CHEMTECH

August 1996

Copyright © 1996 by the American Chemical Society.

Designing polymer blends

Finding the appropriate ternary system using Edisonian methods is time consuming and not always productive. This experimentally proven modeling method, which can be run on fast PCs, will help.

by Timothy J. Cavanaugh, A. Peter Russo, and E. Bruce Nauman

Multiphase polymer blends are of major economic importance in the polymer industry. The most widespread examples involve the impact modification of a thermoplastic by the microdispersion of a rubber into a brittle polymer matrix. Most commercial blends consist of two polymers combined with small amounts of a third, compatibilizing polymer--typically a block or graft copolymer. Multiphase blends can be made using a variety of methods; however, research has focused on compositional quenching (1), melt compounding, and reactive technologies such as precipitation polymerization and reactive extrusion (2, 3).

Multiphase polymer blends can be easier to process than a single polymer with similar properties. The possible blends from a given set of polymers offer many more physical properties than do the individual polymers. Selections for blend production are currently made by Edisonian experimentation. This approach has shown some success but becomes cumbersome when more than a few components are involved.

Here we describe an emerging technique for computer-aided design of multiphase polymer blends. Polymer compositions are chosen to produce interesting and desired morphologies. We also provide a classification scheme for specific morphologies and methods of experimental confirmation with material properties.

Modeling blending processes

Compositional quenching. In this relatively new blending process, two or more incompatible polymers are dissolved in a common solvent to form a homogeneous mixture. The solvent is then rapidly removed by flash devolatilization. The initial, single-phase mixture is plunged quickly and deeply into a multiphase region in which phase separation occurs by a process known as spinodal decomposition. Our discussion begins with this process because it can be accurately modeled. The model consists of three equations: a modified Flory-Huggins equation for the free energy of mixing, the Landau-Ginzburg functional equation, and a modified Cahn-Hilliard diffusion equation.

Many polymer pairs are incompatible and form two phases when mixed. When a common solvent is added, a limited region of two-phase miscibility occurs. The formation of a single-phase solution--two or more polymers in a single solvent--is the starting point for compositional quenching.

The first quantitative theory of polymer-solvent equilibria was presented by Flory (4) and Huggins (5). They modeled the Gibbs free energy of mixing (g_mix) in a binary polymer-solvent system. According to this theory, the Gibbs free energy of mixing per unit volume is written as

The Flory-Huggins theory is similar to regular solution theory except that volume fractions replace mole fractions as the fundamental composition variable. The drawback of this theory is that it cannot predict the low critical solution temperatures that have been observed in many polymer systems. Eliminating this drawback, however, makes the equation more complex. The Flory-Huggins theory is adequate for our purposes.

In compositional quenching, the solvent is removed by flash devolatilization. In a deep adiabatic flash, most of the solvent is vaporized, generating an enormous ratio of vapor volume to liquid volume. The vapor volume that is generated completely disrupts the foam structure, so the partially devolatilized polymer is dispersed in the vapor as fine droplets or as small-diameter fibers (Figure 1) (6).

Figure 1. Flash devolatilization processing forms fibers in polymer blends.

To form a microdispersion of one polymer in another by compositional quenching, a common solvent must be found (e.g., xylene for polystyrene and polybutadiene). A single-phase solution is formed; it typically contains 5-10% total polymer and 90-95% solvent. This mixture is heated under pressure and then flashed.

The rapid removal of solvent by flash devolatilization causes a large displacement from equilibrium. Phase separation occurs by spinodal decomposition (7-9). The free energy of the system depends on the component concentrations and the gradients in concentration. Spinodal decomposition is governed by a fourth-order partial differential equation known as the modified Cahn-Hilliard equation:

This equation applies to a binary blend. Another partial differential equation is added for each additional component. Details have been given in the literature (10). Although these equations are mathematically complex, it is important to remember that they contain few adjustable parameters: the volume fraction of the polymers, the interaction parameter, and the chain lengths of the polymers. Diffusivity is a parameter that disappears on scaling. These equations can, in fact, be solved easily. A two-dimensional simulation of a ternary polymer blend requires a few hours on a fast personal computer.

