RAPID PROTOTYPING POLYMER PARTS VIA SHAPE DEPOSITION MANUFACTURING

A DISSERTATION SUBMITTED TO THE DEPARTMENT OF MECHANICAL ENGINEERING AND THE COMMITTEE ON GRADUATE STUDIES OF STANFORD UNIVERSITY IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

John Kietzman February 1999

Copyright by John Kietzman 1999 All Rights Reserved

ii

I certify that I have read this dissertation and that in my opinion it is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. ___________________________________ Fritz Prinz (Principal Adviser)

I certify that I have read this dissertation and that in my opinion it is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. ___________________________________ Mark Cutkosky

I certify that I have read this dissertation and that in my opinion it is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. ___________________________________ Kosuke Ishii

Approved for the University Committee on Graduate Studies: ___________________________________

iii

Abstract Polymer Shape Deposition Manufacturing is a new Solid Freeform Fabrication process. Shape Deposition Manufacturing (SDM) is a layered manufacturing process involving an iterative combination of material addition and material removal using a part material and a sacrificial support material. Polymer SDM uses such techniques to build polymer parts. In this research, new material combinations were identified, material strength was characterized, and several applications were identified. The constraints on Polymer SDM part and sacrificial support materials were identified, and a variety of compatible material combinations were developed. Ultraviolet-curing part and support materials and fast-curing polyurethane part materials were developed to reduce cycle times. Solvent-removable support materials for use with castable thermoset part materials were developed. Epoxy and polyurethane materials with improved interlayer bonding were identified. Polymer SDM part strength is determined both by the bulk material properties of the part materials and by their interlayer bonding. Tensile strengths of three thermoset part materials were measured. Both monolithic specimens and specimens with two interlayer interfaces were tested. Interlayer strengths ranged from 5-40 MPa. Relative interface strengths varied even between similar materials; such differences may be explained by surface wetting and polymer chain interpenetration. There are many applications for the Polymer SDM process. Mechanisms with moving parts and small clearances can be built already assembled, and objects can be embedded within parts. The Polymer SDM process was compared to commercial rapid prototyping technologies. Polymer SDM can build parts with superior surface finish and comparable strength but has more flexibility for construction of mechanisms and parts with embedded objects.

iv

Polymer SDM can also be used to construct fugitive wax molds for casting other materials. This technique, called "Mold SDM," was developed by other researchers at Stanford using materials identified during the present research. This indirect process allows the use of a wider variety of part materials, and produces monolithic finished parts which are stronger since they lack interlayer interfaces. Comparison parts were made using both the Polymer SDM and the Mold SDM methods.

Acknowledgments I would like to thank my advisor, Prof. Fritz Prinz for his support and advice during this project. I would like to thank Professors Mark Cutkosky and Kos Ishii for serving on my reading committee. Their comments and advice were insightful. I would like to thank Professor Dan DeBra for his guidance and assistance during my time at Stanford. I would like to thank Professors Dave Beach and Ed Carryer for their dedication to their courses, which formed the backbone of my Design Division experience at Stanford. I would like to thank the students of the Rapid Prototyping Laboratory, past and present, for their contributions to this work and the collegial environment necessary for research. Alexander Cooper, Dr. Robert Merz, Eric Chamness, Sangkyun Kang, Dr. Krishnan Ramaswami, Byongho Park and Tom Hasler were particularly involved with this project. I would like to thank our collaborators at Carnegie Mellon University, especially Dr. Lee Weiss, Larry Schultz, and Dr. Gennady Neplotnik, for their assistance and advice. Dr. Mike Uchic, Professor W. D. Nix, Gayle Link, and the General Motors North American Operations Materials Testing Laboratory all assisted with the tensile testing work described in this dissertation. Sylvia Walters and Lynn Hoschek provided invaluable administrative support. My studies were funded by a DoD National Defense Science and Engineering Graduate Fellowship and a DoE Integrated Manufacturing Pre-Doctoral Fellowship. The projects were funded by the Defense Advanced Research Projects Agency, under contracts DABT63-95-C-0026 and N00014-96-1-0625. I would like to thank all of these funding agencies and their administrators and contract monitors. I would also like to thank Judith Haccou, the Mechanical Engineering Student Services Manager, for her assistance with the administration of fellowship research support funds. I would like to thank all of the friends I've met at Stanford University for making this an interesting time. I would especially like to thank my close friends for their support and entertainment. Finally, I'd like to thank my family for their continuing support.

vi

Contents Abstract ............................................................................................ iv Acknowledgments ................................................................................ vi List of Tables ...................................................................................... xi List of Figures .................................................................................... xii INTRODUCTION ................................................................................ 1 1.1. Overview............................................................................... 1 1.2. Thesis Outline ................................................................ ......... 1 1.3. Individual / Group Research Statement ............................................ 2 BACKGROUND ................................................................................. 3 2.1. Rapid Prototyping ................................................................ . . . . 3 2.1.1. Literature on Rapid Prototyping Processes ........................................ 4 2.1.2. Commercial Polymer Rapid Prototyping Systems ................................ 5 2.1.3. Other Polymer Rapid Prototyping Systems........................................ 10 2.1.4. Commercial Rapid Prototyping Materials .......................................... 11 2.1.5. Anisotropic Mechanical Properties in Commercial Rapid Prototyping ......... 14 2.1.6. Secondary Processes for the Production of Polymer Parts ...................... 15 2.2. Shape Deposition Manufacturing ................................................... 16 2.2.1. Process Description................................................................. . . 16 2.2.2. Part Decomposition and Planning................................................... 19 2.2.3. Hardware .............................................................................. 20 2.2.4. Origins ................................................................................. 22 2.2.5. Comparison to Other Rapid Prototyping Processes .............................. 22 2.2.6. Heterogeneous Structures............................................................ 24 2.3. Polymer Shape Deposition Manufacturing at Carnegie Mellon University..... 26 2.3.1. Embedded Electronics ................................................................ 26 2.3.2. VuMan Wearable Computer ......................................................... 28 2.3.3. Waterproof Computers and Thermal Issues ....................................... 29

vii

2.3.4. 2.3.5.

Material Research ................................................................ . . . . . 29 Part and Support Material Deposition Methods ................................. . . 32

POLYMER SHAPE DEPOSITION MANUFACTURING ................................. 33 3.1. Introduction............................................................................ 33 3.2. The Polymer SDM Process .......................................................... 33 3.3. Process Example ................................................................ ..... . 34 3.4. Process Requirements ................................................................ 38 3.5. Automating Process Planning ....................................................... 39 3.6. Automating Part Building ............................................................ 39 MATERIALS FOR POLYMER SDM .......................................................... 42 4.1. Introduction............................................................................ 42 4.2. Material Requirements for Polymer SDM.......................................... 42 4.3. Material Combinations ............................................................... 44 4.3.1. Support Materials for Thermoplastic Part Materials .............................. 45 4.3.2. UV-Curable Part and Support Materials for Fast Cycle Times .................. 45 4.3.3. Completely Removable Support Materials for Castable Part Materials ......... 47 4.3.4. Castable Part Materials for Short Cycle Times ................................. . . . 48 4.3.5. Part Materials for Higher Interlayer Strength ..................................... . 49 4.3.6. Other Materials ................................................................ ........ 50 4.4. Discussion ............................................................................. 52 4.4.1. Categories of Polymer SDM Material Combinations ............................ 52 4.4.2. Material Shrinkage and Part Distortion ............................................. 54 4.5. Conclusions ........................................................................... 57 MATERIAL STRENGTH IN POLYMER SDM.............................................. 59 5.1. Polymer SDM Materials.............................................................. 59 5.2. Metrics for Polymer SDM Material Strength ..................................... . 60 5.3. Tensile Strength of Polymer SDM Materials ..................................... . 62 5.3.1. Test Results............................................................................ 66 5.3.2. Discussion ............................................................................. 69 5.3.2.1. Theories of Adhesion................................................................. 71 5.3.2.2. Self-Adhesion of Crosslinked Polymers in SDM ................................. 73 5.3.2.3. Self-Adhesion Differences between Polyurethanes ............................... 77 5.3.2.4. Self-Adhesion of Polyurethane and Epoxy Materials............................. 81

viii

5.4. 5.4.1. 5.5. 5.5.1. 5.5.2. 5.6.

Effects of Gritblasting on Interlayer Strength ................................. . . . . 81 Test Results............................................................................ 82 Effects of the Support Material Removal Process on Interlayer Strength ..... . 84 Test Results............................................................................ 85 Discussion ............................................................................. 86 Conclusions ........................................................................... 87

APPLICATIONS OF POLYMER SDM ....................................................... 90 6.1. Introduction............................................................................ 90 6.2. Rapid Prototyping ................................................................ . . . . 90 6.3. Multimaterial Structures .............................................................. 91 6.4. Mechanisms ........................................................................... 92 6.5. Embedded Objects ................................................................ . . . . 93 6.6. Mold SDM ............................................................................. 96 6.6.1. The Mold SDM Process.............................................................. 96 6.6.2. Process Automation ................................................................. . 99 6.6.3. Process Comparison................................................................. . 101 6.7. Comparison between Polymer SDM and Commercial RP Systems ............ 105 6.7.1. Planning Automation ................................................................. 105 6.7.2. Part Build Rates ................................................................ ....... 106 6.7.3. Part Properties................................................................ ......... 107 6.8. Preferred Applications of Polymer Shape Deposition Manufacturing .......... 110 6.8.1. Parts Made Faster through Polymer SDM ......................................... 110 6.8.2. Parts with Better Surface Finish through Polymer SDM......................... 111 6.8.3. Parts Manufacturable only through Polymer SDM ............................... 112 6.9. Conclusions ........................................................................... 114 CONCLUSIONS ................................................................................. 116 7.1. Contributions .......................................................................... 116 7.2. Future Work ........................................................................... 118 7.2.1. Polymer SDM Materials Improvements ............................................ 118 7.2.2. Polymer SDM Material Selection Process Improvements........................ 121 7.2.3. Polymer SDM Process Improvements ............................................. 121 7.2.4. Mold SDM Materials Improvements................................................ 123

ix

Appendices MATERIAL COMBINATIONS EXAMINED ................................................ 124 A.1. Support Materials for Thermoplastic Part Materials .............................. 124 A.2. UV-Curable Part and Support Materials for Fast Cycle Times .................. 129 A.3. Completely Removable Support Materials for Castable Part Materials ......... 134 A.4. Castable Part Materials for Short Cycle Times ................................. . . . 140 A.5. Part Materials for Higher Interlayer Strength ..................................... . 143 TENSILE TESTING NOTES ................................................................. . . 144 B.1. Insufficient Cure Time before Tensile Testing ................................. . . . 144 B.2. Material Inconsistencies during Tensile Tests ................................. . . . . 145 B.3. Timing of Tensile Testing Activities ................................................ 147 B.4. Experimental Uncertainty ............................................................ 147

NOTES ............................................................................................. 151

List of Tables Table 2.1. Table 4.1. Table 4.2. Table 4.3. Table 5.1. Table 5.2. Table 5.3. Table 5.4. Table 5.5. Table 5.6. Table 5.7. Table 5.8. Table 5.9. Table A.1. Table A.2. Table B.1. Table B.2. Table B.3. Table B.4. Table B.5. Mechanical Properties of Rapid Prototyping Materials........................ 13 Categories of Polymer SDM Material Combinations.......................... 53 Linear Shrinkage of Material Combinations................................. . . . 57 Useful Materials Combinations for Polymer SDM ............................ 58 Manufacturers' Material Specifications ......................................... 60 Strength Data - Monolithic Tensile Specimens................................. 67 Strength Data - Dual-Interface Tensile Specimens............................. 67 Average Tensile Strengths........................................................ 67 Relative Interface Strengths ...................................................... 70 Tensile Strength of Gritblasted Specimens ................................. . . . . 83 Tensile Strengths - Normal and Gritblasted Specimens....................... 83 Tensile Strength of Specimens Soaked in BIOACT 280...................... 85 Tensile Strengths - Normal and BIOACT-Soaked Specimens ............... 86 Commercially Available Solder Masks.......................................... 126 Acrylate Stereolithography Resin Part Materials............................... 131 Tensile Strength of Incompletely Cured LUC-4180........................... 144 Container Open-Times Before Casting Specimens ............................ 146 Interoperation Delays during Tensile Testing .................................. 147 Statistical Summary of Experimental Results .................................. 148 Estimated Bias Errors ............................................................. 149

xi

List of Figures Figure Figure Figure Figure Figure 2.1. 2.2. 2.3. 2.4. 2.5. A Rapid Prototyping System ................................................... 4 Shape Deposition Manufacturing ............................................... 17 Building a Single Layer of an SDM Part ..................................... . 17 Construction of Simple Features in SDM ................................. . . . . 18 Decomposition into Manufacturable Compacts ............................... 20 Polymer SDM Process Sequence, Part I ..................................... . 35 Polymer SDM Process Sequence, Part II ................................. . . . . 36 Polymer SDM Process Sequence, Part III ................................. . . . 37 Stanford Rapid Prototyping Laboratory ....................................... 40 CMU Polymer SDM Station .................................................... 41 Acrylate Stereolithography Resin Part ......................................... 46 Stainless Steel Part ............................................................... 47 LUC-4180 Polyurethane Assemblies .......................................... 48 TDT 205-3 Polyurethane Flap Mechanisms................................. . . 49 Green Ceramic Part .............................................................. 51 Wax Mold ................................................................ ......... 52 Formation of "Christmas-Tree" Distortion ................................. . . . 55 Distorted Cross-Section of LUC-4180 Polyurethane Part................... 57 ASTM Type I Tensile Specimen Geometry ................................. . . 63 Ideal Planar and Vertical Tensile Specimens ................................. . 63 Monolithic and Simplified Dual-Interface Tensile Specimens............... 64 Dual-Interface Tensile Specimen Geometry ................................. . . 64 Typical Tensile Tests of Monolithic Specimens .............................. 68 Typical Tensile Tests of Dual-Interface Specimens .......................... 68 Tensile Strength Summary - Comparing Materials ........................... 69

Figure 3.1. Figure 3.2. Figure 3.3. Figure 3.4. Figure 3.5. Figure 4.1. Figure 4.2. Figure 4.3. Figure 4.4. Figure 4.5. Figure 4.6. Figure 4.7. Figure 4.8. Figure 5.1. Figure 5.2. Figure 5.3. Figure 5.4. Figure 5.5. Figure 5.6. Figure 5.7.

xii

Figure 5.8. Figure 5.9. Figure 6.1. Figure 6.2. Figure 6.3. Figure 6.4. Figure 6.5. Figure 6.6. Figure 6.7. Figure 6.8. Figure 6.9. Figure 6.10. Figure 6.11. Figure 6.12. Figure 6.13. Figure 6.14.

LUC-4180 Tensile Strength - Normal and Gritblasted Specimens ......... 83 Tensile Strength - Normal and BIOACT-Soaked Specimens ............... 86 Polyurethane Shell with Polyurethane Foam Filler........................... 92 Assembled System of Polyurethane Mechanisms ............................ 93 Polyurethane Dial Input Device ................................................. 94 Part with Embedded Sensor..................................................... 95 Polyurethane Wearable Computer for Divers ................................. 95 The Mold SDM Process ......................................................... 97 Mold SDM Polyurethane Part................................................... 98 Mold SDM Polyurethane "Inchworm" Toy ................................. . . 99 Automated Mold SDM Machine ................................................ 100 Automated Mold SDM Machine Schematic ................................. . . 100 Automated Mold SDM Machine Controls Schematic ........................ 101 Direct SDM Assemblies in LUC-4180 Polyurethane ........................ 104 Top View of Mold SDM Assembly ............................................ 104 Bottom View of Mold SDM Assembly ........................................ 105

Figure A.1. Polyurethane Surface Cured in Contact with Solder Mask .................. 135 Figure A.2. TDT 205-3 Polyurethane Assembly............................................ 142

xiii

CHAPTER ONE INTRODUCTION 1.1. Overview This dissertation describes improvements to the Polymer Shape Deposition Manufacturing (SDM) process. This is a method for rapid prototyping polymer parts which was originally developed at Carnegie Mellon University. The present research resulted in a variety of improved materials for use in the process. Material improvements allow faster production of stronger parts with smaller clearances, and reduce postprocessing labor. New materials have been used to construct parts with intricate mechanisms, built already assembled. These materials have also been used to construct parts with embedded objects, including sensors. Such materials allow the manufacture of "smart structures" which can monitor and react to environmental changes or structural damage. Materials developed during this research have also enabled the development of the "Mold SDM" process, an indirect molding process in which fugitive wax molds are built through Polymer SDM and then used to cast other parts. This process allows the fabrication of monolithic molded parts which could not be removed from conventional molds. It allows the use of final part materials which cannot be processed in "Direct" Polymer Shape Deposition Manufacturing.

