Emulsion characterization by the combined sedimentation field—flow fractionation—photon correlation spectroscopy methods

Emulsion Characterization by the Combined Sedimentation Field-Flow Fractionation-Photon Correlation Spectroscopy Methods KARIN D. CALDWELL AND JIANMIN LI Department of Bioengineering, University of Utah, Salt Lake City, Utah 84112 Received September 12, 1988; accepted December 19, 1988 The combination of sedimentation field-flow fractionation (sedFFF) and photon correlation spectroscopy (PCS) is shown to characterize the size distribution for polydisperse samples, where neither technique by itself is capable of producing this information. PCS, on the one hand, gives accurate diameter assignments for monodisperse suspensions but fails to describe the size composition of mixed samples, sedFFF, on the other hand, is a high-resolution separation technique that is capable of yielding detailed size information, provided the particle density is accurately known. Where this is not the case, the technique fails to evaluate the fractogram in terms of a size distribution. By combining the two characterization methods, the well size-sorted particles that dute from the separator can be evaluated in terms of both size and density. To yield accurate readings in PCS, the particle concentration must exceed a certain critical value; in FFF, however, accuracy is promoted by reduced sample loads. The implication of these constraints for combining the two techniques is discussed, and conditions that will reduce errors in the analysis of particle sizes preSent in a given sample are suggested. The protocol is applied to the analysis of two emulsions (Liposyn and Hercon 85) for which density differences between particle and suspension medium are not known from outside sources. Although these differences are found to be quite small, they can be determined with sufficient accuracy to permit the evaluation of particle sizes present in the two emulsions. © 1989AcademicPress,Inc. INTRODUCTION

Dispersed two-phase systems, consisting of submicrometer solid particles or liquid droplets suspended in a liquid medium, have a wide range of industrial uses. In addition to such classical products as paints, pigments, and cosmetic emulsions, a variety of pharmaceutical dispersions are currently being developed for intravenous use (1). These products include the new generation of slow-release drugs, whose active components are attached to particulate carriers or encased in liposomes, as well as the various vegetable oil emulsions administered as supplementary nutrients to patients incapable of oral intake. In each of these products, the size distribution of the suspended phase influences both function and stability; for those that are to be administered intravenously, it is particularly important that particle sizes at the high end

of the distribution not exceed the dimensions of the blood capillaries, or else they are likely to cause impaired circulation (2). For such suspensions, product evaluation and quality control are therefore highly dependent on methods for accurate determination of both average particle size and the distribution of sizes present in any given sample. A number of commonly used sizing methods are based on the measurement of light scattering, and among these, photon correlation spectroscopy (PCS) enjoys significant popularity (3). Routine measurements at a fixed angle of detection are convenient to carry out and require only a few minutes of analysis time. No sample preparations are necessary, aside from ensuring that the suspensions are kept dust free and their particle concentrations are kept within the range specified by the detection limit, on the one hand, and the occurrence of multiple scattering, on the other.

256 0021-9797/89 $3.00 Copyright© 1989by AcademicPress,Inc. Allfightsofreproductionin any formreserved.

Journalof ColloidandInterfaceScience,Vol.132,No. 1,October1, 1989

EMULSION

CHARACTERIZATION

For monodisperse samples, the fixed angle method yields particle diameters in good agreement both with electron microscopy and with the various methods based on sedimentation in a centrifugal field. Although frequently used also for quality control of polydisperse samples, such measurements are less meaningful, as they offer no information on the shape of the distribution curve. A separate group of size characterization methods is based on the sedimentation behavior in a centrifugal field. These methods involve the observation of either sedimentation rates or the concentration profiles established upon equilibration of the particle population in the field (4). Translation of such observations into values for the average size and size distribution of a sample requires knowledge of the densities both of the suspended particles and of the suspension medium. For suspensions of solid and stable particles, densities are easily measured with good accuracy, and sizing tools such as the JoyceLoebl disc centrifuge are therefore used with significant success (5). The situation is altogether different for emulsions or suspensions of vesicles and other fragile and easily deformable particles, for which an accurate sizing can be obtained only if both density and size are determined on the unperturbed sample. As its name implies, sedimentation fieldflow fractionation (sedFFF) is a member of the group of sedimentation-based characterization methods (6); however, its character of an elution technique sets it apart from the traditional centrifugation methods. Through the sedFFF process, even highly polydisperse samples are separated into fractions of considerable size uniformity (7), provided these fractions are well retained. If the particle density is accurately known, the elution volume for a given fraction is a direct measure of the size of its particle load. In the absence of a value for the particle density, these uniform fractions may be subjected to a PCS analysis for accurate evaluation of the particle diameter at the selected elution position (8). Although, in principle, only one correctly evaluated density