The volume fractions of the components are most important in determining the morphology of a polymer system. For a ternary polymer system, a component with a volume fraction of 0.6 or higher always forms a continuous phase, whereas no continuous phase is observed for volume fractions <0.33. The semicontinuous phase is limited to the range of 0.3-0.5, and dispersed morphologies are confined to volume fractions <0.45.

The dependence of the equations on the interaction parameter is more difficult to describe. We know that the polymers dislike each other more with an increasing values. In determining the overall morphology, the ratio of the interaction parameters is believed to be more important than the absolute values themselves. Simulations performed with high values (e.g., _ab = 6, _ac = 12, and _bc = 6) yield the same morphology but with faster ripening (particle growth) than those performed with low values (e.g., _ab = 3, _ac = 6, and _bc = 3).

Simulations run on the effect of chain length for a binary system have shown interesting results (11). For off-critical quenches, the effects of the chain length of the matrix material on the domain size are negligible. However, a smaller chain length of the minor component greatly increases the ripening rate and therefore the domain size.

Conventional compounding. The most common method for generating multicomponent polymer blends is melt compounding. Even blends produced by compositional quenching must be processed by conventional means to form useful products. This type of processing often exposes the blend to forces that may affect its morphology. He and Nauman (12) addressed this concern by examining spinodal decomposition under shear flows. The geometry used in their simulations is a starting approximation of the region of a single-screw extruder near the barrel wall.

The effect of hydrodynamics on spinodal decomposition was first studied by Farrell and Valls (13). Rousar and Nauman (14, 15) and Vasishtha and Nauman (16) expanded this study by including the continuity equation and flows induced by pressure gradients. The work of He and Nauman (12) was based on all of these results.

The modeling equations included a convective diffusion form of the modified Cahn-Hilliard equation, in which an additional term accounts for flows induced by concentration gradients. This equation was coupled with the two-dimensional equations of motion and the continuity equation. A special volumetric body force term was included in the equations of motion to account for the heterogeneity of the fluid mixture (15, 16).

The geometry of interest consists of two concentric cylinders representing the walls of an extruder. The inner cylinder could be rotated to apply shear in the circumferential direction. Numerical simulations were performed on this geometry for conditions that are initially homogeneous (i.e., extrusion of a compositionally quenched blend) or phase separated (i.e., blend of two homopolymers).

Simulation results showed that including body force hastened phase ripening and that the resultant morphologies were similar to diffusion-only models. Applying shear had a tremendous effect on morphology. The appearance of phase boundaries was delayed, and there was a strong tangential orientation. This preferential orientation occurs because shear effects dominate at short and long times, whereas diffusion effects dominate the intermediate stages. The final morphology appears as alternating concentric regions of each phase (Figure 2). The starting condition of the system seems to have little effect on the final morphology--thus the importance of spinodal decomposition in traditional melt blending and compositional quenching.

Figure 2. Simulation of a two-phase extrusion showing alternating concentric regions of each phase. Shown is the morphology after 1 s (top) and 20 s (bottom).

After the final morphology was established, rotation was stopped. When the internal body force was present, the circumferential orientation began to break up and evolve toward compact domains. Such structures are expected in actual extrusion.

The work done so far provides a starting point for the modeling of melt blending based on the principles of spinodal decomposition. Actual extruders have a three-dimensional (3-D) geometry with more complex flow fields than do the models; the scale of the structures formed in actual extruders and processing times are similar to those used in these studies. Therefore, these results are a reasonable first approximation of the morphology to be expected from polymer blends processed by conventional means.

Reactive techniques. Techniques involving chemical reactions provide another method for producing polymer blends. The classic example is the production of high-impact polystyrene (HIPS) by precipitation polymerization. HIPS is manufactured by polymerizing styrene in the presence of dissolved polybutadiene (PB). Early in the polymerization, the PB phase separates because of the immiscibility of the rubber with the polystyrene (PS). As the polymerization continues, the styrene grafts to the PB, and as the volume ratio of PS to PB in styrene approaches 1, a phase inversion occurs. The rubber-phase morphology is strongly influenced by the shearing agitation during and shortly after phase inversion. During removal of the residual monomer, the PB in the rubber phase is cross-linked.

