1. Introduction. 2. Silica. O)(SiO 2

Transcription

1 Multilevel structure Structure of reinforcing Reinforcingsilica and Silica carbon and Carbon Dale W. Schaefer, a Thomas Rieker, b Michael Agamalian, c J. S. Lin, c Daniel Fischer, d Sathish Sukumaran, a Chunyan Chen, a Greg Beaucage, a Charles Herd e and Jimmy Ivie e a University of Cincinnati, Cincinnati, OH , b University of New Mexico, Albuquerque, New Mexico, 87131, c Solid State Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831, d National Institutes of Standards and Technology, Gaithersburg, MD 20899, and e Columbian Chemicals Co., Marietta, GA dwschae@uceng.uc.edu Using small-angle x-ray (SAXS), neutron (SANS), x-ray diffraction and light scattering, we study the structure of colloidal silica and carbon on length scales from 4 Å < q -1 < 10 7 Å where q is the magnitude of the scattering vector. These materials consist of primary particles of the order of 100 Å, aggregated into micron-sized aggregates that in turn are agglomerated into 100 µ agglomerates. The diffraction data show that the primary particles in precipitated silica are composed of highly defective amorphous silica with little intermediate-range order (order on the scale of several bond distances). On the next level of morphology, primary particles arise by a complex nucleation process in which primordial nuclei briefly aggregate into rough particles that subsequently smooth out to become the seeds for the primaries. The primaries aggregate to strongly bonded clusters by a complex process involving kinetic growth, mechanical disintegration and restructuring. Finally, the small-angle scattering (SAS) data lead us to postulate that the aggregates cluster into porous, rough-surfaced, non-mass-fractal agglomerates that can be broken down to the more strongly bonded aggregates by application of shear. We find similar structure in pelletized carbon blacks. In this case we show a linear scaling relation between the primary and aggregate sizes. We attribute the scaling to mechanical processing that deforms the fractal aggregates down to the maximum size able to withstand the compaction stress. Finally, we rationalize the observed structure based on empirical optimization by filler suppliers and some recent theoretical ideas due to Witten, Rubenstein and Colby. 1. Introduction Reinforcing fillers are used to enhance the mechanical properties of polymers in general and elastomers in particular. In tire applications, for example, reinforcing fillers increase the modulus, strength and wear resistance of elastomers. For decades carbon black has been the dominant material used for tire reinforcement. Recently, however, many tire manufacturers have incorporated colloidal silica, usually in combination with carbon black, to improve the rolling resistance with little or no sacrifice in traction or wear. It is of interest, therefore, to determine the structure of these two materials in order to trace performance characteristics to the physical characteristic of the filler colloids. In spite of decades of study, there is still controversy about the mechanism of reinforcement. The complex structure consisting of primary particles aggregated into micron-sized clusters is known to be necessary for effective reinforcement, but there is little consensus on why such a structure is necessary. Part of the motivation of our work was to search for common structural features in both the silica and carbon systems that might illuminate this issue. Characterization of aggregated structures is a major challenge since these materials show distinct structural features over many decades in length scale, the signature of the complex growth processes active during synthesis. Figure 1, for examples, shows schematically the processes that are believed to be active during synthesis along with the structural features resulting from these processes. These systems display distinct structural features on length scales ranging from Angstroms to 100 microns, presenting a major challenge for small-angle scattering. 2. Silica 2.1. Synthesis and Structure Precipitated silica is prepared by acidification of water glass, (Na 2 O)(SiO 2 ) 3.3, in a stirred reactor. Water glass is silica dissolved in NaOH. As indicated in Figure 1, which is based in Iler (Iler, 1979), particles first nucleate and then grow as acid is added to the strongly basic water glass solution. In these early stages of particle growth, aggregation is suppressed in favor of growth due to the high surface charge on the growing particles. As neutralization proceeds, however, the surface charge is reduced leading to aggregation and agglomeration (Chevallier & Rabeyrin, 1995; Krivak, et al, 1992). The distinction between aggregation and agglomeration is not exactly clear. Neither are the conditions that differentiate between aggregation and Nucleation Growth Aggregation Agglomeration Figure 1 Schematic representation of the formation of reinforcing silica, following Iler. Particles nucleate early and then grow as more precursor silicate is added. Eventually the surface charge drops to the point that the particles become destabilized to aggregation. Further deposition is then favored at the necks between particles leading to bridging. The SAS data indicate that the robust aggregates are clustered into uniformly dense agglomerates with fractally rough surfaces as indicated in the right schematic. Agglomerates are less robust and can be disrupted by shear. J. Appl. Cryst. (2000). 33, 587±591 # 2000 International Union of Crystallography Printed in Great Britain ± all rights reserved 587

