12798
Langmuir 2008, 24, 12798-12806
Aggregation of Nanosized Colloidal Silica in the Presence of Various Alkali Cations Investigated by the Electrospray Technique Ann-Catrin J. H. Johnson,*,† Peter Greenwood,‡ Magnus Hagstro¨m,† Zareen Abbas,† and Staffan Wall† Department of Chemistry, UniVersity of Gothenburg, SE-412 96 Go¨teborg, Sweden, and Eka Chemicals AB, a Business Unit within AkzoNobel, SE-445 80 Bohus, Sweden ReceiVed August 11, 2008. ReVised Manuscript ReceiVed September 1, 2008 The slow aggregation process of a concentrated silica dispersion (Bindzil 40/220) in the presence of alkali chlorides (LiCl, NaCl, KCl, RbCl, and CsCl) was investigated by means of mobility measurements. At intervals during the aggregation, particles and aggregates were transferred from the liquid phase to the gas phase via electrospray (ES) and subsequently size selected and counted using a scanning mobility particle sizer (SMPS). This method enables the acquisition of particle and aggregate size distributions with a time resolution of minutes. To our knowledge, this is the first time that the method has been applied to study the process of colloidal aggregation. The obtained results indicate that, independent of the type of counterion, a sufficient dilution of the formed gel will cause the particles to redisperse. Hence, the silica particles are, at least initially, reversibly aggregated. The reversibility of the aggregation indicates additional non-DLVO repulsive steric interactions that are likely due to the presence of a gel layer at the surface. The size of the disintegrating aggregates was monitored as a function of the time after dilution. It was found that the most stable aggregates were formed by the ions that adsorb most strongly on the particle surface. This attractive effect was ascribed to an ion-ion correlation interaction.
1. Introduction Aqueous colloidal silica dispersions, silica sols, have an extensive range of industrial applications including flocculation applications such as beverage fining and retention aids in paper making, as a binder in high-temperature applications such as foundry production, ceramics, and catalytic applications, as an additive in cementitious applications,1,2 as an abrasive in wafer polishing (WP) and chemical mechanical planarization (CMP),3 as solid state electrochemical devices,4 and as a grouting material when sealing hard-rock tunnels.5 Furthermore, silica sols are used in a vast number of coating applications to improve mechanical properties as well as antiblocking, adhesion, and wetting properties.1 Regardless of whether the aggregation is unwanted as in most coating and WP/CMP applications or is desirable as in foundry and gelling applications, the aggregation behavior as well as the size and especially the distribution of sizes of the silica particles is of vital importance to the performance of the silica sols. The stability of colloidal dispersions is usually described by Derjaguin-Landau-Verwey-Overbeek (DLVO) theory, where the total interaction energy between two particles is given by the sum of the attraction due to van der Waals interactions and the electrostatic repulsion originating from the overlap of the charged diffuse layers.6 At low electrolyte concentrations, the diffuse layer extends far from the particle surface, and the overall * Author to whom correspondence should be addressed. E-mail:
[email protected]. † University of Gothenburg. ‡ Eka Chemicals AB.
(1) Iler, R. K. The Chemistry of Silica; Wiley Interscience: New York, 1979. (2) Bergna, H. E. The Colloid Chemistry of Silica; American Chemical Society: Washington DC, 1994. (3) Lei, H.; Luo, J. Wear 2004, 257, 461. (4) Collinson, M. M.; Zambrano, P. J.; Wang, H.; Taussig, J. S. Langmuir 1999, 15, 662. ˚ . Tunnelling Underground Space Technol. 2006, (5) Funehag, J.; Fransson, A 21, 492. (6) Hunter, R. J. Foundations of Colloid Science; Oxford University Press: Oxford, U.K., 2001.
interaction is repulsive. An increased electrolyte concentration will decrease the extension of the diffuse layer and thereby reduce the repulsive contribution, eventually resulting in reversible or irreversible aggregation.6 In the case of silica particles, this theory is valid only for relatively large particles (several hundred nanometers) at large separation distances in low electrolyte concentrations,7,8 whereas much smaller silica particles demonstrate non-DLVO behavior with a local stability maximum at the isoelectric point (pH ∼2) and a local stability minimum around pH 6.1 The non-DLVO short-range repulsive interaction displayed by nanosized silica particles has been attributed to the existence of a structured layer of water molecules at the surface.9,10 The structured layer is thought to arise from hydrogen bonding between the water molecules and silanol groups on the particle surface. An increase in pH and/or the addition of cations with the ability to exchange the hydrogens on the silanol groups will attenuate the hydration layer.9 Some experimental results contradict the existence of a structured water layer at the silica surface. For instance, the heat of immersion classifies silica as a structure-breaker surface,11 and Horn et al. have shown that the viscosity of an aqueous electrolyte solution close to the particle surface does not differ from the viscosity of the bulk solution.12 Some authors have proposed that the short-range repulsion occurs as a result of gel layers located on the surfaces of the particles.8,13,14 These gel layers can be envisioned as consisting of covalently bonded polymeric chains of silicic acid protruding from the surface14 and possibly polymeric silicic acid hydrogen bonded either to (7) Hartley, P. G.; Larson, I.; Scales, P. J. Langmuir 1997, 13, 2207. (8) Kobayashi, M.; Juillerat, F.; Galletto, P.; Bowen, P.; Borkovec, M. Langmuir 2005, 21, 5761. (9) Allen, L. H.; Matijevic, E. J. Colloid Interface Sci. 1969, 31, 287. (10) Chapel, J.-P. J. Colloid Interface Sci. 1994, 162, 517. (11) Dumont, F.; Warlus, J.; Watillon, A. J. Colloid Interface Sci. 1990, 138, 543. (12) Horn, R. G.; Smith, D. T.; Haller, W. Chem. Phys. Lett. 1989, 162, 404. (13) Yates, D. E.; Healy, T. W. J. Colloid Interface Sci. 1976, 55, 9. (14) Vigil, G.; Xu, Z.; Steinberg, S.; Israelachvili, J. J. Colloid Interface Sci. 1994, 165, 367.
