Chemical Engineering Science 56 (2001) 3011}3020

Shape forming of ceramic powders by manipulating the interparticle pair potential

F. F. Lange*
Materials Department, University of California, Santa Barbara, CA 93106, USA Received 20 May 2000; received in revised form 28 July 2000; accepted 22 August 2000

Abstract New forming technologies have been developed for advanced ceramic powders by manipulating the potential between particles using either the electrostatic or the steric method. The important aspect in either case is the development of a short-range repulsive potential, which is summed with the pervasive attractive van der Waals potential to produce an interparticle pair potential characterized by a well in which particles sit at an equilibrium separation distance. This interparticle potential makes it more di$cult to push the particles into contact during consolidation, and it controls the #ow behavior of the consolidated body. Flow stresses in the range of 0.1 MPa impart clay-like behavior and opportunities for pressure forming. Much smaller #ow stresses allow shape forming via Colloidal IsoPressing, where weakly attractive particles are forced into contact within a rubber mold to produce an elastic body that can be removed from the mold without shape distortion. In these approaches, strength degrading inclusions can be removed prior to consolidation by passing the powder, formulated as a dispersed slurry, through a "lter to produce more reliable ceramic components. 2001 Published by Elsevier Science Ltd.
Keywords: Ceramics; Powders; Interparticle potential; Forming; Shaping; Slurry; Viscosity; Consolidation; Particles

1. Introduction Ceramic components are formed by compacting a powder into the desired engineering shape. The void space is eliminated with a high-temperature heat treatment called densi"cation. Because advance ceramic powders such as silicon nitride and alumina usually lack the plastic properties of traditional clay-based powders when dispersed in water, shape forming is carried out by either pressing a dry powder mixed with a polymer binder or by injection molding the powder mixed into a polymer that imparts plasticity. The reliability of engineering components shaped by these commercial methods su!er from inclusions, present in the powder and retained during shape forming and densi"cation, that concentrate an applied stress to severely degrade the component's strength. In addition, large quantities (up to +40 vol%) of an expensive polymer must be removed very slowly to avoid disruption due to gas evolution. Shape-forming methods can also use slurries (particle/liquid mixtures). A few of these slurry methods have been commercially used for many years to form speci"c
* Tel.: #1-8058938248; fax: #1-8058938486. E-mail address: #ange@engineering.ucsb.edu (F. F. Lange).

engineering components. Slurry methods for forming engineering shapes can be categorized as either consolidation methods or direct shaping methods. Consolidation methods start with a slurry containing a low-volume fraction of powder, which is concentrated by either evaporation or pressure "ltration. Examples include tape casting to form thin sheets (evaporation) (Mistler, Shane"eld, & Runk, 1978), slip casting to form thin walled bodies (capillary suction into a porous mold) (Michaels, 1958), and pressure "ltration (Lange & Miller, 1987). Because of the low-volume fraction of solids, the slurries can "rst be passed through a "lter to remove strength degrading inclusions. Also, these consolidation methods produce a body containing a very high-volume fraction of powder. However, because the liquid removed during consolidation must #ow through the consolidating body, these methods require long periods within the mold. In addition, tape and slip casting are limited, in practice, to thin and thin-walled bodies, respectively. Direct shaping methods start with a slurry containing a high-volume fraction of powder that can be either poured or injected into a mold. In these methods, the volume fraction of powder does not change during molding. Highly repulsive interparticle potentials are needed to formulate slurries containing the high-volume fraction

0009-2509/01/$ - see front matter 2001 Published by Elsevier Science Ltd. PII: S 0 0 0 9 - 2 5 0 9 ( 0 0 ) 0 0 4 8 4 - X

3012

F. F. Lange / Chemical Engineering Science 56 (2001) 3011}3020

of powder. During molding, the slurry must change to an elastic solid so that the body can be removed from the mold. Direct shaping methods include injection molding (Mangels, 1982), gel casting (Omatete, Janney, & Strehlow, 1991), direct coagulation casting (DCC) (Gauckler, Graule, & Baader, 1999) and VibraForming (Franks, Velamakanni, & Lange, 1995). In the case of injection molding and gel casting, the slurry's liquid-phase solidi"es by either freezing or polymerization. For DCC and VibraForming, the highly repulsive interparticle pair potential is converted to a highly attractive and strong particle network, e.g., by changing the pH of the liquid phase via a temperature-induced chemical reaction. Because high-volume fraction of solids are used in the direct shaping methods, removing strength degrading inclusions by "ltration (or sedimentation) is not practical. Although injection molding only requires very short periods within the mold, long periods are needed to remove the polymer without disruption, i.e., cracking, blistering, etc. Gel casting, DCC and VibraForming require long periods (10 min}2 h) within the mold because of the relatively long periods required to induce either gelation or the change in the interparticle potential. After forming, the liquid must be removed from the saturated powder body. Shrinkage occurs because the powder can further consolidate as liquid evaporates, driven by the capillary pressure exerted on the particle network. Surface tensile stresses arise because the shrinkage, which initiates at the surface where the liquid evaporates, is constrained by the interior (Scherer, 1990). The magnitude of the tensile stress, which can induce cracking during drying, depends on a number of factors including the initial relative density achieved during shape forming and the rate of drying. Bodies produced by direct shaping methods (gel casting, DCC and VibraForming) are more prone to cracking because they exhibit greater shrinkage relative to the higher density, consolidated bodies. Although slurry consolidation methods, which allow the removal of strength degrading #aws prior to shape forming, o!er greater structural reliability, their greater cost due to signi"cantly longer forming periods generally favor the use of conventional, dry powder methods that inherently produce unreliable bodies due to their uncontrolled #aw populations. The object here is to review recent work relating the potential between particles and the rheology of consolidated powder compacts. It will be shown that new shape-forming methods have been developed that were enabled by the basic phenomena reviewed here.

