding sodium bis 2-ethylhexyl sulfosuccinate (AOT). ~ A.S. Dukhin et al. / Colloids and Surfaces A: Physicochem.Eng. Aspects 173 (2(}{XJ)127-158 . 3 i ~ e v ~ -'P- -YJ- = I ~~- :2. 2 -+- --- -,6c- Z%.-- W --- 1%""" Q-~__4 -.;.--,-- -'---~ 0 ;;l~;~::::!::!:::!::J:;;I~:J:*;I;=;;;~ to' 10' 10. '~umcy [MHzI Diameter(flm) Fig. 14. Attenuation spectra of the various dairy products and corresponding PSD. topic of research for biological systems and the ent techniques were used: PCS [50-55], classic AOT head groups have been found to influence light scattering [52,54,56), SANS [57-59], SAXS the conformation of proteins and increaseenzyme [51,60,61), ultracentrifugation [49,53,56), and vis- activity [46-49]. The unique environment created cosity [51,53,56). It was observed that the hep- in the small water pools of swollen reverse mi- tane/water/AOT microemulsions have water pools celles allows for increasedchemical reactivity. The with diameters ranging from 2 up to 30 nm. The increase in surface area with decreasein size of water drops are encapsulatedby the AOT surfac- the droplets also can significantly increasereactiv- tant so that virtually all of the AOT is located at ity by allowing greater contact of immiscible the interface shell. The size of the water droplets reactants. can be conveniently altered by adjusting the molar There have been many attempts to measurethe ratios of water to surfactant designated as R droplet size of this microemulsion. Several difTer- ([H20)/[AOT). At low R values (R < 10) the wa- A.S. Dukhin et al. / Colloids and Surfaces A: Physicochem.Eng. Aspects 173 (2(J(XJ)127-158 ter is strongly bound to the AOT surfactant polar tained from Sigma as HPLC grade (99 + % pu- head groups and exhibits unique characteristics rity). Known amounts of 18 Mll-cm water were different from bulk water [56]. At higher water added to the AOT -heptane solution using a 1 ml ratios, (R> 20), free water is predominant in the total volume, graduated glass syringe and then swollen reverse micellular solutions, and at ap- shaken for 30 s in Teflon capped glass bottles. proximately R = 60, the system undergoes a tran- The shaking action was required to overcome an sition from a transparent microemulsion into an energy barrier to distribute the water into the unstable turbid macroemulsion. This macroemul- nano-sized droplets, as it could not be achieved sion separates on standing into a clear upper using a magnetic stirrer. phase and a turbid lower phase. In all cases,the reported R values are based on The increase in droplet size and phase the added water, and were not corrected for any boundary can also be achieved by raising the residual water that may have been in the dried temperature up to a critical temperature of 55°C. AOT or heptane solvent. Karl Fischer analysis of In addition this system has been found to exhibit the AOT -heptane solutions before the addition of an electrical percolation threshold whereby the water resulted in an R value of 0.4. This amount conductivity increasesby several orders of magni- was considered to be negligible. tude by either varying the R ratio or increasing Measurementswere made starting with the pure the temperature [59,60,62,63]. Despite all these water and heptane and then the AOT-heptane efforts, there still remain questions regarding the sample with no added water (R = 0). The sample polydispersity of the water droplets, and few stud- fluid was removed from the instrument cell and ies are available above the R value of 60 where a placed in a glass bottle with a Teflon cap. Addi- turbid macroemulsion state exists. tional water was titrated and the microemulsion Acoustic spectroscopyoffers a new opportunity was shaken for 30 s before being placed back into for characterizing these complicated systems. De- the instrument cell. The sample cell contained a tails of this experiment are presented in the paper cover to prevent evaporation of the solvents. The [64]. The reverse microemulsions were prepared samples were visually inspected for clarity and by first making a 0.1 molar AOT in heptane rheological properties for each R value. These solution (6.1 wt.