10.1021/la8026122 CCC: $40.75 2008 American Chemical Society Published on Web 10/14/2008
Aggregation InVestigated by Electrospray Technique
the surface or to the covalently bonded chains.1,2 At high pH values, the negatively charged chains within the gel layer repel each other, and the chains are repelled by the particle surface; these effects inflate the gel layer. The addition of electrolyte will screen the charge on the chains as well as the charge on the surface, causing the gel layer to deflate.14 Equicharged ions present in the same concentrations influence colloidal interactions differently, a phenomenon known as ion specificity.15 A good example is the specific screening efficiency of alkali counterions on the silica surface. This can be related to the hydrated structures of both the silica surface and the counterions.11,15-17 However, the precise structure of the silica-water interface is not known. In recent years, considerable efforts have been made to elucidate the hydrated structures of alkali ions by both experimental and theoretical methods.18-22 Despite these efforts, a coherent picture of alkali ion hydration has yet to evolve. A recent review of neutron diffraction studies22 has shown that the hydration numbers of Li+, Na+, and K+ are 4, 5, and 6, respectively. However, measurements with dielectric spectroscopy have shown that Li+ has two tightly bound hydration shells and Na+ has only one tightly bound hydration shell and a loosely bound second hydration shell, whereas the dielectric spectra of K+, Rb+, and Cs+ show no such tightly bound water molecules.19 The aggregation of colloidal silica has been extensively studied throughout the years. Various techniques have been utilized, such as static and dynamic light scattering,8 viscosity and turbidity measurements,16,23,24 and single-particle optical sizing.25 In recent years, a limited number of papers have been published that describe the study of colloidal nanoparticles through the application of methods from aerosol science.26-28 These methods are generally not well known within the field of colloids. Notable among these is work by Lenggoro et al., where electrospray aerosol generation was combined with a differential mobility analyzer to evaluate the size distributions of several kinds of particles, including silica, with diameters between 10 and 100 nm.28 Our purpose with this work is to show that it is possible to use this experimental technique for kinetic evaluations of the aggregation behavior. In the present work, we employ electrospray combined with a scanning mobility particle sizer to investigate the reversible formation and disintegration of nanosized colloidal silica aggregates in the presence of various alkali cations. The effect of the type of counterion on the variation of the size distribution during aggregation is shown, and the importance of additional non-DLVO interactions is discussed.
2. Experimental Section Materials. The colloidal silica used for the aggregation experiments in this study is a commercially available amorphous silica dispersion (Bindzil 40/220, Eka Chemicals AB, a business unit within AkzoNobel). This silica sol is stabilized by a low concentration of (15) Bostro¨m, M.; Deniz, V.; Franks, G. V.; Ninham, B. W. AdV. Colloid Interface Sci. 2006, 123, 5. (16) Trompette, J. L.; Meireles, M. J. Colloid Interface Sci. 2003, 263, 522. (17) Lyklema, J AdV. Colloid Interface Sci. 2003, 100-102, 1. (18) Ohtaki, H.; Radnai, T. Chem. ReV. 1993, 93, 1157. (19) Wachter, W.; Fernandez, S.; Buchner, R.; Hefter, G. J. Phys. Chem. B 2007, 111, 9010. (20) Buchner, R.; Hefter, G.; May, P. M. J. Phys. Chem. A 1999, 103, 1. (21) Chen, T.; Hefter, G.; Buchner, R. J. Phys. Chem. A 2003, 107, 4025. (22) Varma, S.; Rempe, S. B. Biophys. Chem. 2006, 124, 192. (23) Franks, G. V. J. Colloid Interface Sci. 2002, 249, 44. (24) Colic, M.; Fisher, M. L.; Franks, G. V. Langmuir 1998, 14, 6107. (25) Barany, S.; Cohen Stuart, M. A.; Fleer, G. J. Colloids Surf. 1996, 106, 213. (26) Chen, D-R.; Pui, D. Y. H.; Kaufman, S. L. J. Aerosol Sci. 1995, 26, 963. (27) Kaufman, S. L. J. Aerosol Sci. 1998, 29, 537. (28) Lenggoro, I. W.; Xia, B.; Okuyama, K.; Fernandez de la Mora, J. Langmuir 2002, 18, 4584.