257

is needed to convert the fractogram into a raw size distribution for the sample [a volume correction and adjustments for Mie scattering are required to transform this distribution into its fully corrected analog, as described in Ref. (9)], the ability to correlate size and retention for several fractions increases the accuracy in the assigned density. It should be noted that some form of PCS-based detection may afford similar information in combination with other centrifugal or sedimentation-based separation techniques. The present study attempts to implement this promising approach to the analysis of polydisperse samples of the emulsion type. The selection of operational conditions results from the following considerations: 1. Since an accurate density assignment requires an accurately determined particle size which, if determined by PCS, requires a uniform population of particles, it is important to understand the relationship between sedFFF retention and the polydispersity of particles present in a given cut. 2. While the strong size selectivity (10) associated with high sedFFF retention is a necessary condition for accuracy in the PCS measurement, it results in small particle loads for any given fraction, and the PCS detection limits may be difficult to exceed. It is therefore necessary to establish the lower limit of concentrations for which the PCS gives reliable size data. 3. Although it is theoretically possible to increase the concentration of particles in any given fraction by increasing the size of the initial injection into the separator, this may lead to overloading (11) and a loss of accuracy in the relationship between retention and size. When present, overloading causes a shift in the elution profile which increases with increasing size of the injection. 4. As with other elution-based separation techniques, operation at slow flow tends to minimize zone broadening in sedFFF (12), and by trading speed for resolution, one may enhance the load of any given fraction beyond Journal of Colloid and Interface Science, Vol. 132, No. 1, October 1, 1989

258

CALDWELL

AND

LI

Boltzmann constant and temperature, respectively. When the sample is deemed to be comTHEORY pletely relaxed, the channel flow is resumed, The theoretical basis for FFF retention and and the various exponential particle clouds zone broadening has been described in detail begin to migrate downstream at rates deterelsewhere (12), and in this section we only mined by their respective X parameters. Parbriefly summarize those relationships that im- ticles small enough to be totally unaffected by pact the analysis of polydisperse samples of the field are associated with X's significantly unknown density. A recent simulation of the larger than unity and will elute at the void sedFFF process has highlighted certain slight volume V°, whereas all others will emerge at discrepancies between the actual particle be- retention volumes V, which exceed V° to the havior in the channel and that described by degree specified by their characteristic Xvalues. the analytical theory (13). These discrepancies, Cast in terms of the retention ratio R, the folwhich stem from the theory's assumption that lowing expression relates Vr to X for particles each particle type has the time to fully ran- assumed to behave as point masses: domize its distribution during channel passage, R = V°/Vr = 6X[coth(1/2X) - 2X]. [3] are present only at weak retentions. After five Particles whose size is comparable in magto eight column volumns of eluate have nitude to the thickness of the zone are miemerged, the coincidence between simulation grating with velocities reflected by both X and and analytical theory is complete. their diameter d (14) Immediately upon injection into the thin FFF channel, the sample begins a radial mi- R = 6y(d/2w - d2/4w 2) gration under the influence of the applied cen+ 6(1 - d/2w){coth[(1 - d/2w)/2X] trifugal field. During a period of interrupted carrier flow each particle size relaxes into its - 2X/(1 - d/2w)}. [4] specific exponential distribution, so that at the end of this equilibration period the sample is In this expression, factor 7 encompasses a vaarranged as a composite of characteristic ex- riety of velocity-dependent effects; however, ponentials at the head of the channel. For any its value is generally close to unity. The size selectivity Sa is a measure of the given particle size d, the concentration distridegree to which an elution based separation bution c(x) is given by (12) technique is capable of differentiating between c(x) = c(O)exp(-x/X(d)w) [1] particles of different diameter d (10). The definition introduced by Giddings where x represents distance along the field axis Sd = [ d i n V J d l n d[ [5] from the accumulation wall, c(0) is the concentration at that wall, w is the thickness of gives a convenient tool for comparisons bethe channel, and X(d) is a (dimensionless) tween techniques, as well as between ranges measure of the extension of the distribution. of particle sizes to which a given technique Parameter X(d) varies both with the strength can be applied. Clearly, the selectivity relates G of the applied field and with Ap, which is to the level of retention; for sedFFF the relathe difference in density characteristic for the tionship has been expressed as particle/cartier system; for spherical particles Sa = 9(R/36X 2 + 1 - 1/R). [6] its functional form is given by X(d) = 6kT/(dBAprcGw) [2] In this case, the size selectivity will therefore vary between zero, for particles eluting at the where k and T have the usual meaning of void, and 3, for those that are highly retained. the critical concentration required for accurate PCS sizing.