% AOT). The heptane was ob- steps were repeated for increasing water weight fraction or R ratios up to R = 100. At R ~ 60 the 4- microemulsions became turbid. At R > 80, the emulsions became distinctly more viscous. -ti7- 6ght el'eam The weight fractions of the dispersed phase -+- light eream -6raeeh were calculated for water only without including 3 -$-- rue" the AOT. Each trial run lasted approximately -.- eh 5-10 min with the temperature varied from 25 to 27°C. A separatemicroemulsion sample for R = = :2.2 j. 40 was madeup a few daysprior to the first = ~ study. For the R = 70 sample, a second acoustic -.c . measurement was made with the same sample = , = 41 used for the first study. The complete set of = 1 ~~::::::~--' -< Hghterea. experiments for water, heptane, and the reverse microemulsions from R = 0-100 was repeated to evaluate the reproducibility. 0 Attenuation spectra measured in the first run 10' 10' l~ up to R = 80 are presentedin Fig. 16. The results Frequency [MHzI for R = 90 and R = 100 are not reported because Fig. 15. Attenuation spectra of ranch salad dressing and light they were found to vary appreciably. As the water cream measured several times for illustrating reproducibility. concentration is increased, the attenuation spec- 152 A.S. Dukhin el aI. / Colloids and Surfaces A: Physicochem.Eng. Aspects 173 (2lXXJ)127-158 4- higher than that of the pure liquids. The increase R4J. _I~SI in attenuation, therefore, is due to this hetero- ~4-R=M, _1482 ,.- geneity of the water in the heptane system. The . extra attenuation is causedby motion of droplets, ";;3 .--'"". == .." . !4-~._I2.97 ",k . ! not separate molecules. The scale factor (size of ~ . . a , . i. ~4-R-so._IIOS droplets) corresponding to this attenuation is ~ . much higher than that for pure liquids (size of ~2 , R=40.wfr=904 = +- molecules). 0 .~ The current systemcontains a third component- = R-)()._94 = . . AOT. A question arises on the contribution of = . ... . ',-- 4; R=20.MI-o47J ...... '. .. .. AOT to the measured attenuation. In order to < .-. . .. ~-~..-11=10._2-4 ' ...... , ",.~. answer this question, measurementswere done on ...... III ..~..~ a mixture of 6.1 wt.% AOT in heptane (R = 0). It 0 ... is the third smallest attenuation curve on Fig. 13. 10' 10' 10' It is seenthat attenuation increasessomewhat due Frequency IMHzI to AOT. However, this increase is less than the extra attenuation produced by water droplets. Fig. 16. Acoustic attenuation spectra measured for water! sodium bis 2~thylhexyl sulfosuccinate (AOT);heptane system The small increase in attenuation is attributed to for differentwater to AOT ratios R. AOT micelles. Unfortunately thermal properties of the AOT as a liquid phase are not known and trum rises in intensity and there is a distinct jump the size of these micelles could not be calculated. in the attenuation spectrum from R = 50 to R = The particle size distributions corresponding to 60 in the low frequency range. This discontinuity the measuredattenuation spectra are presentedin is also reflected in the visual appearance as at Fig. 15. It can be seen that the distribution be- R = 60 the system becomes turbid. The smooth comes bimodal for R ~ 60 that coincides with the shape of the attenuation curve also changes at onset of turbidity. It is to be noted that such a R > 60. The stability and reproducibility of the conclusion could not easily be arrived at