The size and morphology of the rubber phase can be controlled to obtain the desired properties. Besides the shearing action, initiators, chain-transfer agents, and diluents play an important role in controlling the morphology of the rubber phase (17). This third technique has to be modeled to obtain an overall understanding of the production of polymer blends.

Morphologies of interest

Of the morphologies in the work of Nauman and He, three types have significant potential for industrial polymer blends: core-shell, multiple discrete particle, and dual semicontinuous. The core-shell structure (see Sidebar "Classifying Morphology") has the ternary classification cp-dc-si. This morphology can simultaneously increase the volume of the dispersed phase and the matrix-elastomer interface for a given amount of elastomer. If the matrix-elastomer interface adhesion is strong, the material should have a higher impact strength than the same blend without adhesion (18). For component C, commercial blenders use a block copolymer of components A and B to increase matrix-elastomer adhesion, but it is clear from the simulations that many polymer compositions can give this structure.

TO SIDEBAR: Classifying morphology

The phase classification for the multiple-discrete-particle structure is cp-dc-dc; phase A contacts both B and C, and phases B and C do not contact each other (Figure 3). On a commercial scale, it would be possible to produce a bimodal particle size distribution of two different rubbers. It is believed that, for polymers that fail by crazing and shear yielding (such as ABS), the presence of the two particle sizes--one below the critical size and one above--would enhance the toughness of the matrix compared with the same matrix with only one particle size (19).

Figure 3. Ternary blend simulation yield a multiple discrete particle structure. The phase classification for this structure is cp-dc-dc; phase A contacts phases B and C, whereas phases B and C do not contact each other. For this simulation, a = 0.6, b = 0.2, c = 0.2, _ab = 3, _ac = 3, and _bc = 6.

Currently, multiple-discrete-particle blends are produced by compositional quenching of HIPS with a mixture of PS, PB, and PS-PB diblock copolymer. One set of rubber particles is produced by precipitation polymerization and the other by compositional quenching. Excellent impact strength can be obtained by adding small amounts of HIPS to the blend of PS, PB, and diblock (Figure 4). The volume fraction of the rubber phase was maintained at 23%.

Figure 4. Excellent impact strength can be obtained with compression molding by adding small amounts of high-impact polystyrene (HIPS) to the blend of polystyrene (PS), polybutadiene (PB), and PS-PB diblock polymers. The volume fraction of the rubber phase was maintained at a constant 23%. Compression molding was done for 2 h at 200°C.

The particulate morphology of ternary mixtures has been studied; dual semicontinuous morphology prepared by spinodal decomposition, however, has rarely been reported. A dual semicontinuous morphology with component C at the interface is shown in Figure 5. If stabilized, these blends represent a much better trade-off between Izod strength and tensile modulus than the particulate morphology. Besides mechanical properties, these blends can possess other desired features, such as a selective permeability.

Figure 5. A dual semicontinuous morphology with component C at the interface. If stabilized, these blends represent a much better trade-off between izod strength and tensile modulus than the particulate morphology. Besides mechanical properties, these blends can have other desired features, such as a selective permeability. For this simulation, a = 0.45, b = 0.45, c = 0.10, _ab = 6, _ac = 3, and _bc = 3.

Kilogram-scale preparations

Although modeling can give an excellent basis for the design of a polymer blend, experimental verification is still necessary. Also, even though excellent predictions of the morphologies are available, the relationship between the morphology and the physical properties is not always clear.

To produce blends by conventional melt compounding, devices such as twin-screw extruders and Banbury mixers are often used. The Banbury mixer, which consists of two spiral shafts that counter-rotate about a fixed axis, is capable of producing batch blends from a few to several thousand kilograms. The ingredients are pushed into the top of the mixer by an air cylinder ram. The time of mixing and the pressure exerted by the ram can be varied.

Twin-screw extrusion is a continuous process in which the premixed feed material enters the extruder, usually in the form of a solid. The material is subjected to high shear by the co-rotating screws. This high shear aids in melting the material and provides excellent mixing.