they are surrounded by a #uid with a di!erent dielectric constant (Horn, 1990). By itself, the van der Waals potential produces a network of particles in elastic contact; such particles are more di$cult to rearrange and consolidate due to friction (Franks & Lange, 1996). To keep the particles separated, they must be shrouded by matter that causes the system to increase its free energy as the shrouds of two approaching particles interpenetrate one another. Matter comprising the shroud must be of low mass so that the shroud itself does not signi"cantly contribute to the attractive van der Waals potential. In e!ect, the interpenetrating shrouds keep the particles apart and, in e!ect, shield either a portion or all of the attractive van der Waals potential. When the electrostatic double-layer method (approximated by the DLVO theory (Horn, 1990)) is used to produce a repulsive potential, the shroud is comprised of counterions. When the steric approach is used, the shroud is composed of molecules that are of either phys- or chem-adsorbed to the surface of the particles (Napper, 1983). 2.1. The electrostatic approach In the electrostatic method, where the shroud or barrier layer is produced with a cloud of ions, the density and &thickness' of the cloud is controlled in an aqueous slurry by pH and salt concentration (Horn, 1990). In this approach, a dispersed slurry (i.e., particles are highly repulsive) is formed by reacting either H O> or OH\  (acid or base) with neutral surface sites (}M}OH) to produce charged surface sites (e.g. }MOH> or }MO\).  Oppositely charged ions, called counterions, form a cloud around each particle in an attempt to neutralize the surface. As two particles approach one another, the concentration of counterions increase between the particles as the approaching counterion clouds penetrate one another. The increase in counterion concentration between the particles gives rise to an increase in free energy of the system, which, in turn, produces a repulsive interparticle potential. The distance where the repulsion becomes signi"cant depends on the thickness of the counterion cloud, generally described as the Debye length. Large Debye lengths (low counterion concentration) can produce a long-range repulsive potential where particles always repel one another. Addition of salt (adding counterions without changing pH) to a dispersed slurry decreases the Debye length such that strong interparticle repulsion only occurs at small interparticle distances (Velamakanni, Chang, Lange, & Pearson, 1990; Colic, Franks, Fisher, & Lange, 1997). For this case, particles can be attractive to one another by the van der Waals potential, and only feel repulsion at very small separation distances. That is, when salt is added to signi"cantly reduce the Debye length, the particles can reside in a potential well (generally referred to as the secondary minimum, whereas the primarily

2. Interparticle pair potentials The van der Waals potential always causes particles with similar dielectric constants to be attractive when

F. F. Lange / Chemical Engineering Science 56 (2001) 3011}3020

3013

minimum is the condition where particles are in physical contact), where they retain an equilibrium separation distance. Classical DLVO theory, which treats counterions as point charges without dimensions, does not teach this phenomenon. Instead, the DLVO theory teaches that the Debye length decreases to zero with large additions of counterions to produce an interparticle potential identical to that at the isoelectric point (i.e.p.) where only the van der Waals potential is present. Velamakanni et. al. (1990) demonstrated that adding excessive amounts of salt only produces a weaker network than that observed at the isoelectric point where no repulsive potential exists between particles. Evidence suggests that at high salt concentrations the Debye length approaches a minimum value that is related to the "nite size, and possibly the hydration potential of the added counterion (Colic et al., 1997). As discussed below, it has been discovered that the attractive but not touching particle network is of great importance to new shape-forming technologies. Namely, it allows the #ow stress of a slurry and the yield stress of a saturated powder compact to be manipulated by changing the interparticle potential. 2.2. The steric approach In the steric method, two e!ects give rise to the repulsive potential produced by adsorbed molecules (Napper, 1983). The "rst arises when the adsorbed molecules on the surface of two approaching particles begin to mutually con"ne their con"gurational space. Con"nement decreases their entropy, gives rise to an increase in free energy, and thus, a repulsive interparticle potential. The second e!ect is due to the strain energy that arises when the con"ned molecules are further pushed together. The interparticle separation distance where the repulsive potential initiates is generally estimated as twice the size of the adsorbed molecules. Thus, very large molecules generally produce long-range repulsion where the molecular &brushes' extending from the particle surfaces completely shield the attractive van der Waals potential. Whereas small molecules only produce strong repulsion at small interparticle separation distances and can only truncate the van der Waals potential. The short molecules thus produce a weakly attractive, but non-touching particle network, which has been found to be very useful for new shape forming technologies. Many di!erent types of molecules (hydrocarbon chains provided by alcohol molecules (Kramer & Lange, 1994), surfactants (Luther, Kramer, Lange, & Pearson, 1994; Ducker, Luther, Clarke, & Lange, 1997), bi-functional molecules such as aminosilanes (Colic, Franks, Fisher, & Lange, 1998), di-block copolymers (Palmqvist, Sigmund, Sindel, & Lange, etc.), have been identi"ed to produce that short-range repulsive potential which appears to be important for new shape-forming tech-

