Вы находитесь на странице: 1из 11

Colloids and Surfaces A: Physicochemical and Engineering Aspects 172 (2000) 163 173 www.elsevier.

nl/locate/colsurfa

Characterization of concentrated colloidal suspensions using time-dependent photon migration measurements


S.M. Richter a, E.M. Sevick-Muraca a,b,*
a b

The Photon Migration Laboratory, School of Chemical Engineering, Purdue Uni6ersity, West Lafayette, IN 47907 -1283, USA Chemical Engineering Department, The Photon Migration Laboratory, 337 Zachry Engineering Center, Texas A&M Uni6ersity, College Station, TX 77843, USA Received 1 December 1999; accepted 24 March 2000

Abstract A method is presented to characterize concentrated suspensions of particles using frequency-domain photon migration in order to demonstrate its potential for on-line monitoring in the pharmaceutical and chemical based industries. In this study, the isotropic scattering coefcient was determined from FDPM measurements conducted at wavelengths between 448 and 835 nm using surfactant stabilized polystyrene latices of 195 335 nm mean diameter, 719% polydispersity and concentrations between 1 and 40% by volume. Upon using the Percus Yevick model for static structure of monodisperse spheres to account for hindered scattering in the inverse solution, determination of the volume fraction and mean particle diameter were considerably improved size at solids volume fractions of 10% or greater. Owing to (i) the relative refractive index of polystyrene in water, (ii) the inaccuracy of the rst order particle interaction model for suspensions with polydispersity \8%, and (iii) the minimal number of wavelengths used, it was difcult to recover the variance of the unimodal particle size distributions. Nonetheless, this work demonstrates the use of time-dependent measurements of multiply scattered light to characterize suspensions with solids volume fractions greater than 10%. 2000 Elsevier Science B.V. All rights reserved.
Keywords: Particle sizing; Light scattering; Static structure factor; Photon migration; Colloids

1. Introduction While the size characteristics of dense suspensions of colloidal particles signicantly impact the process characteristics of many chemical and pharmaceutical products, there remains no
* Corresponding author. Tel.: +1-979-4583206; fax: +1979-8456446. E-mail address: sevick@che.tamu.edu (E.M. Sevick-Muraca)

method for measuring these characteristics at solids volume fractions greater than 10% without the requirement for sample and dilution or empirical calibration. For example, ensemble techniques such as dynamic light scattering (DLS), angular scatter, and turbidity are restricted to 0.1% solids where multiple scattering and particle interactions do not inuence the detected signal. Fiber optic DLS and diffuse wave spectroscopy provide measurements at solids concentrations as high as 40%

0927-7757/00/$ - see front matter 2000 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 7 - 7 7 5 7 ( 0 0 ) 0 0 5 8 1 - 1

164

S.M. Richter, E.M. Se6ick-Muraca / Colloids and Surfaces A: Physicochem. Eng. Aspects 172 (2000) 163173

by volume, but need to account for the hydrodynamic particle interactions which impact the autocorrelation function in order to provide accurate size information. Diffuse reectance can likewise provide signals from highly concentrated, multiply scattering particle suspensions, but unless particle interactions are otherwise accounted for, will underpredict the scattering capacity owing to particle interactions. Particle interactions will typically occur at solids volume fractions \ 10% or alternatively when the average interparticle spacing becomes comparable or smaller than the wavelength of light. Ensemble techniques must account for particle interactions for accurate sizing in concentrated suspensions. Non-ensemble techniques, such as microscopy and focused beam reection methods reportedly perform well but are unable to provide size information in colloidal particle size ranges. Other ensemble techniques include acoustic or ultrasonic and electrokinetic sonic amplitude (ESA) spectroscopy. Acoustic or ultrasonic spectroscopy techniques are advertised for the determination of size characteristics in concentrated suspension, but practice shows that external calibration for the particular suspension must be performed for accurate prediction of size in concentrated suspensions. In addition, ultrasonic spectroscopy requires knowledge of the thermophysical properties of each phase, such as the density, specic heat capacity, thermal conductivity, thermal expansivity, and viscosity. While signicant progress has been made in the development of forward models to predict acoustic attenuation signals in concentrated suspensions, there has been little work to invert these measurements to obtain sizing information. Nonetheless, the application of ultrasonic spectroscopy to study emulsions is especially valuable (for example, see reference [1]) and the technique shows great promise for industrial sensing applications. Electrokinetic sonic amplitude (ESA) spectroscopy also depends upon the propagation of sound waves but differs in that the waves are generated from the oscillations of charged colloidal particles under the inuence of an alternating potential eld. Conversely, electroacoustic techniques focus on the detection of an alternat-