1.2. Thesis Outline Following the present introduction, Chapter Two provides background material on rapid prototyping, the Shape Deposition Manufacturing process, and initial Polymer SDM work at Carnegie Mellon University. Chapter Three presents an overview of the concept of Polymer SDM, including a process sequence example. Chapter Four describes the polymer material selection process, including the wide variety of part and support materials considered and the constraints on material choices which made the selection process challenging. Chapter Five describes an investigation into process-specific material

strengths of different part materials. Chapter Six focuses on applications of Polymer SDM, including the Mold SDM variant process. A comparison of those two processes is presented, as well as a comparison between Polymer SDM and commercial rapid prototyping technologies for various applications. Finally, Chapter Seven provides a brief conclusion. Additional information about particular part and support materials is provided in Appendix A, and some notes on the tensile testing process are provided in Appendix B.

1.3. Individual / Group Research Statement Part of the present work represents group research. The Stanford University Rapid Prototyping Laboratory is an environment where teamwork and group brainstorming are encouraged, for the purpose of accelerating the rate of research progress. All of the present work was performed in this environment. The part and support material investigation is almost completely my own work, founded on the referenced background work done at Carnegie Mellon University. Eric Chamness did perform several experiments during this investigation. Some of the part and support material constraints were identified by my coauthors on a paper1. The investigation of process-specific material strengths of different part materials is entirely my own work. Finally, I identified the "solder mask" support materials and built parts with them, but Alexander Cooper et al. developed the Mold SDM process based on these materials. Their work is presented here because of its significance as an application of Polymer SDM. The design and construction of the automated Mold SDM machine was a collaborative effort; I assisted with project management and the design and construction of power and control electronics. Any errors in the present work are my responsibility.

CHAPTER TWO BACKGROUND 2.1. Rapid Prototyping In mechanical engineering, rapid prototyping is the process of building prototype objects to see whether a proposed design will be successful. Verifying "successful" designs has several aspects, including correct shapes, correct sizes, adequate strength, and others. Different types of prototypes are required to answer these different questions. Questions of shape can be settled with fragile models; as long as the models can survive normal handling, strong materials are not required. Questions of strength require either models of adequate strength or models with the exact strengths of production parts. Rough shapes can be examined with relatively inaccurate models while precision fits around other parts may require quite accurate and precise models. The field of rapid prototyping has seen the recent introduction of automated systems which can convert a computer solid model into a three-dimensional artifact. These technologies have also been called "layered manufacturing" or "solid freeform fabrication." Layered manufacturing implies that these artifacts are built in two-dimensional layers; this greatly simplifies these processes and enables their automation. Solid freeform fabrication implies that complex shapes are as easy to build as simple shapes, since the planning and manufacturing processes are automated. A variety of processes have been developed to implement these concepts; a prototypical system is illustrated in Figure 2.1. Some of these systems have been commercialized while some only exist as research projects. This thesis will describe research into a new rapid prototyping process called Shape Deposition Manufacturing.

CAD System Automated Process Planner

Automated Fabrication Machine

data exchange format

motion control trajectories

Computerized Solid Model

Physical Object Figure 2.1. A Rapid Prototyping System 2

2.1.1. Literature on Rapid Prototyping Processes There are several good places to begin surveying the literature of rapid prototyping. Review articles include one by Conley and Marcus3 and earlier ones by Kruth4 and Au and Wright5. Burns6 provides an extended overview of the field, describing current processes and applications as well as some related topics. An earlier overview is provided by Kochan7; this work revolves around descriptions of available machines and the processes they use. From a research perspective, some of the most interesting works are those which describe the development of particular processes - particularly noteworthy are the books by Jacobs 8 ,9 . While he does describe other processes, he focuses on the evolution of Stereolithography, with which he was closely involved. These two books describe the incremental process improvements which have improved accuracy and created new applications. The latest monograph on rapid prototyping, focusing more on the Selective Laser Sintering process, was written by Beaman, et al 10. This book also contains a good description of the earliest history of rapid prototyping and layered manufacturing. Its second chapter has the most complete comparison of development and release dates for commercial and research systems. Burns and Kochan have the best descriptions of machines and processes available only in Japan although the JTEC/WTEC report has a more complete list of systems and better materials data11.

While these books represent a starting point, much of the important details of rapid prototyping are available only in papers, patents, or proprietary corporate documents. The most important academic conferences include the Solid Freeform Fabrication Symposium hosted by the University of Texas at Austin, the Inter/National Conference on Rapid Prototyping sponsored by the University of Dayton, and the European Conference on Rapid Prototyping and Manufacturing. These have each occurred for several years but may not be completely indexed for computer searching (the SFF Symposium particularly so). Another conference focused on commercial and industrial users is Rapid Prototyping and Manufacturing, sponsored by the Society of Manufacturing Engineers. The Rapid Prototyping Journal began recently, but rapid prototyping papers have appeared in a wide variety of journals, and will continue to do so. For details about commercial machine, material, and market developments, the Rapid Prototyping Report12 and Wohlers' annual State of the Industry Report13 are good sources. Finally, the patent literature can be a useful resource. Computer databases such as the IBM collection14 allow searches which would be quite tedious using paper sources. For example, the full texts are indexed, and both earlier patents cited in a patent and later patents which cite a given patent can be located.

2.1.2. Commercial Polymer Rapid Prototyping Systems During the last decade (1988-1997), 3289 rapid prototyping machines have been sold commercially throughout the world15. Annual sales have increased to 1057 worldwide in 1997 alone. By number of machines sold in 1997, the leading companies are 3D Systems (278), Stratasys (260), and Sanders Prototype (152). Almost all of these rapid prototyping machines make parts from polymer materials, although the processes involved vary widely. The bulk of the world's machines have been sold into the United States market; the following process overview describes the machines and processes available in this country.

2.1.2.1. Stereolithography This is a process which polymerizes an ultraviolet-curable polymer resin using a laser; it was commercialized by 3D Systems16 ,17 . This process begins with a vat of liquid thermoset polymer with an elevator platform slightly below the surface of the liquid. An ultraviolet laser, aimed by scanning mirrors, draws a layer of the desired part onto the

surface of the vat of resin. The material solidifies where it was illuminated, and adheres to the top of the platform. After the first layer is drawn, the platform moves downward by the height of one layer. The next layer is then drawn onto the surface of the liquid. This material solidifies in contact with the previous layer, and the entire part is built in this manner. After the whole part has been drawn, it is removed from the vat of resin and detached from the platform. Any uncured resin is washed off the part using a solvent, and the part is then completely cured or "postcured" using a chamber with ultraviolet flood lights. The initial resins were acrylic materials, but the pursuit of improved accuracy and strength convinced most vendors to switch to newer epoxy materials. Ciba-Geigy, Du Pont, and Allied-Signal have developed resins for use in this process. Since pieces of a part can only be located by being built in contact with previously cured resin or the support platform, special procedures are required to build layers with islands or overhanging features. In these cases, special support structures are built upward from the platform to support these objects. Support structures are designed to be only as large or as strong as necessary for support since they will later be removed from the parts. After the parts are postcured, these structures are cut off or broken off from the part, leaving only the desired geometry of the part itself.

2.1.2.2. Selective Laser Sintering Developed at the University of Texas at Austin and commercialized by the DTM Corp., this process uses a laser to fuse thermoplastic polymer powders together into a part18. To begin constructing a part, a thin layer of part material powder is spread over a platform and then leveled by a roller. A laser then draws the first layer of the part, fusing powder particles together. The platform indexes downward, after which another layer of powder is spread out and leveled. This sequence is cyclically repeated to build an entire part. The part is built within an environmental chamber which maintains a temperature just below the glass-transition temperature or the melting point of the part material. This allows the laser to scan rapidly across the part since it only needs to provide enough energy to locally raise the temperature of the part material a few degrees to fuse it. The environmental chamber is also filled with nitrogen to avoid chemical reactions in the heated polymer material. The unconsolidated powder in each layer serves as a support material for the part. After the part is complete, this powder must be blown out of internal cavities or brushed away from the surfaces of the part.

DTM has released a variety of materials for this process. Some are designed to make parts with optimal accuracy, others are optimized for part strength, and others are intended for use as investment casting patterns. Recently, a flexible part material from Du Pont was introduced by DTM19.

2.1.2.3. Fused Deposition Modeling This process, developed by Stratasys, Inc., extrudes a bead of thermoplastic polymer with which parts are drawn20. The part material begins as a thin filament wound on a spool. Drive rollers push this filament into a heated chamber where it is liquefied and then extruded through a nozzle. This nozzle is moved around over a platform to draw a layer of a part. Individual layers are normally made by drawing an outline of a cross-section and then filling it in with parallel lines of polymer. The extruded lines of polymer are called "roads" in this process, and the "road width" must be carefully controlled. Undercut features must be supported by a thermoplastic support material deposited in the same manner by another extrusion nozzle. This support material is designed to be mechanically weak so that it can be broken off from the finished part at the end of the process. The part materials have been different ABS polymers (available in colors), an investment casting pattern wax, and a new elastomeric material 21. A soluble support material was recently announced. Additional research has focused on building ceramic green parts by extruding ceramic particles in a polymer binder; this has been called Fused Deposition of Ceramics22.

2.1.2.4. Laminated Object Manufacturing This process laminates adhesive-backed part materials together to form prototypes; it was commercialized in the United States by Helisys23. The part material enters the machine as a roll of adhesive-coated sheet. This is pulled through the machine until new material is present over the build area. A heated platen rolls over this area, laminating the new layer over any previous layers. A laser then outlines the boundary of the desired cross-section of the part. Any new material outside the desired boundary of the part is crosshatched with the laser, so that it can be later removed in small pieces. The process continues as additional layers of part are added until the part is complete. The crosshatched part material serves as a support for undercut features on higher layers within the part. At the end of the

process, a manual "decubing" step is required to remove all of the small cubes of crosshatched extra material from around and within the final part. This step can be quite tedious, and dental tools are often used to remove small pieces. This process originally used Kraft paper as the part material, but a new polymer part material has been developed for the system, justifying its inclusion here24. There has also been independent research into polymer-matrix composite materials for this process25; one material has been released commercially26.

2.1.2.5. Solid Ground Curing This is a complicated process developed by the Cubital Corporation in which thermoset part material is imaged using an ultraviolet light to form layers of a part27. To make a layer, a thin coating of ultraviolet-curable photopolymer is spread over the bottom of the build chamber. An electrostatically-charged roller is used to apply toner to a mask, which will shield all but the desired areas of photopolymer. An ultraviolet flood lamp is shined through this mask to cure the solid parts of this cross-section of the part. The excess, uncured photopolymer is removed from the part. Next, hot wax is used to fill in areas of the layer where cured part material is not present. A cooled platen is used to solidify this wax, which acts as a support material for undercut features. Finally, a milling cutter is used to plane both the wax and the photopolymer down to the desired height of the top of the layer. This is a single cycle, which is repeated many times to build an entire part. At the end of the process, the part is removed from the chamber and the support material is heated to assist manual removal. The company has announced its intention to develop a reverse-cycle process, in which photopolymers would be used as support materials while other polymer, metallic, or ceramic materials are used as part materials28. In contrast to the previous techniques, this process is designed to build an entire vat of parts at once. The individual steps of the process take the same amount of time regardless of whether a single small part, a group of small parts, or a big part is being built. This reduces the average cost and time required to build parts as long as the demand for the machine is high enough to fill its chamber during each build. This machine is the most complicated rapid prototyping machine commercially available, and it has developed a reputation as requiring a lot of maintenance29.

2.1.2.6. Z Corporation Z Corporation is a licensee of the Three-Dimensional Printing30 process developed by researchers at the Massachusetts Institute of Technology. This process begins with a smooth bed of powder. Each layer of the part is made by depositing a liquid binder material over certain regions of the powder bed in order to make a solid part. Additional layers of powder are added to the part and then selectively bound together with additional binder. ZCorp's license limits its use of the process to building appearance models. The materials used are a powder which combines starch and cellulose and a water-based binder. The parts are relatively porous immediately upon completion by the machine. They can either be heated and infiltrated with wax, or they can be infiltrated with a cyanoacrylate material. The wax infiltration method is designed to be usable within an office environment while the cyanoacrylate infiltration produces stronger parts. The primary selling point of the ZCorp machine is its rapid production speed31,32.

2.1.2.7. Sanders Prototype Incorporated The Sanders ModelMaker Pro and its successors are examples of inkjet rapid prototyping machines33. These machines deposit thermoplastic part and support materials through a system of inkjet nozzles. After a layer is covered by these two materials, the Sanders machines use a milling cutter to trim the top of the layer so that its height matches those of other layers. These machines use investment casting waxes as part material and other waxes for support material. The Sanders support material is dissolved in a heated solvent at the end of the process; the solvent and the heating temperature are chosen to avoid damage to the part material wax. The Sanders machines are optimized for producing investment casting patterns - they use very thin layers to create precise geometries on relatively small scales. They are also relatively slow.

2.1.2.8. Actua and Genisys The Actua34 and Genisys 35 machines are inkjet modelers sold by 3D Systems and Stratasys, respectively. They both target the appearance model market, allowing the rapid production of models with inferior physical properties. These models are intended to improve communication in design processes. The machines are relatively similar to one

another and are somewhat similar to the Sanders machines. 3D Systems calls their technology "Multi-Jet Modeling," after the linear array of 96 ink-jet nozzles which deposit material. Stratasys' Genisys machine uses a single inkjet nozzle to deposit thermoplastic part material. Both machines make supports out of part material and perforate the interfaces between parts and their supports. In contrast to the Sanders machines, neither machine uses a milling cutter - this reduces the systems' complexity but may also reduce accuracy in the vertical direction.

2.1.2.9. Ballistic Particle Manufacturing Ballistic Particle Manufacturing is another inkjet rapid prototyping process, commercialized by BPM Technology, Inc.36,37. This process used an inkjet nozzle on a five-axis motion system to spray thermoplastic part material at an object under construction. The five-axis nature of the process was intended to reduce the stairsteps created by other 2-1/2D layered manufacturing processes. The material could be used as investment casting patterns or for the production of soft tooling to make functional plastic parts38. Support structures were made from the same material, designed to break away easily from part surfaces after the part was complete. Machines were sold commercially but BPM Technology has since ceased operation.

2.1.3. Other Polymer Rapid Prototyping Systems The previous list of processes represents all of the processes which have any degree of commercial presence in the United States market. Several different companies have attempted to develop and introduce processes here. Some of these efforts led to technology bought by or licensed and integrated into the described processes. Some companies failed to produce commercially successful systems, or were forced to cease operations for intellectual property reasons. Two examples are Light Sculpting Inc.39 and Quadrax Laser Technologies, Inc.40, both of which were involved with photopolymer materials and systems 41 similar to Stereolithography. There are also a few companies developing processes which have not made any commercial impact. For example, AAROFLEX is developing a competing Stereolithography process based on research by Du Pont42. Other processes are being developed in the academic arena; one such process is Shape Deposition Manufacturing, which is more fully described in the next section of this chapter.

10

Besides the companies which have entered the United States market, there are several companies which are only doing business in other parts of the world. There are a variety of photopolymer-based systems available in Japan, each with slightly different design details. Computer Modeling and Engineering Technology is a joint venture involving Mitsubishi which offers a line of five machines called Solid Object Ultraviolet Plotters 43. Design-Model Engineering Center is a subsidiary of Japan Synthetic Rubber which sells machines made by Sony called the Solid Creation System (SCS or JCS)44. Teijin Seiki sells machines called Soliform which are based on technology licensed from Du Pont45. Other Japanese photopolymer systems include the COLAMM by Mitsui Zosen46, the Solid Laser Plotter by Denken Engineering 47, the Uni-Rapid by Ushio, Inc.47, and a system from the Meiko Corporation48. Japanese researchers have also developed methods for building small, high-resolution structures using variants of the Stereolithography process49,50. In Germany, EOS GmbH51 has produced machines similar to those offered by 3D Systems and DTM in the United States. That company recently sold the part of its business employing photopolymers to 3D Systems. It now focuses exclusively on the EOSINT line of machines, which are similar to SLS machines. The company emphasizes applications involving the direct or indirect production of metallic parts, as well as the production of polymer parts. Also in Germany is Fockele & Schwarze, which has developed a photopolymer-based system, called the Laser Modeling System52. Another European company is Laser 3D of France, which has developed a process called Stereophotolithography53.

2.1.4. Commercial Rapid Prototyping Materials These different rapid prototyping systems use a variety of types of materials. Final part quality in rapid prototyping is dependent both on process and material parameters. System designers have cyclically improved both their build processes and their materials. Improved materials have been an area of competition between different system developers. A great deal of research into photocurable prototyping resins has occurred, conducted by Ciba, Allied-Signal, Du Pont, Zeneca, Coates Brothers, DSM Desotech, Loctite, Asahi Denka, Japan Synthetic Rubber, Teijin Seiki, and other companies. Most of the original resins were acrylic-based, but urethane acrylate, vinylether, and epoxy acrylate materials were developed, and most of the newest resins are epoxies. 3D S y s t e m s ' Stereolithography process was the beneficiary of much of this photopolymer work.