nologies. Kramer and Lange (1994), who reacted simple alcohols with increasing molecular weight with the ceramic particle, have shown that the strength of the particle network is inversely proportional to the length of the attached molecules. Molecules with shorter chains truncate the van der Waals potential at a very short interparticle distance to produce a deeper potential well and a stronger network, whereas longer molecules produce a shallow potential well and a weaker network. Kramer and Lange (1994) have also shown that the molecules must be strongly bound to the surface to avoid being pushed away during consolidation. It has been shown that phys-adsorbed molecules can easily be pushed away during particle packing (Kramer & Lange, 1994; Ducker, Luther, Clarke, & Lange, 1997), whereas chem-adsorbed molecules are more robust and thus, more desirable for producing the repulsive potentials required for ceramic powder processing. It was shown that amino-silane molecules (Colic, Franks, Fisher, & Lange, 1998), terminated at one end with RO-groups (R"alkyl), will hydrolyze in water and react with }Si}OH sites on Si N surfaces to produce strongly   bonded, chem-adsorbed molecules that are di$cult to push away during pressure "ltration. It was demonstrated that a di-block copolymer, PMAA-block-PEO chem-adsorbs on alumina powder to produce well dispersed aqueous slurries over a wide range of pH (Palmqvist, Sigmund, Sindel, & Lange, 2000). At pH'4, the negative sites on the PMAA anchor block over compensate for the positive sites on the alumina surface to reduce the pH (pH of the isoelectric point) of the  coated powder from 9 to +5. At pH"9, the particles are strongly dispersed due to repulsive potential produced by both the electrostatic repulsive potential produced by the net-negative surface charge, and by the steric e!ect of the polyethylene oxide (PEO) steric block of the copolymer. At pH"5, the pH of the coated alumina  particles, the steric e!ect of the PEO block also produces a dispersed slurry.

3. Relation between rheology and interparticle potential Fig. 1 describes the potential and force between weakly attracted particles as a function of their separation distance. The schematic function for the force, shown in Fig. 1b, is the derivative of the particle pair potential shown in Fig. 1a with respect to the separation distance, &h'. The force is zero at the equilibrium separation distance (h , bottom of the potential well). The second CO derivative at h is the &spring' constant between the two CE particles, which is related to the elastic modulus (G) of the attractive particle network. When h(h , the parCO ticles are repulsive, whereas a tensile force is needed to separate the particles for h'h . The maximum force CO needed to separate the particles and the number of

3014

F. F. Lange / Chemical Engineering Science 56 (2001) 3011}3020

Fig. 2. Relative density of AKP-50 alumina (particle diameter +0.2 m) as a function of consolidation pressure. Each data point is the average of between 2 and 5 specimens from the same powder lot (Franks & Lange, 1996).

Fig. 1. Interparticle pair potential (a) and interparticle force (b) for weakly attractive particle (Yanez, Shikata, Lange, & Pearson, 1996).

particles per unit volume are related to the #ow stress of the particle network, i.e., the stress where the network no longer behaves in an elastic manner, but breaks apart to initiate #ow. Thus, by measuring the #ow stress and elastic modulus of the particle network in the slurry state, the "rst and second derivative of the pair potential can be estimated and related to systematic changes made in the chemistry used to form and alter the shroud around the particles for a given powder. The interparticle potential shown in Fig. 1a is the sum of the van der Waals attractive potential and a shortrange repulsive potential produced by the penetration of either two counterion clouds with short Debye lengths, or two brushes of chem-adsorbed short molecules. Assuming that the van der Waals potential is not strongly a!ected by changes in what is done to the surface to alter the short-range repulsive potential, it can be seen that the force needed to pull two particles apart will scale inversely to the depth of the potential well, viz., for a "xed van der Waals potential, a deeper potential well will produce a steeper slope at the in#ection point. Thus, for a "xed volume fraction of particles, the strength of the attractive network can be changed by altering either the Debye length (adding more salt) or the length of the chemadsorbed molecules. For a given interparticle potential, the #ow stress and elastic modulus of the attractive network will increase with the number of particles per unit volume, i.e., the volume fraction of powder within the slurry. Systematic measurements for slurries formulated with di!erent salt