Langmuir, Vol. 24, No. 22, 2008 12799 sodium ions and contains 40% SiO2 by weight. The particle surface is fully hydroxylated with a specific surface area, as measured by titration,29 of 220 m2 g-1. Other physical parameters of importance are pH 9.8, density 1.3 kg dm-3, and viscosity 15 mPa s. The total sulfate and chloride content is about 600 ppm.30 The dispersion was supplied by the manufacturer and used as received. Stock solutions of the electrolytes (LiCl, NaCl, KCl, RbCl, and CsCl, typically 10% by weight) and of the buffer solution (ammonium acetate, 20 mM, pH 8.0) were prepared. Water purified by a Milli-Q purification system (Millipore Corp., Synergy 185) was utilized in the preparation of all stock solutions. The alkali chlorides (suprapur, Merck), ammonium acetate (pro analysi, Merck), and ammonia (pro analysi, Scharlau) were used as received. Size Distribution Measurements. A combination of an electrospray (TSI Inc., ES model 3480) and a scanning mobility particle sizer (TSI Inc., SMPS model 3936) was employed to determine the particle mobility in air at atmospheric pressure. This method has been described in detail elsewhere,26-28,31 so only a short description will be given here. The method is based on transferring colloidal nanoparticles from the liquid phase to the gas phase by means of ES. The particles are subsequently size selected with the SMPS, containing a nano-differential mobility analyzer (TSI Inc., nDMA model 3085) for particles smaller than 120 nm, or a long differential mobility analyzer (TSI Inc., DMA model 3081) for particles larger than 120 nm and smaller than 1000 nm. After size selection with the DMA, the particles are counted with a condensation particle counter (TSI Inc., CPC model 3010). The obtained results yield the aerodynamic mobility diameter of each individual particle, that is, the diameter of a sphere with equivalent mobility in air. It should be noted that the method is independent of particle density. A capillary with an inner diameter of 30 µm was chosen to ensure the free passage of large aggregates up to several micrometers. The capillary inlet was positioned at the bottom of the sample vial to avoid discrimination against large aggregates due to sedimentation. The upper size limit for analysis was therefore not dependent on the spraying process but on the type and geometry of the DMA and on the settings of the SMPS. The combination of ES with SMPS has been shown to be capable of sizing colloidal particles in the size range from 3 to 1000 nm with high accuracy and reproducibility.27,28 The default lower detection limit (50% detection efficiency) of this particular CPC is 10 nm. To attain higher detection efficiency for particle diameters below 10 nm, the CPC was modified according to the description by Mertes et al.,32 which resulted in a lower limit of 100% detection at approximately 6 nm and 50% at 5 nm. During preliminary experiments, no significant number of particles or aggregates smaller than 7 nm or larger than 100 nm in diameter could be detected, and a somewhat wider scan interval between 3 and 120 nm was therefore chosen. A schematic picture of the experimental setup is shown in Figure 1. A size distribution of the unaggregated silica dispersion was obtained using the ES-SMPS system, and the result is shown in Figure 2. The bars represent size intervals into which the particles were categorized in the SMPS. All size distributions in the article were normalized; the number of particles counted in a size interval was divided by the total number of particles in all size intervals. Hence, a fraction of the total number of particles was obtained for each size interval. A sum of two Gaussian distributions was fitted to the distribution (solid line), and the individual Gaussian distributions are shown as dashed lines. To verify the size distributions, micrographs of the particles and aggregates were obtained using a scanning electron microscope (LEO 55 Ultra field emission SEM). In preparation for the SEM analysis, the sprayed particles were collected on a substrate. An aerodynamic lens system was utilized to aid the formation of a collimated particle beam, and the particles in the beam were allowed to impact on the substrate. (29) Sears, G. W. Anal. Chem. 1956, 28, 1981. (30) Information provided by the manufacturer, Eka Chemicals AB. (31) Song, D. K.; Lenggoro, I. W.; Hayashi, Y.; Okuyama, K.; Kim, S. S. Langmuir 2005, 21, 10375. (32) Mertes, S.; Schro¨der, F.; Wiedensohler, A. Aerosol Sci. Technol. 1995, 23, 257.
12800 Langmuir, Vol. 24, No. 22, 2008
Johnson et al.
Figure 1. Schematic of the experimental setup. Route 1: sampling for size classification. Route 2: sampling for SEM analysis. See the text for an explanation of the abbreviations.
Figure 2. Size distribution of pure Bindzil 40/220. A sum of two Gaussians were fitted to the data: (---) fitted Gaussians, (s) sum of the fitted curves, and (|) size intervals. The fraction plotted on the ordinate axis represents the number of particles in a size interval divided by the total number of particles in all size intervals.
The theory concerning the aerodynamic lens system is described elsewhere.33,34 An SEM image of the unaggregated silica particles is shown in Figure 3a. Aggregation Experiments. The addition of a sufficient amount of electrolyte, such as KCl, to the undiluted silica sol results in the aggregation of the particles as a result of the decrease in electrostatic repulsion. After a brief viscosity decrease mainly due to dilution effects, this leads to a viscosity increase and eventually the formation of a gel because the system has formed a single aggregate, or a contiguous network, that spans the entire volume.16 In these experiments, the point of gelation was determined visually as the time at which the system was no longer fluid in a normal gravity field. The time between salt addition and gelation depends mainly on the type of silica sol, counterion valence and concentration, and temperature. In this series of experiments, we chose one type of sol (Bindzil 40/220) and constant temperature (20 °C). The salts chosen were all monovalent alkali chlorides (LiCl, NaCl, KCl, RbCl, and CsCl). The addition of a highly concentrated electrolyte solution to a concentrated silica dispersion can result in local precipitation. In all performed experiments, the salt solutions were of sufficiently low concentrations and were added under good agitation so as to not cause local precipitation. Salt-induced aggregation is usually studied by increasing the electrolyte concentration until the fast aggregation regime is reached, thereby establishing the critical coagulation concentration (CCC).9,11 At the CCC, the energy barrier preventing aggregation is negligible (33) Liu, P.; Ziemann, P. J.; Kittelson, D. B.; McMurry, P. H. Aerosol Sci. Technol. 1995, 22, 293. (34) Svane, M.; Hagstro¨m, M.; Pettersson, J. B. C. Aerosol Sci. Technol. 2004, 38, 655.