11

Cubital's Solid Ground Curing process especially has been limited by the lack of availability of advanced materials; only very recently has an epoxy material finally been announced. DTM has introduced a wide variety of polymer materials for Selective Laser Sintering. Some are optimized for strength while others are optimized for surface finish or ease of finishing; several of the newer materials are glass-filled resins. The Fused Deposition Modeling process has experienced a complete turnover in materials intended for functional prototypes. The original modeling materials were polyolefin and polyamide54; the current materials are ABS and a modified, high-impact ABS material55. The newest commercial systems are Actua, Genisys, and Z Corporation. These are intended for appearance models rather than functional prototypes, but materials or processes intended to make stronger parts have been released for all of these systems. Besides materials for functional prototypes, most of the commercial systems also have specially designed materials or processes to construct investment casting patterns. Sanders' Model Maker, Stratasys' FDM, and DTM's SLS machines can use investment casting waxes directly. 3D Systems has build styles optimized for investment casting, and the standard paper material for LOM has also been used for patterns. The EOS polystyrene material is commonly used for investment casting in Europe. Most information about the mechanical properties achievable using these commercial systems comes from the system vendors since relatively few studies of mechanical properties have been conducted independently. A summary of current material properties claimed by manufacturers is presented in Table 2.1.

12

Material

Tensile Strength (MPa)

Tensile Modulus (MPa)

Elongation (%)

Notched Izod (J/m)

Flexural Strength (MPa)

Flexural Modulus (MPa) 3100 2800 124 630-720 2910 2300 16 870 1050 4300 2618 1771 138

Ref.

Ciba (for SLA) SL 5510 77 3300 5.4 27 99 SL 5410 72 3100 4.2 37 127 Du Pont (for SLA) Somos 2100 9 80 47 90 Somos 8120 23-27 470-700 18-30 48-59 25-28 Somos 7100 66 2324 2.7-7.8 32 105 Somos 6110 69 2800 10 27 DTM (for SLS) Somos 201 20 111 Nylon 36 yield 1400 32 69 Polycarbonate 23 yield 1200 5 53 "Protoform" 49 yield 2800 6 67 Stratasys (for FDM) ABS 34 2480 >10 107 65 ABSi 37 1963 >10 176 59 E20 6.4 69 >10 347 (NB) 5 Cubital (for SGC) XA7501 31 700 20 Z9501 50 1150 10 34 Stratasys (for Genisys) P1500 19 827 <10 32 28 Asahi Denka (for CMET SLA) HS 673 67 3334 3.1 HS 663 12 481 16 Japan Synthetic Rubber (for DMEC SLA) SCR 200 59 1400 15 29 SCR 310 39 1200 30 147 Teijin Seiki (for Soliform SLA) TSR 752 76 14000 1 97 Comparison Materials: Nylon 6-6 50-65 yield 1600-2000 50-200 112 50 yield Med.-Impact ABS 40-60 2000-2600 20 100-300 70-80 Polystyrene 40-50 3300 <3 11-21 70-90 Polycarbonate >65 2300 >100 700-900 100 yield Note: All materials have been postcured before these properties were measured.

56 57 58 59 60 61 62 63 63 63 64 64 64 65 66,67

827

68

69 69

69 69

14000

69 70 70 70 70

Table 2.1. Mechanical Properties of Rapid Prototyping Materials

13

2.1.5. Anisotropic Mechanical Properties in Commercial Rapid Prototyping While most manufacturers have published test data specifying tensile, flexural and impact properties, the data usually assume that rapid prototyped parts would have isotropic properties. Given the layered nature of all of these properties, this is obviously naive. Material deposited in successive layers tends to have properties which are different in the "build direction" than in-plane. This is an area which is beginning to be examined by the research community. Different authors have examined anisotropy in FDM. Fodran et al. tested tensile specimens made by extrusion either parallel or diagonal to the test direction71. They measured tensile strengths 18% lower when the patterns were not aligned with the testing direction, indicating that interbead bonding may be a limitation. Their testing relied on very few samples and did not produce tensile strengths nearly as high as manufacturer specifications. Other researchers at the Institut fuer Maschinenelemente in Stuttgart72 and the University of Illinois at Chicago73 have also investigated anisotropy in FDM parts. Several authors have published data on anisotropic properties in Selective Laser Sintering. Properties examined include polymer part strengths, green strengths of ceramic parts made with polymer binders74,75, and strengths of copper-infiltrated RapidSteel parts76. Both Thompson77 and Gibson and Shi78 have investigated anisotropy in polymer parts made via SLS. Thompson performed a combined experiment in which laser power, layer thickness, and part orientation were varied simultaneously while building polycarbonate tensile bars. His data shows that samples built out-of-plane have at best 56% the tensile strength of samples built in-plane. This result cannot be assumed to be valid for commercial machines and typical settings, however, since an old machine and unacceptably high power settings were used for these tests. Gibson and Shi built tensile bars in a variety of orientations using DTM's "fine nylon" material on a commercial Sinterstation 2000 machine, and measured out-of-plane strength to be about 40% of in-plane strength. As with Thompson's data, however, the parameters used may have been different from the standard set since even the in-plane strength was below half the manufacturer's specification for this material. Helisys has published data on in-plane versus out-of-plane data for Laminated Object Manufacturing of paper materials but none for polymer materials. Compressive strength was measured both in-plane and transverse; the out-of-plane compressive strength was found to be only 15% of the in-plane value79. The very different bonding mechanisms

14

between in-plane material and successive out-of-plane layers in this process suggest that strength anisotropy might be more pronounced in LOM than in the other commercial processes. No published work analyzes anisotropy in parts made via the Stereolithography, Solid Ground Curing, or Sanders processes, or in Actua, Genisys, or Z Corporation machines. Unpublished work does suggest that the Stereolithography process may produce more isotropic parts, especially with newer epoxy materials and scanning techniques. Both "green" parts and postcured Stereolithography parts are said to have relatively isotropic tensile strengths80.

2.1.6. Secondary Processes for the Production of Polymer Parts In general, rapid prototyping parts with these commercial machines is expensive and time consuming. The technologies which are optimized for producing somewhat strong parts can require tens of hours to build complicated parts. Only recently have the specialized appearance modelers appeared. These machines (Actua, Genisys, and ZCorp) are much faster but produce parts with much poorer physical properties. This means that if a number of functional parts are required, none of these machines are economically viable. These machines are, however, used to produce different kinds of "soft tooling" to allow the production of small runs of functional parts. These commercial rapid prototyping processes can be used to make polymer parts indirectly. This is accomplished by casting a silicone mold around a rapid prototyped pattern, cutting the mold into halves and removing the pattern, and casting polymer into this mold81. This allows the use of a variety of castable materials but introduces a transfer or replication step. That is, rather than making a mold directly, a mold is replicated from a pattern. An advantage of this technique is that several parts can be made from a single rapid prototyped pattern. The resulting parts can be made from stronger materials than most rapid prototyping machines can use directly. The drawback of this process is that tolerances grow larger during each replication step, so that parts made this way will be less accurate than parts rapid prototyped directly. Also, there are some restrictions on molds with deep undercuts, since removing parts from these molds makes them wear rapidly. Handbooks are available which provide advice on optimal mold design and casting techniques for such applications82.

15

Another indirect method for making polymer parts is injection molding using rapid prototyped molds. Polymer parts made from standard injection moldable thermoplastics can be made this way, allowing functional test parts to be made from production materials using production methods. Three methods like this are 3D Systems' ACES Injection Mold, DTM's Selective Laser Sintering RapidTool, and Keltool (now owned by 3D Systems). The ACES Injection Mold83 (AIM) process uses epoxy molds made in Stereolithography machines as mold inserts in conventional injection molding machines. This process produces short-lived tooling with much faster lead-times compared to conventional metallic injection molds. DTM has developed the "RapidTool" process for making metal parts. This process uses an iron powder coated with a polymeric binder to make green metallic parts directly in an SLS machine 84. The polymer binder is melted to adhere particles of iron to one another. The parts are then sintered in a furnace while in contact with copper. The copper infiltrates the network of iron particles while the polymer binder is burned out in this process. The resulting parts have properties similar to aluminum and can be polished and used as injection molding tools. These molds take a few days to make, but can produce large numbers of molded parts before deteriorating. The Keltool process is another way of building metallic injection molds85. This is a two-step transfer process, going from a rapid prototyped pattern to a silicone mold to a sintered metal mold. The drawback of this process is that the successive transfer steps can degrade the accuracy of the resulting molded parts. Other pre-commercial processes have also been used to build injection molding tools via rapid prototyping - these include 3D Printing86,87, Shape Deposition Manufacturing, and others.

2.2. Shape Deposition Manufacturing 2.2.1. Process Description Shape Deposition Manufacturing (SDM) is a patented rapid prototyping process developed at Carnegie Mellon University in the early 1990's88 ,89 . The SDM process involves material deposition followed by shaping to create geometry in three dimensions. Separate part and support materials are used in SDM. The support material is used to support undercut features and is etched away after a part is completed. The overall process is illustrated in Figure 2.2. The construction of a single layer of a more complex part (a sphere inside a box) is shown in Figure 2.3.

16

Deposit

Remove

Part with Support Material

Remove Support Material

Figure 2.2. Shape Deposition Manufacturing

1. Deposit Part Material

2. Shape Part Material

3. Deposit Support Material

4. Plane Support Material

Figure 2.3. Building a Single Layer of an SDM Part90 SDM parts are made by a combination of direct shaping of part material and replication of support material features into part material. If a surface is visible from the build direction, then it can be directly shaped into the desired geometry. This process is shown in Sequence A of Figure 2.4. If the part material features in question are undercut or face

17

away from the build direction, then these features will be constructed indirectly. First, support material will be directly shaped into the negatives of the features; this support material will serve as a mold. Then, part material will be deposited over these negatives, replicating their geometries and forming the desired positive features. This indirect construction method is shown in Sequence B of Figure 2.4. "Combination" features which have both undercut and non-undercut faces are built using a combination of these methods. This combined technique is illustrated in Sequence C of Figure 2.4. A. No Undercuts B. Undercuts C. Combination

Deposit Part Material

Deposit Support Material

Deposit Support Material

Shape Part Material

Shape Support Material

Shape Support Material

Deposit Support Material

Deposit Part Material

Deposit Part Material

Plane Support Material

Plane Part Material

Shape Part Material

Part Material Support Material

Deposit Part Material

Plane Support Material

Figure 2.4. Construction of Simple Features in SDM91 The SDM process begins with the construction of a support material cavity shaped like the underside of the bottom of the desired object. Part material is deposited into this cavity; this will be the bottom layer of the object. If necessary, additional upward-facing features are milled into the exposed part material. Additional support material is deposited to surround the bottom layer of part material. If there are any undercut features in the second layer of the part, then additional support material will be deposited and shaped to support them. Part material is deposited to construct the second layer of the part. It is machined into the proper geometry and then surrounded by support material. This process is cyclically repeated until all layers of the part have been built. At this point, the whole object

18

has been constructed, but it is embedded in a block of support material. To complete the process, the sacrificial support material is removed, leaving the finished part.

2.2.2. Part Decomposition and Planning The process of planning a part's construction sequence in Shape Deposition Manufacturing is divided into two steps92. First, the part is decomposed into manufacturable "compacts" segments of a single material which have no undercut surfaces93. Second, the material deposition paths and cutting toolpaths required to make the sequence of compacts are generated. The decomposition step requires a computer solid model of the part to be made. The model is subtracted from a block of solid material to produce a model of the complementary support material required to construct the part. The part and support models are then searched for the locations of "silhouette loops," paths which divide a solid into undercut and non-undercut regions. The separate models are completely subdivided into nonundercut compacts. This planning operation is shown in Figure 2.5. For most reasonable part, these compacts can be ordered in a manufacturable sequence. Such a sequence requires that for any given compact, all compacts which support it can be built previously. Pathological parts may have groups of compacts with cyclic dependencies, preventing such a sequence from existing. In this case, some of the compacts must be further divided to produce a manufacturable sequence. These decomposition steps require various tools available in solid modeling geometry engines. The original SDM planner used the NOODLES solid modeling system, developed at CMU94. Later planners used the ACIS solid modeling kernel, available commercially95. Current efforts include porting the current planner to run within the Unigraphics CAD/CAM system.

19

Figure 2.5. Decomposition into Manufacturable Compacts96 After a manufacturable sequence of compacts has been generated, the machine instructions for generating each compact must be determined. This requires the creation of both deposition paths and toolpaths for the material shaping system. Early SDM planners used an offsetting strategy to generate both deposition paths and cutter toolpaths97 from the outlines of compacts. Later research has centered around the use of an algorithm called the Medial-Axis Transform to make space-filling deposition paths which are optimized for the material deposition systems involved98 . The current move to a planner based upon Unigraphics is also expected to allow the use of commercially-available toolpath generation strategies.

2.2.3. Hardware Several systems are required to build parts via Shape Deposition Manufacturing. Part material and support material deposition systems, a material shaping system, and a materials handling system are all required by this process. Material deposition systems vary greatly according to the materials in question. Metallic materials have been deposited by several systems. The original metal deposition system used at Carnegie Mellon University was the "microcaster." A microcaster is a patented

20

device which feeds material in wire form into an arc, where it is melted and allowed to fall as droplets99,100. Another material deposition system used for metals is a laser combined with a powder feeder101. The laser creates a melt-pool on a substrate; the powder feeder deposits powders into this melt pool for deposition onto the substrate. A combination of powder feeders with different materials can be used to make multimaterial or functionally graded materials102; this process can also be used for ceramic powders. More recent material deposition techniques have included sputtering and etching, to make thin features for embedded sensors. Polymeric materials, the focus of the remainder of this thesis, are extruded or cast using deposition systems which vary according to the exact material involved. The original materials shaping system for Shape Deposition Manufacturing was a computer-controlled 3-axis vertical milling machine. A later implementation of the process used a five-axis milling machine for the creation of more general geometries. Metallic parts have also been shaped using electrical-discharge machining. Since the Shape Deposition Manufacturing process requires cyclic material deposition and shaping to make a part, all of these activities must occur in a common, repeatable reference frame. This can be accomplished by performing all of these processes on a single machine or else ensuring repeatable placement on the different machines involved. Pallet systems with repeatable positioning systems have been used by the SDM systems at Carnegie Mellon University and Stanford University. Parts are built on metal pallets which attach to pallet receivers mounted on all deposition and shaping machines. This palletizing system has the additional advantage of allowing parts to be removed from machines unless they are being actively manipulated. That is, if parts are cooling or otherwise waiting between process steps, they can be pulled from machines to allow other parts to be processed. Material handling for Shape Deposition Manufacturing can be manual or automated. In the CMU and Stanford systems, robots can be used to transfer pallets between individual machines. The CMU robot is a rotary system placed in the center of a group of machines, while the Stanford robot is placed on a linear track which allows it to move between machines arrayed in a line. The resulting automatic rapid prototyping system is patented103.

21

2.2.4. Origins Shape Deposition Manufacturing was developed at Carnegie Mellon University as an improvement upon the MD* process. MD* stands for "recursive mask and deposit". This is a patented layered manufacturing process in which parts are made by spraying layers of material through disposable masks104,105. Either the masks themselves or separately deposited support materials are used as support materials during the process106. The MD* process was used to make parts from sprayed zinc, steel, or nylon materials, but the mask generation and placement steps were relatively cumbersome. The disadvantages of the MD* method are the unreliability of complete automation of the process, the presence of stairsteps in the resulting parts, and the relative inaccuracy of parts produced this way107. The Shape Deposition Manufacturing process was originally a combination of weld deposition and computer-controlled milling108. This pairing allowed engineering alloys to be deposited to "near-net" but somewhat rough shapes and then be machined accurately to the desired geometries. The deposition of bulk material with good bonding to a substrate improved upon the relatively poor adhesion possible through thermal spraying. At the same time, the machining step allowed the creation of geometries an order of magnitude smoother than conventional rapid prototyping machines. As additional trials were made, the process evolved into a cycle of deposition and three-dimensional shaping steps, building parts through direct shaping or replication of geometry from shaped support material. A variety of material deposition methods were integrated into the process, superseding the original "weld deposition" method.