contents and di!erent volume fractions of powder show that for a given interparticle pair potential, the #ow stress and elastic modulus are power law function of the volume fraction, with exponents of +3.5 and 4.5, respectively (Yanez, Shikata, Lange, & Pearson, 1996). That is, small changes in the volume fraction produce very large changes in slurry rheology. Although much more can be said about the relation between interparticle potentials and the rheology of the slurry state, the important features reported here will concern particle packing and the mechanical properties of the consolidated body still saturated with the liquid used to formulate the slurry. That is, the consolidated state is of greatest concern here since it must be formed into an engineering shape. It will be shown that the rheological behavior of a consolidated body not only depends on the interparticle potential and the volume fraction of powder, but also on the consolidation pressure (Franks & Lange, 1996; Palmqvist et al., 2000; Klein, Fisher, Franks, Colic, & Lange, 2000). It was known that the particle packing density achieved during consolidation was related to the interparticle potential. Low relative densities result from slurries formulated for conditions where only the van der Waals potential prevailed, e.g., formulated at the i.e.p., due to their resistance to rearrangement caused by the friction between the touching particles. Also as shown in Fig. 2, the relative density of bodies consolidated from these #occed slurries are also very pressure sensitive. On the other hand, slurries &lubricated' by a highly repulsive interparticle pair potential pack to the highest relative density that can be achieved for a given powder. In addition, the relative density is not as sensitive to the applied pressure. Pair potentials that produce weakly

F. F. Lange / Chemical Engineering Science 56 (2001) 3011}3020

3015

attractive particle networks in the slurry state, e.g., by adding excess salt to a dispersed slurry, result in a relative density that is intermediate to the strongly repulsive and strongly attractive pair potential. The relative density of these weakly attractive particle networks is much less sensitive to the applied pressure relative to the strongly attractive networks. These e!ects are illustrated in Fig. 2, where the interparticle potential was produced by the electrostatic double-layer method, where salt was added to the dispersed slurry to produce the weakly attractive particle network. Franks and Lange (1996) performed axial compression experiments on cylindrical specimens that were used to accumulate the relative density data shown in Fig. 2. They were able to classify the plastic and elastic nature of the di!erent bodies with regard to the interparticle pair potential, relative density, and the applied pressure used to consolidate the bodies from the slurry state. The stress}strain behavior typical of bodies consolidated from slurries formulated to contain a weakly attractive particle network, is shown in Fig. 3. As shown, stress}strain behavior is elastic to a maximum stress called the peak stress. Below a critical consolidation pressure, the bodies were plastic, where, after achieving a peak stress, the stress would drop to a #ow stress. If the plastic body was unloaded and then reloaded, it would again exhibit the same #ow stress, without a peak stress, as shown in Fig. 3, which could continue up to a strain of 50% or more before strain hardening would occur due to the constraint imposed by the loading platens due to the decreased height to diameter ratio of the cylindrical body. When the consolidation pressure exceeded the critical value, the body was elastic; a load drop would only occur when the body cracked to break into pieces. In the plastic regime, i.e., for consolidation pressures less than the critical value, Fig. 4 illustrates that the peak stress increased with the consolidation pressure, whereas

Fig. 4. Peak stress on initial loading and #ow stress plotted as a function of consolidation pressure for a typical weakly attractive system (pH 4.0, 2.0 M NH Cl) (Franks & Lange, 1996). 

Fig. 3. Typical stress}strain curve for plastic specimen comparing peak stress and #ow stress during initial and repeat loading. Note that the lack of peak stress on the second loading, while the #ow stress remains unchanged (Franks & Lange, 1996).

both the #ow stress (see Fig. 4) and relative density (see Fig. 2) were relatively independent of the consolidation pressure. Bodies consolidated with #occed slurries (formulated at the i.e.p.) were observed to change from a plastic to an elastic behavior at a relative density of +0.52. Over the entire range of consolidation pressures, the peak stress increased without a relaxation to a #ow stress. If the body was plastic (relative densities )0.52), the stress} strain relation transitioned from elastic to plastic behavior at a yield stress much like that of a metal. For bodies that were elastic, the body would fail by crack extension prior to any #ow. Bodies consolidated from dispersed slurries were elastic and failed by fracture. To understand the reason for the relaxation from a peak stress to a #ow stress, one should know that particles that are held apart in the slurry state by either a short- or long-range repulsive force can be pushed together during consolidation. The mean force pushing particles together will depend on the consolidation pressure and the number of particles per unit volume. Numerical simulations (Thornton & Antony, 1998) and photoelastic experiments (Kuhn, McMeeking, & Lange, 1991) have shown that the force between particle pairs is widely distributed because of the large number of percolating particle paths that support the applied pressure. That is, the force exerted between some pairs of particles is very high, whereas other particles can be completely shielded from the applied stress by their neighbors. Thus, only a fraction of the particles form percolating particle paths that support much of the applied stress. Particles that are pushed together during consolidation form a very strong touching network; if a large number of particles are pushed together, the body will be elastic and fracture. Franks and Lange (1996) suggested that when the particles are held apart in the slurry state by a shortrange repulsive potential (e.g., for the case of pH