2 R G = 89 µ Light LS230 SAXS BH (UNM) SAXS PH (UNM) Diffraction q (Å -1 ) Figure 2 Scattering profile for precipitated silica. These data were accumulated on four instruments: 10-7 < q < 10-3 Å -1 = Coulter LS230 light scattering; 3 x 10-4 < q < 10 2 Å -1 = UNM Bonze Hart SAXS; 0.03 < q < 0.5 Å -1 = UNM Pinhole SAXS; 0.5 < q < 5 Å -1 = UC x-ray diffraction. The diffraction data was taken on a dry powder whereas the small-angle data were observed on slurries. agglomeration. In the present analysis, we view aggregates as rigid clusters of primary particles characterized by extensive bridging between particles (i. e. thick necks). Agglomerates, on the other hand, are weaker structures held together by van der Waals forces. It is also not clear how one distinguishes between aggregates and agglomerates using either microscopy or scattering. Here we claim that by examining the scattering profile over many decades in q it is possible to separate and characterize both the aggregates and the agglomerates. Figure 2 shows the structure factor for precipitated silica over an extensive range of momentum transfer, q (10-7 < q < 4 Å -1 ). The observed profile reflects the multilevel structure described above. At q > 1 Å -1, the diffraction regime, broad maxima are observed, reminiscent of bulk amorphous silica. For 0.03 < q < 0.3 Å -1, a powerlaw profile is seen with an exponent near the Porod value of -4.0, consistent with smooth-surfaced primary particles. Another powerlaw regime characteristic of fractal clusters extends from < q < 0.02 Å -1. This second power-law regime indicates that the primary particles are aggregated into fractal clusters of fractal dimension D f = 2.0, the observed slope. Finally, in the light scattering regime below Å -1, a huge feature is observed that we attribute to agglomerates, weakly bonded clumps of the strongly bonded aggregates. The data in Figure 2 were not measured on an absolute scale and were shifted with respect to each other to make them match in the overlap regime. We, therefore, do not include numbers on the ordinate since such numbers have no meaning. Each major tick mark, however, represents one decade on the log scale, so the intensity drops thirteen orders of magnitude over the q range studied. No multiple scattering corrections were needed since the samples that were used for the light scattering were diluted until the scattered intensity was less than 10% of the incident beam. At larger q, the scattering is reduced by many decades in intensity, so multiple scattering is not a problem. We verified this fact using samples of different thickness. In the next section each structural regime is analyzed in greater depth Atomic level structure. At the shortest length scale corresponding to large q, we find that the first diffraction peak in Figure 2 at q = 1.5 Å -1 is very broad compared to bulk amorphous silica. The deviation from the expected pattern for bulk silica is due to the fact that precipitated silica never experiences the high temperatures found in melt processing, so residual -OH groups act as network modifiers, primarily found on the surface of the growing colloidal particles. It is the predominance of non-bridging oxygens that accounts for the near absence of sharp diffraction peaks. Campbell, et al. (Campbell, et al, 1999) recently simulated the structure of high surface-area silica and found that the primary contribution of the interfacial atoms was to decrease the first sharp diffraction peak (FSDP). The FSDP arises from contributions from Si-Si, O-O and Si-O distances. In our case, the FSDP is reduced compared to bulk silica. Campbell also finds that the FSDP shifts toward smaller q with increasing surface area. In our case, however, the peak is close to that of bulk SiO D. W. Schaefer et al. J. Appl. Cryst. (2000). 33, 587±591