Journal of Colloid and Interface Science, Vol. 132, No. 1, October 1, 1989

EMULSION CHARACTERIZATION

259

The processes of zone broadening in FFF have been examined in some detail (12). For particulate samples in the colloidal size range, longitudinal diffusion effects are vanishingly small. In addition, systematic plate height studies on monodisperse latex standards retained in sedFFF systems have shown that instrumental band broadening can be neglected in comparison with the broadening caused by nonequilibrium effects for all but the most highly retained samples. Thus, ifa polydisperse sample is injected as a plug of negligible width and transported the full length L of the channel, the zone occupied by each particle size within the sample will have broadened in proportion to this size and the level of retention according to

The choice of cartier velocity will impact not only the polydispersity of any collected fraction, but also the concentration of material in this fraction. The degree of sample dilution which occurs as a function of flow rate is most easily analyzed for monodisperse samples where fractionation effects are absent, and all zone broadening stems from the nonequilibrium term of Eq. [7]. In this case, the sample will reach the channel exit as a Gaussian distribution along the distance coordinate z, with the maximum concentration Cmaxlocated at z =z0

H = X(~)w2(v)/D

a2r = HL.

[7]

where (v) is the linear cartier velocity and X(?0, the retention-dependent nonequilibrium coefficient, is approximated by (15) X(X) = 24X3(1 - 8X + 12X2)

[81

for R values below 0.7. By applying the Stokes-Einstein relationship between diffusion coefficient D, cartier viscosity n, and particle diameter d (D = kT/37r~ld), together with Eq. [2] which relates d to X, it is possible to express the plate height H entirely in terms of experimentally observed quantities (16):

H = X(X)(6/XApG)l/3(Tr/kT)2/3wS/33~7(v).

[11]

This distribution has a variance (in cm 2) specified by plate height H and column length L: [121

The relationship between Cmaxand the total amount a of sample injected into the column is found by integration of Eq. [ 11] between the limits of z = + Cmax = a/(2rcHL) 1/2

[ 13]

where a is a constant. Given the linear relationship between plate height and velocity expressed in Eq. [7], the maximum concentration is seen to vary inversely with ( v ) ~/2. EXPERIMENTAL

sedFFF System and Procedure [9]

In a study of reinjection procedures for the FFF analysis of polydisperse systems (7), Giddings and Yang developed an expression to account for the polydispersity found in a cut of volume A V, collected at a specific elution volume Vr. By introducing the symbol O-dfor the standard deviation in particle diameter within such a cut, and performing a slight rearrangement of the Giddings-Yang expression, one arrives at an equation suitable for the present discussion:

(ad/d) = [(H/L)Sd 2 × (1 + (AV/Vr)Z(L/12H)] ~/2.

c(z) = Cmaxexp[-(z - Zo)2/2aL2].

[10]

The general construction of the sedFFF system has been described in detail previously (6); the unit used in the present study had a length L of 96 cm, a thickness w of 0.0254 cm, and a void volume of 4.87 ml. This volume was determined from the elution pattern of a nonretained compound (acetone) observed at slow carrier flows, and agrees well with the value determined from the channel geometry. Carrier was pumped into the system by means ofa Minipulse II peristaltic pump from Gilson. A 0.2-1zm line filter was inserted between the pump and channel to ensure the delivery of dust-free medium to the system. Journal of Colloid and Interface Science, Vol. 132, No. 1, October 1, 1989

260

CALDWELL AND LI

Slight variations in the p u m p rate during the runs necessitated measuring the effluent volu m e by a graduated cylinder. The volumes of collected fractions were carefully measured and added to the cumulative volume in the cylinder. Prior to collection, the effluent was routed through an L D C U V monitor with a 254-nm light source and a cell volume of 8/A; the signal from this detector was fed to an Omniscribe flatbed recorder from Houston Instrument. Injection volumes ranged from 5 to 25 ~zl. Samples were injected into the channel with the carrier flowing at the rate selected for the analysis. Immediately after the injection the p u m p was turned off. The at-times significant afterflow was collected in a graduated pipet and subtracted from the known void volume to yield a corrected void volume for the particular run. The centrifugal field was then applied, and the samples were allowed to relax into their equilibrium distributions in the absence of flow. After relaxation periods of 2 0 40 min the flow was resumed and the effluents collected. Fractions of about 2 ml were collected (manually) directly into the disposable cuvettes used for the PCS measurement.

Photon Correlation Spectroscopy The PCS spectrometer used in this study was a 90 ° fixed angle unit of Model BI-90 from Brookhaven Instruments. This instrument has a H e - N e laser light source of 632.8-nm wavelength. Throughout this study, the BI-90 was operated in the standard mode, whereby the data analysis assumes a lognormal particle size distribution in the sample. Measurements were made at 25°C, and at least ten readings were taken for each sample. The filtered carrier ensured a low background of dust in these measurements.