3016

F. F. Lange / Chemical Engineering Science 56 (2001) 3011}3020

4, 2 mol NH Cl added salt), the fraction of particles  pushed together during consolidation increases with the consolidation pressure. They suggested that the stress relaxes from its peak value to a #ow stress as observed in Fig. 3 for a plastic body when the strong network of particles produced during consolidation is broken apart by the applied stress. The broken network of touching particles does not reform, and thus, a peak stress in not observed when the body is loaded again. The peak stress increases with the consolidation pressure and eventually becomes so large that the body fails by crack extension before the stress can relax by plastic #ow. Thus, for a given interparticle pair potential, the transition between the plastic and elastic behavior occurs at a critical consolidation pressure where a su$cient number of particles are pushed into contact to form a strong network that will fracture before it breaks apart to #ow. The plastic/elastic data for bodies consolidated from slurries formulated at the i.e.p., where particles are known to touch one another, was used to estimate the volume fraction of touching particles needed to form an elastic body. This apparently occurs at a volume fraction of 0.52 for the speci"c powder studied by Franks and Lange (1996). Since the initial report of the plastic to brittle transition, several studies have shown that the transition pressure and the volume fraction of particles where the transition occurs depends on the details of the interparticle potential. For example, di!erent counterions appear to produce either stronger or weaker attractive networks, and change the consolidation pressure associated with the plastic to brittle transition (Franks & Lange, 1996). In addition, these di!erent pair potentials also produce different #ow stresses that can be smaller or larger than those shown in Fig. 4. It has been shown that one can achieve a #ow stress between 0.01 and 0.5 MPa for di!erent commercially important powders such as alumina (Franks & Lange, 1996) and silicon nitride (Colic, Franks, Fisher, & Lange, 1998), which causes the consolidated bodies to behave like clay. Di!erent commercial throwing clays have #ow stress between 0.01 and 0.1 MPa (Franks, 1999). It was reasoned that the force needed to push particles into touching contact, i.e., from the secondary minimum (the potential well shown in Fig. 1a) to the primary minimum would depend on the slope (i.e., resisting force) of the pair potential located between these two minima (Franks & Lange, 1996). When the pair potential is composed of a short-range repulsion, the slope would be steep and the force large, relative to the case where the repulsive potential is long range, the slope shallow, and a small force needed to push the particles together. Thus, as observed, bodies consolidated from slurries formulated with highly repulsive particles are expected to be brittle after consolidation. The short-range repulsive potential needed to produce a plastic body with a high relative density can also be

achieved using the steric approach, i.e., by chem-adsorbing molecules to the surface of the particles. Two di!erent amino-silane molecules (Klein et al., 2000) and a diblock copolymer (Palmqvist et al., 2000) have been used to produce plastic bodies composed of either zirconia or alumina, respectively. Fig. 5 illustrates the relative density and plastic to elastic transition for an aqueous ZrO (i.e.p."pH 7.5) powder that was formulated with  three di!erent interparticle pair potentials. Two aminosilanes, N-(triethoxysilylpropyl)-O-polyethylene oxide urethane (peg-silane), and N-(3-triethyoxysilylpropyl) gluconamid (glucol-silane) where reacted with the powder to formulate two of the three slurries (Klein et al., 2000). The peg-silane molecules are approximately twice as long (22 carbon units) as the glucol-silane molecules. One terminal group, }Si}(OR) , of each silane hydrolyzes  and reacts with the Zr}OH surface sites to chemically bond to the powder surface. The third slurry was formulated without either amino-silane. All three slurries were formulated at their respective isoelectric point so that each would produce an attractive particle network. Namely, for the unreacted powder, the particles would form a touching network, whereas for the powder reacted with either of the two amino-silanes, the particles would form weakly attractive networks, held apart by their respective amino-silane molecules. Rheological measures of the slurries used to consolidate the bodies represented in Fig. 5 showed that the strongest network was produced with the unreacted powder, and the weakest network was produced when the powder was reacted with the longer, peg-silane molecules. Likewise, consolidated bodies with the lowest relative density and lowest critical transition pressure (7.5}10 MPa) were produced with the unreacted powder, which produced the strongest particle network. Whereas, Fig. 5 shows that consolidated bodies with the highest

Fig. 5. Peak stress vs. relative density plots for uncoated-, glucol- and peg-silane powder compacts each formulated from a slurry containing 1.0 M TMACl. The numbers on either side of the vertical lines indicating the plastic-to-brittle transitions correspond to the consolidation pressures in MPa (Klein et al., 2000).

F. F. Lange / Chemical Engineering Science 56 (2001) 3011}3020

3017

relative density and the highest critical transition pressure (35}50 MPa) was produced with the powder reacted with the largest amino-silane molecule, which is expected to produce the shallowest potential well. Each series of consolidated bodies, each produced with a di!erent interparticle pair potential, had their own #ow stress, largest for the unreacted powder, and smallest for the powder reacted with the largest amino-silane molecule. These results are consistent with those discussed above where the electrostatic double-layer (counterion shroud) approach was used to produce the interparticle potential.

