Ceramics Manufacturing-from Science to Practice

Transcription

1 Int. J. of The Soc. of Mat. Eng. for Resources Vol. 3 No, 1 5 `10 (1995) Plenary Lecture Ceramics Manufacturing-from Science to Practice by Michael J. CIMA õ Background The following is a brief history of the development of ceramics processing science. It is a somewhat personal perspective, but I think it gives the reader a sense of what spurred research in ceramics processing, both structural and electronic ceramics. The contrasts between structural ceramics processing research and that for electronic ceramics are significant because they illustrate important differences in approach to materials research. I also present a case to show that the title of this paper is backwards. Most existing ceramics manufacturing methods began with practice. A processing science arose from careful observation of existing processes. I end this article with an example from my laborartory that illustrates what may an important direction for ceramics processing research and technology. Ceramics technology has its roots in antiquity. Ancient cultures throughout the world developed processes by which naturally occurring materials could be transformed into useful articles. Advances in ceramics technology can be found in the archeological record and are the subject of several reviews and proceedings. Many developments were important, but the discovery of triaxial porcelains, colorful glazes, and high temperature firing methods are notable. The aesthetic quality of fine porcelain undoubtedly was a major factor for world-wide demand. Triaxial porcelains also had a much wider firing range than earlier ceramic bodies. Thus, increased yields could be obtained even with rather crude furnace technologies. This is just one example where the manufacturability of a material may have as much to do with its acceptance as do the properties of the material. Physical ceramic science developed in the early 1960's and coincided with growth in nuclear, space, and defense industries. These industries were primarily interested in ceramics for hostel environments. Ceramic components were critical structural elements for many of these systems. The cost of the ceramic component was, therefore, secondary to function and properties during use. They were typically required in only very low production volume and, therefore, the fabrication methods were highly specialized and not designed for reproducibility. General acceptance of structural ceramics in larger markets did not occur. Component reliability was certainly an issue. The strategy to overcome poor reliability seemed to one focused on new ceramic compositions rather than understanding the manufacturing process. Received September 7, 1994 Ceramics Processing õ Research Laboratory, Materials Processing Center MA 02139, U. S. A., Massachusetts Institute of Technology, Cambridge, 5

2 6 Michael J. CIMA This effort paralleled successful strategies to design metal alloys, where subtle changes in composition and heat treatment could be used to achieve broad ranges in material properties, such as toughness and modulus. No such broad spectrum ceramic material systems were discovered and the focus switched by the 1980's to composites where increased toughness seemed likely to emerge. Ceramic matrix composites were an extension of the materials design concepts of metallurgists where composition and structure could overcome difficulties in manufacturing. The Griffith failure criterion for brittle materials is often cited as justification for this approach. The fracture stress is proportional to a materials toughness and inversely proportional to the square root of the defect size found within the component. The notoriously low toughness of ceramic materials can be used to show that defects in the range of 10 to 100 microns were often responsible for component failure. Many felt it unlikely that ceramic components could be produced without flaws of this size. Thus, the only viable approach was to seek improved toughness. Ceramics processing science developed later and for different reasons than that for physical ceramics. Manufacturing of ceramic components until the late 1970's and early 1980's was largely an art even when practiced at large scale. Besides the traditional ceramic industries, it was only electronics that depended on high volume ceramic components. Electronic ceramics manufacture often requires processing control that is more stringent than structural ceramics. Indeed, the problems faced by electronics industry may be more responsible for the development of ceramics processing science than any other factor. The explosion in consumer electronics and computing was driven by miniaturization and system integration. Electronic ceramic component design required new materials and component design to meet simultaneous demands of smaller size and lower cost. The problems created by this rapid change could not be addressed by empirical process development. Radically new process concepts and understanding were required. Electronic ceramics are required for the manufacture of a number of discrete components, such as capacitors, varistors, and inductors and as insulators in electronic packaging. These "functional ceramics" are distinct from structural ceramics. The composition of the material is largely determined by the desired device characteristics. The mechanical properties of the ceramic are important, but are secondary to the bulk electronic or magnetic properties of the material. Defects are nonetheless, very important to control. The dimensions of relevant defects are governed by factors other than the Griffith flaw criterion. Instead, the basic structure of the device determines the relevant scale. Multilayer ceramic capacitors (MLC) are an example. Billions of these devices are produced world-wide annually. The size of a given capacitor depends on both the dielectric constant of the ceramic and the thickness of the dielectric layers. Decreasing capacitor dimensions required either larger dielectric constant or thinner dielectric layers. The layer thickness was commonly 25 microns by the end of the 1970's and continued to decrease during the 1980's. Dielectric thickness' of 12 micron or less are now mass produced. Defects in MLCs result in regions of decreased dielectric strength and ultimately electrical shorting between the electrodes. Defects of the order of the dielectric thickness can not be tolerated anywhere in the device. Thus, while developers of structural ceramics had not focused on defect-free manufacturing, electronic ceramic manufacturers had to build parts with an eye toward controlling defects only one order of magnitude larger than the raw material powder size. This daunting task can be illustrated by considering the volume of a defect