with system was questioned due to the irregular nature other techniques. However, Fig. 17 illustrates a of the curve so the experiment at R = 70 was peculiarity of this system that can be compared with independent data from literature: mean par- repeated and gave almost identical results. An ticle size increases with R almost in a linear additional experiment was run at R = 40 for a fashion. This dependencebecomes apparent when separate microemulsion prepared a few days ear- mean size is plotted as a function of R as in Fig. lier. This showed excellent agreement with the 18. results for freshly titrated microemulsion. It is seenthat mean particle size measuredusing For R values> 70, an increase in the viscosity acoustic spectroscopyare in good agreementwith and a decreasein the reproducibility of the atten- those obtained independently using the neutron uation measurementwere observed. This could be scattering (SANS) and X-ray scattering (SAXS) due to the failure of the model for this system as techniques [46,51,57]for R values ranging from 20 a collection of separatedroplets at high R values. to 60. A simple theory based on equi-partition of The two lowest attenuation curves correspond water and surfactant [36] can reasonably explain to the attenuation in the two pure liquids; water the observed linear dependence. and heptane. This attenuation is associated with At R = 10 the acoustic method gave a slightly oscillation of liquid molecules in the sound field. larger diameter than expected. This could be due If these two liquids are soluble in each other, the to the constrained state of the 'bound water" in total attenuation of the mixture would lie between the swollen reverse micelles. The water under these two lowest attenuation curves. But it can be these conditions may exhibit different thermal seen that the attenuation of the mixture is much properties than the bulk water used in the particle A.S.Dukhin et aI. /Colloidv and Surfaces A: Physicochem.Eng. Aspects 173 (2fXXJ)127-158 LS3 size calculations. Also at the low R values (R < 10 content. An increased concentration of water or < 2.4% water), the attenuation spectrum is not resulted in higher ~-potentials. However, the water very large as compared to the background heptane content was not the most important factor. This signal. Contribution of droplets to attenuation experiment was performed at two different AOT spectrum then may become too low to be reliably concentrations and the ratio of water to AOT (R) distinguished from the background signal coming was discovered to be the key parameter. When the from heptane molecules and AOT micelles. ~-potential was plotted versus the R values, the In addition to particle size, the Colloid Vibration samecurve was obtained for both AOT concentra- Current was also measuredfor calculating ~-poten- tions. This demonstrates that the ~-potential de- tial. The results are presented in Fig. 19 and the pends on the degree of the water surface coverage ~-potential was found to depend on the water by AOT molecules. .~ ; .Q ~ ~ .~ 2 Q ~ o. 10" 10" 10" 10' DiameterI..m) Fig. 17. Drop size distribution for varying R [H20]/[sodium bis 2-ethylhcxyl sulfosuccinate (AOT)l from 10 to 50 and from 50 to 80. 1$4 A.S. Dukhin el a/. / Co/wid\' and Surfaces A: Physicochem.Eng. Aspect.r 173 (2OfXJ)127-158 4.6. Latex There have been many successful experiments that have characterized latex systems using both acoustics and electroacoustics.For instance, Alle- gra and Hawley measured polystyrene latex. There is another successfulapplication, this time with neoprene latex, which is described in the paper [15]. The accuracy and precision of the 0-1200 Acoustic Spectrometerhave also been tested using . 