4. Shape-forming technologies enabled by short-range repulsive potentials Above, it was shown that plastic powder compacts could be consolidated from slurries formulated to contain attractive, but non-touching particle networks. The short-range repulsive potential needed for these weakly attractive networks could be achieved with either the electrostatic double layer approach by adding excess counter ions to decrease the Debye length, or the steric approach by using relatively short, chem-adsorbed molecules. An experienced person would suggest that the use of chem-adsorbed molecules, which are di$cult to push away from between the particles during consolidation, is the more robust approach. As shown by Colic et al. (1998), who used peg-silane molecules to produce plastic silicon nitride powder bodies from aqueous slurries, the chem-adsorbed molecules can also protect the surface from water molecules that would normally react with the Si N . Chem-adsorbed molecules can also be syn  thesized to have other important functions, e.g., they can be used either to sense speci"c phenomena during processing or induce particle bonding via a photo-sensitive reaction. Although not reviewed in detail here, the cited literature (Franks & Lange, 1996; Colic, Franks, Fisher, & Lange, 1997, 1998) will show that the #ow stress for a given powder system can be altered by changing the strength of the weakly attractive particle network. For the electrostatic approach, the network strength can be altered by changing the surface charge density (pH of slurry), the concentration of added counterions, and the type of counterion (e.g., Li> through the much larger tetramethyl ammonium ion, TMA> (Colic et al., 1997)). For the steric approach, the network strength can be altered by using a di!erent chem-adsorbed molecule, or by changing the apparent length of a given chem-adsorbed molecule by adding salt to the slurry (Colic et al., 1998). When both the electrostatic and steric approach are combined, as for example, by the use of a diblock copolymer where the &anchor' block dissociated to become charged with changing pH and can overcompensate the surface charge of the particle, while at the same

time, the steric block can produce a strongly repulsive network even when the pH is changed to where the surface is neutral (Palmqvist et al., 2000). Using these methods of changing the strength of the weakly attractive particle network, one can produce consolidated bodies that can have a #ow stress too low to retain its shape due to the action of gravity, just right to allow clay-like plastic forming, to a plastic body with a yield stress too high for any forming technology other than injection molding. Also noteworthy is that the methods described above to produce a plastic body are also consistent with the need to "rst produce a well-dispersed slurry (highly repulsive particle network) needed to remove strength degrading inclusions by passing the slurry through a "lter prior to consolidation and shape forming. The remainder of this review will describe three emerging shape-forming technologies enabled by the science of plastic particle networks described above. 4.1. Shape forming via plastic deformation A tour of a factory that manufactures electrical insulators for power transmission will enlighten the visitor to the three processing steps; slurry formulation, consolidation, and pressure forming, that have been used for millennia, in one form or another, to shape clay-based products. Slurry formulation initiates with the mixing and milling of three raw materials as an aqueous suspension: clay (e.g., kaolin), a "ller (e.g., quartz or alumina), and a #uxing agent (e.g., feldspar). Slurry formulation includes additives that control the interparticle potential. Each constituent has its processing role; after consolidation, the clay imparts the plasticity needed for pressure molding, the "ller limits the shrinkage of the clay when it dehydrates, i.e., loses water molecules that are a part of the crystalline structure at high temperatures, and the feldspar contains potassium that reacts with the clay at high temperatures to produce a de"ned volume fraction of melt that aids densi"cation. After milling, the slurry is pumped to a dewatering press (pressure "lter) that increases the solid fraction and forms a water saturated "lter &cake'. The cakes are placed into an auger, de-aired, and extruded to form a long cylinder. The cylinder is cut into sections with a wire. In the "nal step, the cylinders are placed in one half of a die, the other half is pressed onto the cylinder to squeeze the clay into all corners of the cavity to form the desired shape. Without the need for the "ller and #uxing agent, powders used to produce advanced ceramics, with the clay-like plasticity reviewed above, can now be shaped by the forging method used for clay. The workability of a plastic material is de"ned as how much shape change can be accomplished before tears and cracks appear. Workability not only depends on the yield stress of the material, but also on the e!ect

3018

F. F. Lange / Chemical Engineering Science 56 (2001) 3011}3020

of strain rate on the yield stress and the exact shape that is being formed. The reason why the workability is dependent on the shape is that the stress and strain rate distribution is not uniform within the component as its being formed. Recently, Hbaieb et al. reported the workability of a cylindrical Al O powder bodies, saturated with water,   and formed by pressure "ltration from a slurry formulated to have a weakly attractive particle network (dispersed at pH"12, and coagulated with 0.5 mol of KCl). For this method, the short-range repulsive potential must be chosen such that after the slurry is consolidated below the critical consolidation pressure, the yield stress of the body is similar to that found for clay bodies. The cylindrical bodies, with a yield stress of +0.1 MPa (similar to commercial throwing clay) were subjected to both axial and diametrical loading as shown in Fig. 6. Stress}strain data obtained during axial loading was used to determine the elastic modulus of the body during its initial elastic regime, and the yield stress. These data were used by a commercial "nite-element program (elastic, perfectly plastic deformation model) to predict the shape change that would be expected in diametrical loading. The "nite-element program was also used to determine the stress where cracks were "rst observed in both types of experiments. For axially deformed bodies, shear bands were "rst observed on the surface at +453 from the loading direction prior to the formation of tear-like cracks parallel to the shear bands at strains between 0.25 and 0.35 (Fig. 6a) regardless of aspect ratio (height/diameter ratio). The formation of shear bands was consistent with the surface stresses (tensile hoop plus axial compression) determined in the simulations. Data obtained from the axial experiments (elastic modulus, #ow stress and coe$cient of friction) simulated the diametrically loaded specimens very well. Cracking in these bodies occurred at an apparent strain between 0.2 and 0.3 (Fig. 6b). Unlike the axially loaded specimens, shear bands were not observed and the crack initiated at the edge of the specimen and propagated axially along the unconstrained surface. These observations were consistent with the surface stresses (tensile hoop and smaller