Samples and Carriers Polystyrene latex spheres of known diameters were acquired from Seragen; their density was assumed to be 1.05 g/ml. All studies involving these samples utilized a 0. 1% aqueous Journal of Colloid and Interface Science, VoL 132, No. 1, October 1, 1989

solution of the FL-70 detergent from Fisher Scientific as suspension m e d i u m and/or cartier. The Liposyn II 10% intravenous fat emulsion is a product of Abbott Laboratories. This product contains equal amounts (5%) of safflower and soybean oil, emulsified with 1.2% egg phosphatides in water containing 2.5% glycerin. Hercon 85 is a ketene dimer aqueous emulsion used for paper sizing. This product was a gift from the manufacturer, Hercules, Inc. Both emulsions were analyzed using distilled water as the carrier. RESULTS AND DISCUSSION The accuracy and reproducibility of the sedimentation FFF measurement of particle size has been discussed in detail in previous work (16). For the purpose of the present study, a series of polystyrene latex particles with nominal diameters ranging from 278 to 672 n m were examined under a variety of field strengths and carrier velocities. This was done mainly to probe for systematic errors in the size assignment, such as would result from incorrect values for the rotor radius (which determines G in Eq. [2]) or the system's void volume (V ° in Eq. [3]). Elution volumes were collected and converted into values for the respective reduced layer thickness X by means of Eq. [3]; Eq. [2] was then used, in conjunction with the known value for the density difference between particle and cartier, to convert this X value into the corresponding particle size. The data are presented in Table I, and standard deviations for the set of FFF-based TABLE I Measured Diameters for Polystyrene Latex Standards Nominal diameter (nm)

sedFFF (nm)

PCS (nm)

173 278 394 672

183 + 2 277 + 4 383 + 4 599 + 4

171 + 2 265 + 3 403 _+9 623 + 9

EMULSION CHARACTERIZATION diameters are the result of five to ten determinations for each particle type. The same particles were also measured by PCS, using fixed angle detection. The standard deviations for these measurements represent at least ten measurements per particle size. Both sets of data are in general agreement with each other and with the size assignments given by the manufacturer. One notable exception is the largest bead of the set, whose nominal diameter is quoted as 672 rim. Although there is a virtual coincidence between the sedFFF and PCS measurements for this particle, there is a severe discrepancy between these values and the value given by the manufacturer. An error in the PCS value could possibly stem from the strong variations in scattering intensity with particle size in the range 600-700 nm (17) which are predicted by Mie theory and which, if unaccounted for, would lead to a low value for the diameter of a less-thanuniform population of particles. Steric effects and particle-wall or interparticle repulsion are known to cause lower-than-actual diameter assignments in sedFFF. Errors of the latter type are reduced or eliminated by operation at weaker fields, which give rise to less compressed particle zones at the analytical wall of the separator and therefore cause less concentration-related nonidealities. Despite a fivefold variation in field strength, the retention derived values for the diameter agreed to within 1% and showed no discernible trend toward increased values with weaker fields. Electron microscopic observations on these beads indicated values even lower than those obtained by the two methods compared here (550 rim). There is a clear difference between the sedFFF and fixed angle PCS methods in terms of their ability to identify particle diameters in bimodal or more complex mixtures, as illustrated in Fig. 1. While the B[-90 instrument in its standard operating mode, i.e., without forcing the data to display bimodality, assigns average diameters for the two mixtures shown in the figure, these values are less meaningful than in the case of the monodisperse samples. The sedFFF system, by contrast, has no dif-

261

ficulty resolving either of the test mixtures into their components with the correct diameter assignments. Although the two smallest beads (Fig. la) differ in diameter by only 17%, their baseline separation is accomplished at a modest field strength of 1100 rpm (2 l0 gravities). The high resolving power of the sedFFF system makes it ideally suited for the analysis of even highly polydisperse samples. If the density of a sample is accurately known, the fractogram can readily be converted into a representation of its size distribution. An actual distribution curve is obtained by applying the selectivity related scale correction, as has been described in detail elsewhere (9); if optical detection is used, the detector response also needs to be corrected for Mie scattering (9). From an inspection of Eqs. [2] and [3], it is evident that two sample specific parameters together determine the retention volume at a particular field strength: size d and density p; the latter appears in Eq. [2] as a difference in density between sample and carrier. The need to know the particle density, to interpret a sample's fractogram in terms of its size distribution, is a major drawback for those samples whose density is difficult to determine with the required accuracy. Previous work has established procedures for the sedFFF-based determination of size and density for stable colloids through a systematic variation in the density of the carrier solution (18). This approach is clearly not suitable for samples that undergo conformational changes or changes in composition as a result of changes in their environment. By determining the size of particles present in a specific sedFFF fraction through some other means, e.g., electron microscopy or PCS, it is possible to derive a Ap value for the particle suspension using Eqs. [2] and [3] or [4]. Once this value is accurately established, it may be applied to convert the fractogram into a distribution curve in the usual manner. A preliminary illustration of this approach employed samples of monodisperse polystyrene particles and polydisperse PVC beads, all of known density (8). Journal of Colloid and Interface Science, V ol. 132, No. 1, October 1, 1989

262 8,

CALDWELL AND LI

~ZO I 01 d = 290 :L4 n m

Ix

d = 385

nm

= 320 n m Voi,

;

s'o

DIAMETER (nm) Void

soo

d = 280

4

nm

0 d = 605

~

sl0 I I

500 DIAMETER (nm) nm

I=

i I

I 0

30 f~O 9'0 E L U T I O N V O L U M E (mL)