3 -3.1 Native Sheared Fit induce directional aggregation, leading to aggregate asymmetry. The observed value of D f = 2 is probably a fortuitous result of the competition between random growth, shear-biased growth, restructuring, and breakage. Data in the regime were collected on a slit-smeared Bonze- Hart camera and were desmeared using the Lake procedure as described by Long, Jemian and Wertman (Long, et al, 1991). R g = 2.08 µ q(å -1 ) Figure 3 Comparison of the desmeared USANS profile of sheared and unsheared precipitated silica. On vigorous agitation, the agglomerates break down to 2µ aggregates. The final size is comparable to the crossover length scale found in the unsheared material Particulate Regime The perspective presented here, based on growth of particulates, follows the ideas of Iler (Iler, 1979): Nucleation and growth of particles whose surface is smoothed by surface tension. The growth phase of this model does seem applicable once fully formed, smooth surfaced particles are present. The genesis of these primaries, however, may be substantially more complicated than simple nucleation. We have observed rather complicated SAXS profiles in the nucleation stage of particle formation. A single-step nucleation model does not seem to be adequate to account for these early stage processes. These issues will be addressed in a future publication. We recognize that alternative explanations for the multilevel structure, based on phase separation, have been advanced to explain the structure of aerogels derived from silicate esters (Schaefer, 1994) as well as Resorcinol-Formaldehyde (Schaefer, et al, 1995). In addition, Butler, Muzny and Hanley present evidence that phase separation may account for silicates prepared under shear (Butler, et al, 2000). Their data show the same qualitative features that we find in precipitated silica Aggregate Regime The aggregate regime (0.002 < q < 0.02 Å -1 ) is typical of colloidal silica aggregated in solution. Kinetic processes such as diffusionlimited cluster aggregation or chemically limited cluster aggregation lead to fractal clusters with fractal dimension, D f, between 1.7 to 2.1 depending on the aggressiveness of the aggregation process. Stirring, however, certainly alters this simple aggregation process. Stirring not only leads to restructuring, which increases the fractal dimension, but to shear-induced rupture, that has the opposite effect. Shear can also 2.5. Agglomerate Regime We attribute the broad feature in the regime 10 7 < q < 10-3 Å -1, to agglomerates, non-fractal clusters of aggregates, into a porous structure similar to a bunch of grapes where each grape is a fractal aggregate. The agglomerates are weakly bonded, uniformly dense, highly porous and non-mass-fractal. The slope of -3.1 means the surface of the agglomerates is fractally rough as one might anticipate if mass fractal aggregates were jammed together. There is, after all, no active process to smooth the agglomerate surfaces. Alternatively, the -3.1 slope could result if the distribution of agglomerate sizes were powerlaw polydisperse (Schmidt, 1989). Agglomeration need not occur during the chemical synthesis, but could well be driven by capillary forces due to drying. We will show, for example, that agglomerates are also found in carbon black where post synthesis processing, designed to increase the powder density, is probably responsible for agglomerate formation. Agglomerates have also been reported in other commercial precipitated silicas (Chevallier & Rabeyrin, 1995). Others and we have often seen the rising intensity at small q, but to our knowledge, this is the first observation of a Guinier regime corresponding to a radius of gyration of the order of 100 µ. Sorensen, et al. (Sorensen, et al, 1998) found a rising small-q intensity due to Porod scattering from the edges of the illuminated region. In our case, however, the length scale of the illuminated region is an order of magnitude larger than the observed R g, so we have not reached the Sorensen limit. The observed R g is, on the other hand, comparable to agglomerates seen by electron microscopy. To investigate agglomerates further, we subjected a second precipitated silica to vigorous shear in an attempt to break up such agglomerates. Unfortunately, we could not quantify the magnitude of the shear. Using ultra small-angle neutron scattering (USANS) we observed the profiles in Figure 3 shown before and after shear (the sample was sheared and studied several days after synthesis). These data span a q region that covers the transition region between agglomerates and aggregates. Although this crossover region is shifted a decade toward smaller q compared to the materials used in Figure 1, the qualitative features are identical. The origin of the shift in the crossover q is not known but is presumably due to differences in synthesis and processing of the two samples, which are optimized for different applications. Figure 3 shows that on shearing the suspension the small-q agglomerate feature is eliminated and a new Guinier regime appears at the same q where the transition from slopes of 4.0 to 2.0 is observed in the unsheared sample. We take this observation as evidence that the agglomerates are broken up by the shear whereas the aggregates, being smaller and stronger, survive. The Guinier radii calculated in Figure 3 were extracted using the unified model of Beaucage (Beaucage, 1995; Beaucage, 1996; Beaucage & Schaefer, 1994). 3. Carbon Figure 4 shows that the generic structure observed in precipitated silica is also observed for pelletized carbon blacks. The carbons studied are J. Appl. Cryst. (2000). 33, 587±591 D. W. Schaefer et al. 589