axial tension) determined from the "nite-element simulation. As shown in Fig. 6c, the data obtained from axial compression experiments could not only be used to determine the stress where cracks were "rst observed in diametrical compression, but could also be used to determine the shape of the deformed specimen as a function of the applied load and loading rate. With further information, e.g., the e!ect of strain rate on the yield stress, simulations could be used to determine the loading condition required to form engineering components within the workability limits of the plastic powder compact. 4.2. Shape forming via colloidal isopressing (Yu, & Lange) Colloidal isopressing is a rapid method to form an engineering shape from the slurry state, previously "ltered to remove strength degrading inclusions. A slurry is formulated to produce a weakly attractive particle network with a short-range repulsive interparticle pair potential. The short-range repulsive potential is chosen such that the slurry can be consolidated to a high particle density that is easily #uidized by vibration. The role of the short-range repulsive potential is two-fold. First, it prevents particle from being pushed into contact during consolidation while the particles are consolidated to a high relative density. Second, the interparticle pair potential is chosen such that the yield stress is very low, i.e., so low that after consolidation, the body does not retain its shape but still appears #uid-like, albeit with a much higher viscosity relative to the initial slurry state. The #uid-like body is injected into a rubber mold and subjected to a larger isostatic pressure to force particles into contact. The rubber mold contains a porous region so that the particle network within the mold can be compressed and the particle pushed into contact. Pushing the particles together causes the molded material to become a strong elastic body with the shape of the rubber mold. Because the particles are forced into contact at a high pressure, the liquid remaining within the component can be removed by evaporation without shrinkage.

Fig. 6. Axial compressive experiments (a) were used to obtain properties of the plastic powder compact and to simulate the expected results, via "nite-element analysis, of diametrical compressive experiments (b and c). The "nite-element analysis was also used to determine the stresses where shear banding was observed in the axial compressive experiments, and where cracks were observed in the diametrical experiments (Hbaieb et al.).

F. F. Lange / Chemical Engineering Science 56 (2001) 3011}3020

3019

Fig. 7 shows a metal car valve, the silicon rubber mold that replicates the valve, and the aluminum oxide valve produced from the rubber mold via the colloidal isopressure method. Forming the valve within the rubber mold requires less than one minute at an isopressure of 100 MPa. 4.3. Fabricating ceramic matrix composites via vibraintrusion (Haslam, Beroth, & Lange, 2000) An important attribute of any ceramic, reinforced with very strong ceramic "bers, generally called a ceramic matrix composite (CMC), is that its strength be relatively insensitive to the presence of notches and holes. That is, the strength of a notched CMC component should be essentially identical to that of the un-notched body with the same, reduced cross-section. Since the failure strain of a strong "ber is generally much larger than the surrounding matrix, crack extension initiates within the matrix. In terms of crack extension, one requirement for notch insensitivity is that the "bers must be isolated from the very high-stress "eld of a matrix crack. A "ber within a good CMC is only expected to fail when the applied stress exceeds its strength. The processing of porous matrix CMCs starts by packing a powder around the "bers within a preform, e.g., stacked sheets of cloth, a woven or knitted fabric, etc., and ends with a step that makes the powder matrix sti!er and strong. Haslam et al. (2000) have introduced a rapid method to pack powder around the "bers called vibraIntrusion. VibraIntrusion also utilizes a consolidated body that can be #uidized as described above. The slurry containing the weakly attractive particle network is "rst consolidated by pressure "ltration and then packed around the "bers by vibration-assisted intrusion. As described above, it requires a slurry formulated with a special interparticle pair potential and knowledge of how this potential is a!ected by the consolidation pressure. With this special interparticle potential, a consolidated body can be formed with a high particle packing density with the rheological properties of a #uid. To prevent evaporative drying, the #uidized body is rolled between sheets of plastic to form a thin layer, frozen to aid handling, placed between two layers of ceramic cloth, and vibrated between the "bers after it thaws. Since the #uidized body exhibits shear-rate thinning, the vibration reduces the viscosity and allows rapid intrusion of the particles into the "ber preform. This process can be repeated, with intermediate freezing of the previously intruded cloth layers, to quickly fabricate a multi-layered CMC. The intermediate freezing is needed to aid in handling thin layer of #uidized body. Since the thawed CMC is #exible, it can be bent, cut and formed, similar to a thawed, epoxy/"ber prepreg used to fabricate wings on an airplane. After shaping, the liquid within the powder matrix is removed by evaporation,

Fig. 7. Examples of alumina components produced by the colloidal isopressing method. Metal engine valve (left), silicone rubber mold cast from the metal part (center), and sintered alumina valve (right) (Yu & Lange).

and the matrix is made strong by one of a number of methods described elsewhere (Lange, Levi, & Zok, 2000).

Acknowledgements This research was supported by the Army O$ce of Research, under Grant DAAG55-98-1-0455.