ing current in response to a propagating sound wave. While ESA approaches require empirical calibration for sizing in suspensions greater than 5% by volume, Dukin and Goetz [2] recently showed the ability to recover mean size of Ludox and titanium dioxide suspensions up to 45 volume percent using electroacoustic spectroscopy. Additional a priori information or measurements such as the zeta potential and uid permittivity are required for ESA and electroacoustic approaches. Finally, since these techniques measure attenuation, they require external calibration with a reference. In the following, we demonstrate the use of another potential colloidal sizing technique, frequency-domain photon migration, which is capable of recovering volume fraction and mean diameter from measurements in concentrated, surfactant-stabilized polystyrene suspensions. Our results are possible owing to (i) the accurate determination of the isotropic scattering coefcient, (1 g)vs, from self-calibrating, frequency domain photon migration measurements and (ii) the incorporation of the static structure factor, which accounts for particle interactions. In contrast to intensity or attenuation based techniques, FDPM measures the time-dependent propagation of light instead of the amount of light that reaches a detector. Since the phase-delay rather than the attenuation of the intensity modulated light is measured, instrument calibration to an external standard is not necessary. Also, the only a priori information required for FDPM sizing is the relative refractive index of the colloid. In the following, we briey describe the recovery of size and volume fraction information from light scattering measurements in concentrated, unimodal suspensions in which PercusYevick hard sphere models can accurately predict particle interactions for suspensions with low polydispersity (B 8%). Next, the experimental details of frequency-domain photon migration measurements of the isotropic scattering coefcient and the preparation of colloidal suspensions are described. Finally, we present our results, which illustrate the ability to characterize concentrated suspensions and the necessity to account for particle interactions.

S.M. Richter, E.M. Se6ick-Muraca / Colloids and Surfaces A: Physicochem. Eng. Aspects 172 (2000) 163173

165

2. Theory As the concentration of colloidal scatterers increases within a suspension, the average distance between scattering particles decreases. This reduction in the separation of scattering particles leads to particle particle interactions, resulting in spatial ordering of the particles. The isotropic scattering coefcient, (1 g)vs, for spherical particles is given by the expression [3]: (1 g)vs (u)
y

&&
0

3 q (n, x, q, u)(1 cos q) 2x scat

suspension of hard-spheres, the non-random arrangement arises primarily from the inability of particles to occupy the same volume. While the attractive and repulsive interactions owing to depletion, electrostatic, or van der Waals forces also contribute to alter scattering, these contributions are secondary and may be an order of magnitude smaller than the interactions that result from volume exclusion. The PercusYevick hard-sphere model for S(q) provides an expression for the dependence of structure on volume fraction, , and size, x, of a monodisperse suspension of hard-spheres [4]: 1/S(q)

4yn q sin qS sin f(x) dq dx u 2

(1)

= 1+

24 A(sin cos ) x3

where S(4y/u sinq/2) or S(q) is the static structure factor, and q is the magnitude of the wave vector, q=(4y/u sin(q/2)). Here qscat(n, x, l) is the angular scattering efciency for a particle of diameter, x, with relative refractive index, n, at wavelength, u, and q is the angle of scatter. In the absence of particle interactions, such as in a dilute suspension, S(q) is equal to unity. For a dense

+B +

A 24 6 + 4 1 2 sin 3 2 12 24 + cos 2 4

 

2 2 1 cos + 2 sin 2

 