4 commercial grades supplied by Columbian Chemicals Co. These materials are synthesized in a furnace reactor, but the detail of how the reactor is controlled to produce the various grades is proprietary. As discussed below, it is important to note that these materials are pelletized by a compaction process following flame synthesis. We found that the curves in Figure 4 can be superimposed by plotting I/I 0 vs. qr g where I o and R g are the intensity and radius of gyration resulting from a Guinier-type analysis primary particle region. The scaled data are shown in Figure 5. Such superposition indicates that the correlation range associated with the aggregates scales linearly with the primary particle size. That is, superimposing the curves in the Guinier region of the primary particles also superimposes the crossover from the aggregate to the agglomerate region. We associate this latter crossover with the size of the hard agglomerates based on the shear experiment reported in Figure 3. Our conclusions, therefore, assume that we have correctly interpreted this aggregate-agglomerate crossover. Analysis of these multiple length-scale systems is not straightforward and alternate interpretations of features are always possible. Sorensen et al., for example, associate the crossover with scattering from the finite scattering volume (Sorensen, et al, 1998). One could, on the other hand, attribute the small q scattering to some unspecified largescale inhomogeniety following Debye, Anderson and Brumberger (Debye, et al, 1957). Witten, Rubenstein and Colby (Witten, et al, 1993) have argued that aggregate stiffness scales with the primary particle size, and showed that there is a limiting size for aggregates beyond which they will not survive a given stress. Aggregates larger than the limit break or collapse; large aggregates can even collapse under gravitational forces. Since the blacks in Figure 4 are pelletized by a mechanical process, it is likely that they are subjected to a critical shear which reduces them to a collection of robust aggregates small enough to resist the imposed stress. Since the critical size scales linearly with the particle radius for a given stress, larger cluster will survive only when made up of large heavily bridged (and therefore stiff) primaries, consistent with the scaling we observe. That is, larger primaries lead to a stiffer aggregate backbone that, below some length scale, is able to resist the compaction stress. At larger scales, the aggregates break or yield to interpenetration. For the case of fluffy carbon black (as opposed to pelletized) or fluffy fumed silica (as opposed to precipitated silica), the rising intensity at small q is not observed. At least the limit of q reported by Schaefer, et al. (Schaefer, et al, 1991) and Sorensen, et al. (Sorensen, et al, 1998), fluffy systems simply show fractal aggregates with no transition to agglomerates at small q. This result is not surprising given the radical difference in the synthetic process compared to solution grown precipitated silica. Rieker, Misono and Ehrburger-Dolle did see some deviation from power-law behavior at small q for a series of fluffy blacks (Rieker, et al, 1999). These fluffy systems are very low density due to the intrinsic vacuous nature of low-dimensional fractals. In the intermediate q range (aggregate regime) the pelletized carbons display a power-law regime with a slope of about 1.0, substantially above that observed in fluffy systems where slopes near 1.7 have been observed for both fluffy carbon black and fumed silica. One might be tempted to attribute the decrease to a stripping of the side branches and resulting linearization of the clusters. Sorensen, et al. (Sorensen, et al, 1998), however, attribute the change in slope to the compaction process itself, which effectively crushes the aggregates until the entire system is uniformly dense. Rieker, Ehrburger-Dolle and Hindermann-Bischoff (Rieker, et al, 2000) find that power laws more typical of diffusion-limited aggregates are restored if the pelletized is re-fluffed implying that the suppressed slope for the pelletized blacks is due to interpenetration of aggregates q(å - 1 ) R-5000 R-2000 N-121 N-339 N-630 Figure 4 SAXS profiles for a series furnace blacks. The BET surface areas from bottom to top are 33, 94, 131, 190 and 632 m 2 /g. These surface areas are taken from the manufacturer s product literature. On the scale of the primary