3 Vol. 3 No, 1 (1995) Ceramics Manufacturing-from Science to Practice 7 compared to the size of the device. A typical MLC may be 2 mm on a side and defects of the order of 5 microns can not be tolerated. The defect volume contains roughly 1000 particles, but the device is 64,000,000 times the volume of a single defect. The same proportions applied in two dimensions to 2" silicon wafer processing yields control of defects less than 5. 6 microns in size. This is easily achieved in modern silicon wafer processing lines. Two additional factors, however, must be considered. First, high yields of MLCs may demand that defects smaller than 5 microns have to be controlled. Second, MLCs have to be produced at three to four orders of magnitude lower cost than a typical processed wafer. Fine powder synthesis and application of colloid science principles were among the first research areas in ceramic processing during the late 1970's. Finer dimensions of electronic structures demanded smaller starting particle sizes. Developers soon found that blind use of submicron powders met with many new difficulties. Green densities were dramatically reduced, slip rheology changed, and binder removal became even more difficult. Many researchers realized that interaction between the particle surfaces dominated the behavior of the ceramic suspensions. The ceramic suspension were true colloidal suspensions governed by Hamaker forces, surface charge, and steric interation of absorbed molecules. Similar concerns had been the domain of clay-based ceramic manufacturing for many years. The availability of new ceramic materials in the colloidal size range, however, created new demands for specialized dispersants and process formulations. Often these new systems were solvent-based rather than the more familiar aqueous systems used for traditional ceramics. Thus, new processing aids had to be developed rapidly and made to work under new constraints. Tape casting was a particularly challenging problem. The suspensions were required to have a rather narrow range of rheolodical characteristics. The cast tapes required very specific mechanical properties so that they were strong enough to not distort on handling, laminate easily without change in density or dimension, and fire without deposition of carbon during removal of the processing aids. Very often a change in formulation to address a problem in one area causes new problems to arise. The most appropriate powder size distribution was and still remains a very controversial topic in the advanced ceramic community. There are two sides to this issue. Many powder metallurgy studies show that the packing density of a green body increases with the breadth of the powder size distribution. Computer simulation of random packing of hard spheres indicates that this phenomena can be explained by imagining the small particles as occupying the interstitial spaces between larger particles. The resulting higher packing density reduces the total shrinkage during sintering. Low shrinkage is thought to make it easier to hold tolerance to a desired dimension in the final component. Experimental studies of this phenomena use rather large particles. The particles are typically 10 microns or greater. Most powders of interest to advanced ceramics are, however, submicron and experiments to determine the effect of powder size distribution on packing density invariably show no effect on packing densities that are less than that obtained by a narrow size distribution. The high surface area of submicron powders increases the cohesive strength of particle agglomerates. Fine powders are, therefore, not free flowing and do not exhibit ideal packing. Proponents of narrow size distribution powders contend that high packing density is secondary to uniformity and reproducibility of green density. Many ceramic processes apply shear or deform the particulate mass during forming. Other processes involve casting or drying slurries against a surface. All of these processes can segregate particles based on size and, therefore, create a

4 8 Michael J. CIMA nonuniform distribution in density. An extreme case is observed during injection molding of particulate suspension in plastic binders. Shear rates can be very high in narrow sections of the die. The dilatant behavior of the suspension may cause the particulate suspension to bridge across the gap in the mold surfaces. The plastic with fine powders may continue to pass which effectively creates segregation. Tape casting is another process which may segregate particles according to size. Conditions may exist where large particles settle during drying of the cast slurry or small particles are drawn to the surface by surface tension driven flows. New directions Insertion of advanced materials in commercial and DOD applications is complicated by the typically long lead-times and high expense required for the production of prototype components. Tooling for constructing these components is often expensive and difficult to make for all but the simplest shapes. Secondly, the properties of many advanced materials are sensitive to the forming process used to manufacture the required shape. Strength limiting flaws within structural ceramics, for example, are often found to be introduced by the forming process. The situation is made even more difficult because defects introduced by powder forming technologies are very dependent on the shape of the component. Thus, significant time and expense are required to optimize the forming process for each new component. Designers must have confidence in their designs and component performance before introduction of products which incorporate advanced materials. Many materials developers view this stage of the product development cycle as having the largest activation energy. The field of solid free form fabrication (SFF) methods provides an interesting alternative to the common methods for making complicated shapes and specifically addresses the problem of Figure 1 Alumina components prepared by 3 DP. The part on the left is as it emerges from the 3 DP process. The part on the right is after subsequent firing to sinter to a density greater than 99% or theoretical.