10 20 3J «) s) «I M Standard Oow Latex with expected median parti- R fkJ&: ~d~Ml) cle size of 0.083 ~m. Results are shown in the 1"--- -,ft -- ~- I Table 3. These successfulexamples of characterizing la- Fig. 18. Comparison of mean droplet size measured using tex systemsare possible only when thermal expan- acoustic spectroscopy,neutron scattering and X-ray scattering. sion coefficients are known. Unfortunately, this parameter is not known for many latex polymers. This problem becomeseven more complicated for latex systemsthan for emulsions becausethe value of the thermal expansion depends strongly on the This experiment allows one to suggesta mecha- chemical composition of the polymer. Fig. 20 nism of electric charge formation on the surface illustrates this fact for severalethylene copolymers of the water droplets in the oil phase. This is a with different ethylene content. Variation of the field of great interest in modem emulsion science. ethylene content from 5 to 100/0was found to According to the experiment, the ~-potential ap- causes significant change in attenuation spectra. pears when there is a deficit of AOT molecules for This change is associatedwith the thermal expan- complete coverageof the water droplets. As more sion coefficient, but not the particle size. elements of the water phase become exposed to The uncertainty related with the thermal expan- the oil, higher values of the ~-potential are mea- sion coefficient makes latex systems the most sured. The water phase also contains a consider- complicated systems for acoustics. This is impor- able concentration of sodium ions that originate tant to keep in mind for testing a particular model from the AOT and serve as counter-ions to the of an acoustic instrument. Latex dispersions that are used as standards for light based methods negatively charged sulfosuccinatehead groups. As a result of decreasedsurface coverage, the water should be used with caution as in many casesthe thermal expansion properties of these standards droplets gain surface charge when they are in are not well known. contact with oil. This surface charge can appear becauseof ion exchangebetween the water and oil phase caused by the difference in standard chemi- 5. Conclusions cal potentials in each phase. Molecules of AOT do not create surface charge, but conversely It is hoped that it is proven with this short screen the surface charge of the initial water review that acoustics and electroacousticscan be droplets. At the same time these AOT molecules extremely helpful in characterizing particle size, change the interfacial tension creating conditions ~-potential and some other properties of concen- for a thermodynamically stable microemulsion. trated dispersed systems. Both methods are com- This is only a hypothesis so far and further inves- mercially available already. There are still some tigation is required for confirmation. problems with theoretical background for elec- A.S. Dukhin et al. / Colloids and Surfaces A: Physicochem.Eng. Aspects 173 (2lXXJ)127-158 15S: troacoustics of emulsions but analysis of the liter- separately. Electroacoustic phenomena is more ature shows gradual improvement in this field. complicated to be interpreted when comparing the The combination of the acoustic and electroa- acoustic ones becausean additional field (electric) coustic spectroscopy provides a much more reli- is involved. This problem becomes even more able and complete characterization of the disperse pronounced for a concentrated system. It makes system than either one of those spectroscopes acoustics favorable for characterizing particle size, > .i :s . a 8. i Fig. 19. ~-Potential measured electroaooustically for water droplets covered with sodium his 2-ethylhexyl sulfosuccinate (AOT) in heptane vs. water content. 