Anal. Chem. 1997, 69, 3442-3450
Compositional Effects in the Retention of Colloids by Thermal Field-Flow Fractionation Sun J. Jeon, Martin E. Schimpf,* and Andrew Nyborg
Department of Chemistry, Boise State University, Boise, Idaho 83725
The retention of polystyrene and silica colloids that have been chemically modified is measured in several aqueous carrier liquids. Retention levels are governed by particle size and composition but are also sensitive to subtle changes in the carrier. Size-based selectivities are higher in aqueous carriers compared to acetonitrile. In aqueous carriers, retention varies dramatically with the nature of the additive, and for a given additive, retention increases with ionic strength, regardless of modifications to the particle surface. The role played by electrostatic effects in retention is studied by varying the ionic strength of the carrier, estimating electrical double layers, determining particle-wall interaction parameters, and calculating the coefficients of mass diffusion and thermal diffusion. Although electrostatic phenomena can affect mass diffusion and particle-wall interactions in carriers of low ionic strength (20
1 in aqueous carriers and Sd < 1 in ACN. According to eq 5, Sd will equal unity if DT is independent of size. From retention ratios, we can calculate DT values from eqs 1 and 2, using eq 4 to calculate D. Figure 2 plots the DT values as a function of particle diameter for most of the systems represented in Figure 1. (Silica in phosphate buffer is not included because the particle must be retained in order to calculate DT.) Consistent with previous reports, DT generally increases with particle size in the aqueous carrier liquids, while it decreases with size in ACN. The dependence of DT on particle size is in contrast to the situation with dissolved polymers, where DT is independent of size. However, the DT values in Figure 2 were calculated without considering the potential effect of the electrical double layer on the hydrodynamic diameter; that is, D values were calculated from eq 4 using the nominal diameter of the particle. Errors in D directly affect the calculation of DT from eq 1. Since the doublelayer thickness is independent of particle diameter, ignoring it will affect the dependence of DT on particle size. We can estimate Analytical Chemistry, Vol. 69, No. 17, September 1, 1997
3445
Figure 3. Elution profiles of silica before (no. 12) and after (nos. 16 and 17) silanization. The aqueous carrier liquid contained 3 mM NaN3 and 0.1 wt % FL-70; ∆T ) 40 K. Figure 2. Plots of DT versus particle diameter.
the double-layer thickness with the Guoy-Chapman theory, which gives the expression22
1/κ ) (kT/2e2I)1/2
(7)
Here, is the dielectric constant of the liquid, e is the electronic charge, and κ is the reciprocal double-layer thickness. Doublelayer thicknesses (1/κ) for the various carrier liquids are included in Table 2. While it is not possible to precisely quantify the increase in hydrodynamic size due to the double layer, the increase is certainly less than 1/κ. Thus, in the worst case, the calculated DT value for the smallest PS particle (91 nm) would be 10% high in the aqueous carrier containing 3 mM NaN3 and 0.1 wt % FL-70, while the largest PS particle would be only 3% high in the same carrier. Therefore, double layers cannot explain the 2-fold difference in DT values for these two particles. Neither can the double layer account for increases in DT when the NaN3 concentration is raised from 3 to 9 mM. The presence of particle-wall interactions will also affect calculations of DT because such interactions are not accounted for in eq 1. Such interactions are considered below, but first we consider the effect of serveral surface modifications on retention. Retention of Modified Silica. In the fractionation of dissolved copolymers by ThFFF, the outer region of the polymer-solvent sphere dominates the compositional dependence of retention. Previous work14,15,21 indicates that surface composition dominates the retention of particles as well. Figure 3 compares the retention of silica before and after silanization with either octyl- or octadecyltrichlorosilane. Silanization greatly alters the surface properties of silica, changing it from hydrophilic to hydrophobic. The result is a significant increase in retention, yet the length of the attached alkyl groups has little influence on the magnitude of the increase. The asymmetry in the elution profiles displayed in Figure 4 is not a consequence of overloading, since the profiles do not shift when the sample load is varied. The tail in the profile of unmodified silica increases with sample age, but it can be reduced somewhat by sonication of the particles prior to injection. This indicates that the tail contains aggregated material. Regarding the elution profiles of modified silica, peak fronting is probably due to variations in the extent of silanization. (22) Antonoff, J. Zh. Russ. Fiz.-Khim. O-va. 1907, 39, 342.