5 Vol. 3 No, 1 (1995) Ceramics Manufacturing-from Science to Practice 9 Figure 2 Schematic of the first 3 DP-derived ZTA part is shown on left. 5vol. %ZrO2 slurry was printed on Al2O3 powder. A cross-section of the CIPed and fired sample is also shown. Each bright feature is a cross-section of a printed line. product development. SFF methods create parts by laminating thin sections of the component. Each layer is defined by determining the intersection of a plane with a computer model of the desired shape. Several of these types of processes are used, such as stereo lithography, selective laser sintering, and 3 D Printing. The 3 D Printing process was discussed at the first ICMR conference and is described in several publications [1-3] Briefly, 3 DP creates parts by a layered printing process. The infomation for each layer is obtained by applying a slicing algorithm to the computer model of the part. An individual two-dimensional layer is created by adding a layer of powder to the top of a piston and cylinder containing a powder bed and the part being fabricated. The new powder layer is selectively joined where the part is to be formed by "ink-jet" printing of a binder material. The piston, powder bed and part are lowered and a new layer of powder is spread out and selectively joined. The layering process is repeated until the part is completely printed. Removal of the unbound powder reveals the fabricated part. The process has been primarily applied to ceramic molds for metal casting as discussed in the first ICMR conference. More recently, however, it has been used to make metal parts directly [4], structural ceramic parts[5], and polymer parts[6]. Shown in Figure 1 is a structural ceramic component prepared by 3 DP. The ability to control local composition is a unique feature of 3 DP. This type of control permits fabrication of materials with computer derived microstructures or spatially controlled compositions (SCC). These are components in which their microstructure is designed on a computer and built via the 3 DP process. 3 DP can selectively deposit matter within the structure of a component so that composition can vary from point to point. The macroscopic shape of the component can, however, be specified completely independently. Thus, we envision components where both the macrostructure and microstructure are designed by computer and constructed by 3 DP. Potential applications of such a technology are numerous, such as components with anisotropic mechanical properties, microengineered porosity, or constructing composite multilayer modules for electronic packaging. Conventional powder forming technologies cannot provide simultaneous microstructure and macrostructure control. Thus, demonstration of this approach will be a quantum leap beyond current material fabrication and will create a technology that is not unlike the control that

6 10 Michael J. CIMA photolithography provided the electronics industry. An example of this approach is shown in Figure 2, where a zirconia toughened alumina component has been prepared by 3 DP. Note that the placement of zirconia precipitates within the alumina matrix is precisely controlled by the 2 DP machine. SFF methods for making components with complex microstructures will only be worthwhile if the production rate is high enough to actually manufacture the component using the SFF method. Components with unique microstructures will only be useful if thousands of parts can be made. 3 DP may be uniquely qualified to address this manufacturing issue. Multi jet printing is now possible on 3 DP machines as has been demonstrated at Soligen[7] and at MIT. This has dramatically increased the production rate of the 3 DP process. Commercial ink-jet print heads are available with thousands of individually controlled jets. Thus, future 3 P machines may be closer to production tools rather than tools for prototyping. References 1. E. M. Sachs, M.J. Cima, P. Williams, D. Brancazio, J. Cornie, "Three-Dimensional Printing Rapid Tooling and Prototypes Directly from a CAD Model," J. Eng. Ind., 114 pp (1992). 2. E. Sachs, M.J. Cima, J. Bredt, A. Curodeau, "CAD-Casting: The Direct Fabrication of Ceramic Shells and Cores by Three Dimensional Printing," Man. Rev. 5 [2] pp , (1992). 3. M. J. Cima and E. M. Sachs, "Three Dimensional Printing: Form, Materials, and Performance," in Proceedings of the Solid Freeform Fabrication Symposium (8/12/91-8/14/91, Austin, TX). Edited by J. J. Beaman, H. L. Marcus, D. L. Bourell, J. W. Barlow, and T. Crawford. University of Texas, Austin, TXpp S. Michaels, E. M. Sachs, M. J. Cima, "Metal Parts Generation by Three Dimensional Printing," in proceedings of the Solid Freeform Fabrication Symposium (8/3/92-8/5/92, Austin, TX). Edited by J. J. Beaman, H. L. Marcus, D. L. Burrell and J. W. Barlow. University of Texas, Austin, TX, pp J.Yoo, M. J. Cima, S. Khanuja, and E. Sachs, "Structural Ceramic Components by 3 D Printing," SFF Symposium Proceedings, Univ. of Texas, p , S. W. Borland, B. M. Wu, L. G. Cima, R. A. Giordano, E. M. Sachs, and M.J. Cima, (submitted for publication in Biomaterials) 7. Soligen Inc., Northridge, CA