2.2.5. Comparison to Other Rapid Prototyping Processes Compared to most other rapid prototyping techniques, SDM has the advantage of decoupling material properties from deposition characteristics. This decoupling occurs because part features are created in SDM through material addition and removal rather than material deposition alone. Processes such as Stereolithography, Selective Laser Sintering, Fused Deposition Modeling, and Three-Dimensional Printing are purely additive; material must be deposited in the desired final or "net" shape. This requires compromises between materials with optimal properties and materials which can be deposited to net shape. SDM allows materials to be optimized for excellent final-part properties, without regard to whether they can be deposited to net-shape. This is possible because the final geometry

22

will be machined into the part material, or replicated from machined support material. As a result, a broader range of materials can be processed using SDM than is possible through purely additive rapid prototyping techniques. While most rapid prototyping processes involve only material deposition, there are some processes that involve depositing and shaping material on a layer by layer basis. Laminated Object Manufacturing adds additional paper layers to a part and then outlines the periphery of those layers using a laser. The Solid Ground Curing process involves planing a layer of part and support material to an exact thickness. The Sanders process works the same way although a different style of milling cutter is used. All of these processes first deposit an entire planar layer of material and then outline or plane that layer. Either one or two dimensions of geometrical information in such processes is completely determined by the material deposition process involved. In contrast to these processes, Shape Deposition Manufacturing uses the successive deposition and shaping of a series of "compacts" to produce a part. Several compacts, each with fully three-dimensional shapes, generally compose a layer of an SDM part. These SDM layers need not have a constant thickness or even be planar at all. SDM can deposit and shape geometries which are purely threedimensional, in contrast to the 2-1/2-dimensional stacked layers produced by the other more limited additive and subtractive processes. All of the commercially available rapid prototyping processes manufacture parts from a series of planar layers; these layers are usually the same thickness throughout a part109. This use of planar layers produces stairsteps on any slanted surfaces of the resulting parts. This is undesirable because of the resulting reduced part accuracy and poor surface finish. If a smooth slanted surface is desired, the part can be reoriented so that this surface will be vertical or horizontal while the part is built. If multiple smooth surfaces are desired, such a reorientation is generally impossible, and manual postprocessing such as sanding is required. Such handwork produces smooth surfaces, but at the cost of whatever geometrical accuracy is achievable by a machine. Since SDM compacts' shapes can be defined by three or five-axis milling, slanted surfaces can be accurately built without stairsteps. Besides Shape Deposition Manufacturing, there are two other rapid prototyping processes which can produce parts with fully three-dimensional features: Controlled Metal Buildup and Multiple Material Machining. Controlled Metal Buildup is a technique being developed by the Fraunhofer Institute of Production Technology110 . This process uses laser

23

deposition of metal powders to build parts. Three-axis milling defines the outlines and top surfaces of each layer. Researchers plan to use five-axis machining to produce smooth oblique sides on parts. This process is the very similar to Shape Deposition Manufacturing although support materials have not been used with the technique, limiting the possible part geometries. Multiple Material Machining is also very similar to SDM; separate part and support materials are used (although they are named differently), and features are created via machining and replication from support material111. While the use of milling to define geometry allows the creation of fully three-dimensional features in SDM, using milling cutters for shaping does place some limits on producible features. If a square pocket into part material is desired, rounded interior corners will generally be produced by the milling cutters used. Cooper is researching the problem of generating arbitrary geometrical shapes using round milling cutters112.

2.2.6. Heterogeneous Structures One of the advantages of Shape Deposition Manufacturing compared to most other rapid prototyping techniques is the ease with which heterogeneous structures can be manufactured. Multimaterial parts and parts with embedded objects are both relatively easy to construct because of the modularity of the process sequence113. Since there are frequent breaks between material deposition and shaping steps, the process can be interrupted often and part materials changed or objects inserted. Changing part materials is as simple as switching between material dispensers in a single machine or moving a pre-referenced pallet from one machine to another. Embedding objects is as simple as laying objects into cavities between one machining step and the next deposition step. In most other processes, this is much more difficult. Most rapid prototyping processes are designed to operate with a very limited set of materials, some of which may even be incompatible with one another. Changing part materials in a machine can be relatively easy or quite tedious. An example of an easy task is replacing the spool of filament on an FDM machine. A difficult task is changing an entire vat of SLA resin and cleaning the whole machine before introducing an incompatible resin. In some processes, changing materials within a part might require purging one part material, inserting another, and resetting software parameters. In other processes, instant changes in material may be impossible. Changing materials in the Selective Laser Sintering

24

process would probably lead to blended material parts, if the environmental and laser parameters for one material didn't damage or destroy the other material involved. These examples illustrate that using different materials within a single part is often more difficult in other processes than in SDM. A similar observation can be made about the difficulty of building parts with embedded objects. Some processes can be used to build such parts, but many cannot. A new process which allows the creation of parts with embedded electronics is Offset Fabrication, being developed by Ennex Fabrication Technologies. This process is a hybrid of Laminated Object Manufacturing and signmaking - layers are cut and then bonded to the rest of the part. A battery-powered model car was built using this process114. This process is still under development and uses paper part materials. It could presumably be extended to use polymer part materials in the same way that the Laminated Object Manufacturing process was enhanced. The Extrusion Freeform Fabrication process being developed by Advanced Ceramics Research has also been used to build objects with embedded sensors 115. However, most rapid prototyping processes are designed to add material to a flat, planar surface produced by previous deposition. The recoating blades in Stereolithography, the laminating platen in Laminated Object Manufacturing, and the material dispersal roller in Selective Laser Sintering are machine components which have difficulty accommodating embedded objects. In spite of these obstacles, researchers have made Stereolithography parts with embedded sensors 116,117. The milling cutters in the Solid Ground Curing and Sanders processes are even more difficult machine elements from this point of view. Objects could be embedded into such parts, but the parts would require substantial manual manipulation in order to create a suitable cavity and insert an object. This might even require the removal of the parts from machines, with consequent fixturing and referencing problems. Some processes also rely upon continuous thermal conditioning while building a part - if a part were allowed to cool while an object was embedded, the part would be distorted or warped by the temperature change. Shape Deposition Manufacturing allows easier production of parts with embedded objects than these other rapid prototyping methods. In conclusion, multimaterial parts and parts with embedded objects can be built more easily through Shape Deposition Manufacturing than through most other rapid prototyping systems. This is a principal advantage of the process, and allows the production of parts which could be made only with great difficulty through other methods.

25

2.3. Polymer Shape Deposition Manufacturing at Carnegie Mellon University Researchers at Carnegie Mellon University began to explore the use of MD* and Shape Deposition Manufacturing to make polymer artifacts as the processes were being developed. The initial goal was the rapid production of objects with embedded electronics.

2.3.1. Embedded Electronics The first electronic parts made were made from metal and ceramic materials deposited via MD* 118. An early example was a humidity sensor manufactured in three layers. The components were a resistive heater, a thermocouple, and another resistor. The resistive heater and the thermocouple were connected in a feedback loop to maintain a constant temperature at the other resistor. Its resistance then changed according to the ambient humidity, providing an output which measured humidity. The first polymer parts made via MD* used plasma- or flame-sprayed nylon. These parts were made for a DARPA project, and involved the use of individually embedded electronic devices and sprayed interconnect traces between them. The use of the MD* process to make embedded electronic parts via thermal spraying was patented119. This technology was seen as especially useful for ongoing research into wearable computers at Carnegie Mellon120,121. CMU had developed a series of wearable computers intended to assist people while navigating an area or accomplishing a task requiring one or both hands (for example, heavy equipment maintenance). To speed this series development task, rapid prototyping technologies were used. A housing made through Stereolithography was used as the case for the "VuMan 2" computer, completed in December 1992122. Soon after the completion of this project, the idea of rapid prototyping an entire wearable computer as a monolithic structure with embedded electronics arose. The MD* process was originally envisioned for this purpose123, but MD* gradually evolved into Shape Deposition Manufacturing as different process steps replaced earlier subprocesses. The MD* process used one set of masks for spray deposition of part material and another complementary set of masks for spray deposition of support material. Different thermal spray processes were used: arcspraying was used for metallic materials, flame-sprayed nylon was used for early polymer devices, and plasma spraying was also considered for polymer use. A drawback of the flame-spray process for polymers was that

26

the flame tended to destroy some mask materials and was expected to damage or destroy embedded electronic components. For this reason, castable polymeric materials were soon used for embedded electronics although such parts were originally cast into previously constructed open molds. An early part constructed in this manner was a "Simon" game packaged within a brick of polymer material124. This device used buttons and LEDs as its user interface. Individual integrated circuits and passive components were connected by sprayed zinc traces on two planar layers. The housing was made upside down in an aluminum mold, using a castable polymer material. After the first layer or faceplate was cast, components and vias were embedded into and through that layer. Another layer of polymer was cast to encapsulate those devices, leaving only their leads exposed. The top of this layer was allowed to cure and then gritblasted to improve adhesion of sprayed traces. A thermal spray process was used to draw a layer of zinc traces to connect devices to one another. These traces were then covered by another layer of castable polymer, leaving only the tops of vias exposed. A battery cable was installed, and another layer of zinc traces were sprayed onto the part. Finally, the last layer of polymer was cast over these traces, completing the part. There were some drawbacks to the use of sprayed traces. It was difficult to draw fine traces using the spraying process, and there were problems with shadows cast b y protruding vias and component leads. These issues drove the adoption of traditional circuit boards to provide interlayer interconnections. In later designs, populated circuit boards were simply aligned on a layer and then embedded by casting polymer material over them. Separately constructed vertical vias were still required to connect different layers of circuitry. As soon as the Shape Deposition Manufacturing process was developed, its use for the production of wearable computers seemed like an obvious match125. The Simon game previously built as a brick of material was redesigned using circuit boards and rebuilt in a wristwatch form-factor. A two-part castable polyurethane resin was used for this newer part126. While the outside geometry of the "brick" Simon had been generated by casting the whole part into an aluminum mold, the geometry of the wristwatch was shaped by successive machining of the separate layers of the device.

27

2.3.2. VuMan Wearable Computer The first major wearable computer built using Shape Deposition Manufacturing was the "VuMan 3R" or "VuMan SDM"127. Two circuit boards containing the electronics were sandwiched between three solid layers of polymer. The circuit boards were connected by pins and pin receptacles128, which could have been stacked to allow the use of more than two boards in a single computer. Buttons, LEDs, and a jack for the attachment of an headmounted display constituted the user interface for the computer. A socket for a PCMCIA card was included to allow the computer to be reconfigured. Standard consumer batteries powered the computer; they were located in battery compartments on both ends of the housing. The overall size of the computer was about 5" x 5" x 2". The VuMan was built as three layers of polyurethane part material. An investment casting wax was used as the support material. The process began with the deposition of wax to support the bottom layer of the computer. This wax was heated, manually poured into an aluminum frame, and allowed to cool. It was then sculpted by three-axis milling into a mold for the bottom surface of the housing. A two-part castable polyurethane was mixed and manually cast into this mold. After it cured, it was planed flat and then pocketed for the first circuit board. This circuit board was manually placed, and its pin receptacles were covered by plastic caps. The circuit board was sealed using a silicone resin deposited manually, and the second layer of polyurethane was cast over it. This layer was planed flat and pocketed for the second circuit board. Then, the pin receptacles of the previous circuit board were uncovered by drilling through their plastic caps. The second circuit board was manually placed into its pocket, and its pins were pressed into the pin receptacles from the previous board. Before casting the third layer, buttons and LEDs for user interaction were attached to the second circuit board. The buttons were covered with caps like those of the pin receptacles. Light pipes were glued to the LEDs and allowed to extend past the final boundaries of the housing. The third layer of polyurethane was then cast and allowed to cure. It was machined to form the external geometry of the wearable computer. The LED light pipes were machined at the same time, so they became flush with the outside of the housing. Finally, the buttons and other interface jacks were uncovered. At this point, the computer was finished but partially embedded in a wax mold. The original aluminum frame was disassembled and removed, and the wax mold was cut off its base. The wax was melted by heating it in an oven, and a heatgun was used to melt residual wax for removal from the housing.

28

2.3.3. Waterproof Computers and Thermal Issues Concurrent with the present work, researchers at Carnegie Mellon continued to build wearable computers using SDM; this was their major application for Shape Deposition Manufacturing using polymer materials. Following the construction of the VuMan SDM wearable computer, their next project was the design of a waterproof version of the same system, called the "FrogMan." This computer was intended for military divers, so it was required to operate underwater. This required the use of materials which are relatively impermeable to water, to prevent water from damaging internal electronics either directly or through corrosion. New cable entrances and seals and new seals for battery compartments were also required. A castable polyurethane material was selected as the part material for this design. A detailed description of the manufacturing process has been published129, including methods for interconnecting circuit boards. Later research at Carnegie Mellon University focused on thermal issues for wearable computers. As computational requirements for wearable computing applications rise over time, microprocessors run faster and hotter than before. This produces large amounts of heat, which must be removed from internal electronics to prevent them from overheating and failing. The construction of wearable computers must be optimized to ensure heat can be removed from internal spaces and dissipated efficiently to the environment. This requires careful attention from a materials selection viewpoint, and may require the inclusion of special heat transfer components like heat spreaders or heat pipes. The preceding needs are shared by virtually all high performance computers although wearable computers present additional limitations. A requirement unique to wearable computers is the need that a computer remain cool enough for operator safety and comfort. CMU researchers examined heat transfer issues for conventionally manufactured wearable computers 130 , 131 , 132 as well as special considerations for wearable computers manufactured through Shape Deposition Manufacturing133.

2.3.4. Material Research Several different part materials were used for early work in Polymer Shape Deposition Manufacturing at Carnegie Mellon University. Since the flame-sprayed nylon used with the MD* process was incompatible with electronics, other materials were investigated. This experimentation occurred during the transition from MD* to Shape Deposition

29

Manufacturing. The first trial material was a conventional ultraviolet-cured Stereolithography resin. This material was manually deposited and cured using an ultraviolet flood light. That process was relatively tedious since only very thin layers could be cured. After some experimentation with this material, researchers switched to castable thermoset polymer materials. One of the first castable materials was a two-part polyurethane called LUC-4102, from Adtech Plastic Systems (Charlotte, Michigan). This is a filled thermoset resin and hardener combination which cures relatively rapidly. It becomes relatively solid within forty-five minutes of casting, and can be machined in about an hour. The fillers reduce shrinkage and heat generation during the curing process, but have several disadvantages for machinability. First, the aluminum oxide filler is very hard and abrasive, wearing cutting tools rapidly. Secondly, dust is produced when machining the material - this dust is a nuisance for operators and machines (since it spreads the abrasive particles throughout machine tools). This material has a moderate thermal conductivity but has relatively poor impact strength and moisture resistance. It was used for the construction of the wristwatch Simon game and the first VuMan wearable computer. The next material used for embedded electronic parts was formulated by researchers at Carnegie Mellon. This was another two-part castable urethane polymer, with a special filler added. The base resin was TDT 205-3 from Ciba Specialty Chemicals (East Lansing, Michigan) or a special unfilled grade of LUC-4102 from Adtech. Both of these materials are fast-curing urethane resins. A nylon powder was added to this resin to improve the impact properties of the polymer material. A version of the VuMan wearable computer was made using this material. There were some problems with this combination material, probably because the nylon powder tended to absorb water which interfered with the curing mechanism of the polyurethane. After this experience, it was decided that only commercial products would be used rather than formulating custom mixes. For the production of the FrogMan wearable computers, a broader materials search was conducted. The main goal was to find a material with minimal moisture absorption (to ensure the packaged electronics remain dry) and high impact strength (to prevent damage to the electronics during handling or rough use). A variety of castable epoxy and urethane materials were examined, both alone and in combination with additional coatings. Besides moisture absorption and impact strength, there were several lesser materials requirements involving physical properties, thermal issues, and dispensing issues. The desired physical

30

property attributes included ruggedness, waterproofing, machinability, moderate hardness, dielectric strength, and light weight. The thermal attributes included good heat transfer, high glass-transition temperature, low exotherm, reasonable cure time, low shrinkage, and low CTE. The dispensing attributes included low viscosity, easy degassing, and safe for operators. At this point, water-absorption tests were conducted. Test specimens were cast from a variety of materials; they were scaled to have the same surface to volume ratio as the Frogman design. These coupons were tested for water absorption by weighing them after exposing them to water at a pressure equivalent to a depth of 30 m. Weighings were conducted at various intervals, up to a day of exposure. The best results were obtained from Adtech 501 epoxy and Adtech LUC-4180 urethane. Moisture contents for these parts were 0.17%-0.2% after 24 hours. Those two materials were then coated with a variety of epoxy and rubberized paints, and the tests were repeated. The results were moisture contents of 0.07%-0.10% after 24 hours. Next, machinability was evaluated; both materials had low tool wear and good surface finish, and did not easily chip. To assess impact performance, drop tests of weighted samples were performed; the urethane material performed better. Finally, linear shrinkage of the materials was evaluated. At the end of these tests, the LUC-4180 urethane from Adtech was chosen for the Frogman device. This is an unfilled two-part castable urethane resin with cures sufficiently for machining after about twelve hours. Although a series of different part materials were tested, less experimentation with support materials was performed. While the MD* process was evolving into SDM, the first cast polymer parts were contained in aluminum molds - no "support material" in the SDM sense was used. The first material tested as a support material for castable thermoset polymer part materials was a machinable wax. This was a product called "Master File-A-Wax" from the Kindt-Collins company, used for checking computer controlled machining programs before attempting to make metal parts. This material was tested as a support material, but it shrinks too much upon cooling to be useful in this application. The next support material tested was more successful. This was "Master Protowax", also from the Kindt-Collins company. This is an investment casting pattern wax, intended to be easily machinable. It shrinks significantly less than the File-A-Wax material upon cooling. The wristwatch Simon, the VuMan wearable computers, and the FrogMan wearable computers were all made using this support material.