120

, 3'0 6*0 9'0 E L U T I O N V O L U M E (mL)

120

FIG. 1. Comparison of sedFFF and PCS analyses of pairs of polystyrene latex standards. (a) Samples with nominal diameters of 278 nm and 330 nm were mixed and separated by sedFFF at a field strength of 210 gravities and a flow rate of 1.41 ml/min. (The baseline shift results from a change in detector sensitivity.) (b) Samples with nominal diameters of 394 and 672 nm, mixed and separated under a field of 43.3 gravities and a cartier flow of 2.50 ml/min. The diameters indicated above each peak are calculated from their FFF retention, using a density difference of 0.05 g/ml. Inset in each figure: a PCS analysis of the unfractionated mixture. In the present study we have extended this approach to the analysis o f two emulsions whose densities were k n o w n to be close to that o f the suspension m e d i u m . According to Eq. [2], the crucial retention parameter ~ is inversely proportional to Ap, and for particles u n d e r nearly isopychnic conditions, any error in their density assignment will result in large shifts in the distribution calculated f r o m the fractogram. Thus, for particles suspended in distilled w a t e r at r o o m temperature (0 = 0.9964 g/ml), a mere 1% error in the particle density would lead to a 10% error in the density difference between particle a n d m e d i u m if this difference is a r o u n d 0.1 g/ml; for a A p o f a r o u n d 0.01 g/ml the same 1% error in particle density would give a 100% error in the density difference, and this error would increase tenfold for every tenfold decrease in z~p. Journal of Colloid and Interface Science, Vol. 132, No. 1, October 1, 1989

Effects of Concentration sedFFF is n o t a preparative technique, and for the highest accuracy in retention-based size assignments, sample a m o u n t s are kept below a hundred micrograms. While passing through the channel, a polydisperse sample undergoes significant dilution both due to the nonequilibrium effects described by Eq. [7] and as a result o f the actual fractionation o f its constituents. A sample o f this kind will often elute as a tailing peak whose concentration maxim u m falls near the void volume. Exactly where to sample this type o f a peak for subsequent analysis by PCS requires some consideration. First, it is necessary to establish the lower limit o f sample concentrations that can be sized by the spectrometer. This detection limit

263

EMULSION CHARACTERIZATION

will vary with particle size and index of refraction, as predicted by Mie theory (17). Figure 2 illustrates the effect of sample concentration on the PCS size assignments for several polystyrene standard particles using the BI-90 fixed angle instrument. Concentrations below those indicated in the figure failed to give a reading in this instrument. Clearly, in the case of this sample valid measurements are obtained only for concentrations above 1-2 ~tg/ ml for the smaller beads; for diameters above 400 nm the lower limit increases rapidly to reach 12 #g/ml for the 672-nm standard particles. With the loads normally used in sedFFF, these concentrations are reached only at or near the peak maximum for samples with an appreciable polydispersity. The relationship between peak concentration and plate height for a monodisperse sample is specified by Eq. [12]. Since zone broadening is a linear function of carder velocity, as indicated in Eq. [6], operation at high flow rates may result in excessive dilution of the sample during the separation process and thereby reduce the likelihood of an accurate PCS size determination anywhere in the elution peak. To avoid overloading effects (11), it may therefore be desirable to lengthen the

analysis time in return for higher accuracy in the sizing step. The second question concerns the polydispersity of particles present in a particular cut from the FFF fractogram. Figure 1 indicates that accurate size data are difficult to obtain for polydisperse samples using the PCS technique. For maximum accuracy in the sizing of any given fraction, one should therefore seek experimental conditions that minimize the polydispersity in that fraction. Equation [ 10] expresses the polydispersity in a given volume element, AV, as a function of the plate height H and selectivity Sd associated with its particular elution position, Vr. Although the volume of a cut clearly relates to the polydispersity of its content, A V is not truly a variable in this context, since it is specified by the volume of the cuvette used in the PCS measurement. For the present discussion, AVis assumed to equal 0.5V ° (2.4 ml). Figure 3 is an illustration of the relationship between polydispersity and elution position predicted for a field of 111 gravities and a density difference of 0.05 g/ml. From Eq. [7] it is clear that the H t e r m depends linearly on the cartier velocity (v); the flow rate is therefore affecting the polydispersity in each fraction, as shown in the figure. Thus,

14 ~

12

,~

lO

Z

o

8

~

6

[.,. Z r~

Z 0

4 2

0 ~A--~=.A IO0

I

200

i

•

I

300

I

I

400

m

t

500

I

I

600

J

700

d (nm)

FIG. 2. Effect of particle size and concentration on the detection limit in PCS. The data are collected for aqueous suspensions of polystyrene latex standards.