3446 Analytical Chemistry, Vol. 69, No. 17, September 1, 1997
Figure 4. Elution profiles of silica with and without different surface modifications: (a) 300 nm particles (nos. 14 and 18); (b) 400 nm particles (nos. 15 and 19). The aqueous carrier liquid contained 3 mM NaN3 and 0.1 wt % FL-70; ∆T ) 40 K.
Although the increase in retention with silanization is consistent with previous reports,15 the magnitude of the shift is much larger in this work. Thus, Shiundu et al. reported a 50% increase in the tr/t° value of 300 nm diameter silica upon silanization with octadecyltrichlorosilane. Using the same carrier liquid, tr/t° values increase by 330% with silanization in this work. The greater shift is not related to particle size, as we observe an identical shift of 330% with 200 nm particles (not shown). The discrepancy is due to the higher cold-wall temperature used in this work and the fact that the temperature dependence of thermal diffusion differs depending on the surface tension of the suspended particle. This was demonstrated in the work of Shiundu et al.,13 where it was found that in systems where the surface tension is high, retention increases with cold-wall temperature, while in systems of low surface tension, retention decreases with cold-wall temperature. The cold-wall temperature used in this work was 20 K higher than that used by Shiundu and co-workers. As a result, the retention of unmodified silica is lower in our work compared to that of Shiundu et al. while the retention of silanized silica is higher.
Figure 5. Elution profiles of unmodified (no. 4), carboxylated (no. 9), and aminated (no. 10) polystyrene of similar size in (a) 0.1 mM TBAP (acetonitrile), (b) 3 mM NaN3 and 0.1 wt % FL-70 (aqueous) and (c) 9 mM NaN3 and 0.1 wt % FL-70 (aqueous). ∆T ) 40 K.
Another example of the effect of surface tension on the retention of particles in water is demonstrated with silica that has been modified with bromopropyl versus aminopropyl groups. On the basis of the dipole moments of Si-OH, C-NH2, and C-Br bonds,19 the surface polarity, which is inversely proportional to the surface tension in water, is lower for unmodified silica compared to bromopropyl-modified silica and higher compared to aminopropyl-modified silica. Figure 4 compares the elution profiles of these three particles obtained under identical conditions. Retention increases when aminopropyl groups are attached to the surface, but not as much as the increase observed with silanization. When bromopropane groups are attached, retention decreases. Thus, the retention order for silica and the three derivatized forms of silica follows the order of decreasing surface polarity or increasing surface tension. Retention of Modified Polystyrene. In the silica experiments, we increased the retention of particles with hydrophilic surfaces by decreasing the surface polarity. In the next set of experiments, we study changes in the retention behavior of particles having hydrophobic surfaces as the surface polarity is increased and even charged at the proper pH. In order to make these comparisons, we utilize the availability of monodisperse polystyrene latex colloids that have been chemically modified with a high density of either carboxylate or amine groups. These functional groups impart surface charge to the colloids at the
appropriate pH. Figure 5 compares the elution profiles of carboxylated (PS/COOH) and aminated (PS/AB) polystyrene lattices in aqueous and organic carriers to the elution profiles of an unmodified latex (PS) of similar size. As illustrated in Figure 5a, retentions of the three samples in ACN are not significantly different. Although PS/COOH is retained slightly longer than the other two, it has a larger diameter, and the increase is no greater than expected from the size dependence illustrated in Figure 1. This behavior is in contrast to the trend observed in aqueous carriers, where PS/COOH elutes much earlier than the other two samples. All three samples elute earlier in 3 mM NaN3 compared to ACN. However, retention increases with the concentration of NaN3, so that in 9 mM NaN3 both PS and PS/ AB reach nearly the same level of retention as that measured in ACN. The increase in retention with NaN3 concentration is consistent with an increase in thermal diffusion with surface tension, since surface tension increases with salt concentration.22 However, we must also consider that retention can be affected indirectly by interactions between double layers. For example, repulsion of double layers may force particles into faster-moving flowstreams located away from the wall.18 The repulsion could be between particles or between particles and the accumulation wall. In either case, an increase in ionic strength would decrease repulsion and therefore increase retention. Interparticle repulsion is generally manifested as overloading, which can be eliminated by reducing the sample load. Since retention did not change when the sample load was reduced, we can rule out interparticle repulsion. To look at wall interactions, we utilize the semiempirical interaction parameter δw recently introduced by Williams et al.23 The interaction parameter is formally defined by
δw )
∫
∞
0
[1 - exp(-Vw(δ)/kT)] dδ
(8)