(2)

n

Fig. 1. Plot of the PercusYevick model of the static structure factor as a function of the size parameter for various volume fractions.

where A= (12)2/(1 )4, B= 3(+ 2)2/ 2(1 )4, and , the size parameter, = 2yx/u. Fig. 1 shows the effect of the volume fraction on structure factor. As the volume fraction increases, long range order increases until the particles form a colloidal crystal, at which point Bragg diffraction occurs and the structure factor becomes a series of regularly spaced delta functions at x = 2u. S(q) approaches unity for high values, or alternatively large particles or small wavelengths, since the spatial correlation of the particles does not occur on the length scale of the wavelength of light. While the PercusYevik model is strictly valid for a monodisperse suspension, it provides an accurate prediction of static structure factor when polydispersity is less than 10% [5]. At values greater than 10%, polydiversity effects can be modeled using the approaches outlined by Lebowitz [6] and others. The static structure factor is typically measured in dense suspensions by monitoring the angular dependence of X-ray and neutron scattering. Small angle light scattering is also employed, but is restricted to dilute suspensions

166

S.M. Richter, E.M. Se6ick-Muraca / Colloids and Surfaces A: Physicochem. Eng. Aspects 172 (2000) 163173

Fig. 2. Plot of experimental measurements (points), predictions neglecting structure, and predictions determined by the incorporation of the PercusYevick model for the static structure factor versus volume fraction for a sample of 500 run polystyrene microspheres at 670 nm. (From Richter et al. [14]).

In the absence of structure formed by excluded hard sphere volume, the relationship between isotropic scattering and volume fraction is linear, as depicted by the dashed line in Fig. 2. The inuence of static structure profoundly affects the isotropic scattering coefcient as indicated by the experimental data points (lled circles) and as predicted by Eqs. (1) and (2) (solid line). Clearly, if colloidal particle sizing is to be accomplished in concentrated suspensions, particle interactions must be accounted for. Herein, we seek to demonstrate the ability to recover volume fraction and size information for concentrated suspensions from FDPM measurements of isotropic scattering as a function of wavelength.

3. Materials and methods

or very short path-lengths since multiple scattering confounds the collection of light scattered in direction q. In addition to hard-sphere interactions, attractive and repulsive forces can also contribute to the structure of the suspension. The static structure of a suspension impacted by electrostatic repulsive or attractive depletion forces can be computed by the mean sphere approximation, in which the comparatively small attractive and repulsive contributions to structure are added to those predicted by the Percus Yevick model. Using the approaches outlined herein, we have assessed changes in static structure owing to electrostatic and depletion forces [7 9]. Other investigators have assessed the impact of static structure on multiply scattered light. Kaplan et al. [10] investigated the binary hard-sphere static structure factor using diffuse transmittance measurements, while Garg et al. [11] and Shinde et al. [7] employed time-resolved and frequency-domain studies, respectively to show that the isotropic scattering coefcient of dense polystyrene latices can be accurately predicted by Eqs. (1) and (2). Fig. 2 illustrates the isotropic scattering coefcient measured for a 500 nm diameter surfactant stabilized polystyrene suspensions as a function of solids volume fraction at a wavelength of 670 nm.

3.1. Frequency-domain photon migration measurements


Owing to its ability to determine absorption and isotropic scattering properties without the need for external calibration, we chose to employ frequency-domain photon migration measurement of phase-delay of a propagating photon density wave for assessing the optical properties of our dense surfactant stabilized suspensions. Briey, FDPM consists of launching intensity modulated light ( f=50200 MHz) into a sample via a ber optic and then detecting the intensity modulated light collected by another ber optic some distance z away from the source. Due to the absorption and isotropic scattering properties of the suspension, the propagating intensity wave is phase-delayed and amplitude attenuated relative to the incident wave. The collected light is detected using a photomultiplier tube whose gain is modulated at the frequency of the source light modulation plus a small offset frequency of typically 100 Hz. The mixed signal is collected via standard A/D and evaluated for amplitude and phase delay using a fast Fourier transform. Details of FDPM heterodyne detection used in these experiments are described elsewhere [7]. Measurements were performed using sources at 458, 488, 514, 670 and 830 nm. Intensity modulation of the