Figure 3. Micrographs of Bindzil 40/220 particles and aggregates. (a) Unaggregated silica particles and (b) aggregates formed in the presence of CsCl.
regardless of the type of counterion. This will yield consistent kinetics for gel formation and hence consistent gel structure for the various ions. In this work, we have instead chosen to focus on the slow aggregation regime, and we aimed to create conditions where the energy barrier between the silica particles in various electrolytes was comparable or equal. This was achieved by adjusting the bulk electrolyte concentration in the aggregation mixtures to match a chosen gel time. The gel time was set to an experimentally feasible time of 60 min. This approach has the advantage that we can be confident that important parameters for the aggregation kinetics, such as the effective surface charge which in turn strongly influences
Aggregation InVestigated by Electrospray Technique
Langmuir, Vol. 24, No. 22, 2008 12801
Table 1. Electrolyte and Silica Concentrations in the Aggregation Mixtures salt
bulk conc (mol dm-3)
wt % SiO2
LiCl NaCl KCl RbCl CsCl
0.6 0.28 0.2 0.18 0.15
34 36 36.3 34.4 34.9
Table 2. Measured Electrophoretic Mobilities salt
electrophoretic mobility (H2O) × 10-8 (m2 V-1 s-1)
electrophoretic mobility (buffer) × 10-8 (m2 V-1s-1)
no salt LiCl NaCl KCl RbCl CsCl
-3.53 -2.42 -2.52 -2.24 -2.47 -2.44
-1.51 -1.22 -1.56 -1.56 -1.61 -1.69
the electrostatic repulsion, are kept reasonably constant between experiments. Also, the kinetics of formation of siloxane bonds depends on the particle contact time. Equivalent energy barriers result in similar particle contact times in the various mixtures. When calculating the bulk electrolyte concentrations in the aggregation mixtures, the volume of the particles was taken into account; for instance, the bulk electrolyte concentration in the LiCl mixtures was 0.6 M, and the SiO2 content was 34% by weight. Bulk electrolyte concentrations and the weight percentage of SiO2 in the various aggregation mixtures are listed in Table 1. To verify that similar surface charge screening was obtained for the various coagulation mixtures, the electrophoretic mobility of the particles in dilute samples (0.5 wt %) was measured (Malvern Instruments Ltd., zeta sizer model Nano ZS). The coagulation mixtures were diluted in Milli-Q water and buffer solution; the electrophoretic mobilities of the particles in the various coagulation mixtures are listed in Table 2. The experiments were conducted in the following manner. An appropriate amount of electrolyte solution was added to the dispersion, and the aggregation mixture was stirred for 20 s. Prior to mixing and throughout the aggregation sequence, the temperature of the system was held constant at 20 °C. At intervals of 10 min, samples of the mixture were analyzed with the ES-SMPS system. ES is widely used in mass spectrometry as a very gentle ionization method, yielding minimal molecular fragmentation in weakly bound biomolecules such as noncovalently associated protein complexes.35,36 It is therefore likely that the aggregates formed in the liquid phase will be transferred intact to the gas phase. To ensure that no aggregates were formed as a result of the spraying itself, the samples were extensively diluted. The dilution was carried out in a three-step sequence from 36-34 wt % SiO2 depending on the cation via 5 and 0.25 wt % to a final concentration of approximately 0.0025 wt % SiO2 (25 ppm). A concentration this low is necessary to ensure that the fraction of droplets emanating from the electrospray containing more than one particle or one aggregate is negligible; most droplets will contain no particles or aggregates. We are aware that this concentration may be lower than the solubility of colloidal silica in water; solubilities of typically 0.01 wt % (100 ppm) for similar systems have been reported in the literature.37,38 However, it is known that ammonia, which is present in our buffer solution, stabilizes the silica particles.1 Thus, we will show that the rate at which the particles dissolve to monosilicic acid is so low that it does not affect our measurements.
3. Results and Discussion Aggregation. The gradual buildup of aggregates upon electrolyte addition to the silica dispersion was monitored. Figure (35) Veenstra, T. D. Biophys. Chem. 1999, 79, 63. (36) Loo, J. A.; Berhane, B.; Kaddis, C. S.; Wooding, K. M.; Xie, Y.; Kaufman, S. L.; Chernushevich, I. V. J. Am. Soc. Mass Spectrom. 2005, 16, 998. (37) Alexander, G. B.; Heston, W. M.; Iler, R. K. J. Phys. Chem. 1954, 58, 453. (38) Gunnarsson, I.; Arno´rsson, S. Geochim. Cosmochim. Acta 2000, 64, 2295.
Figure 4. Size distribution variation during aggregation: (---) fitted Gaussians, (s) sum of the fitted curves, and (|) size intervals. (a) Prior to electrolyte addition, (b) 20 min after electrolyte addition, and (c) 65 min after electrolyte addition (5 min after the gelation point).