Journal of Colloid and Interface Science,

Vol. 132, No. 1, October 1, 1989

264 70

CALDWELL AND LI i-- I

~

i

I

i

I

I

I

I

I

J

I

it is helpful to refer to Eq. [2] which specifies the relationship between these properties and the experimentally determined retention parameter ~. Assuming that values for X can be accurately determined (to within 1%), it is immediately obvious that a 10% error in the measured d translates into a 30% error in the calculated Ap. If the sedFFF-PCS combination is to yield reliable values for Ap, it is therefore important that the sizing be made on fractions with the lowest possible polydispersity, yet the highest possible particle content to ensure an accurate reading in the PCS.

60-

~'~30

-

Application to Emulsions 10

0

~

i

41

I

~

,

~

I

1I

I1~

I

Vr/V° FIG. 3. Polydispersity in fractions collected from sedFFF at different levels of retention and different flow rates (Eq. [10]). The fraction volume is assumed to be 0.5 V°, the field strength 111 gravities, the density difference 0.05 g/ ml, and the viscosity 1.0 cP.

operation at slow flow is advantageous in terms of both enhancing the concentration of each particle type, as discussed above, and the related effect of reducing the polydispersity in each eluting fraction. Although, from the standpoint of the subsequent PCS measurement, the particle concentration is desirably high at or near the void volume, Fig. 3 gives a clear indication that this portion of the fractogram is associated with extremely high polydispersities. Fractions collected in this region are therefore not suitable for sizing by PCS, since they will lead to unacceptable errors in the determined density which is to form the basis for a subsequent transformation of the fractogram into its corresponding size distribution. To estimate the effect of errors in either of the two particle properties, size and density, Journal of Colloid andlnterface Science, Vol. 132, No. 1, October 1, 1989

Liposyn II. The Liposyn II fat emulsion is used routinely as a nutritional supplement, and is given intravenously to patients with impaired oral intake. This mode of administration requires the emulsion droplets to be significantly smaller than the blood capillaries to avoid clogging and circulatory disorders. Due to the clinical importance of such emulsions, we devoted our attention to their size analysis in an earlier sedFFF study (19). At that time, we were forced to base our size evaluation on an estimated density difference between carrier and droplet of 0.0827 g/cm 3, and we found the droplet size to range between 0.1 and 0.4/zm. At this time, 6 years later, a commercial sample of Liposyn II was obtained offthe shelf and subjected to analysis using the combined sedFFF and PCS methods. Prior to any fractionation attempt, the emulsion was found by PCS to have an average droplet diameter of 277 + 3 nm. To force as many sedFFF fractions as possible to exceed the lower PCS detection limit, a relatively large injection was needed. A series of five injections, ranging from 1 to 10 tzl of emulsion, was first analyzed at the chosen field strength of 390 gravities to examine the system for overloading effects. In all cases, the peak fraction emerged with an elution volume of 33 ml, and contained particles with an average

EMULSION CHARACTERIZATION

diameter of 266 n m as determined by PCS. Due to the absence of visible overloading effects, a 10-/A injection was thus considered acceptable, and a representative fractogram recorded for this sample size is shown in Fig. 4. The indicated nine cuts contained particle loads in excess of the PCS detection limit for that particle size, and the PCS evaluated diameter associated with each cut is listed in Table II. For each fraction, the average elution volume was converted into the appropriate retention ratio R. Equation [3] was then used to convert R values into the corresponding values for parameter ~, from which/xp could be calculated using Eq. [2]. Table II lists the density difference characteristic for this system, as determined for each fraction. These values are

2345678

0 '

20

40

60

80

100 120 140 Vr(raL)

245 315 364 404 440 471

d(nm)

FIG. 4. Fractogram of Liposyn II collected under a field of 390 gravities and a flow of 2 ml/min. Fractions were collected at the indicated retention volumes and analyzed by means of PCS (see Table II). The particle size axis is based on the average value for Ap (0.0146 g/ml) calculated for the cuts.

265

all significantly less than the value of 0.0827 g/ml assumed in our earlier work. The Ap values in Table II should presumably have been identical, yet there is a clear trend toward smaller values with higher retention. Presumably, the observed variation resuits from the errors in the measurement which were discussed above, although the possibility of additional systematic errors cannot be excluded at this point. Fractions 1 and 2 were collected in a range where polydispersities are relatively high, and where the analytical retention theory summarized above tends to somewhat overestimate the particle size, as shown in a recent simulation study (13). By contrast, fractions 4-9 are collected in a retention range where the theory yields accurate size assignments, where the polydispersity is low, and where the concentration is high enough to give PCS readings with a precision of +3%. Based on the average density difference calculated from fractions 4-9 (Ap = 1.42 × 10 -2 g/ml), in conjunction with Eqs. [2] and [3], it is now possible to convert the elution volume axis into an axis indicating the distribution of particle diameters within the sample, as shown in Fig. 4. Note that this fractogram needs a scale correction as well as a correction for Mie scattering to give a true representation of the sample's size distribution (9). However, for the purpose of evaluating an emulsion from the standpoint of its physiological suitability, it may be more important to demonstrate the presence or absence of droplets above a certain critical size, rather than to accurately quantify their amount. Hercon 85. Hercules, Inc. produces a series of emulsions for use as paper sizing agents. Among these, the Hercon 85 is of particular interest as a model substance for the development of sedimentation-based characterization techniques, since its particle density is very close to that of its aqueous suspension medium, and errors in the assigned particle density, therefore, would lead to particularly serious errors in the sizing of this sample. Journal of Colloid and Interface Science,