S.M. Richter, E.M. Se6ick-Muraca / Colloids and Surfaces A: Physicochem. Eng. Aspects 172 (2000) 163173

167

blue and green light was accomplished by external modulation (EOM model 350 160, Conoptics, Danbury, CT.) of the 5 100 mW CW output of an argon ion laser (Model 543, Omnichrome, Chino, CA). At the red wavelengths, 500 mW of 670 nm light and 40 mW of 830 nm light were obtained from laser diodes (SDL-7432-H1 and Thorlabs EM325G, respectively) with DC bias provided by drivers (SDI, model SDU820, and Melles Griot 06 DID 203). RF modulation was accomplished by superimposing the amplied (ENI, RF amplier, 604L) output of a frequency synthesizer (Marconi Instrumets Signal Generator 2022A, MountainView, CA) through a bias network. Measurements of phase delay were made as a function of source intensity modulation frequency. Measurements of phase-delay resulting from light propagation between the source and detecting ber optics (1000 microns, HCP-M1000T-08, Spectron, Avon, CT) were made with optical bers, immersed in the concentrated polystyrene suspensions. However, since the measured phase delay reports both the electronic delays associated with the instrumentation as well as the optical delay due to the propagation of light in the sample, we conducted measurements of relative phase-shift, Dqrel. Relative phase-shift measurements consist of conducting phase-delay measurements at two different sourcedetector separations, z1 and z2. The relative phase-shift, Dqrel(z1 z2) as a function of modulation frequency is then t to the innite media solution to the FDPM diffusion equation to determine the isotropic scattering coefcient, (1 g)vs, and absorption coefcient, va (for example, see reference [12]).

ultraltered water were combined and allowed to equilibrate to 70C. 0.36 g of potassium persulfate (Aldrich c 21,622-4) were added as an initiator. When the reaction was complete, approximately half of the reactor volume was removed. Proper proportions of the reactants were added to replace the removed latex and the reaction was continued. This procedure was repeated three times to obtain progressively larger sizes. The resulting mean particle diameters were 196, 281, and 334 nm, as determined by DLS analysis of diluted samples (see below). Reaction times of up to 12 h resulted in latices of up to 40% solids volume. Dilutions of the reactor product were made using deionized, ultra-ltered water.

3.3. Dynamic light scattering measurements of f(x)


To provide a gold-standard measurement of particle size and size distribution, samples were diluted until transparent and then measured using dynamic light scattering (Coulter N4Plus, Beckman Coulter, Fullerton CA).

3.4. E6aporation studies for the determination of 6olume fraction


Volume fractions were determined by drying a known mass of the latex, Mwet at 98C. After 24 h, the mass of the dried sample, Mdry was measured. was then determined from the mass balance given by: Mdry (3) zPS (M Mdry)+ Mdry zH2O wet where zps is the density of polystyrene and zH2O is the density of water. =

3.2. Samples and sample preparation


Owing to the large volumes of surfactant stabilized polystyrene suspensions required, we produced our own using the emulsion polymerization recipe described by Miller et al. [13]. In a 2 l jacketed reaction ask (AceGlass Inc., Vineland, NJ), 1.4 g sodium soap (Sodium lauryl sulfate, Aldrich c 43,614-3), 0.11 grams sodium bicarbonate (Aldrich c 22,535-0), 660 ml of styrene (Aldrich c S497-2), and 1000 ml of deionized

3.5. Strategy for in6erting isotropic scattering coefcients to characterize dense suspensions
The goal of this research is to demonstrate the ability to obtain size information in concentrated colloidal suspensions and to show the utility of the hard-sphere interaction model as a rst order correction for structure effects. Our approach was