4a-c shows an example of the formation and size increase of aggregates produced during the aggregation process, in this case, in the presence of CsCl. Figure 3b shows a micrograph of the aggregates in contrast to the nonaggregated colloidal silica shown in Figure 3a. Initially, independent of the type of alkali ion present, small aggregates with a diameter of about 40 nm started to form. These aggregates contained on average three to four particles. As the aggregation proceeded, larger aggregates were formed, with broad size distributions of around 50-70 nm depending on the type of cation. The tendency to form both larger and more abundant aggregates increased in the following order: Li+ < Na+ < K+ < Rb+ < Cs+. As can be seen from Figure 4c, the formed gel contained a substantial number of unaggregated particles. The relative number of nonaggregated particles present at the point of gelation for the various mixtures decreased progressively while moving to heavier alkali cations. The size distributions of the particles as well as of the small and large aggregates can be represented by Gaussian distributions, and the data sets can be fitted to a sum of two or three Gaussians depending on the number of aggregate modes present in the sample. We suggest that these small and large aggregates comprise the building blocks from which the 3D gel structure is formed. Because a gel structure is fractal, it seems likely that aggregates representing parts of the structure would be nonspherical and quite large, a few hundred nanometers or larger. Because no significant number of aggregates larger than about 100 nm was
12802 Langmuir, Vol. 24, No. 22, 2008
Johnson et al. Table 3. Tabulated and Calculated Cation Properties ion
polarizability in water, R(0) (Å3)a
calculated electrostatic free energy of adsorption/kBT
Li+ Na+ K+ Rb+ Cs+
0.0285 0.1485 0.7912 1.3411 2.2643
+ 8.38 + 5.81 + 4.28 + 3.85 + 3.41
a
Figure 5. (a) Increase in aggregate mean mobility diameter during aggregation for (9) LiCl, (O) NaCl, (2) KCl, (]) RbCl, and (b) CsCl. (b) Increase in mean aggregate volume where the aggregate volume was calculated from the fitted mean mobility diameters. The aggregate volumes were corrected for the SiO2 and electrolyte concentrations. Straight lines were fitted to the initial increase in mean aggregate volume.
detected and the small and large aggregates were confirmed by SEM imaging to be roughly spherical, we conclude that the 3D structure disintegrated rapidly during the dilution steps prior to the ES-SMPS analysis and it was therefore not possible to measure the size of these aggregates. The aggregate volumes were calculated from the arithmetic mean diameter given by the Gaussian distributions of the small and large aggregates where the aggregates were assumed to be spherical. Because the gelation time was kept constant, the particle interaction energies should be approximately the same in all aggregation mixtures. The electrophoretic mobilities of particles and aggregates in the samples of coagulation mixtures diluted in Milli-Q water as well as in buffer solution are listed in Table 2. The samples diluted in buffer solution showed a higher degree of charge screening owing to the fact that the ionic strength is higher in the buffer solution compared to that in Milli-Q water. Nevertheless, the measurements of the samples diluted in Milli-Q water show that similar surface charge screening was achieved in the various mixtures and that the aggregate growth was quite similar. This behavior is depicted in Figure 5a where the increase in aggregate mean diameter during aggregation is shown. Because the aggregation mixtures of the various ions contained electrolyte and silica concentrations to match a gelation time of 1 h, the aggregate volumes were multiplied by a factor inversely proportional to these concentrations. The initial linear increase in aggregate volume is associated with the formation of dimers, and the initial aggregation rate constant is the slope of the straight lines fitted to these data points as shown in Figure 5b. Data points corresponding to time zero are not included because no electrolyte had been added and these cannot be normalized with respect to electrolyte content. As the aggregation proceeds, it becomes more complex; dimers start to attach to each other as well as to particles, and the increase in aggregate volume becomes
From ref 44.
more pronounced. However, toward the end of gelation the increase rate levels off. The initial aggregation rate constants for slow aggregation in the presence of the various ions were all on the order of 10-25 m3 s-1. Kobayashi and co-workers have determined a fast aggregation rate constant on the order of 10-20 m3 s-1 for a similar silica system.8 Hence we conclude that our experiments were conducted well within the slow regime. The initial aggregation rate constants increase in the following order: Li+ < Na+ < K+ < Rb+ < Cs+. Because the volumes of the aggregates were normalized with respect to the electrolyte content, an experiment with an ion with a lower aggregation rate required a higher electrolyte concentration to achieve the same gelation time compared to that for an ion with a higher aggregation rate. Thus, a lower rate should correspond to a less closely adsorbed ion because a higher electrolyte concentration was required to reach the gelation point in the same time. This can be compared to the concept of a critical coagulation concentration, where a higher electrolyte concentration is needed to reach the fast aggregation regime for an ion that is less closely adsorbed.11 It has been suggested that the closer adsorption of cesium ions on the silica surface could simply be attributed to the ion size.39,40 Cesium ions, being the least hydrated ions, would accordingly adsorb more closely to the surface than would the more hydrated ions such as Na+ and Li+. We have calculated the free energies of adsorption of alkali ions on silica surface by using a model developed by Kharkats and Ulstrup41 and corrected for typographical errors by Markin and Volkov.42 In this model, the ion solvation energy according to the Born model and an interaction energy of an ion with its image are included. In these calculations, crystallographic ion radii43 were used. The relative dielectric constants of silica and bulk solution were assumed to be 20 and 78.36, respectively. Electrostatic free energies of adsorption obtained for the alkali ions increased in the following order: Cs+ < Rb+ < K+ < Na+ < Li+ (Table 3). The given free-energy value of an ion corresponds to the difference between the energy of the ion in the bulk phase, which is the reference state, and the energy of the ion at zero distance from the surface, which is the adsorbed state. These results indicate that the larger ions adsorb more favorably on the silica surface than do the smaller ions. Furthermore, our measurements of the electrophoretic mobility indicated that the bulk electrolyte concentration required to reach the same degree of surface charge screening was lower for cesium ions than for the other alkali counterions. Thus, the aggregation rate results, along with the free-energy calculations and the electrophoretic mobility measurements, clearly show that Cs+ ions adsorb more closely on this silica surface than do the other alkali counterions. These results are in good agreement with X-ray scattering measurements40 and zeta potential measurements.23 (39) Tadros, TH. F.; Lyklema, J. J. Electroanal. Chem. 1968, 17, 267. (40) Tikhonov, A. M. J. Phys. Chem. C 2007, 111, 930. (41) Kharkats, Y. I.; Ulstrup, J. J. Electroanal. Chem. 1991, 308, 17. (42) Markin, V. S.; Volkov, A. G. J. Phys. Chem. B 2002, 106, 11810. (43) Conway, B. E. Ionic Hydration in Chemistry and Biophysics; Elsevier: Amsterdam, 1981.