particles, the scaled data in Figure 5 show that the limiting slope is near, but not equal to four (Schaefer, et al, 1991). Although slopes less steep than the Porod value of 4.0 can be interpreted as surface roughness, they can also be attributed to internal intermediate range structure. Since these carbons look smooth when viewed in an electron microscope, we attribute the deviation from 4.0 not to surface roughness, but to contributions to the scattering from in internal, partially graphitic (so-called turbostratic) structure of carbon. Whereas fumed silica shows systematic deviation from 4.0 with surface area (Schaefer & Hurd, 1990), there is no correlation between BET surface areas and the limiting exponent for the case of these blacks (Schaefer, et al, 1991). That is, fumed silica surfaces are smoother (Porod exponents are closer to -4.0) when processed at higher temperature as one would expect if viscous sintering were actively smoothing the surface. No such trends are known for carbon black, so the observed deviations are not likely due to surface effects. 590 D. W. Schaefer et al. J. Appl. Cryst. (2000). 33, 587±591

5 R-5000 R-2000 N-121 N-339 N-630 Agglomeration therefore serves a processing need since agglomerated powders are more granular, more dense and less dusty. The structure of the aggregates can be rationalized given a model of reinforcement. Witten, Rubenstein and Colby (Witten, et al, 1993), for example, have argued that the bending modulus of an aggregate drops precipitously with its size. At some size the polymers reinforce the aggregates rather than visa versa. Therefore, there is a maximum size above which aggregates do not increase the modulus of the composite material. It stands to reason, therefore, that empirical optimization of reinforcing fillers would find this size. Our analysis indicates that this optimum size is about 2000 Å, consistent with Witten, Rubenstein and Colby. I/I 0 4. Conclusion qr G Figure 5 Scaled carbon black scattering curves. The data were scaled by fitting a Guinier function in the particulate region. That is R g value was obtained from a local Guinier fit in the region of the knee of the curve. In spite of drastic differences in the synthetic protocol for precipitated silica and pelletized carbon black, the resulting structure of particles, aggregates and agglomerates is surprisingly similar, not only in qualitative features (e. g. particles clustered into aggregates clustered into agglomerates) but also in their quantitative aspects. All three structures are similar in size and the fractal nature of each structural level is similar. This similarity in structure between carbon and silica must be a somehow be related to the physics of reinforcement. The similarity of structure undoubtedly arises from decades of empirical optimization for reinforcing applications. There must be an underlying reason, however, why the hard aggregates are sub micron, why the particles are about 200 Å and why the agglomerates are tens of microns in size. The existence of large-scale agglomerates can be understood from an engineering prospective. The unagglomerated fluffy powders are difficult to handle, and difficult to incorporate into polymers. References Beaucage, G. (1995). J. Appl. Cryst. 28, Beaucage, G. (1996). J. Appl. Cryst. 29, Beaucage, G. & Schaefer, D. W. (1994). J. Non-Cryst. Solids 172, Butler, P. D., Muzny, C. & Hanley, H. (2000). J. Appl. Cryst. This volume. Campbell, T., Kalia, R. K., Nakano, A., Shimojo, F., Tsuruta, K., Vashishta, P. & Ogata, S. (1999). Phys. Rev. Lett. 82, Chevallier, Y. & Rabeyrin, M. (1995). Dispersible Silica Particulates. U. S. Patent # , Assigned to Rhone-Polenc Chimie. Debye, P., Anderson, H. R. & Brumberger, H. J. (1957). J. Appl. Phys. 28, 679. Iler, R. K. (1979) The Chemistry of Silica. New York:John Wiley & Sons. Krivak, T. G., Okel, T. A. & Wagner, M. P. (1992). Reinforced Precipitated Silica. U. S. Patent # , Assigned to PPG Industries. Long, G. G., Jemian, P. R. & Weertman, J. R. (1991). J. Appl. Cryst. 24, Rieker, T. P., Ehrburger-Dolle, F. & Hindermann-Bischoff, M. (2000). Poster Presentation at SAS99, to be published. Rieker, T. P., Misono, S. & Ehrburger-Dolle, F. (1999). Langmuir 15, Schaefer, D. W. (1994). MRS Bull. 19, Schaefer, D. W. & Hurd, A. J. (1990). Aerosol Sci. & Tech. 12, Schaefer, D. W., Olivier, B. J., Hurd, A. J., Beaucage, G. B., Ivie, J. J. & Herd, C. R. (1991). J. Aerosol. Sci. 22, S447-S450. Schaefer, D. W., Pekala, R. & Beaucage, G. (1995). J. Non-Cryst. Solids 186, Schmidt, P. W. (1989). The Fractal Approach to Heterogeneous Chemistry, edited by D. Avnir, pp New York: John Wiley & Sons. Sorensen, C. M., Oh, C., Schmidt, P. W. & Rieker, T. P. (1998). Phys. Rev. E. 58, Witten, T. A., Rubenstein, M. & Colby, R. H. (1993). J. Phys. II (France) 3, 367. J. Appl. Cryst. (2000). 33, 587±591 Received 12 June 1999 Accepted 19 January