where δ is the distance from the particle surface to the accumulation wall and Vw is the energy of interaction between the particle and the wall as a function of the distance between them. The parameter δw is positive and negative for repulsive and attractive interactions, respectively. Its value is obtained by measuring the retention of a given particle over a range of field strengths and plotting the function
f(R,R,λ) )
RP - 6R(1 - R) 6λ[(1 - 2R) coth[(1 - 2R)/2λ]
)1+
δw l
(9)
Here, RP is the retention ratio perturbed by the particle-wall interaction, λ is defined by eq 1, R is the ratio of the particle diameter to the channel thickness, and l ) wλ. A plot of f(R,R,λ) versus 1/l yields a slope equal to δw and an intercept of unity. Plots of f(R,R,λ) versus 1/l are illustrated in Figure 6 for PS (no. 4), PS/COOH (no. 9), and PS/AB (no. 10) in aqueous solutions containing 0.1 wt % FL-70 and either 3 mM or 9 mM NaN3. The three points in each plot were obtained using ∆T values of 20, 30, and 40 K. For PS and PS/AB, the value of f(R,R,λ) is close to unity, regardless of field strength, indicating no wall interactions. This is consistent with previous studies in sedimentation FFF,18,23 where interactions of PS with the channel wall were (23) Williams, P. S.; Xu, Y.; Reschliglian, P.; Giddings, J. C. Anal. Chem. 1997, 69, 349-360.
Analytical Chemistry, Vol. 69, No. 17, September 1, 1997
3447
Figure 6. Plots of f(R,R,λ) versus 1/l. The particle-wall interaction parameter is obtained from the slope.
Figure 8. Plots of tr/t° versus FL-70 concentration in aqueous carrier liquids for unmodified (no. 4), carboxylated (no. 9), and aminated (no. 10) polystyrene particles of similar size: (a) without NaN3; (b) containing 3 mM NaN3. ∆T ) 30 K.
Figure 7. Plots of tr/t° versus NaN3 concentration in aqueous carrier liquids for unmodified (no. 4), carboxylated (no. 9), and aminated (no. 10) polystyrene particles of similar size. ∆T ) 30 K.
absent with I > 10-3 M. In contrast, a significant repulsive interaction is indicated in Figure 6 for PS/COOH in 3 mM NaN3. The repulsion is greatly diminished when the NaN3 concentration is increased to 9 mM. To more fully examine the effect of ionic strength on retention, we conducted a separate set of experiments where the NaN3 concentration was varied over a wide range, in both the presence and absence of FL-70 surfactant. For this set of experiments, ∆T was lowered from 40 to 30 K in order to maintain adequate but not extraordinarily high levels of retention over the entire range of electrolyte concentrations. The results are displayed in Figure 7. In the presence of FL-70, retention of each latex sample reaches a plateau as the NaN3 concentration is increased; the plateau concentration is approximately 30 mM. Plots of tr/t° versus NaN3 concentration for the three samples are generally parallel, with PS and PS/COOH exhibiting the highest and lowest levels of retention, respectively, in a given carrier liquid. The exception to this trend occurs in the absence of NaN3, where PS/ COOH is the only sample of the three that is retained at all. The minimum in PS/COOH retention with increasing NaN3 concentration was subsequently reproduced with freshly prepared and deionized carrier liquid. Retention levels are generally higher in the absence of FL-70, but the samples are more difficult to handle. Without FL-70, 3448 Analytical Chemistry, Vol. 69, No. 17, September 1, 1997
samples require periodic sonication to counteract their tendency to aggregate over time. When the NaN3 concentration is raised above 5 mM, retention times vary and elution profiles become asymmetric; sample recovery also diminishes. With further increases, the samples fail to elute. This behavior indicates sample adsorption to the accumulation wall, which is confirmed in the disassembled channel. In the case of PS and PS/AB, adsorption occurs before the plateau concentration of NaN3 is reached. The retention level where adsorption occurs is significantly below the levels achieved (tr/t° ) 30) with FL-70 (see Figure 1). Thus, adding electrolyte increases retention, but too much electrolyte induces sample adsorption unless a surfactant is used. To further characterize the effect of FL-70, we increased its concentration to 0.5 wt % in carrier liquids containing 0 and 3 mM NaN3. The results are displayed in Figure 8. In 3 mM NaN3, retention continues to decrease with increasing FL-70 concentration. In the absence of NaN3, only PS/COOH is retained, and its retention increases with FL-70 concentration. Considering the effect of surface tension on thermal diffusion, the decrease in retention of PS and PS/AB can be explained by the increase in surface polarity accompanied by the adsorption of FL-70 to the hydrophobic surface. With PS/COOH, on the other hand, retention increases with FL-70 in the absence of NaN3 due to the formation of ion pairs between the aliphatic amines in FL-70 and the carboxylate groups on the particle surface, which attenuates particle-wall repulsion and increases surface tension. Although FL-70 influences retention, the effect is not as strong as that of NaN3. Furthermore, the fact that FL-70 influences the retention of PS/COOH in an opposite manner depending on the presence of NaN3 is consistent with the observed minimum in this particle’s retention as the NaN3 concentration is increased (see Figure 7). Thus, the presence of NaN3 inhibits ion-pair formation, which is the mechanism by which FL-70 increases retention, but the resulting loss of retention is reversed by the growing benefit associated with further increases in electrolyte concentration.