168

S.M. Richter, E.M. Se6ick-Muraca / Colloids and Surfaces A: Physicochem. Eng. Aspects 172 (2000) 163173

to (i) measure relative phase-shift, Dqrel(u, ), as a function of both wavelength, u, and modulation frequency, (ii) determine the isotropic scattering coefcient, (1 g)vs, from regression of Dqrel(u, ) data to the solution of the optical diffusion equation, and (iii) recover the size distribution and volume fraction from the inverse solution of Eq. (1) (with S(q) set equal to one) for non-interacting suspensions and of Eqs. (1) and (2) for interacting suspensions. The direct inversion of Eq. (1) for recovery of f(x) and is difcult due to the fact that the number of measurements is less than the number of unknowns. This renders non-unique solutions. In addition, the scattering property matrix is ill-conditioned, resulting in magnication of measurement error into non-physical and spurious recovered values for f(x) and . Owing to the small number of wavelengths employed in this study, we opted to regularize our inversions by simply assuming a Gaussian form of the size distribution. We also assume negligible polydispersity effects on the static structure factor as suggested by DAguanno and Klein [5] for suspensions with polydispersity less than 10%. Future work involves increasing the number of measurements and regularizing with a more general form of a distribution, i.e. a series of Weibull distributions and relaxing the static structure factor to include polydispersity effects. Nonetheless, by reducing the number of parameters to be recovered, the inverse problem is reformulated as an optimization problem, where the objective function to be minimized is expressed as: = %((v %(ui))0 (v %(ui))c)2 s s (4)

structed by xing the value of one of the parameters, either mean particle size, x, variance, |, or volume fraction, . The value of 2 is then calculated for the combinations of the remaining two parameters in the neighborhood around the correct values as predicted by Eqs. (1) and (2). We use the refractive index of polystyrene reported by Apfel et al. [14]. The resulting values of the 2 form the error surface. A well on the surface indicates a minimum in 2, and therefore, a possible solution to the inverse problem. In this rst study, we assume that the polydispersity of the suspension is less than 10% and that the Percus Yevick model is accurate.

4. Results and discussion

4.1. Error surface analysis


Fig. 3 illustrates the ln( 2) plots versus (a) mean, x and variance, |, (b) mean x and volume fraction, , and (c) variance | and volume fraction for an aqueous polystyrene suspension with x =300 nm, |= 20 nm, and =31%. The inverse solution was performed neglecting particle interactions, i.e. S(q)= 1, and with ten wavelengths available in the laboratory 325, 425, 488, 514, 633, 670, 730, 760, 780, 808, 835 nm. As with the error surfaces created for dilute (B 10% solids) aqueous polystyrene suspensions (data not shown for brevity), there is well dened minimum for a mean and volume fraction parameters, but not for the variance. When the PercusYevick model is incorporated into the inversions, similar error surfaces are obtained, as shown in Fig. 4. As can be inferred from the error surface plots at both high and low concentrations, accurate recovery of variance will be difcult due presumably to the small wavelength dependence on the refractive index of polystyrene. This is in contrast to metal oxides whose refractive index exhibit a strong wavelength dependence. Previously, we have shown the ability to reconstruct f(x) from as many as 15 wavelength measurements for both diluted, but multiply scattering polystyrene and titania mixtures [15,16]. In the dilute cases, it is comparatively more difcult to

where (v %(ui))0 is the experimentally observed s value for the isotropic scattering coefcient at wavelength ui and (v %(ui))c is that calculated from s Eq. (1). The minimization is performed using a simplex routine in the Matlab optimization toolbox (Natwick, MA). To better understand and evaluate the inverse results, error surfaces were constructed in order to examine both its effectiveness in the presence of noise and the sensitivity for recovery of parameters. The surfaces are con-

S.M. Richter, E.M. Se6ick-Muraca / Colloids and Surfaces A: Physicochem. Eng. Aspects 172 (2000) 163173

169

Fig. 3. Error surfaces determined without the incorporation of the static structure factor (a) Mean-variation (b) Mean-volume fraction and (c) Variation-volume fraction. The simulated measurements were for a Gaussian distribution of polystyrene with a mean of 300 nm, a variation of 20 nm and a volume fraction of 31%. 5% random Gaussian noise was added to simulate experimental measurments.