31

2.3.5. Part and Support Material Deposition Methods For all of the parts built so far, material deposition was performed manually. The part material components were mixed together and vacuum degassed before being cast onto the parts. The support material was usually heated using a hotplate and then manually poured over the part. Before applying wax support material this way, the part material was sometimes preheated to 38-43C using a heatgun. All of these material deposition processes can be automated. Part material mixing and deposition can be automated using a meter-mixing system. The material would be degassed before mixing, so the mixed material would not require degassing. Separate supplies of the resin and hardener would be combined in a manifold and fed through a static mixing nozzle onto the part. The support material deposition can be automated through the use of a "glue pot" or other system for the dispensing of hot-melt adhesives or waxes. Such a system would heat a volume of material until molten and then dispense it under pressure through a nozzle.

32

CHAPTER THREE POLYMER SHAPE DEPOSITION MANUFACTURING 3.1. Introduction The Polymer Shape Deposition Manufacturing process is an implementation of the SDM process for the direct production of polymer parts. Polymeric part and support materials are alternately deposited and shaped to produce objects. This section provides an example of the basic process and describes both process and automation requirements.

3.2. The Polymer SDM Process Regardless of the materials being used, the Shape Deposition Manufacturing process revolves around the deposition and shaping of part and support materials. If a part material feature can be machined or shaped directly, then this is the preferred practice. If undercut surfaces or other features which cannot be machined directly are required, then the complements of these features are machined into support material. Part material is then deposited on top of this support material, replicating its surfaces and producing the desired geometry. Any transitions between undercut and non-undercut surfaces require interlayer boundaries along the path of the transition. The resulting layers can be non-planar and variable thickness, with curved surfaces. If this splitting scheme results in layers too thick to be deposited in a single layer, then the thick layers must be split into layers as thin as necessary for deposition. It is desirable to have the smallest number of layers possible, since each interlayer interface is a source of potential weakness or geometrical error in finished parts. Polymer SDM has been used to build a variety of assemblies and objects with internal moving parts. Such parts normally require clearances between different regions of part material; the clearances are filled with support material during the build process. If such a

33

clearance is large, then it can be produced by machining gaps inside a single region of part material, separating it into two separate regions. Support material is then backfilled into the newly created clearance region. Another method for producing a fine clearance is to machine a thin rib of support material before casting part material around it. This is often possible, but relies upon the use of relatively high strength and machinable support material. However, some clearances would require ribs to be too thin to support themselves during machining, and must be created using a different operation sequence. In this case, one region of part material is deposited and machined into the proper shape. A quantity of support material much bigger than the desired clearance is then deposited next to this part material. This support material is machined into the desired shape of the clearance. Finally, more part material is deposited next to the clearance, representing the second required region of part material. This allows clearances thinner than a given cutting tool to be machined, although it requires thin layers of support material to adhere very well to part material during machining.

3.3. Process Example The following example build sequence will help clarify the Polymer SDM process. This sequence represents the building of a flap mechanism intended for aerodynamic tests 134. Figure 3.1 shows the first steps in the process. The first step is the creation of a cavity which complements the geometry of the first layer of the part. Since the part will be a mechanism, there are two separate cavities separated by a thin rib of support material. The left-hand cavity will be the base of the housing for a flap, and the right-hand cavity will be the bottom of the flap. These cavities are machined into a support material substrate which is mounted to a pallet for registration purposes. The "Step 1" picture of Figure 3.1 illustrates the two support material cavities needed. Since the rib of support material is so thin, this step is actually performed slightly differently. First, the right-hand cavity is milled into the support material, and part material is cast into this cavity. Then , the lefthand cavity is machined into the support material. Since the thin rib of support material is now backed by part material, the machining process will not destroy the rib. After the lefthand cavity is machined, part material is cast into this cavity as well. The second step is planing the newly cast part material flat on top. The "Step 2" picture of Figure 3.1 shows what the resulting part would look like if removed from the support material at this stage. While different shades are used in that picture to represent the bottom of the housing and

34

the bottom of the flap, the same material was actually used for both; the pictures use different shades to differentiate between different parts in close proximity.

Step 1

Step 2

Step 3

Step 4 Figure 3.1. Polymer SDM Process Sequence, Part I135

The third step is to machine a cavity for a layer of the flap's pivot and flap and a cavity for another layer of the flap housing. The "Step 3" picture of Figure 3.1 shows how the part and support material would look if the cavities could be machined at the same time. As with the previous step, however, the required support material rib between the two cavities would be too thin to survive the machining process. Therefore, the cavities for the flap's pivot and the end of the flap are machined first, and part material is cast into those cavities. Next, the cavity for a layer of the bottom of the flap housing is machined, with the thin support material rib again backed by part material. After this cavity is machined, part material is cast into it. Rather than machining this new material flat as had occurred in

35

previous steps, some features are machined into the top surface of the part material overhang at this point. The "Step 4" picture in Figure 3.1 shows what the part would look like if removed from support material after the completion of this step. The remainder of the process steps are shown in Figures 3.2 and 3.3. The fifth step in the process is to build the cavity for the side walls of the housing. This is accomplished by casting support material over the whole part and then machining trenches down to previous part material where the walls of the housing, the flap's pivot, and the flap itself are located. The sixth step in the process is casting part material into these trenches to build the required part material structures. The "Step 5" picture in Figure 3.2 shows the desired cavities. However, the use of thin ribs of support material which act as clearances requires these steps to be combined and performed slightly out of order. First, the cavity for the flap's pivot is machined, and part material is cast into this cavity. Second, the cavity for the outer wall of the flap housing is machined and cast into. Finally, cavities for the central part of the flap housing and the flap itself are machined and deposited into. The "Step 6" picture in Figure 3.2 shows the appearance of the part if removed from support material after these steps.

Step 5

Step 6 Figure 3.2. Polymer SDM Process Sequence, Part II

The seventh step is shaping a cavity for the underside of a large overhang on the housing, as well as building a cavity for another layer of the flap's pivot and the flap. The "Step 7" picture in Figure 3.3 shows the desired cavities for this step. As before, this is accomplished by machining the cavities for the flap and its pivot first and casting part material into them. Then, the cavity for the next layer of the flap housing is machined and cat into. The no-support-material view after the completion of this phase is shown in the

36

"Step 8" picture in Figure 3.3. The ninth step in the process is the construction of cavities for the top of the housing and for the reconnection of the flap to its pivot. The "Step 9" picture in Figure 3.3 shows the needed cavities. The flap cavity is machined first and cast into, followed by the cavity for the top of the flap housing. The completed part is shown in the "Step 10" picture of Figure 3.3. Actual completed flaps are shown in Figures 4.4 and 6.2.

Step 7

Step 8

Step 9

Step 10 Figure 3.3. Polymer SDM Process Sequence, Part III

While the preceding description was the way that particular part was made, some of the steps were atypical. During the sequence, part material was very seldom directly machined; in most Polymer SDM parts, this is a more common occurrence.

37

3.4. Process Requirements The Polymer SDM process needs several elements for successful manufacture of parts. Both material selection, part planning, and part manufacturing requirements exist. Most of the planning and manufacturing elements can be implemented in a manual or an automatic fashion although steps like material shaping virtually require automated solutions. Before parts can be completely planned, much less manufactured, materials for their construction must be identified. One of the most important elements of Polymer SDM is a system of compatible part and support materials. Closely related to this element is the need for a support material removal process to allow the excavation of finished parts embedded in support material. These material aspects of the process will be discussed in detail in Chapter Four. Before parts can be made, they must be planned. In Polymer SDM, planning begins with decomposing parts into layers and sublayer geometries called "compacts." These compacts are regions of single materials which can be deposited and shaped in a single step. After these compacts have been identified, their construction can be planned. This includes the determination of paths for material deposition equipment and material shaping processes. These paths will depend upon both the geometry of the previously identified compacts and the material properties of the part and support materials to be used. All of these planning steps can be accomplished manually, although even manual planning requires support from sophisticated computer environments. Some or all of the steps could be accomplished automatically although additional software development would be required. After the parts have been planned, they can be constructed. Since SDM uses a cycle of material deposition and shaping, these processes are needed next. While polymer deposition processes vary somewhat, the shaping is performed via machining for the polymeric materials discussed in this thesis. As with the software requirements, some of these processes can be performed manually and some require automation. Material deposition can be accomplished manually, although the improved repeatability of an automated process and the tedium of manual deposition both favor automated systems. Shaping processes, however, virtually require automation; this is normally accomplished through CNC machining of surfaces. The material removal process can be either manual or automated; for solvent-based processes, automatic agitation systems may shorten the times required.

38

3.5. Automating Process Planning Automating Polymer SDM process planning requires automating both the part decomposition and the deposition and shaping path planning steps. It is not yet clear whether this will be possible for parts of arbitrary geometry. Automated part decomposition requires the ability to divide parts at undercut/non-undercut transitions, which may occur along branching non-linear curves. Deposition path planning usually requires the generation of a space-filling curve. Current research at Stanford in this area is directed toward determining paths to produce metallic parts with better properties136. These improved paths prevent incomplete deposition at the center of regions; this is a problem whose solution may also improve the quality of polymer parts. Path planning for shaping requires the generation of CNC machining code to cut arbitrary surfaces. The ability to generate multiaxis CNC code automatically does not yet exist for general parts. However, the shaping steps of the SDM process have some properties which simplify this planning task. For example, each compact is defined to require only the machining of nonundercut surfaces. If undercut surfaces of a compact required machining, then that surface of the compact would have been generated by replication from support material rather than direct machining. Therefore, the general problem of tool reachability is simplified, and refixturing will never be required. However, an additional challenge of toolpath generation for Polymer SDM is the production of interior sharp corners. Such sharp corners can only be milled directly using arbitrarily small-diameter cutting tools. To circumvent this restriction, complicated deposition, cutting, and redeposition strategies are required. These strategies are currently being identified and developed during manual process planning. Achieving the same functionality during automated planning will require significant additional effort. At the present time, part planning for Polymer SDM parts is performed manually using commercially available CAD solid modeling packages while planning automation research occurs in parallel using custom software.

3.6. Automating Part Building Part building requires material deposition and shaping processes. Automated part building requires automated material deposition and shaping as well as automated material handling and control. Automating material shaping is relatively easy since common CNC milling machines can be used; this is the only reasonable way to shape Polymer SDM parts anyway. Automating material deposition requires part and material deposition systems

39

whose paths can be controlled as they traverse parts. This can be accomplished through dedicated motion systems, with general-purpose robots, or by integrating deposition equipment onto CNC milling machines. Automated material handling is desirable since deposition and shaping cycles must be repeated many times to make a single part. Automated material handling requires either moving parts to different deposition and shaping equipment, or moving both deposition and shaping equipment into contact with a part. Either of these arrangements can be expanded to include multiple parts in a system at the same time, allowing parallel part production. There are two different possible approaches to the manufacturing automation issue. Either a group of specialized machines can be integrated together using a part transfer system or all equipment can be combined into a single standalone machine. The first approach was followed during the construction of the Stanford Rapid Prototyping Laboratory. This laboratory was intended for research into a variety of materials and deposition and shaping processes, requiring different pieces of specialized hardware for these purposes. A transfer robot was installed to tie these processes together for automatic use. Parts are built on pallets which can be moved from machine to machine for processing; the pallets fit into specialized pallet receivers to assure repeatability of installation. The layout of the laboratory can be seen in Figure 3.4. A 5-axis CNC mill, a washer, a plasma microcaster, and a laser deposition station are all linked by a transfer robot, allowing the automated production of metallic parts.

Figure 3.4. Stanford Rapid Prototyping Laboratory137 The use of multiple pallets and pallet receivers adds significant flexibility to this type of system. Since machines operate independently, different machines can work on different

40

parts at the same time. This allows the integration of more than one of a given type of machine, allowing line balancing to accommodate different operation durations. If deposition is three times faster than shaping, then three times as many shaping machines can be installed and connected. Additionally, automated storage and retrieval systems can be added for times when curing or cooling operations require a long wait between operations. Since pallets can be removed from machines, parts can be warehoused until they are ready for machine operations again. While such a system will not necessarily decrease the production time for a single part, it can greatly increase the overall production rate of groups of parts The original Shape Deposition Manufacturing Laboratory at Carnegie Mellon University also used the first automation approach for the production of metallic SDM parts, using a transfer robot to move pallets between stations. However, for the manufacture of polymeric parts at CMU, the second approach was taken. The resulting machine is pictured in Figure 3.5; it is an integrated machine built around a Fadal CNC vertical milling machine. High-pressure thermoplastic extrusion heads were mounted on the machine for deposition of part and support materials. The material shaping step was accomplished using the original functionality of the milling machine.

Figure 3.5. CMU Polymer SDM Station

41

CHAPTER FOUR MATERIALS FOR POLYMER SDM 4.1. Introduction The Polymer Shape Deposition Manufacturing process has applications in a variety of areas, but it requires systems of compatible materials and processes for successful use. Both a part material and deposition technique, a support material and deposition technique, and a support material removal technique are required. There are a variety of constraints on the selection of such materials and processes, which the following section describes. Over a period of time, several systems of compatible materials have been identified. While most of these systems were developed for particular applications, some have the potential for broader use. Finally, some spin-offs occurred, in which materials developed for a single purpose were found to be even more useful in other areas.

4.2. Material Requirements for Polymer SDM Polymer SDM requires compatible part and support materials, subject to the constraints of the deposition and removal processes. Extrusion and casting are the current deposition methods for Polymer SDM. Part and support materials must be physically and chemically compatible. These materials must be deposited or cured at temperatures which do not damage previously shaped features. Part and support materials must have minimal intersolubility, and neither material may inhibit curing processes of the other. In order to build strong parts using SDM, successive layers of part and support materials must exhibit high degrees of cohesion and adhesion. The part and support materials must solidify without the formation of defects, e.g. internal voids. Layers of cast material have the same intralayer properties as in any conventional casting application. Extruded material, however, may exhibit poorer mechanical properties within a layer due to interbead

42

interfaces. To prevent voids during extrusion, either low viscosity material or high extrusion pressure is required to fill sharp interior corners and interbead gaps. In order to obtain high strength parts, the part material must also have good interlayer adhesion. Some degree of adhesion between part and support materials is also required, or part and support material may separate during machining, destroying the parts. As discussed in greater detail in the next chapter, achieving such adhesion requires complete wetting of previously machined surfaces by any newly deposited material. Distortion and warpage must be minimized to build accurate parts. Shrinkage is commonly encountered during layer solidification or curing, and it is often nonuniform. This shrinkage must be minimized to reduce residual stresses which can result in warpage or delamination138. Choosing amorphous rather than crystalline materials helps minimize shrinkage 139. Preheating previous layers before depositing new material minimizes temperature differentials between old and new material, reducing differential shrinkage of new material as the layers cool. Coefficients of thermal expansion of part and support materials should be similar and as low as possible to reduce thermal distortion. Besides the material constraints imposed by deposition processes, there are also constraints associated with support material removal. Support material must be removed carefully to avoid damage to part material. If solvents are employed, they must induce minimal swelling of the support material before dissolving it, or dissolution stresses may crack a part. It is advantageous to use solvents to remove the support material from internal cavities; manual removal processes are difficult or impossible in such locations. Manual removal processes also risk damage to fragile parts. Most of the aforementioned constraints on Polymer SDM are shared with the other solid freeform fabrication extrusion or jetting processes, e.g. FDM, BPM, Sanders, Actua, and Genisys. These other processes, however, deposit material to net shape. Therefore, their materials' viscosity and solidification characteristics must be optimized for material to maintain its shape while still bonding to previously deposited material. SDM avoids some of these linked constraints by including a material shaping step. This requires the SDM part and support materials to have sufficient strength and stiffness to enable accurate machining. The support material must exhibit smooth surfaces after machining, since its surfaces will be replicated onto later-deposited part material. The part material must exhibit smooth surfaces after either replication or direct machining processes.