156 A.s.DIIkJtbI et aJ./Co/Joids and Slufaces A: PhysicocAem.Eng. Aspects 173 (2tXXJ)127-151 Fig. 20. Attenuation spectra for latex dispersions with different content of ethylene. whereas electroacoustics yields electric surface ~ hi(b) bh.(b)- 21.3 properties. 1 It is believed that these ultrasound based tech- 'bl+2(12113-1.1~ niques is a very valuable addition to the tradi- tional colloid chemical arsenal of tools designed where b is radius of the cell: for characterizing surface phenomena. J b3=~ According to the paper [18] drag coefficient can be expressed in the following form general for both Kuvabara and Happel cell models: 0= _~2{d(C1h1+C~2+~~' h(x) = hl(x)h2 Kuvabara ( X2 3x Happel x e-~1 +/{ ~~ +j p+2p; 2p3 hib) c. e-.- [.lO] J.R. Allegra, S.A. Hawley, J. Acoust. Soc. Am. 51 (1972) 1545-1564. [II] J.D. McClements, Colloids Surf. 90 (1994) 25-35. [12] D.J. McClements, J. Acoust. Soc. Am. 9l (2) (1992) 849-854. [13] A.K. Holmes, R.E. Challis, D.J. Wedlock, J. Colloid Interface Sci. 156 (1993) 261-269. [14] AK. Holmes, R.E. Challis, D.J. Wedlock, J. Colloid Interface sa. 168 (1994) 339-348. h)(x) [15) A.S. Dukhin, PJ. Goetz, C.W. Hamlet, Langmuir 12 (21) (1996) 4998-5004. [l6] L.W. Anson, R.C. Chivers, Ultrasonic 28 (1990) 16-25. [17] N.L. Catton, The Neoprenes, Rubber Chemical Division, E.I. Du Pont De Nemours, Wilmington, DW, 1953. X+l . +j (-;-cosx+smx )] [18] A.S. Dukhin, PJ. Goetz, Langmuir 12 (21) (1996) 4987- 4997. [19] A.H. Harker, J.A.G. Temple, J. Phys. D.: Appl Phys. 21 h2(x) (1988) 1576-1588. [20] R.L. Gibson, M.N. Toksoz, J. Acoust. Soc. Am. 85 (l989) 1925-1934. [21] U. Riebel, et aI., Part. Part. Syst. Charact. 6 (1989) 135-143. [22] A.S. Dukhin, H. Ohshima, V.N. Shilov, PJ. Goetz, Elec- troacoustics for concentrated dispersions, Langmuir 15, Cell model in Happel version is more suitable 10 (1999) 3445-3451. for acoustics becauseit reflects properly hydrody- [23] A.S. Dukhin, V.N. Shilov, H. Ohshima, PJ. Goetz, Elec- troacoustics phenomena in concentrated dispersions. New namic dissipation within the cell. Cell model in theory and CYI experiment, Langmuir 15, 20 (1999) Kuvabara version is better for electroacoustics 6692-6706. because it automatically yields Onsager relation- [24) P.S. Waterman, RJ. Truell, Math. Phys. 2 (1961) 512. ship [30]. [25] R. Chanamai, J.N. Coupland, DJ. McCtements, Colloids Surf. l39 (1998) 241-250. [26] S. Temkin, Phys. Fluids 4 (11) (1992) 2399-2409. [27] S. Temkin, J. Acoust. Soc. Am. 103 (2) (1998) 838-849. References [28] R.W. O'Brien, J. Fluid Mech. 190 (1988) 71-86. [29] R. W. O'Brien, Determination of Particle Size and Electric [I] I.R. Pellam, I.K. Galt, I. Chern. Phys. 14 (10) (1946) Charge, US Patent 5,059,909,Ocl 22 (1991). 608-613. [30) A.S. Dukhin, V.N. Shilov, Yu. Borkovskaya, Dynamic [2) C.T.J. Sewell, Phil. Trans. R. Soc. Lond. 210 (1910) electrophoretic mobility in concentrated dispersed sys- 239-270. tems. Cell model. Langmuir 15, 10 (1999) 3452-3457. [3] P.S. Epstein, R.R. Carhart, I. Acoust. Soc. Am. 25 (3) [31] H. Ohshirna, A.S. Dukhin, J.Colloid Interface Sci. 2l2 (1953) 553-565. (1999) 449-452. [4] D.I. McClements, Adv. Colloid Interface Sci. 37 (1991) [32] L.L. Foldy, Propagation of sound through a liquid con- 33- 72. taining bubbles, OSRD Report No. 6.l-srlI30-1378, [5] V.A. Hackley, I. Texter (Eds.), Ultrasonic and Dielectric 1944. Characterization Techniques for Suspended Particultes, [33] E.L. Carnstein, L.L. Foldy, J. Acoust. Soc. Am. 19 (3) The American Chemical Society, Ohio, 1998. (1947) 48l-499. [6] J. Lyklema, Fundamentals of Interface and Colloid Sci- [34] F.E. Fox, S.R. Curley, G.S. Larson, J. Acousl Soc. Am. ence, vol. I, Academic Press, New York, 1993. 