In compositional quenching, the blends of interest are produced by dissolving the polymers in a solvent at room temperature or, in some cases, slightly elevated temperatures. Typically, solutions of 3-10 wt% are made. The solution is pumped through a heat exchanger and then flashed across a needle valve into a low-pressure chamber. The needle valve provides back pressure so that the solution does not boil in the heat exchanger. The inlet temperature and flash pressure are monitored to control the domain size of the morphology. Commercially, the effluent polymer blend would be a pumpable liquid. The apparatus shown in Figure 6 can flash to a nearly dry solid and produces 5- to 10-kg samples.

Figure 6. This compositional quenching apparatus can flash to a nearly dry solid and produces 5- to 10-kg samples.

Analysis and images

After a blend is produced, we analyze the material to determine physical properties and confirm the predicted morphology. Mechanical properties such as impact strength and tensile strength must be measured. In most cases, we want increased impact strength over a wide temperature range without a significant loss in tensile modulus. Thermal properties such as melt flow rate are tested to determine the required processing parameters of the new blend. The combination of these properties determines the blend's marketability.

Because the microstructure of these blends can be on the order of 0.1-10 µm, it is often necessary to combine several microscopic techniques, including optical, scanning, and transmission electron microscopy. An optical phase-contrast microscope can distinguish the refractive indices of the three phases (20); Figure 7 was created using this technique. It offers ease of sample preparation, but there are two main drawbacks. First, if the refractive indices of two of the phases are close, it is difficult to distinguish the difference. Second, the optical resolution is poor for objects smaller than ~1 µm. Therefore, optical phase-contrast microscopy is unsuitable for the morphological study of materials with small particle sizes (21).

Figure 7. An optical-phase contrast microscope can distinguish among the refractive indices of three phases. The dark, continuous phase is polymethyl methacrylate (PMMA). The lighter, irregular regions are PS. The dark particles inside the PS are PB.

Scanning electron microscopy (SEM) offers much better resolution (~10 nm) than optical microscopy, and large samples may be examined with only minimal preparation. However, SEM is limited to the examination of sample surfaces, making it difficult--if not impossible--to distinguish the morphology of a ternary blend. SEM is used extensively to examine fracture surfaces. The fracture surface of a blend of PS, PB, and polypropylene (PP) is shown in Figure 8.

Figure 8. Scanning electron microscopy shows the fracture surface of a ternary blend of PS, PB, and polypropylene (PP). The large particles consist of PP surrounded by smaller PB spheres.

Transmission electron microscopy (TEM) is analogous to optical microscopy except that a beam of electrons, rather than light, is used as the illuminating source. Typical resolutions are on the order of 1 nm. TEM requires tedious sample preparation and phase-contrast enhancement by staining methods (22). Typical specimens for conventional TEM are thin films <100 nm thick.

High-voltage electron microscopy (HVEM) can penetrate thicker specimens than a conventional TEM can because its beam is stronger. Figure 9 was obtained using HVEM on a 0.75-µm-thick sample of a PS-PB-PP blend. With HVEM, stereoscopic images of the sample can be made and then reconstructed into a 3-D image. Several systems are available to reconstruct from serial sections; however, STERECON (STEReoscopic RECONstruction) has the unique ability to reconstruct a 3-D image by stereoscopic contour imaging (23). The photographs are scanned into a computer, and the digitized images are displayed on two monitors, arranged at 45° from a half-silvered mirror. The mirror combines the two images into a stereo pair. The contours are drawn with polarizing glasses in stereo overlay bit planes using a digitized pad. The reconstruction can be displayed many ways, the simplest of which is a contour drawing.

Figure 9. High-voltage electron microscopy (HVEM) of a core-shell structure. HVEM can penetrate much thicker sections than transmission electron microscopy because of its stronger beam. We obtained this image of a PS, PB, and PP blend on a sample 0.75 µm thick. The PB forms the dark rings around the white spheres, which are PP.

A valuable tool

It is important to confirm predicted morphologies with experimental results. A correlation must also be made between blend properties and blend morphology. When this is done, we have a powerful means with which to design marketable blends in the polymer community.

REFERENCES

ACS Publications Division Page ACS Home Page