Fig. 4. Error surfaces determined with the incorporation of the static structure factor (a) Mean-variation (b) Mean-volume fraction and (c) Variation-volume fraction. The simulated measurements were for a Gaussian distribution of polystyrene with a mean of 300 nm, a variation of 20 nm and a volume fraction of 31%. 5% random Gaussian noise was added to simulate experimental measurments.

170

S.M. Richter, E.M. Se6ick-Muraca / Colloids and Surfaces A: Physicochem. Eng. Aspects 172 (2000) 163173

Table 1 Mean diameter and variance of suspensions used as determined by dynamic light scattering on dilutions of the stock suspensions Sample Suspension 1 Suspension 2 Suspension 3 Mean 28192 19593 33492 Variance (nm) 55 9 29 19 9 13 269 6

4.2. Independent measurements of Gaussian f(x) and 6olume fraction


Tables 1 and 2 list the values of mean9 SD, x, and variance9 SD, | (N=3) determined through DLS and the solids volume fraction as determined independently the evaporation procedure described above. While the standard deviation in the variance is rather large for the 280 nm latex, the 200 and 350 nm latices have polydispersities less than B 10%. As reported from Garg et al. [11] and Shinde et al. [7], we expect the PercusYevick model to capture the impact of particle interactions at least to the rst order.

accurately recover the variance of the distribution of polystyrene than that of titanium dioxides. However, there is currently no model for the static structure factor of non-spherical particles, as is the case for titanium dioxide. For this reason, polystyrene latex is used for these experiments, since it is comprised of spherical particles and is well-described by the Percus Yevick (P Y) model for static structure factor. In the formalism for sizing concentrated suspensions, we assume a unimodal suspension with low polydispersity. Hence, at this point in our study, we seek primarily the mean diameter and volume fraction measurement from a concentrated suspension.

4.3. In6ersion results


Tables 35 list the recovered parameters for mean, variance, and volume fraction when particle interactions are assumed negligible, i.e. S(q)= 1, or when they are signicant and modeled by the PercusYevick model Eq. (2). For the dilute cases involving the 280 and 350 nm latex, successful recovery of the variance was achieved.

Table 2 Volume fractions of samples used for this work as determined by evaporation studies Suspension 1 Sample Dilute Stock A C Volume fraction 1.28 17.2 25.7 31.5 Suspension 2 Sample Dilute 10% 20% Stock Volume fraction 2.22 8.02 17.5 33.0 Suspension 3 Sample Dilute 10% 20% Stock Volume fraction 2.30 10.7 17.2 33.8

Table 3 FDPM particle sizing results for a 281 nm latex Sample Mean (nm) S(q)= 1 Dilute (1.28%) Stock (17.2%) A (25.7%) C (31.5%) 275 258 252 249 PY 273 254 260 262 Variance (nm) S(q) = 1 25 1 1 1 PY 30 3 1 1 Volume fraction (%) S(q)= 1 1.19 13.7 17.6 21.5 PY 1.20 16.8 24.3 36.6

S.M. Richter, E.M. Se6ick-Muraca / Colloids and Surfaces A: Physicochem. Eng. Aspects 172 (2000) 163173 Table 4 FDPM particle sizing results for a 195 nm latex Sample Mean (nm) S(q)= 1 Dilute (2.22%) 10% (8.02%) 20% (17.5%) Stock (33.0%) 198 193 170 145 PY 199 201 190 206 Variance (nm) S(q)= 1 2 2 1 1 PY 2 1 2 1 Volume fraction (%) S(q)= 1 2.20 7.27 13.7 24.6

171

PY 2.26 8.31 17.8 39.8

Table 5 FDPM Particle sizing results for a 334 nm latex Sample Mean (nm) S(q) =1 Dilute (2.30%) 10% (10.7%) 20% (17.2%) Stock (33.8%) 350 318 329 330 PY 351 327 339 342 Variance (nm) S(q)= 1 25 1 1 1 PY 26 22 15 1 Volume fraction S(q)= 1 2.34 8.22 14.9 22.7 PY 2.39 9.19 18.1 31.3