43

As in other solid freeform fabrication processes, there are several practical manufacturing requirements which affect material selection. Cost, ease of use, and robustness must be considered. Deposition processes must produce well-controlled deposits over a wide range of material and process variations. Part throughput and process speed requirements encourage the use of rapid deposition processes with rapid solidification or curing as well as rapid machining rates. Inexpensive materials and simple deposition methods (with inexpensive deposition equipment) are preferable. SDM materials should have long shelflives, long pot-lives, as well as low toxicity. The support material and its removal solvents should have low toxicity. All of the these and the preceding process constraints cannot necessarily be satisfied, and compromises must sometimes be made. In spite of this variety of constraints, several compatible systems of materials and processes have been developed. These are described in the next section of this chapter.

4.3. Material Combinations A variety of material and process combinations has been developed to satisfy the preceding list of constraints. Support materials for thermoplastic part materials were investigated, resulting in the identification of ultraviolet-curable solder masks. Ultraviolet-curing part materials were investigated to reduce process cycle times. One successful material combination used acrylate Stereolithography resin part materials with solder mask support materials. This was intended for more rapid testing of process sequences to be used for the production of metallic SDM parts. After this combination was developed, the solder mask support materials were used by Cooper et al. to develop an indirect rapid prototyping process called Mold SDM. Next, in-depth research targeting completely removable support materials for use with castable thermoset part materials was conducted. A major limitation of the materials combination developed by researchers at Carnegie Mellon University was that the support material required some manual work for removal and tended to leave residue on completed parts. A variety of approaches were examined, but the most desirable solution was better support material removal solvents for use with polyurethane part materials and wax support materials. Objects with moving parts were built using this new support material removal method. Castable part materials with shorter processing times were investigated. This work resulted in the selection of a polyurethane with one sixth the cure time and four times the interlayer strength of the previous part material. Finally,

44

materials with improved interlayer strengths were pursued, resulting in the identification of an epoxy part material with twice the interlayer strength of the preceding material and eight times the interlayer strength of the original castable part material. After discussing these material combinations, this section will end with a brief description of Polymer SDM materials developed by other researchers for applications besides polymer parts.

4.3.1. Support Materials for Thermoplastic Part Materials The first research effort was devoted to the development of support materials for injection molded thermoplastic part materials. The initial motivation for this material combination was the creation of parts in strong, tough injection moldable polymers such as Ultem and PEEK. These materials have properties superior to all current polymer rapid prototyping materials, and would be very attractive as materials for the Shape Deposition Manufacturing process. However, these materials must be heated to high temperatures for deposition, and high-temperature-tolerant support materials are required to complement them for use in SDM. Identifying removable support materials with high-temperature survivability was a difficult process. Several classes of materials were examined: ultraviolet-curable Stereolithography resins, ultraviolet-curable adhesives, integrated-circuit photoresists, printed-circuit-board plating resists, and printed-circuit-board solder masks. Although none of these support materials proved compatible with high-temperature injection molding polymers, the search did result in the identification of solder mask materials. These ultraviolet-curable, water-soluble thermoset materials are useful as support materials for ultraviolet-curing part materials, allowing the later development of a rapid-curing SDM materials combination for testing purposes. These solder masks were also used later as support materials for thermoplastic wax part materials, in the Mold SDM process developed by Cooper et al. More detailed information about these materials can be found in section A.1 in the Appendices.

4.3.2. UV-Curable Part and Support Materials for Fast Cycle Times Since adequate support materials could not be found for use with injection-molding thermoplastic materials, other applications of Shape Deposition Manufacturing using polymers were investigated. The earliest successful application was providing a prototyping environment for the manufacture of metallic parts via SDM. Complicated

45

metallic parts have relatively long build times. The material deposition process is relatively rapid, but the material shaping step is much slower since hard metals require low machining feed rates. The use of polymer materials allows more rapid prototyping of complex geometries since much higher milling feed rates are possible for these materials. To further improve the process speed for such prototyping, polymer materials with fast cycle times are desirable. The part materials evaluated for this system were ultravioletcuring resins; ultraviolet-curable solder masks were used as support materials. The best materials combination found was Somos 2100 acrylate Stereolithography resin part material with mixed solder mask support material, using water for support material removal. A copy of an injection mold was made using this materials combination. This copy is shown in Figure 4.1, and the original stainless steel SDM injection mold is shown in Figure 4.2. There were some drawbacks involving strength and shrinkage of this part material, so other ultraviolet-curable part materials were examined as well. No superior materials were identified, although one was marked for future study. The insufficient strength of these part materials forced a return to the part materials originally used at Carnegie Mellon University. More detailed information about these initial materials can be found in section A.2 in the Appendices.

Figure 4.1. Acrylate Stereolithography Resin Part

46

Figure 4.2. Stainless Steel Part140

4.3.3. Completely Removable Support Materials for Castable Part Materials Castable thermoset part materials were originally identified by researchers at Carnegie Mellon University. These materials had adequate strength and good toughness, but the performance of this material system was limited by available support materials. The wax support materials used required significant amounts of manual effort for removal and usually left thin residues on all part material surfaces. While this was acceptable for the large, boxy parts being built at the time, it prevented the manufacture of parts with small internal cavities or fine mechanisms. For these applications to be possible, completely removable support materials are required - this was the next research avenue pursued. Several potential support materials and support material removal methods were identified and evaluated. These included solder masks dissolved by water, water-soluble waxes dissolved by water, epoxies dissolved by solvents, and waxes dissolved by organic solvents. A cleaning solvent called BIOACT 280 was found to dissolve wax support materials completely. This resulted in a successful materials combination of LUC-4180 polyurethane part material and Protowax investment casting wax support material, with BIOACT 280 for support material removal. Several parts were made using this materials combination; two are shown in Figure 4.3. More detailed information about these materials can be found in section A.3 in the Appendices.

47

Figure 4.3. LUC-4180 Polyurethane Assemblies

4.3.4. Castable Part Materials for Short Cycle Times While LUC-4180 polyurethane and Protowax support material is a useful material combination, it is not perfect. One drawback of this combination is the relatively long cure time required by the polyurethane part material - twelve hours are required between casting and machining. This limits part production to roughly a 1-1.5 layers per day on a single part, since more than twelve hours are required between successive castings141 . To shorten the production cycle for Polymer SDM parts, faster-curing thermoset resins were investigated. A variety of faster-curing polyurethane resins were tested, leading to the selection of Ciba TDT 205-3 polyurethane as a part material. This material can be machined within two hours after casting, allowing drastic reductions in cycle time compared to the previous material system. Chemical compatibility requirements forced a change in support materials for use with TDT 205-3. The new support material is KindtCollins KC3230A, which is still removable by BIOACT 280 without harming polyurethane part material. The resulting material system of TDT 205-3, KC3230A, and BIOACT 280 has been used to build small assembled mechanisms for use in aerodynamic tests; these parts are shown in Figure 4.4. More detailed information about these materials can be found in section A.4 in the Appendices.

48

Figure 4.4. TDT 205-3 Polyurethane Flap Mechanisms142

4.3.5. Part Materials for Higher Interlayer Strength LUC-4180 polyurethane has relatively poor interlayer strength; this was discovered when parts such as those shown in Figure 4.3 were made. The external frames of these assemblies were made in three layers, and the interface area between successive layers is relatively small. When the frames were stressed during handling, they delaminated at the interfaces between layers. To overcome this problem, materials with higher interlayer strength were investigated. Two materials identified for this application are Ciba TDT 2053 and Adtech EE-501/530 epoxy. The TDT 205-3 is described in the previous section. It was originally used because of its rapid cure speed, but its interlayer strength is also superior to LUC-4180. As described in the next chapter, its interlayer strength is four times that of LUC-4180. The interlayer strength of the 501/530 material is twice as high as that of TDT 205-3. This makes the 501/530 epoxy material the strongest material identified for Polymer SDM of multilayer structures. However, the material is highly filled, increasing processing difficulties, and requires twenty-four hours between casting and machining. These issues have kept 501/530 from being used much for Polymer Shape Deposition Manufacturing although it has the highest interlayer strength of any material in current use. More detailed information about these materials can be found in section A.5 in the Appendices.

49

4.3.6. Other Materials The preceding materials combinations and systems were developed for the production of polymer parts via SDM. Other researchers have developed polymeric material combinations for different applications of Shape Deposition Manufacturing. Researchers at Advanced Ceramics Research and Carnegie Mellon University have developed a ceramicfilled wax part material for the production of ceramic green parts. Researchers at Stanford University have developed a technology called "Mold SDM," in which other materials are cast into wax molds made via SDM. This technology is described in more detail in Chapter Six, but the materials used to make those molds are described in this section. "Green" ceramic parts are made of ceramic powders bound together by other materials. They are shaped in this state, and then the binders are burned out and the parts fired to sinter the particles together. This process is used because fired ceramic parts are very difficult and costly to shape. There is a large amount of activity in the Solid Freeform Fabrication community focused on the rapid prototyping of green ceramic parts. One method demonstrated at Carnegie Mellon University is the use of Shape Deposition Manufacturing to build such parts.143 The part material in this system is a silicon-nitridefilled wax which is deposited via high-pressure extrusion. This material is shaped by three- or five-axis milling to define geometry. The support material is ACR200 or ACR50 thermoplastic polymer, removable via dissolving in water. Both the ceramic-filled wax and the support materials were developed by Advanced Ceramics Research. A green part made in this manner is shown in Figure 4.5. More detailed information about ACR50 and ACR200 can be found at the end of section A.1 in the Appendices.

50

Figure 4.5. Green Ceramic Part144 The Mold Shape Deposition Manufacturing process is based on the use of wax molds made via SDM. Other materials are then cast into these wax molds to make parts. The polymeric wax molds are made through Shape Deposition Manufacturing - wax is cast and then machined to shape. Support material is used to shape overhanging features and occupy the space in the mold, acting as "temporary part material." The original identification of the solder mask support materials for SDM use allowed the development of this process. Investment casting waxes, machinable waxes, or combinations of the two are used as part materials for construction of the molds. The current part material is Kindt-Collins KC3230A, a material developed for this particular application. Ultraviolet-curable solder masks are used as support materials and are dissolved in water for support material removal. The current support material is a 4:1 mix of ElectroLite ELC4497 and Dymax 920311-F solder masks. This use of solder masks as support materials for waxes represents their actual use as support materials for thermoplastic materials, an early goal of the present materials search. The wax part material and solder mask support material combination is used to make polymeric wax molds such as the sectioned mold shown in Figure 4.6. While this material system has only been used for the manufacture of molds for use in the Mold SDM process, it could also be used directly for other applications, such as wax investment casting patterns. The use of solder mask support material would allow the construction of patterns with complicated core structures, since solder mask can be completely removed after the construction of patterns.

51

Figure 4.6. Wax Mold145

4.4. Discussion The preceding materials combinations can be divided into classes based on the type of polymer materials used as part and support materials. Thermoset and thermoplastic materials have different properties which can be exploited in successful material combinations. A major issue for any type of part and support material combination is the issue of part distortion and inaccuracy. Part distortion is greatly affected by the shrinkage of part and support materials, in both obvious and subtle ways. These issues will be explored in the following section.

4.4.1. Categories of Polymer SDM Material Combinations Possible part and support material combinations for SDM of polymers can be categorized by whether materials are thermoplastic or thermoset. As shown by the matrix in Table 4.1, there are four possible categories of materials for SDM. A discussion of each is presented below.

52

Thermoset Support Material Thermoset Part Material Thermoplastic Part Material Category A Part: SLA Resin Support: Solder Mask Category C Part: Wax Support: Solder Mask

Thermoplastic Support Material Category B Part: Polyurethane or Epoxy Support: Wax Category D Part: Polycarbonate or Filled Wax Support: ACR200

Table 4.1. Categories of Polymer SDM Material Combinations Category A includes combinations of thermoset part materials and thermoset support materials. In this category, thermoset part materials must be more highly crosslinked than the support materials so that the support materials will be removable via solvents which will not harm the part materials. The only major SDM combination from this category is acrylate Stereolithography resin part material with solder mask support material. Category B includes combinations of thermoset part materials and thermoplastic support materials. Thermoplastic support materials can be removed by heating or heated solvents, since thermoset part materials will not melt when heated. If thermoset part materials cure exothermically, however, their heats of polymerization must not cause thermoplastic support materials to melt or flow. SDM material combinations in this category include polyurethane or epoxy part materials and wax support materials. Heat resistance of support materials limits the cure speed of part materials since fast-curing part materials may generate enough heat to melt support materials. Therefore, most of the SDM thermoset part materials in this category cure at intermediate speeds. However, these part materials tend to have low heat resistance, making them soften at the deposition temperatures required by the support materials. This can weaken parts during construction, leading to greater distortion as described in the next section. Despite these heat-resistance issues, most of the successful SDM material combinations have been drawn from this category. Category C includes combinations of thermoplastic part materials and thermoset support materials. In this category, thermoset support materials must be chosen so that heat generated during curing does not melt thermoplastic part materials. Support materials must also be chosen to have sufficient heat resistance to withstand the deposition of hot, liquid thermoplastic part materials. Since thermoset materials do not melt, solvents are required for their removal. This requirement also limits the degree of crosslinking allowed in the support materials; heavily crosslinked materials will be very difficult to dissolve using

53

solvents. The only major SDM material combination in this category is wax part material and solder mask support material, used for the production of Mold SDM molds. The solder mask is highly-branched but only lightly-crosslinked, allowing it to dissolve in water. Category D includes combinations of thermoplastic part materials and thermoplastic support materials. This is a difficult combination to develop because of the different temperature requirements for adhesion and geometry preservation. For optimal adhesion between similar materials, surface remelting of previously deposited material is required. However, if bulk remelting occurs, then previously shaped geometry may be distorted. When depositing over a different material, low viscosity materials which flow easily into fine features are desirable, but the higher temperatures required for reduced viscosities may cause melting of previously defined geometry. Materials with compatible melting and softening temperatures must therefore be chosen, and deposition temperatures must be carefully controlled. The support materials can be removed by dissolving them or melting them at temperatures below the melting points of the part materials. SDM material combinations in this category include polycarbonate part material with ACR200 support material and ceramic-filled waxes with ACR200 support material. Unfortunately, the high temperatures required for extrusion of polycarbonate tend to cause significant softening of ACR200, producing replicated geometry with poor surface quality.

4.4.2. Material Shrinkage and Part Distortion Part distortion can be a serious issue for Polymer SDM. This problem is similar to the part distortion encountered in Shape Deposition Manufacturing of metallic parts. When using metallic materials, distortion occurs because of shrinkage which occurs when molten metal part material solidifies and cools to room temperature. This causes two problems: the "Christmas-tree effect" and overall part distortion. When making metal parts in layers, this shrinkage causes a surface finish defect called the "Christmas-tree effect"146 ; its development is shown in Figure 4.7. First, part material is deposited, cooled to room temperature (Figure 4.7, sequence 1A), and then machined to the desired final geometry (1B). This geometry is then surrounded by support material, and the whole area is planed flat (1C). Next, additional part material is deposited on top of the previous part material. This new part material remelts and bonds to the previous material (1D), and then shrinks as it cools to room temperature (1E). Since the new material is bonded to the older material,

54

its shrinkage also causes the top of the previously deposited part material to shrink. Finally, the new material is machined to the final desired shape (1F). At this point, one layer of the Christmas-tree effect has been created, since the new material is at the proper position while the old material has been pulled away from its previously correct position (1G). In cross-section, sequential layers made this way look like the branches of a Christmas tree. Besides this Christmas-tree effect, overall part distortion can also be a problem for the layered deposition of metallic parts. As newly deposited material shrinks during cooling, this shrinkage is resisted by the previous layers of material to which it is bonded. This results in residual stresses, which accumulate progressively as layered parts are built147. After the part is released from whatever substrate it was built upon, these residual stresses can cause the whole part to distort and warp.

Christmas-Tree from Part Material Shrinkage 1A 1B 1C 1D 1E 1F 1G

Christmas-Tree from Support Material Shrinkage 2A 2B 2C 2D 2E 2F 2G

Figure 4.7. Formation of "Christmas-Tree" Distortion Very similar processes cause distortion of polymer parts manufactured via Shape Deposition Manufacturing. This is because Polymer SDM almost always involves materials which shrink after deposition, either as they chemically cure or simply solidify and cool from a molten liquid state. When the polymer copy of a stainless steel injection mold was produced, it showed the same surface finish defects as the Christmas-tree defects observed on the original SDM stainless steel mold. This occurred because the Stereolithography resin part material shrinks greatly upon curing; this is an example of the

55

same Christmas-tree process which affects metallic parts. On the other hand, LUC-4180 polyurethane parts made with Protowax support material suffer from a different Christmastree process. For this materials combination, the problem is that the support material shrinks greatly even though the part material shrinks very little. Successive deposition and machining of part material produces very little distortion, but deposition of support material around part material causes major distortion. The support material cools from the outside, solidifying and bonding to polyurethane part material there first. The support material then shrinks inward as it cools, pulling the part material into a distorted geometry (Figure 4.7, sequence 2D). This effect can be seen in Figure 4.8, which shows a cross-section of a part after deposition and planing of support material. All lines in this picture should be straight and perpendicular, but have been distorted by the deposition of support material. When the next layer of part material is deposited and machined, a Christmas-tree effect will become quite visible at layer boundaries. While the Christmas-tree effect occurs via different mechanisms in polymeric and metallic materials, the overall part distortion process in both material systems is the same. As polymer part material shrinks after layered deposition, residual stresses accumulate in parts, leading to distortion either during the build process or after parts are removed from support material. Comparative linear shrinkage values for Polymer SDM materials are shown in Table 4.2; however, these numbers should only be considered rough estimates because of variations in testing techniques. The experience with LUC-4180 indicates that low-shrinkage support materials as well as low-shrinkage part materials are required to achieve optimal part surface finishes and geometrical accuracy. If such materials cannot be found, then more complicated compensation strategies must be developed to improve part quality. An alternative approach is to avoid depositing material in layers; this is the method followed in the Mold Shape Deposition Manufacturing variant process.