Aggregation InVestigated by Electrospray Technique
Figure 6. Initial aggregation rate constants as a function of ion polarizability in water. Error bars represent the 95% confidence interval for the rate constants obtained from Figure 5.
The initial aggregation rate constants correlated well with the polarizability of ions in water. Values of the ion polarizabilities44 are listed in Table 3, and the obtained correlation is shown in Figure 6. Recently, Bostrom et al.15 developed a modified DLVO theory that incorporates the dispersion interactions, including Keesom, Debye, and other contributions, for ion-ion and ion-surface interactions. Using this extended theory, they obtained a higher pressure between two silica particles in the presence of less polarizable counterions, such as Li+, compared to more polarizable counterions, such as Cs+. This indicates that the more polarizable ions adsorb closer to the silica surface. Clearly, the ion dispersion interactions influence the overall interaction between the silica particles. Nonetheless, the effect of the ion dispersion interactions on their own, as reported by Bostrom and co-workers, is not sufficient to account for our observed differences in the aggregation rate in the presence of different alkali counterions. One way of rationalizing the linear correlation depicted in Figure 6 is to combine the modified DLVO theory with the Gurney description11,16,17,23 of ion interactions in solution extended to interactions between the ions and the solid surface. According to the Gurney description, particles can be regarded as macroions with the ability to promote or destabilize the water structure surrounding them. These surfaces are also known as structure makers and structure breakers, respectively. Amorphous silica is characterized as having a structure breaker surface.11,16 Correspondingly, ions can promote or destabilize the water structure surrounding them, and it follows that ions with water structures similar to the water structure of the silica surface will have a shorter distance of closest approach to the surface. The water structures surrounding Cs+ and the silica particles are similar so that when the cesium ion is attracted to the surface by the dispersion interaction it is able to substitute its own water structure in favor of the water structure of the surface. For the rubidium ion, the combined effect of the water structure interaction and the dispersion interaction gives rise to an ion that is only moderately integrated with the water structure of the silica surface. In the cases of K+, Na+, and Li+, there does not seem to be any overlap or sharing of water structures between the silica surface and the ions. From the perspective of the system, these ions appear to be completely hydrated and therefore reside a longer distance away from the silica surface than do the cesium or rubidium ions. The ion that fully penetrates the water structure (44) Tavares, F. W.; Bratko, D.; Blanch, H. W.; Prausnitz, J. M. J. Phys. Chem. B 2004, 108, 9228.
Langmuir, Vol. 24, No. 22, 2008 12803
appears as an unhydrated ion because it is sharing the water structure with the particle surface. This corresponds to an effective ion radius for Cs+ that is equal to its crystallographic radius.43 The corresponding effective radius for Rb+ is larger than its crystallographic radius43 but smaller than its hydrated radius.43 As mentioned in the Introduction, dielectric spectroscopy measurements and neutron diffraction studies give different hydration structures for the same ion. Moreover, the change in the state of a hydrated ion approaching a surface is yet to be described exactly. Even so, in addition to the linear correlation shown in Figure 6, a linear correlation between the abovedescribed effective radii and the aggregation rate constants could be observed. Thus, we conclude that the aggregation behavior of silica particles seen in this work is due to the combined effects of the dispersion interactions and the interactions between the hydrated silica surface and the hydrated ions. Aggregate Disintegration. The electrophorectic mobility measurements showed that the surface charge was more effectively screened in suspensions diluted with ammonium buffer than in suspensions diluted with Milli-Q water (Table 2). The achieved screening of the particles and aggregates was similar among the various reaction mixtures, and the screening was not sufficient to stabilize the aggregates against disintegration. Hence, dilution with buffer solution caused the aggregates to disintegrate, where disintegration refers to the detachment of particles or smaller aggregates from the larger aggregates. The overall 3D structure of the gel, formed through interactions between aggregates, disintegrated faster than we were able to measure. Figure 7a shows the decrease in the mean aggregate volume of samples stored at 25 ppm SiO2. We will show below that this decrease was not due to the dissolution of silica but to the detachment of particles. As previously mentioned the aggregate volumes were calculated using the mean aggregate diameters obtained from the Gaussian distributions, assuming spherical aggregates. Within 24 h after dilution, all aggregates had disintegrated completely, independent of the type of counterion present. The plateau in Figure 7a corresponds to the mean volume of the silica particles as calculated from the arithmetic mean diameter, 25 nm. Our results show that the particle aggregation, at least initially, is reversible so that a sufficient decrease in the electrolyte concentration will redisperse the particles. This suggests that the interaction energy well in which the aggregates are formed is shallow compared to the primary energy well predicted by the DLVO theory. We propose that the collapsed gel layer covers a large part of the primary DLVO well, thus preventing the silica particles from aggregating irreversibly. Because of the weak van der Waals interactions of silica,45 a possible secondary DLVO well is unlikely to be sufficiently deep so as to cause reversible aggregation. A schematic representation of the interaction energy as a function of interparticle distance, with added electrolyte and after dilution, is shown in Figure 8. The total interaction energy can be divided into two main parts, a DLVO-type energy, VDLVO, and a steric energy, Vster, arising when the two gel layers on the particle surfaces interact. For simplicity, it is assumed that the total interaction energy, Vtot, is given by the sum of the two contributions
Vtot ) VDLVO + Vster
(1)
When the system is diluted, the electrolyte concentration decreases. The long-range electrostatic repulsion energy between the particles increases as a result of decreased screening. In addition, the decreased screening will reinflate the gel layers14 (45) Depasse, J. J. Colloid Interface Sci. 1997, 188, 229.