References
Colic, M., Franks, G. V., Fisher, M. L., & Lange, F. F. (1997). E!ect of counterion size on short range repulsive forces at high ionic strengths. Langmuir, 13(12), 3129}3135. Colic, M., Franks, G. V., Fisher, M. L., & Lange, F. F. (1998). Chemisorption of organofunctional silane on silicon nitride for improved aqueous processing. Journal of the American Ceramic Society, 81(8), 2157}2163. Ducker, W. A., Luther, E. P., Clarke, D. R., & Lange, F. F. (1997). E!ect of zwitterionic surfactants on interparticle forces, rheology and particle packing of silicon nitride slurries. Journal of the American Ceramic Society, 80(3), 575}583. Franks, G. V., Velamakanni, B. V., & Lange, F. F. (1995). VibraForming and in-situ #occulation of consolidated, coagulated alumina slurries. Journal of the American Ceramic Society, 78, 1324. Franks, G. V., & Lange, F. F. (1996). Plastic to brittle transition of saturated, alumina powder compacts. Journal of the American Ceramic Society, 79(12), 3161}3168. Franks, G. V. (1999). Ceramic powder processing and clay-like #ow stress of saturated alumina powder compacts. Proceedings of the 27th Australasian Chemical Engineering Conference (Chemeca 99), Newcastle, Australia. Gauckler, L. J., Graule, T. J., & Baader, F. H. (1999). Ceramic forming using enzyme catalyzed reactions. Mater. Chem. Phys., 61, 79. Haslam, J. J., Beroth, K. E., & Lange, F. F. (2000). Processing and properties of an all-oxide composite with a porous matrix. Journal of the European Ceramic Society, 20, 607}618. Hbaieb, K., Franks, G. V., Lange, F. F., & McMeeking, R. M. Simulating the plastic deformation of saturated alumina powder compacts with the "nite element method, to be published. Horn, R. G. (1990). Surface forces and their action in ceramic materials. Journal of the American Ceramic Society, 73(5), 1117}1135.

3020

F. F. Lange / Chemical Engineering Science 56 (2001) 3011}3020 Mistler, R. E., Shane"eld, D. J., & Runk, R. B. (1978). Tape casting of ceramics. In G. Y. Onoda & L. L. Hench (Eds.), Ceramic processing before xring (pp. 411}448). New York: Wiley-Interscience. Napper, D. H. (1983). Polymeric stabilization of colloidal dispersions. London: Academic Press. Omatete, O. O., Janney, M. A., & Strehlow, R. A. (1991). Gelcasting: a new ceramic forming process. American Ceramic Society Bulletin, 70, 1641. Palmqvist, L. M., Sigmund, W., Sindel, J., & Lange, F. F. Dispersion and consolidation of alumina using a bis-hydrophilic diblockcopolymer. Journal of the American Ceramic Society, in press. Scherer, G. W. (1990). Theory of drying. Journal of the American Ceramic Society, 73, 3. Thornton, C., & Antony, S. J. (1998). Quasi-static deformation of particulate media. Philosophical Transactions of the Royal Society of London Series A, 356(1747), 2763}2782. Velamakanni, B. V., Chang, J. C., Lange, F. F., & Pearson, D. S. (1990). New method for e$cient colloidal particle packing via modulation of repulsive lubricating hydration forces. Langmuir, 6(7), 1323}1325. Yanez, J. A., Shikata, T., Lange, F. F., & Pearson, D. S. (1996). Shear modulus and yield stress measurements of attractive alumina particle networks in aqueous slurries. Journal of the American Ceramic Society, 79(11), 2917}2924. Yu, B. C., & Lange, F. F. Colloidal isopressing: A new ceramic processing method, to be published.

Klein, S., Fisher, M., Franks, G. V., Colic M., & Lange, F. F. E!ect of the interparticle pair potential on the rheological behavior of zirconia powders: Part II * the in#uence of chem-adsorbed silanes. Journal of the American Ceramic Society, in press. Kramer, T. M., & Lange, F. F. (1994). Rheology and particle packing of chem- and phys-adsorbed, alkylated silicon nitride powders. Journal of the American Ceramic Society, 77(4), 922}928. Kuhn, L. T., McMeeking, R. M., & Lange, F. F. (1991). Modelling powder consolidation. Journal of the American Ceramic Society, 74(3), 682}685. Lange, F. F., Levi, C. G., & Zok, F. W. Processing "ber reinforced ceramic composites with porous matrics. In R. Warren (Ed.), Comprehensive composite material, A. Kelly, Editor-in-Chief, Ceramic, carbon and cerment matrix composites, vol. 4, Amsterdam: Elsevier Science Ltd. Lange, F. F., & Miller, K. T. (1987). Pressure "ltration: Consolidation kinetics and mechanics. American Ceramic Society Bulletin, 66, 1498. Luther, E. P., Kramer, T. M., Lange, F. F., & Pearson, D. S. (1994). Development of short range repulsive potentials in aqueous colloidal processing of silicon nitride. Journal of the American Ceramic Society, 77(4), 1047}1051. Mangels, J. A. (1982). Injection molding ceramics. Ceramic Engineering Science, 3, 529. Michaels, A. S. (1958). In W. D. Kingery (Ed.), Rheological properties of aqueous clay systems, Part I: Slip Casting, Ceramics Fabrication Processes (pp. 23}31) Cambridge, MA: MIT Press.