Vol. 132,No. 1, October1, 1989

CALDWELL AND LI

266

TABLE II Droplet Diameters Determined for Discrete Fractions of Liposyn II~ No,

Vr (ml)

~

Diameter, PCS (nm)

A# (g/ml)

Diameter, FFF b (rim)

1 2 3 4 5 6 7 8 9

16.0 21.5 26.0 32.0 43.0 51.0 57.5 64.0 69.0

0.04849 0.03500 0.02842 0.02274 0.01654 0.01375 0.01207 0.01074 0.00988

164 200 249 266 304 340 364 381 395

0.04883 0.03528 0.02877 0.02310 0.01697 0.01423 0.01258 0.01127 0.01044

227 253 271 292 323 343 357 370 380

a Conditions for the sedFFF run: field strength, 390 gravities; flow rate, 2 ml/min; 1/~¢o.,4.23 ml. Fractions of 2 ml were collected. PCS indicated an average diameter of 277 + 3 nm for the whole sample. b The FFF-based diameter is calculated from the experimental value for ~ (Eqs. [2] and [3]), in combination with the density difference of 0.0146 g/ml (average of the last six entries in the table). P r i o r to the s e d F F F analysis, the average d i a m e t e r o f H e r c o n 85 was m e a s u r e d b y PCS to be 573 + 8 n m , T h e s e d F F F s t u d y o f this s a m p l e was s u b s e q u e n t l y carried o u t at a field strength o f 420 gravities. As in the case o f the L i p o s y n e m u l s i o n , a series o f injections o f increasing size were p e r f o r m e d at this field, a n d the p e a k fraction was collected to ascertain the absence o f overloading. Again, n o systematic v a r i a t i o n in either the p o s i t i o n o f the p e a k fraction o r the P C S - b a s e d particle size associated w i t h this fraction was seen w i t h i n this series o f injections, a n d t h e largest s a m p l e o f the series, c o n t a i n i n g 15 #1 o f e m u l s i o n , was the basis for the a c t u a l sizing o f its d r o p l e t distribution. T h e f r a c t o g r a m in Fig. 5 is a representative picture o f t h e d i s t r i b u t i o n . O n l y the t e n first cuts i n d i c a t e d in t h e figure e x c e e d e d t h e lower d e t e c t i o n l i m i t o f t h e PCS i n s t r u m e n t ; the m e a s u r e d d i a m e t e r s for these cuts are listed in T a b l e III. T h e large size o f these particles necessitates a c o r r e c t i o n for steric effects o n the s e p a r a t i o n process. T h e m e a s u r e d d i a m e t e r s were therefore inserted in Eq. [4], together with the c o r r e s p o n d i n g R values, to yield the values for )~ r e p o r t e d in the table. F o r these calculations, p a r a m e t e r "y in Eq. [4] was given the value 0.7 (20). Journal of Colloid and Interface Science. Vol. 132, No. 1, October I, 1989

A t the selected field, the b u l k o f the s a m p l e elutes in the low r e t e n t i o n region, w h i c h is c h a r a c t e r i z e d b y relatively high polydispersi-

1

34567

Void *x\

/ / 10

(ii 2

\

=0

315

70

90-105

'425560650730768800

852

120 Vr(mL) 9]0 d(nln)

FIG. 5. Fractogram of Hercon 85 collected under a field of 417 gravities and a flow of 2 ml/min. Fractions were collected at the indicated retention volumes and analyzed by PCS (see Table III). The particle size axis is based on an average density difference of 0.00185 g/ml.

EMULSION CHARACTERIZATION

267

TABLE III Particle Diameters Determined for Discrete Fractions of Hercon 85a

No.

Vr (ml)

X

Diameter, PCS (nm)

Ap (g/ml)

Diameter, FFFb (nm)

1 2 3 4 5 6 7 8 9 10

23.6 27.0 29.6 34.6 40.0 42.0 50.0 55.0 62.0 67.0

0.03540 0.03056 0.02762 0.02323 0.01983 0.01870 0.01541 0.01385 0.01211 0.01110

470 455 473 531 541 621 661 679 697 718

0.002039 0.002603 0.002564 0.002155 0.002386 0.001674 0.001685 0.001729 0.001828 0.001825

490 515 532 564 594 606 646 670 700 721

a Conditions for the sedFFF run: field strength, 417 gravities; flow rate, 2 ml/min; V~o~r,4.74 ml; fraction volume, 2 ml. PCS indicated an average diameter of 573 + 8 nm. b The FFF-based diameter is calculated from respective ~ values (Eqs. [2] and [4]), in combination with the PCSbased diameter), using a density difference of 0.00185 g/ml (average of the last six entries in the table).