12804 Langmuir, Vol. 24, No. 22, 2008
Johnson et al.
Figure 9. Initial disintegration rate constants as a function of ion polarizability in water. Error bars represent the 95% confidence interval for the rate constants obtained from Figure 7.
Figure 8. Schematic illustration of the proposed interaction energies between colloidal silica particles. The solid and dashed lines represent a generic DLVO interaction and the steric repulsion due to an inflatable gel layer, respectively. The dotted line is the sum of these two interactions, and its associated potential wells are indicated by the shaded areas. Panel a shows the interaction energies subsequent to electrolyte addition with arrows indicating the directions of change upon dilution. The result of these changes is depicted in panel b. The graphs are not to scale, and the features are exaggerated for clarity.
to become much more shallow so that the thermal motion is sufficient to separate the particles. However, a small energy barrier that retards the disintegration remains as the overall repulsion between the particles increases as a result of the decrease in electrolyte concentration. Non-DLVO interactions other than the steric interaction are not considered in the model. We are aware that the proposed model is a simplification that does not contain all details of the individual contributions. The depth of the remaining potential well subsequent to dilution and thus the stability of the aggregates in the dilute system seem to depend on how close the counterions have been adsorbed and on the duration under which the particles have been kept together. The initial decrease in aggregate volume is shown in Figure 7b. Straight lines were fitted to the data sets, and the slopes of the lines were taken as the initial rate constants of disintegration. As shown in Figure 9 the initial rate constants decay exponentially with the ion polarizability in water. The cesium ions that displayed the largest effect on the aggregation rate also exhibited the most stable aggregates. This is probably linked to the fact that cesium ions adsorb more closely on the silica surface and thus give rise to a stronger attractive short-range interaction between the silica particles, also known as an ion-ion correlation interaction.23,46 Consequently, the particles within these aggregates were kept together more strongly, and a deeper penetration of the collapsed gel layers could therefore occur in this mixture. Furthermore, the hydration of desorbed ions will also contribute to the disintegration process. Lithium ions, which are much more strongly hydrated than cesium ions, have a larger excluded volume, and as a result, LiCl solutions have a higher osmotic pressure than CsCl solutions at a specific salt concentration.47 As a result, there will be a stronger repulsion between the silica particles in the case of Li+ than for the less hydrated alkali counterions. This is depicted in the exponential decay seen in Figure 9. Because it was possible to measure the disintegration of the small and large aggregates but not the disintegration of the 3D gel structure, it is concluded that these aggregates were more stable than the 3D structure. This enhanced stability of aggregates formed early in the aggregation process is thought to occur as a result of the increased duration of the gel layer overlap. In our view, the strength of the aggregates mainly depends on the extent of the penetration of the gel layers. The older aggregates have had more time to produce deep penetration, and the particles in aggregates formed in the presence of ions that give rise to
and thus increase the steric repulsion between the particles in the aggregate. The situation after dilution is illustrated in Figure 8b. The swelling of the gel layer causes the resulting potential well
(46) Guldbrand, L.; Jo¨nsson, B.; Wennerstro¨m, H.; Linse, P. J. Chem. Phys. 1984, 80, 2221. (47) Abbas, Z.; Ahlberg, E.; Nordholm, S. Fluid Phase Equilib. 2007, 260, 233–247.
Figure 7. (a) Decrease in mean aggregate volume during aggregate disintegration for (9) LiCl, (O) NaCl, (2) KCl, (]) RbCl, and (b) CsCl. The aggregate volume was calculated from fitted mean mobility diameters. (b) First 104 seconds showing the initial decrease in mean aggregate volume. Straight lines were fitted to the data sets.