27 (3) (1957) 534-539. [7) R.I. Hunter, Foundations of Collid Science,Oxford Uni- [35] S. Ljunggren, J.C. Eriksson, Colloids Surf. 129-130 versity Press,Oxford, 1989. (1997) 151-155. [8] R.I. Hunter, Colloids Surf. 141 (1998) 37-65. [36] A.S. Dukhin, PJ. Goetz, Colloids Surf. 144 (1998) 49-58. [9] T .A. Strout, Attenuation of Sound in High-Concentration [37] A.S. Dukhin, PJ. Goetz, Method and device for charac- Suspensions:Developmentand Application of an Oscilla- terizing particle size distribution and zeta potential in tory Cell Model, A Thesis, The University of Maine, concentrated system by means of acoustic and electroa- 1991. coustic spectroscopy, patent USA, pending. .." A.S. Dukhin el aI. / CQl/oidsand Surfaces A: Physicochem.Eng. AofpeCIs173 (2OlXJ)127-158 [38] A.S. Dukhin, P.J. Goetz, Method and device for deter- [50] V. Crupi, G. Maisano, D. Majolino, R. Ponterio, V. mining particle size distribution and zeta potential in Villari, E. CaponeUi, J. Mol. Struct. 383 (1996) 171. concentrated dispersions, patent USA, pending. [51] B. Bedwell, E. Gulari, in: K.L. Mittal (Ed.), Solution [39] Shin-ichi Takeda, Characterization of ceramic slurries by Behavior of Surfactants, vol. 2, Plenum Press,New York, ultrasonic attenuation spectroscopy, In: V.A. Hackley, J. 1982. Texter (Eds.), Ultrasonic and Dielectric Characterization [52] E. Gulari, B. Bedwell, S. A1k:hafaji, J. Colloid Interface Techniques for Suspended Particulates, American Ce- Sci. 77 (I) (1980) 202. ramic Society, Westerville, OH, 1998. [53] M. Zulauf, H.-F. Eicke, J. Phys. Chern. 83 (4) (1979) 480. [40] A.S. Dukhin, P.J. Goetz, Characterization of chemical [54] H.-F. Eicke, in: I.D. Rob (Ed.), Microernulsions, Plenum Press, New York, 1982, p. 10. polishing materials (monomodal and bimodal) by means [55] J.D. Nicholson, J.V. Doherty, J.H.R. Clarke, in: I.D. of acoustic spectroscopy, Colloids and Surfaces 158 Rob (Ed.), Microemulsions, Plenum Press, New York, (1999) 343-354. 1982, p. 33. [41] J.P. Wilcoxon, R.L. Williamson, in: C.R. Safinya, S.A. [56] H.-F. Eicke, J. Rehak, Helv. ChilD. Acta 59 (8) (1976) Safran, P.A. Pincus (Eds.), Material Research Society 2883. Sympo~;um Proc., Macromolecular Liquids, vol. 177. (57] P.D.I. Fletcher, B.H. Robinson, F. Bermejo-Barrera, Materials Research Society, Pittsburg, 1990. D.G. Oakenfull, J.C. Dore, D.C. Steytler, in: I.D. Rob [42] F. Candau, in: C.R. Safinya, S.A. Safran, P.A. Pincus (Ed.), Microemulsions, Plenum Press,New York, 1982,p. (Eds.), Material Research Society Symposium Proceed- 221. ings: Macromolecular Liquids, vol. 177, Materials Re- [58] P.C. Cabos, P. Delord, J. Appl. Cryst. 12 (1979) 5Ol. search Society, Pittsburgh, 1990. [59] S. Radiman, L.E. Fountain, C. Toprakcioglu, A. de [43] L. Motte, A. Lebrun, M.P. Pileni, Progr. Colloid Polym. Vallera, P. Chieux, PrOgr. Colloid Polym. Sci. 81 (1990) Sci 89 (1992) 99. 54. [44] D. Ichinohe, T. Am, H. Kise, Synth. Met. 84 (1997) 75. [60] J.P. Huruguen, T. Zemb, M.P. Pileni, Progr. Colloid [45] K.V. Schubert, K.M. Lusvardi, E.W. Kaler, Colloid Polym. Sci. 89 (1992) 39. Polym. Sci. 274 (1996) 875. [61] M.P. Pileni, T. Zemb, C. Petit, Chern. Phys. Lett. 118 (4) (1985) 414. [46] F.M. Menger, K. Yamada, J. Am. Chern. Soc: 101 (22) [62] W. Sager, W. Sun, H.-F. Eicke, PrOgr. Colloid Polym. (1979) 6731. Sci. 89 (1992) 284. [47] D. Chatenay, W. Urbach, A.M. Cazabat, M. Vacher, M. [63] S.A. Safran, G.S. Grest, A.L.R. Bug, in: H.L. Rosano, Waks, Biophys. J. 48 (1985) 893. M. Oause (Eds.), Microemulsion Systems, Marcel [48] G.S. Timmins, M.J. Davies, B.C. Gilbert, H. Caldararu, Dekker, New York, 1987, p. 235. J. Chern. Soc. Faraday Trans. 90 (18) (1994) 2643. [64] T.H. Wines, A.S. Dukhin, P. Somasundaran, Acoustic [49] A. V. Kabanov, Makromol. Chern. Macromol. Symp. 44 spectroscopyfor characterizing heptane{H20{AOT reverse (1991) 253. microemulsions. J. Coil. Interfce Sci. 216 (1999) 303-308.