56

Figure 4.8. Distorted Cross-Section of LUC-4180 Polyurethane Part

Part Material Somos 2100 SLA Resin LUC-4180 Polyurethane TDT 205-3 Polyurethane 501/530 Epoxy

Linear Shrinkage (m/m) 0.01 0.002 0.0022 0.000546

Support Material Solder Mask Mix Protowax KC3230A Wax Protowax

Linear Shrinkage (m/m) Not Available 0.05 0.05-0.06 (est.) 0.05

Table 4.2. Linear Shrinkage of Material Combinations148

4.5. Conclusions There are several constraints on the selection of material combinations for use in Polymer Shape Deposition Manufacturing. These arise because the process relies upon successful deposition and shaping of a pair of complementary materials, one of which must then be removed without damage to the other material or its features. A difficult condition to satisfy is the requirement for chemical and physical compatibility during deposition processes; many potential material combinations failed to satisfy this requirement. Excellent interlayer bonding between layers of part material is required for strong parts; this requirement will be addressed in more detail in the following chapter. Subject to these constraints, several useful materials combinations for Polymer SDM were developed. These combinations are summarized in Table 4.3. The first combination is the original one developed at CMU; the remainder were developed during the present materials search at

57

Stanford. A variety of parts have been made to test the performance of these different material systems. Of the combinations developed, the most successful is TDT 205-3 polyurethane part material, KC3230A wax support material, and BIOACT 280 solvent for support material removal. This part material can be machined as soon as two hours after casting, even in thin layers, and has an interlayer strength four times higher than the earlier LUC-4180 part material. This material combination has been used for the production of small assembled mechanisms. The solder mask materials identified as early Polymer SDM support materials also allowed the development of the Mold SDM process. Characteristics: Part Material: Support Material: Removal: Characteristics: Part Material: Support Material: Removal: Characteristics: Part Material: Support Material: Removal: Characteristics: Part Material: Support Material: Removal: Characteristics: Part Material: Support Material: Removal: Baseline System LUC-4180 Polyurethane Protowax Partial Removal via Melting Fast-Curing, UV-Curable Materials Somos 2100 SLA Resin Solder Mask Mix Complete Removal via Water Complete Support Material Removability LUC-4180 Polyurethane Protowax Complete Removal via BIOACT 280 Fast-Curing, Good Interlayer Strength TDT 205-3 Polyurethane KC3230A Wax Complete Removal via BIOACT 280 High Interlayer Strength 501/530 Epoxy Protowax Complete Removal via BIOACT 280

Table 4.3. Useful Materials Combinations for Polymer SDM

58

CHAPTER FIVE MATERIAL STRENGTH IN POLYMER SDM 5.1. Polymer SDM Materials Several materials have been used for the production of polymer parts via Shape Deposition Manufacturing. Most of the part materials are castable thermoset polymers, and most of the support materials are machinable waxes. The part materials include both two-part castable polyurethane and two-part castable epoxy materials. While polyurethanes and epoxies are both thermoset polymers, there are several differences between the materials. Many different fast-curing room-temperature-cure polyurethane materials are available, while most fast-curing epoxy materials require elevated temperatures for curing. For a given cure time, polyurethanes will also have lower peak exotherms149. In SDM applications, roomtemperature curing and low peak exotherms are desirable since they reduce thermal distortion and prevent melting of support materials. In general, polyurethanes are more flexible while epoxies are more rigid. As adhesives, epoxies and rigid polyurethanes are recommended for structural applications using butt joints; they cannot support peel loads. Softer, more flexible adhesives (including some polyurethanes) are more often used for peel loads where their flexibility prevents a concentrated load from causing complete failure of a joint150. Epoxies usually have lower impact strengths than polyurethanes, although epoxy performance can be improved using fillers. Since Shape Deposition Manufacturing is a layered manufacturing process, the mechanical properties of SDM parts are dependent upon both the bulk material properties and the quality of the resulting interlayer bonds. The successive deposition of cast layers also produces anisotropic mechanical properties in Polymer SDM parts; these properties are investigated in the present study. Particular parameters of interest are the tensile strengths of parts made via Polymer SDM and the degree of anisotropy of these strengths.

59

Three part materials currently used for Polymer Shape Deposition Manufacturing were investigated in the present study. The first part material is LUC-4180 polyurethane from Adtech Plastic Systems (Charlotte, Michigan). This material is reasonably strong, has good impact properties, and can be machined about twelve hours after casting. It was initially chosen for its low water absorption for use in wearable computers for divers 151. However, later experiments indicated that this material suffers from poor interlayer bonding. The second material is TDT 205-3 polyurethane from Ciba Specialty Chemicals (East Lansing, Michigan). This material was chosen for its rapid cure speed; successive layers can be machined two hours after casting. Finally, the third material is 501/530 epoxy, also from Adtech. This material is a highly filled encapsulation resin which was chosen for its good adhesion to other materials; it was hoped that it would also adhere well to previous layers of the same material. This material requires a significantly longer cure time than either of the other materials; twenty-four hours' delay is required between casting and machining. Manufacturer's specifications for each of the materials are summarized in Table 5.1. The reported data is not necessarily strictly comparable; testing procedures vary by manufacturers, especially for parameters like shrinkage. Adtech Ciba LUC-4180 TDT 205-3 Polyurethane Polyurethane Ultimate Tensile Strength (MPa) 55 23 Elongation to Failure 15% 9% Hardness (Shore D Scale) 78-80 70 Mixed Viscosity (cPs) 800-900 80 Shrinkage (m/m) 0.002 0.0022 Reference 152 153 Cure Time before Machining 12 hours 2 hours Material Table 5.1. Manufacturers' Material Specifications Adtech 501/530 Epoxy 42 1% 86-89 3500 0.000546 154 24 hours

5.2. Metrics for Polymer SDM Material Strength A metric for material strength in Polymer Shape Deposition Manufacturing was needed in order to compare different materials. These comparisons must include the effect of the actual processing conditions in SDM, particularly the use of multiple layers to build parts progressively. This results in parts with interlayer interfaces which may be weaker than the bulk material. Therefore, a metric which can quantify strength differences between monolithic and multilayer parts was needed. Tensile strength, flexural strength, and impact strength were identified as potential metrics. 60

Tensile strength is very commonly used as a measure of the strength of a material. Tensile stresses are present within parts subjected to most practical loading conditions. One author calls tensile strength the "most important single indication of strength"155. The ASTM has developed a tensile testing standard for polymeric materials, ASTM D 648. Tensile strength tests of ductile materials are relatively accurate, but tensile tests of brittle materials can lead to underestimates of the failure stress of a specimen. The reason is that any misalignment between machine grips produces additional bending stresses which are combined with the intended tensile stresses. Also, local imperfections in such specimens lead to stress concentrations around them. Both of these effects result in the measurement of lower than actual failure stresses in brittle materials. Ductile materials are not affected this way since they can deform to match load and specimen axes and also locally yield to reduce the effect of stress concentrations156. Tensile strength measurements are commonly reported in the rapid prototyping community as indications of the quality of materials developed or parts produced. In an effort to formalize this use, a "Draft Specification for Tensile Testing of Rapid Prototyped Specimens" was developed by the E28.16 Rapid Prototyping Subcommittee of the ASTM E-28 Mechanical Testing Committee. However, this initial draft was not finalized into an actual ASTM standard, and it appears that this effort has been discontinued157. Flexural strength was also considered for use as a metric for Polymer SDM materials. Flexural strength is defined by ASTM standard D790-96a to be the stress on outer fibers of a specimen when it breaks in bending at a maximum fiber strain below 5%158. If a material can withstand strains above 5%, then a flexural strength can only be reported if the specimen yielded below this strain. In this case, the resulting strength is reported as a flexural yield strength. The equation for fiber stress recommended by the standard also depends on several assumptions, including both small strains and a completely linear relationship between stress and strain. Therefore, flexural strength measurements can be inaccurate for highly ductile materials. Flexure tests of brittle materials can be more accurate than tensile tests since there is less opportunity for misalignment of specimens159. Flexural strength testing is sometimes used industrially for property comparisons or quality control since the equipment is relatively simple160. This parameter is also recommended by authors affiliated with 3D Systems; they believe that most rapid prototyping parts are loaded in bending rather than uniaxial tension161. Impact strength is another possible metric. Some degree of impact strength is required for prototype parts to survive handling. However, impact strength is very sensitive to

61

specimen preparation techniques, and is not a true material property162. Regardless of this, impact strength is sometimes used to measure the effect of changes in material processing conditions on the strength of polymer materials. An example is its use in measuring the effect of injection molding knit-lines on the strength of thermoplastic materials. It is used because impact testing results reveal bigger strength differences between specimens with and without knit-lines than other tests show163. The differences are especially pronounced for ductile thermoplastic materials, whose impact strengths are greatly reduced by knit-lines while their tensile strengths are hardly affected. In this situation, impact strength tests are more sensitive than tensile tests. Both the more frequent use of tensile strength as a metric in the academic rapid prototyping community and better access to tensile testing equipment and fixtures led to the selection of tensile strength as a metric for Polymer SDM material strength. As the remainder of this chapter will describe, tensile strength proved to be a useful metric for Polymer SDM materials. Tensile strengths for monolithic specimens and multilayer specimens were measured for three materials. One material showed large differences between the strengths of the two types of specimens, while the other materials had relatively similar strengths for either type of specimen. Since such differences in process-dependent material behavior were quite apparent using tensile strength, it proved to be a useful metric in this application. Materials with higher multilayer tensile strengths seemed less prone to delamination during handling, indicating that this metric may correspond to actual part performance. No single metric can completely describe a polymer material, but the choice of tensile strength allowed material selection efforts to proceed efficiently. If materials are improved to the point where tensile strength can no longer discriminate between materials with better and worse interlayer strength, then more sensitive metrics may need to be developed.

5.3. Tensile Strength of Polymer SDM Materials After identifying a metric, the next step was to develop tests for the tensile strength of parts made via SDM with the three materials. Since SDM is a layered manufacturing process, test specimens constructed in different orientations may have different tensile strengths. This difference might be particularly pronounced when comparing test specimens built within a single layer to those which extend over multiple layers, since interlayer bonds may be weaker than the bulk material. ASTM D 638-96 specifies shapes and sizes for plastic tensile specimens, as well as conditioning and testing procedures 164 . This standard

62

describes methods of measuring material properties, and was used as a baseline for the present study. The recommended dimensions for ASTM Type I dogbone tensile specimens were used: a 13.0 mm x 57.0 mm narrow section of 3.2 mm thickness, with 20 mm-wide grip ends and a total length of 170 mm (as shown in Figure 5.1). 170 mm 57 mm 20 mm 13 mm

Figure 5.1. ASTM Type I Tensile Specimen Geometry Ideally, bulk and interface tensile strengths would be compared by building standard ASTM tensile specimens both perpendicular and parallel to the build direction. The specimens built perpendicular to the build direction would be planar and monolithic, built horizontally within a single layer of material. The specimens built parallel to the build direction would be vertical, composed of many layers of material. Both ideal specimens are shown in Figure 5.2.

Figure 5.2. Ideal Planar and Vertical Tensile Specimens Building the vertical tensile specimens, however, would require a very long time since single layers of one of these materials require 24-hour cure times. Therefore, a simplified dual-interface test specimen was developed which can be made in two cure cycles; this design is shown in Figure 5.3. This design is very similar to the tensile testing specimens used by Stokes to measure polymer weld strength in a variety of thermoplastic 63

polymers165. These specimens are constructed by casting a complete layer of material, machining a transverse trench through the layer, casting material into that trench, and finally machining the outlines of individual tensile specimens. The trench is made by slotting with a quarter-inch endmill, and the trench is located at the middle of the narrow sections of the specimens. The geometry of the dual-interface specimen is shown in Figure 5.4. The resulting specimens have two interlayer interfaces, rather than the n-1 interfaces present in an n-layer vertical tensile specimen. The tensile strengths of these dual-interface specimens will be expected to represent an upper bound on the tensile strengths of actual vertical n-layer tensile specimens made via SDM.

Monolithic Specimen

Dual-Interface Specimen

Figure 5.3. Monolithic and Simplified Dual-Interface Tensile Specimens

170 mm 57 mm 20 mm 13 mm

6.35 mm Figure 5.4. Dual-Interface Tensile Specimen Geometry To begin the testing process, material was cast into separate substrates for monolithic and dual-interface specimens. After the initial casting, the monolithic tensile specimens were cut out, and trenches were cut into the dual-interface specimens. Additional polymer material was then cast into the trenches and allowed to cure. The dual-interface specimens were cut out, and all specimens were aged for two weeks to ensure complete curing. As described in Appendix B.1, other tests performed one day after casting had resulted in inconsistent results, so this long delay was used. LUC-4180 specimens were machined

64

18-26 hours after casting, except for those of Series II (which were machined four days after casting). The TDT 205-3 polyurethane specimens were machined between 4-6 hours after casting, and the 501/530 epoxy specimens were machined between 24-46 hours after casting. As would be expected during part production, very little surface preparation was performed between the initial machining step and the second casting step. Chips and any residue from the machining process were cleared with compressed air, and the surfaces were wiped with dry Kimwipes; no cleaning solvents were used. Several different batches of LUC-4180 were used, including both material from newly opened containers and older material from containers which had been opened before. While this may result in variations in material properties, it is consistent with part production methods. The TDT 205-3 and 501/530 material used for tensile specimens came from newly opened containers. Since the objective was to determine the strengths of parts made through SDM, several specimen preparation steps normally recommended by the ASTM were skipped because they might overestimate actual strengths. The D 638 standard specifies smoothly polished surfaces, but the machined or replicated surfaces resulting from the SDM process were tested without any additional polishing. On dual-interface tensile specimens, there were slight thickness variations between the central trench area and the surrounding specimens. These artifacts were left undisturbed rather than attempting to remove them. The tensile tests for LUC-4180 were conducted in four rounds, split between General Motors and Stanford University. Series I of the LUC-4180 tests was conducted at GM, while the remaining series were conducted at Stanford University. All of the tests of TDT 205-3 and 501/530 were conducted at Stanford. The tests at GM were performed on an Instron machine with an extensometer, while the Stanford tests were performed on an Instron machine without an extensometer. The Series I LUC-4180 tests were conducted at a constant crosshead speed of 50 mm/min., while all other tests were conducted at 5.1 mm/min. The 5.1 mm/min. test speed was recommended by ASTM D 638. That standard specifies the use of the slowest speed (out of the choices 5, 50, and 500 mm/min.) which breaks specimens in more than thirty seconds but fewer than five minutes. At the 5.1 mm/min. speed, monolithic specimens of all materials except 501/530 epoxy required more than thirty seconds to break. If there were a slower recommended testing rate, it could have been used for the 501/530 epoxy specimens and for the dual-interface specimens of all materials. During the tests, the crosshead speed was constant, but the initial preload caused by tightening the specimens in the grips was not applied at the same speed. For most of the tests at Stanford, this preload seemed to result in a stress below 0.65 MPa. This pre-stress

65

is included in the tensile strengths reported, but the different initial crosshead speed may have affected the measured tensile strengths. This is especially true for the dual-interface LUC-4180 specimens which broke at stresses relatively close to the preload value. No specimen ever broke during the tightening of the grips, however. At Stanford, data from LUC-4180 Series II and III tests was collected using a chart recorder while Series IV, TDT 205-3, and 501/530 data was digitized and stored by a LabVIEW program. The LabVIEW program sampled load data at 14 Hz and averaged three such samples for each reported load value. Some additional background information about the tensile testing conditions can be found in sections B.2 - B.4 in the Appendices.