Figure 10. Elution profiles of unmodified (no. 4), carboxylated (no. 9), and aminated (no. 10) polystyrene particles of similar size in 10 mM phosphate buffer at pH 4.7. ∆T ) 45 K.
Figure 9. Elution profiles of 280 nm diameter particles obtained in aqueous carrier liquids containing 0.1 wt % FL-70 and different concentrations of NaN3: (a) PS (no. 7); (b) PS/AA (no. 11). ∆T ) 20 K.
Since the amine groups in PS/AB are attached directly to aromatic rings, they are not protonated above pH 5, hence their low surface charge (see Table 1). PS/AA, on the other hand, carries a high charge below pH 9. Figure 9 compares the elution profiles of similarly sized particles of PS and PS/AA in aqueous carriers containing 0.1 wt % FL-70 and various concentrations of NaN3 (pH 9). The PS/AA profiles are broad but unimodal, indicating that either the size distribution is relatively broad or the size-based selectivity of the separation is unusually high. More significant to this work is the fact that the retention of PS/AA is not greatly affected by the concentration of NaN3, regardless of the presence of surface charge. This result further implies that while ionic strength affects retention, it does not dominate retention in aqueous carriers. If ionic strength does play a role in the retention of positively charged particles such as PS/AA, the role is not reduced by the addition of electrolyte. The results presented above demonstrate that retention can be manipulated by adjustments in the carrier liquid in order to separate similarly sized particles of polystyrene having subtle differences in their surface composition. To summarize, PS/ COOH can be separated from PS or PS/AB in solutions of FL-70, but resolving PS/AA from PS and PS/AB requires the addition of electrolyte. PS, PS/AA, and PS/COOH can be separated from each other using the proper concentration of NaN3 and FL-70. The separation of PS and PS/AB has proven more difficult. Although a carrier liquid containing 6 mM NaN3 and no FL-70 can achieve nearly base line resolution of these two particles, we found that phosphate buffer works better. As illustrated in Figures 10 and 11, phosphate buffer adjusted to pH 4.7 completely separates PS from PS/AB or PS/AA, as both of the aminated particles are poorly retained. Figure 10 compares the retention of PS and PS/COOH in the phosphate buffer. Note that the retention of PS/COOH, which is expected to carry less charge at pH 4.7, approaches that of PS. This is not surprising when all the results are considered. Certainly the electrostatic properties of the surface of PS and PS/
Figure 11. Elution profiles of unmodified (no. 7) and alkyl-aminated (no. 11) polystyrene particles of similar size in 10 mM phosphate buffer at pH 4.7. ∆T ) 40 K.