However, in all inversions involving concentrated suspensions, in which the Percus Yevick model assumption of monodisperse particles is assumed, the inversion procedure failed predictably to recover the variance. In light of the error surfaces described in Figs. 3 and 4, it is not surprising that we encountered difculties recovering variance from the small number of wavelength measurements conducted. The data show that the recovered parameters of mean size tended to be more accurate when the PercusYevick hard-sphere was employed in 10, 20, and 30% solids volume suspensions. Since the suspension is surfactant stabilized, one would expect the mean size to remain constant as the suspension was diluted. Indeed for the 2819 2 nm mean diameter beads, the average of the recovered values were 25395 nm and 2599 4 nm with S(q)=1 and with P Y prediction of S(q), respectively. These results show inuence of structure on in recovery of the mean size that cannot be predicted by Percus Yevick. Since the large polydispersity of the 280 nm suspension may render the PercusYevick prediction of static structure in-

valid, the failure to recover the mean size in concentrated suspensions may not be surprising. It is noteworthy in the diluted 1% case, recovery of the mean size was closer to the value reported from DLS. On the other hand, for the stabilized polystyrene suspensions of low polydispersity (B l0%) where PercusYevick is expected to apply, the average recovered values of mean size for the 1959 3 nm polystyrene were 1699 24 nm and 19998 nm when S(q) was assumed to be unity and predicted by the Eq. (2), respectively. Clearly there is an improvement in the recovered mean size with the incorporation of particle interaction effects. For the 3349 2 nm suspension, the average recovered values of mean size were 325 9 7 nm and 3369 8 nm when S(q) was assumed to be one and predicted by the Eq. (2), respectively. Again, there is an improvement in the recovered mean size with the incorporation of particle interaction effects. Also, at 2.3 volume percent, it is noteworthy that the recovery of mean size using S(q)= 1 and the PY model provide comparable results as one would expect.

172

S.M. Richter, E.M. Se6ick-Muraca / Colloids and Surfaces A: Physicochem. Eng. Aspects 172 (2000) 163173

The most striking improvement with the incorporation of P Y in our inversion algorithm comes in the prediction of volume fraction as shown by the percent deviation from the value obtained from weight measurements before and after evaporation. Clearly in all cases the recovery of volume fraction is improved upon taking particle interactions into consideration.

Acknowledgements This work is supported in part by the National Science Foundation (CTS-9876583), the National Institutes of Health Research Career Development Award (K04 CA68374) and the DuPont Young Faculty Fellow Program.