Aggregation InVestigated by Electrospray Technique
Langmuir, Vol. 24, No. 22, 2008 12805
stronger attractive interactions will be forced together more strongly, thus increasing the extent of penetration. This implies that the collapsed gel layers do not constitute impenetrable physical barriers but rather that they prevent the particles from coming into immediate contact as these gel layers overlap. Covalent siloxane bonds formed within the aggregates48 would then primarily form between the polysilicic chains in the outermost part of the gel layers and not directly between the particle surfaces. Particle detachment is therefore a possibility, provided that the energy well is shallow. However, aggregates kept together for a sufficient duration will eventually coalesce permanently.48 At present, it is not clear whether the disintegration is primarily due to the desorption of ions from the surface, to the cleavage of siloxane bonds, or to a combination of the two. Preliminary experiments indicate that both ion desorption and pH influence the disintegration rate. These phenomena merit further investigations. Particle Dissolution. Amorphous silica is slightly soluble in water (∼100 ppm37,38), hence silica particles stored in the buffer solution will start to dissolve to silicic acid monomers.49,50 During the first 24 h, the mean particle diameter decrease because silica dissolution alone is on the order of approximately 2 nm for silica particles stored at a total SiO2 concentration of 25 ppm. It is noteworthy that all aggregates were completely disintegrated to discrete particles within 24 h after dilution. This corresponds to a decrease in the aggregate mean diameter of 25-45 nm. Hence, we argue that it is possible to distinguish between the aggregate disintegration due to the low electrolyte concentration in the dilute samples and particle dissolution due to the low silica concentration. Silica dissolves in water to monosilicic acid as described by the following reaction:49
SiO2(s) + 2H2O(l) T H4SiO4(aq)
(2)
The net change in particle volume will be given by50
-
dVp M ) (k1Sp - k2CSp) dt F
(3)
where Vp is the mean volume of the particle, Sp is the mean surface area of the particle, and k1 and k2 are the rate constants for dissolution and redeposition, respectively. C is the concentration of monosilicic acid in the solution, M is the molecular mass of SiO2 (60.1 g mol-1), and F is the density of the particle (2.2 g cm-3). Initially, the concentration of monosilicic acid is low; hence the redeposition term k2CSp can be neglected, and the net change will be given by
-
dVp M ) k1Sp dt F
(4)
This expression can be rewritten as
-
drp M ) k1 dt F
(5)
As indicated by eq 5 and shown in Figure 10, the mean particle radius, rp, showed a linear initial decrease with time after dilution. As the experiment progressed, the dissolution rate decreased, and the mean radius reached a plateau value. It could be argued that this was due to an increased rate of redeposition on the particle surfaces as the concentration of monosilicic acid increased. However, because the total concentration of silica was well below the equilibrium concentration of monosilicic (48) Depasse, J.; Watillon, A. J. Colloid Interface Sci. 1970, 33, 430. (49) Greenberg, S. A.; Price, E. W. J. Phys. Chem. 1957, 61, 1539. (50) O’Connor, T. L.; Greenberg, S. A. J. Phys. Chem. 1958, 62, 1195.
Figure 10. Decrease in the mean particle radius as a function of time after dilution.
acid in water the redeposition term could be neglected at all times. The observed change in the dissolution rate was instead attributed to the dissolution of two distinctly different silica phases. Initially, the slightly less polymerized gel layers surrounding the particles dissolved at a relatively high rate. When the gel layers were consumed, the denser particle cores started to dissolve at a slower rate. This is consistent with earlier silica dissolution results.13,50,51 A straight line was fitted to the data points corresponding to the initial radius decrease, and the initial dissolution rate was calculated using eq 5. The dissolution rate constant for the silica particles in the ammonium acetate buffer solution (pH 8) was on the order of 10-10 mol m-2 s-1. Considering that ammonia is known to stabilize the silica surface and thus reduce the dissolution rate1 and that our measurements were conducted at a lower pH, the results correspond reasonably well with data reported in the literature.8
4. Conclusions The time-dependent size distribution variations of silica nanoparticles during aggregation in the presence of various alkali cations have been monitored by means of a combination of an electrospray and a scanning mobility particle sizer. The aggregation was found to proceed via small and subsequently large aggregates that eventually formed a gel structure. At the point of gelation, there was still a substantial number of free silica particles remaining in the gel. The largest, most stable, and most abundant aggregates were produced in the mixtures containing cesium ions. These mixtures also contained the lowest relative number of free particles, and this coincided with cesium ions displaying the strongest adsorption on the silica surface. The strong adsorption is possibly due to the ion dispersion interaction with the silica surface in combination with the ability of the cesium ions to share a water structure with the particles. For the other alkali counterions, the dispersion interactions as well as the similarity between the water structures of the ions and the surfaces decrease progressively when moving from Rb+ to Li+. As a result, these ions adsorb a longer distance away from the surface. The aggregation was, at least initially, completely reversible upon dilution, and the disintegration of the aggregates was observed. These results confirm the existence of an additional short-range repulsive interaction between the particles in addition to the usual DLVO interactions. The fact that this additional repulsion was present at a pH value as high as 9.8 suggests that (51) Holt, P. F.; King, D. T. J. Chem. Soc. 1955, 3, 773.
12806 Langmuir, Vol. 24, No. 22, 2008
it was not due to a hydration force. Instead, it is believed to originate from an overlap of the gel layers that surround the particles. The rate of aggregate disintegration after dissolution was also monitored, and the obtained results showed that the relative stability of the aggregates varied depending on the type of counterion, although all ions induced reversible aggregation. The aggregate stability decreased following the direct Hofmeister sequence: Cs+ > Rb+ > K+ > Na+ > Li+. The aggregation behavior of silica particles observed in the present study is
Johnson et al.
attributed to the combined effect of dispersion interactions and the interaction between the hydrated silica surface and the hydrated ions. Acknowledgment. We thank Eka Chemicals AB for financial support, Itai Panas for fruitful discussions, and Benny Lo¨nn for technical support. The MACH institute and especially Stefan Gustavsson are acknowledged for help with SEM analysis. LA8026122