COOH are expected to converge as the carboxylate groups are protonated. However, it is significant that while the retention of PS/COOH approaches that of PS when the pH is reduced, it is still significantly lower than that observed in solutions of NaN3 at a higher pH. Thus, neutralizing surface charge does not necessarily increase retention, rather the effects of surface charge can be used to separate two particles that are otherwise similar in size and composition. In order to consider the effect of pH independent of other changes in the carrier liquid, we varied the pH in a series of NaN3 solutions by the addition of HCl. Figure 12 plots the trend in retention with pH for PS, PS/COOH, and PS/AB in these carriers. Consistent with the results presented above, the retention of PS/ AB is dramatically reduced by lowering the pH, particularly below pH 5. However, the retention of PS and PS/COOH also decreases with pH, even though the ionic strength of the carrier increases (see Table 2). While the effect of pH is more dramatic for PS/ AB, the fact remains that retention of PS/COOH continues to decrease as the pH is lowered well below the pKa (4.9) of benzoic acid. We note that FL-70 could not be included in the pH experiments because when the pH is forced below 8, the FL-70 no longer disperses in the carrier liquid. We were concerned that, without buffering the carrier liquid, the pH near the particle surface can be different from that in the bulk solution. To obtain more convincing evidence that protonation of the carboxylate groups does not in itself increase retention, we prepared a final set of carrier liquids consisting of Analytical Chemistry, Vol. 69, No. 17, September 1, 1997
3449
Figure 12. Plots of tr/t° versus pH for unmodified (no. 4), carboxylated (no. 9), and aminated (no. 10) polystyrene particles of similar size in aqueous carrier liquids containing 3 mM NaN3. ∆T ) 30 K. The pH was adjusted with HCl.
Figure 13. Plots of tr/t° versus pH for unmodified (no. 4), carboxylated (no. 9), and aminated (no. 10) polystyrene particles of similar size in 10 mM phosphate buffers. ∆T ) 30 K.
phosphate buffers with pHs ranging between 5 and 9. The effect of pH on retention of PS, PS/COOH, and PS/AB in these liquids is illustrated in Figure 13. The effect is similar to that in unbuffered NaN3 solutions. Regardless of pH or surface modifications, the retention of colloidal polystyrene is lower in phosphate buffer compared to other carrier liquids. An added disadvantage of the phosphate buffer is that when retention levels are increased by raising ∆T, particles begin to interact with the channel wall, with resulting distortion of the elution profiles. For the particles examined in this work, retention in phosphate buffers is limited to tr/t° < 10 at pH 9 and tr/t° < 5 below pH 7. CONCLUSIONS ThFFF is capable of separating particles by their composition as well as their size. Even subtle differences in surface composi-
3450 Analytical Chemistry, Vol. 69, No. 17, September 1, 1997
tion yield significantly different levels of retention in aqueous carrier liquids. Retention can also be manipulated by additives to the carrier. The flexibility of additives is clearly demonstrated by their ability to change the elution order of carboxylated, aminated, and unmodified polystyrene colloids of similar size. As a result, we are able to separate the modified particles from their unmodified counterparts with high resolution by simply manipulating the pH of the carrier. Such behavior is an important advantage in the compositional separation of particles by ThFFF. On the basis of the work outlined above, we offer the following guidelines for manipulating the compositional dependence of particle retention in ThFFF. (1) As is the case with dissolved polymers, the compositional dependence of retention is governed by interactions between the polymer and carrier liquids. For particles, that interaction can be summarized by stating that retention increases with the surface tension of the suspended particle. As a result, retention levels are profoundly affected by the nature of any additives to the carrier liquid. (2) A small amount of electrolyte is generally required to retain particles by ThFFF. Electrolytes tend to promote particle aggregation and adsorption to the accumulations wall, but such interactions can be avoided with the proper surfactant. A surfactant that has been used with great success for particle analysis by other FFF subtechniques is FL-70. Although ThFFF retention decreases when FL-70 is added to the carrier, higher levels of retention can ultimately be achieved with FL-70 through increases in ∆T without adsorption of particles to the accumulation wall. (3) For carriers with a fixed concentration of FL-70, retention increases with ionic strength. This effect is not explained by an attenuation of doublelayer repulsion, as suggested in previous reports. However, the effect is consistent with guideline 1, since surface tension increases with ionic strength in aqueous solutions. (4) For particles with a high surface charge, wall repulsion can be a factor, but wall repulsion is of no greater concern in ThFFF than in other FFF subtechniques and is easily attenuated by neutralizing the charge through adjustments in pH or an increase in ionic strength. (5) In carriers with a fixed ionic strength, retention increases with pH, regardless of the surface composition, provided wall repulsion is not a factor. ACKNOWLEDGMENT This work was funded by the Idaho EPSCoR and Grant CHE9634195 from The National Science Foundation.
Received for review December 31, 1996. Accepted June 3, 1997.X AC9613040 X
Abstract published in Advance ACS Abstracts, July 15, 1997.