References 5. Summary There is no doubt that the considerable improvement in the recovery of parameters describing the size distribution and volume fraction can be achieved with (i) a greater range and number of wavelength measurements, (ii) a more exible description of the size distribution such as a series of Weibull functions, and (iii) the incorporation of polydispersity effects on the static structure factor. Nonetheless, herein we have presented the concept of using a rst order model of particle interactions in combination with accurate isotropic scattering measurements using FDPM for on-line analysis of concentrated suspensions. As with all sizing approaches, FDPM has its limitations. First, because the measurement is theoretically limited to multiply scattered light, only samples with high scatter and low absorbance are suitable. However, advances by Venugopalan and co-workers, [17] promise to lessen this restriction. The second, albeit temporary limitation, is the lack of instrumentation which includes a multiwavelength source(s) which can be efciently intensity modulated at MHz frequencies and with sufcient optical power. Nonetheless, FDPM theoretically requires no additional information beyond the relative refractive index. In this initial study of polystyrene whereby measurements were limited to specic wavelengths, we assumed a normal, unimodal distribution of our colloid in our inversion approach. Certainly, more measurements and more sophisticated optimization will lessen this restriction. Owing to its self-calibrating nature (or the lack of need for calibration against an external standard), FDPM may in be especially suited for industrial applications [18,19].
[1] P. Kippax, J.D. Sherwoon, D.J. McClements, Ultrasonic spectroscopy study of globule aggregation in parenteral fat emulsions containing calcium chloride, Langmuir 5 (1999) 1673 1678. [2] A.S. Dukhin, P.J. Goetz, Acoustic spectroscopy for concentrated polydisperse colloids with high density contrast, Langmuir 12 (1996) 4987 4997. [3] M. Kerker, The Scattering of Light and other Electromagnetic Radiation, Academic Press, New York, 1969. [4] R.L. Hunter, Foundations of Colloid Science, vol. 2, Oxford University, Oxford, 1989. [5] B. DArguanno, R. Klein, Structural effects of polydispersity in charged colloidal suspensions, J. Chem. Soc. Faraday Trans. 87 (1991) 379 390. [6] J.L. Lebowitz, Exact solution of generalized Percus Yevick equation for a mixture of hard-spheres, Phys. Rev. 133 (1964) 895 899. [7] R.R. Shinde, G.V. Balgi, S.M. Richter, S. Banerjee, L.S. Reynolds, J.E. Pierce, E.M. Sevick-Muraca, Investigation of static structure factor in dense suspensions using multiply scattered light, Appl. Optics 38 (1999) 197 204. [8] S. Banerjee, R. Shinde, E.M. Sevick-Muraca, Probing static structure with multiply scattered light, J. Colloid Interface Sci. 209 (1999) 142 153. [9] S. Banerjee, R. Shinde, E.M. Sevick-Muraca, Determination of S(0,) in concentrated colloidal suspensions using multiply scattered light, J. Chem. Phys. 111 (1999) 9133 9136. [10] P.D. Kaplan, A.G. Yodh, D.J. Pine, Diffusion and structure in dense binary systems, Phys. Rev. Lett. 68 (1992) 393 396. [11] R. Garg, R.K. Prudhomme, I.A. Aksay, F. Liu, R.R. Alfano, Optical transmission in highly concentrated dispersion, J. Opt. Soc. Am. A 15 (1998) 932 935. [12] J. Fishkin, P. So, A.E. Cerussi, S. Fantini, M.A. Franceschini, E. Gratton, Frequency-domain method for measuring spectral properties in multiple scattering media: methemoglobin absorption spectrum in a tissue-like phantom, Appl. Opt. 34 (1995) 1143 1155. [13] C. Miller, P. Clay, R. Gilbert, M. El-Aasser, An experimental investigation on the evolution of the molecular weight distribution in styrene emulsion polymerization, J. Poly. Sci. A (1997) 989 1006.

S.M. Richter, E.M. Se6ick-Muraca / Colloids and Surfaces A: Physicochem. Eng. Aspects 172 (2000) 163173 [14] U. Apfel, R. Grunder, M. Balluff, A turbidity study of particle interaction in latex suspensions, Coll. Poly. Sci. 272 (1994) 820 829. [15] S.M. Richter, R.R. Shinde, G.V. Balgi, E.M. SevickMuraca, Particle sizing using frequency-domain photon migration, Part. Part. Syst. Charact. 15 (1998) 918. [16] E.M. Sevick-Muraca, J.E. Pierce, H. Jiang, J. Kao, Photon migration measurement of latex size distribution in concentrated suspensions, AIChE J. 43 (1997) 655 664. [17] V. Venugopalan, J.S. You, B.J. Tromberg, Radiative transport in the diffusion approximation: an extension for

173

highly absorbing media and small source-detector separations, Phys. Rev. E 58 (1998) 2395 2407. [18] G. Balgi, J.S. Reynolds, R.H. Mayer, R. Cooley, E.M. Sevick-Muraca, Measurements of multiply scattered light for on-line monitoring of changes in size distribution of cell-debris suspensions, Biotech. Prog. 15 (1999) 1106 1114. [19] R.R. Shinde, G.V. Balgi, S. Nail, E.M. Sevick Muraca, Frequency-domain photon migration measurements for quantitative assessment of powder absorbance: a novel sensor of blend homogeneity, J. Pharm. Sci. 88 (10) (1999) 956 966.

Вам также может понравиться