ties. In this region, one also introduces slight errors by using the analytical retention theory reviewed above. Therefore, we based our assigned density difference on data for fractions 5-10: Ap = 1.85 × 10 -3 g/ml. This value was subsequently used in conjunction with Eqs. [2] and [4] to establish the particle size axis in the figure. Unlike the Liposyn analysis discussed above, data for Hercon 85 (Table III) show no systematic decrease in the calculated Ap values with increasing retention, and fears of unknown systematic errors in the procedure appear unwarranted. CONCLUSION

The "hyphenation" of the sedFFF technique with photon correlation spectroscopy offers a viable approach to the otherwise difficult task of determining the particle size distribution in emulsions and other fragile colloidal suspensions. By a combination of these two independent size measurements, it is possible to accurately determine even very small density differences between colloid and suspension medium. Such density data are required to interpret any observation of the sample's sedimentation behavior, whether

obtained in an ultracentrifuge, a disc centrifuge, or a sedFFF separator, in terms of particle sizes present in the sample. The ability to collect a series of fractions whose particle load can be sized separately gives significant confidence both in those values for the density difference, which are derived from the combined sedFFF-PCS measurements, and in the analytical validity of the fractionation process itself. ACKNOWLEDGMENTS The authors are grateful for a Grant-in-Aid from Hercules, Inc., as well as for support from the National Institutes of Health through Grant 1 R01 GM38008-01A1. Cooperation from Brookhaven Instrument Corporation is likewise gratefully acknowledged.

REFERENCES 1. Groves, M. J., in "Modern Methods of Particle Analysis" (H. G. Barth, Ed.), p. 43. Wiley-lnterscience, New York, 1984. 2. Fujita, T., Samaya, T., and Yokoyama, K., Eur. Surg. Res. 3, 436 (1971). 3. Weiner, B. B., in "Modem Methods of Particle Analysis" (H. G. Barth, Ed.), p. 93. Wiley-Interscience, New York, 1984.

Journal of Colloid and Interface Science, Vol. 132, No. 1, October 1, 1989

268

CALDWELL AND LI

4. Maechtle, W., Makromol. Chem. 185, 1025 (1984). 5. Koehler, M. E., Zander, R. A., Gill, T., Provder, T., and Nieman, T. F., in "Particle Size Distribution: Assessment and Characterization" (T. Provder, Ed.), ACS Symposium Series No. 322, p. 180. Amer. Chem. Soc., Washington, DC, 1987. 6. Giddings, J. C., Yang, F. J. F., and Myers, M. N., Anal, Chem. 46, 1917 (1974). 7. Giddings, J. C., and Yang, F.-S., or. Colloid Interface Sci. 105, 55 (1985). 8. Caldwell, K. D., Jones, H. K., and Giddings, J. C., Colloids Surfaces 18, 123 (1986). 9. Yang, F.-S., Caldwell, K. D., and Giddings, J. C., J. Colloid Interface Sci. 92, 81 (1983). 10. Giddings, J. C., Caldwell, K. D., and Jones, H. K., in "Particle Size Distribution: Assessment and Characterization" (T. Provder, Ed.), ACS Symposium Series No. 332, p. 215. Amer. Chem. Soc., Washington, DC, 1987. 11. Caldwell, K. D., Brimhall, S. L,, Gao, Y., and Giddings, J. C., J. Appl, Polym. Sci. 36, 703 (1988).

Journalof ColloidandInterfaceScience,Vol.132,No. 1,October1, 1989

12. Karaiskakis, G., Myers, M. N., Caldwell, K. D., and Giddings, J. C., Anal Chem. 53, 1314 (1981). 13. Schure, M. R., Anal, Chem. 60, 1109 (1988). 14. Myers, M. N., and Giddings, J. C., Anal, Chem. 54, 2284 (1982). 15. Giddings, J. C., Yoon, Y. H., Caldwell, K. D., Myers, M. N., and Hovingh, M. E., Sep. Sci. 10, 447 (1975). 16. Giddings, J. C., Karaiskakis, G., Caldwell, K. D., and Myers, M. N., J. Colloid Interface Sci. 92, 66 (1983). 17. Bott, S. E., in "Particle Size Distribution: Assessment and Characterization" (T. Provder, Ed.), ACS Symposium Series No. 332, p. 74. Amer. Chem. Soc., Washington, DC, 1987. 18, Giddings, J. C., Karaiskakis, G., and Caldwell, K. D., Sep. Sci. Technol. 16, 607 (1981). 19. Yang, F.-S., Caldwell, K. D., Myers, M. N., and Girldings, J. C., J. Colloid lnterface Sci. 93, 115 (1983). 20. Caldwell, K. D., Nguyen, T. T., Myers, M. N., and Giddings, J. C., Sep. Sci. Technol, 14, 935 (1979).