5.3.1. Test Results The tensile testing produced some interesting results for both monolithic and dual-interface specimens. The raw data is presented in Tables 5.2 and 5.3, and a summary of the test data is presented in Table 5.4. The monolithic LUC-4180 polyurethane specimens failed very differently from monolithic specimens of the other materials. LUC-4180 specimens yielded and formed necks before finally breaking; both TDT 205-3 polyurethane and 501/530 epoxy failed in a more brittle fashion. Typical plots of applied stress versus crosshead position for monolithic specimens are shown in Figure 5.5. LUC-4180 failed at the highest stresses, followed by 501/530, while TDT 205-3 failed at much lower stresses. Dual-interface specimens produced more surprising results. There were great difficulties manufacturing dual-interface LUC-4180 specimens; 8 out of 18 were destroyed during removal from their support material substrates. The LUC-4180 specimens always failed exactly at the interfaces, at very low stresses. The handling difficulties indicate that the average tensile strength would have been even lower if all LUC-4180 specimens had been successfully tested. The TDT 205-3 polyurethane specimens, on the other hand, always failed away from the interfaces, and failed at nearly the same stresses as the monolithic specimens had. The 501/530 epoxy specimens usually failed very close to or at the interfaces, and also failed at a large fraction of their monolithic failure stresses. Typical plots of applied stress versus crosshead position for the dual-interface specimens are shown in Figure 5.6. Overall, the 501/530 epoxy dual-interface specimens failed at the highest stresses, while the TDT 205-3 specimens failed at half those stresses, and LUC-

66

4180 specimens failed below one eighth of the 501/530 values. A plot of all test data is shown in Figure 5.7. LUC4180 I 56.3 57.3 57.5 56.8 54.2 LUC4180 II 58.0 57.5 58.2 58.0 57.3 56.7 LUC4180 III 60.1 59.7 59.5 59.0 59.0 59.3 TDT 205-3 17.0 19.1 20.0 23.9 20.7 31.3 501/ 530 48.7 49.3 49.2 47.5 48.4 48.3

# 1 2 3 4 5 6

Average 58 22 49 LUC-4180 specimens: engineering stress at yield (MPa) All other specimens: engineering stress at break (MPa) LUC-4180 average strength includes all three series of data "-" means no specimen made Table 5.2. Strength Data - Monolithic Tensile Specimens

# 1 2 3 4 5 6 Average

LUC4180 I D D D 6.4 8.2 8.7

LUC4180 II 2.5 2.1 2.9 3.0 3.6 3.9

LUC4180 III D 4.2 D D D D

TDT 205-3 19.2 20.1 19.4 20.2 20.0 19.6

501/ 530 36.9 37.1 41.1 42.4 47.6 34.1

4.6 20 40 Engineering stress at break (MPa) LUC-4180 average strength includes all three series of data "D" means destroyed in handling Table 5.3. Strength Data - Dual-Interface Tensile Specimens

LUC-4180 Monolithic Specimens Dual-Interface Specimens 58 MPa 4.6 MPa

TDT 205-3 22 MPa 20 MPa

501/530 49 MPa 40 MPa

Monolithic LUC-4180 specimens: engineering stress at yield All other specimens: engineering stress at break Table 5.4. Average Tensile Strengths

67

60 Engineering Stress (MPa) 50 40 30 20 10 LUC-4180 0 Crosshead Position TDT 205-3 501/530

Figure 5.5. Typical Tensile Tests of Monolithic Specimens166

60 Engineering Stress (MPa) 50 40 30 20 10 LUC-4180 0 Crosshead Position TDT 205-3 501/530

Figure 5.6. Typical Tensile Tests of Dual-Interface Specimens167

68

70.0 Tensile Strength (MPa) 60.0 50.0 40.0 30.0 20.0 10.0 0.0

LUC-4180 501/530

TDT 205-3

Circles = Monolithic Specimens, Triangles = Dual-Interface Specimens

Figure 5.7. Tensile Strength Summary - Comparing Materials

5.3.2. Discussion The large differences between tensile strengths of monolithic LUC-4180 specimens and dual-interface LUC-4180 specimens indicate that tensile strength is a sufficiently sensitive metric to determine the effects of SDM processing on materials. The plots of typical monolithic specimens show that LUC-4180 samples have much more area under their load versus crosshead position plots. This greater area is evidence of significantly greater toughness, which provides greater survivability of impact or shock loads in service168. While monolithic LUC-4180 polyurethane is reasonably strong, dualinterface specimens of this material failed at very low stresses during tensile testing. This demonstrates that high interlayer strength does not necessarily accompany high bulk strength. The LUC-4180 experiments also show that poor interlayer bonding can negate the advantages of good bulk toughness properties. Some layered LUC-4180 parts have suffered from delamination during handling, indicating that low interlayer strengths also occur during actual use of the material. These results indicate that LUC-4180 is only useful for monolithic parts, such as those produced in the Mold SDM process.

69

Dual-interface specimens of both TDT 205-3 polyurethane and 501/530 epoxy failed at much higher tensile stresses - these two materials are better suited for producing layered parts. While TDT 205-3 has a lower interlayer strength than 501/530, it does have some advantages over that material for use in the SDM process. It has very low viscosity and good air-release properties, both of which help prevent voids in interior corners of layers. More importantly, it can be machined over ten times sooner than 501/530, in two hours rather than twenty-four. For these reasons, TDT 205-3 is currently the polymer material of choice except for situations where the strongest possible parts are needed. These test results demonstrate that there can be a very substantial anisotropic variation of mechanical properties in Polymer Shape Deposition Manufacturing. A "relative interface strength" can be defined as the ratio of interlayer tensile strength to bulk tensile strength169; the relative interface strengths of the materials tested are reported in Table 5.5. Relative interface strength is a measure of strength anisotropy since it should be roughly proportional to the difference in strengths between tensile tests parallel to and perpendicular to the build direction. For LUC-4180, tensile strength in the build direction may be twelve times lower than tensile strength within a single layer. LUC-4180 Relative Interface Strength 8% TDT 205-3 90 % 501/530 82 %

Table 5.5. Relative Interface Strengths Both bulk strength and relative interface strength must be considered when selecting polymer part materials for SDM. However, neither a relative interface strength nor an absolute interlayer strength is normally tabulated by manufacturers since few applications require materials to adhere well when cast over machined surfaces of the same material. Maximizing interlayer strength results in the strongest objects although it can produce parts with greater anisotropy. An example is that 501/530 forms stronger interlayer interfaces than TDT 205-3, even though its tensile strength is more anisotropic. As a comparison between LUC-4180 and TDT 205-3 indicates, relative interface strengths can vary significantly even within a class of materials. Both of these materials are twopart, castable, MDI-based polyurethane resins, but their relative interface strengths are 8% and 90%, respectively. The reasons for such differences are not well understood although literature in the fields of adhesion and polymer interfaces suggests several reasons for the

70

differences in relative interface strengths and absolute interlayer strengths among the three materials tested.

5.3.2.1. Theories of Adhesion There is a large literature in the area of adhesives, but there are no unified theories of adhesion. A handful of theories have been proposed as "universal," but additional research has suggested that different theories are required to explain various experimental observations. One of the problems which limits the development of a unified theory of adhesion is that most experimental techniques used for evaluating the strength of adhesive joints are "not well suited to theoretical analysis"170. The following descriptions of theories of adhesion are taken from articles by Kinloch171, Fourche172, and Allen173: The "Mechanical" Adhesion model was one of the earliest theories of adhesion, originally proposed by McBain174. Mechanical interlocking between adherends and adhesives was thought to be the basis of all adhesive processes. Mechanical interlocking is no longer considered a valid general theory for adhesion, especially since smooth-surfaced adherends can be successfully attached to one another. On the macroscopic scale, mechanical interlocking is currently only thought to explain adhesion to porous or fibrous materials (wood, paper, and textiles). On a much smaller length scale, mechanical interlocking is thought to explain the improved bonding of anodized aluminum as its tiny pores allow adhesives to penetrate the surface175. The "Electrical" or "Electronic" model of adhesion was developed by Deryaguin and coworkers176. It describes the attachment between two materials as if they were plates in a capacitor, arguing that electrostatic forces are a cause of high joint strength, rather than an effect. This model only applies to dissimilar materials, and is thought to be useful mainly in the context of adhesion between metals and polymers. While this theory is not very popular, it has been used to explain some current work177. The "Diffusion" model, developed by Voyutskii178, argues that adhesion is the result of interdiffusion between the two substrates. The theory was developed to explain polymer adhesion, and only applies to mutually soluble materials at a high enough temperature for molecular mobility. This theory has been modified to include the idea of interpenetration of polymer chains, with diffusion occurring mainly at the ends of the chains. Diffusion is

71

important for "autohesion" (self-adhesion of solids) of elastomers and solvent bonding of amorphous plastics, but is not important for highly crosslinked materials, materials of differing solubility, materials below their glass-transition temperatures179, or crystalline materials. Bikerman's Weak Boundary Layer model claims that adhesive joints never fail exactly at interfaces, but only within weak boundary layers inside either the adhesive or the adherend180. Bikerman identified seven kinds of weak boundary layers, including trapped air, contaminants, and reaction products evolved at material interfaces 181. This theory caused a great deal of controversy, with several authors attempting to discredit it completely. Current research suggests that adhesive joints can fail exactly at interfaces, but that weak "interphases" within adherends can limit adhesive strength. This theory also emphasizes the negative effects of surface contamination on adhesive strength. The "Adsorption" model is the "most generally accepted theory"182, arguing that adhesion results from intermolecular forces between materials in intimate contact. This theory was initially developed by Sharpe and Schonhorn183. This theory also includes elements that some authors refer to a separate theories: the Wetting model, the Rheological model, and the Chemical Adhesion model. The Wetting model claims that good adhesion requires complete wetting of a liquid on a surface to allow the required intimate contact for surface forces to act. A liquid can spread on a surface only when its surface tension is equal to or lower than the critical surface tension of the surface. In general, clean metals, oxides, and ceramics are high-energy surfaces, with high critical surface tension values; this means that liquid adhesives can completely wet such surfaces. On the other hand, most polymer surfaces are low energy, with low critical surface tensions so that adhesives may have difficulty completely wetting such surfaces. The adhesive strengths resulted from surface forces predicted by the Wetting theory are much smaller than those experimentally observed. This theory has been improved upon by the Rheological model which adds assumptions about the way substrates separate based upon viscoelastic bulk properties of the adhesive involved. Those rheological assumptions allow better estimation of the peel strengths of real adhesive joints. The Chemical Adhesion model states that adhesion can be greatly improved by chemical bonds formed at interfaces between adhesives and substrates. Some aspects of this theory have been verified for glass-adhesive and metal-polymer bonds.

72

5.3.2.2. Self-Adhesion of Crosslinked Polymers in SDM In Polymer Shape Deposition Manufacturing, the adhesive process of interest is selfadhesion of crosslinked polymers. The principal questions are how strong a joint is possible between two successively cast and machined layers of a given thermoset polymer, and what factors affect the strength of this joint. Questions like this have received relatively little interest from the adhesives and polymer science communities. Somewhat related work includes self-adhesion of thermoplastic polymers, bonding partially crosslinked elastomers, adhesive bonding techniques, and techniques for casting thick thermoset parts. None of these processes involve depositing material over machined layers of similar material. Very recently, some more closely related work on thermoset adhesion to ground thermoset surfaces has been published. In the following section, these related areas are described, and the issues important in the particular case of Polymer SDM are explored. Much work has been done to investigate the self-adhesion of thermoplastic polymer materials. This process is important in injection molding and welding of thermoplastic materials. Self-adhesion is important in injection molding because of the presence of knitlines where injection wavefronts join together. If molds and processing conditions are not optimized, knit-lines can cause weaknesses in molded parts. Self-adhesion of thermoplastics is also important for joining processes for thermoplastic polymers; such joining methods have been reviewed in the literature184. Thermoplastic polymers can be welded through the use of heat, friction, or ultrasonic energy; the additional energy causes remelting and resolidification at interfaces. Researchers have found that optimal interlayer strength is only achieved when the average polymer chain crosses a welded interface sixteen times185. However, the literature on thermoplastic self-adhesion does not apply to the present situation since Polymer SDM involves thermoset polymers. Crosslinked thermoset polymers decompose upon heating instead of melting and reforming, so welding them is impossible. Another related process is the bonding of partially crosslinked materials to one another. One such process is the bonding of two identical partially crosslinked elastomeric substrates to one another; this subject has received attention because of its application to bonding rubber materials186. However, those results do not apply to the Polymer SDM process since bonding in SDM occurs under quite different conditions. First, any previous layers in SDM are cured and crosslinked before the next layers are applied; nearly complete curing and crosslinking are required for machinability in SDM. Second, the SDM casting

73

processes occur at room temperature (well below the glass-transition temperatures of the materials) while the elastomeric bonding processes are performed well above the glasstransition temperatures of those materials. On the other hand, the adhesives literature does provide practical surface preparation and adhesive selection techniques for the bonding of thermoset polymer materials like epoxy and polyurethane. To bond such substrates, a three-step process is recommended. First, the surface is wiped with a degreasing solvent to remove any contaminants. Second, the surface is physically abraded, using sandpaper or abrasive blasting. Third, the surface is again cleaned with a degreasing solvent. Different solvents are recommended for different materials. For elastomeric (rubbery) polyurethane substrates, methanol is recommended. For hard materials like thermoset epoxies or polyurethanes, either acetone or methyl ethyl ketone is recommended187. After surface preparations like this are accomplished, the use of epoxy, thermoset acrylic, or polyurethane adhesives are recommended to bond such thermoset substrates188. Manufacturers' literature for castable thermoset materials also provides some guidance for ensuring adhesion between successive immediate castings. Such situations arise when sections too thick to be cast in single layers are desired. A single layer would be "too thick" if gellation would begin before the required amount of material could be mixed and dispensed. To build such thick objects, a process called stepcasting or stage pouring is used189; successive thin layers are cast before underlying layers begin to gel190. This normally ensures that good chemical bonding will occur between the successive layers. While this practice is recommended by the literature, it cannot be integrated into the SDM process. The reason is that SDM requires each previous layer to cure almost completely before it is machined, otherwise the resulting surface quality and shape definition will be too poor. Later SDM layers must always be cast long after previous layers have gelled, so the literature advice about stepcasting cannot be followed. The only published work describing the casting of thermoset polymers over machined surfaces of thermoset polymers was conducted recently by authors Raghavan and Wool. Some results were presented at a conference191, and the journal publication of the work is in progress192. The authors made test specimens of crosslinked vinyl ester, ground the surfaces, cast fresh vinyl ester monomer over the specimens, and allowed it to solidify and crosslink. They measured strain energy release rates for compact tensile specimens made this way and through monolithic casting. They expected equally high fracture strengths

74

from both the ground and monolithic specimens, but measured a relative value of only 43% for the ground specimens. Through some additional experiments, the authors concluded that the most important factor resulting in reduced strength was poor interpenetration of the newly cast material. That is, the new monomer needed to diffuse into the old crosslinked structure and then crosslink, so that the new polymer chains would cross and recross the interface, providing strength. Since the required interpenetration did not occur, the interfaces were weaker than expected. Other rejected explanations included the development of an interphase, poor wettability, high crosslink density of the surface, and gel-time issues. The authors were able to improve the strength of ground and recast specimens by performing chemical treatments of the ground surfaces before casting new material, but the resulting specimens were still weaker than monolithically cast specimens. However, there were issues which may have confounded the results of the experiments 193. Other differences between this work and the present Shape Deposition Manufacturing process include the use of grinding rather than machining and the casting of short-chain monomers rather than long-chain isocyanate and polyol oligomers. As a result, the conclusion that interpenetration and connectivity are the most important requirements may or may not apply to the SDM process. Since no previous work has addressed the self-adhesion of crosslinked polyurethane and epoxy polymers as it occurs in SDM, the basic theories of adhesion will be examined for their relevance to this process. The Mechanical Adhesion model does not directly explain differences in thermoset materials' self-adhesion because their surfaces are not fibrous and their interface surfaces have relatively consistent machined surface finishes. The Electronic model applies only to adhesion between dissimilar materials, so it should not be applied to adhesion between identical polymeric materials. Likewise, the Diffusion model does not apply to this situation because it does not explain the adhesion of crosslinked materials whose molecular chains are pinned to their surfaces and incapable of diffusing across interfaces. The Weak Boundary Layer model does apply to this situation, in that weak interfaces must be prevented in order for strong adhesive bonds to develop. Such weak interfaces could be caused by trapped air, contaminants, or reaction products. The Adsorption model also applies to this situation; its requirement for good adhesion is that an adhesive can wet an adherend completely. Therefore, according to these theories of adhesion, strong bonds between successive layers of castable thermoset materials require an absence of contaminants or trapped air, complete wetting of interface surfaces, and strong intermolecular attraction between the layers.

75

This raises the question of how well adhesives wet crosslinked surfaces. Earlier review articles concluded that crosslinking had an inconsistent relationship to adhesive bond strength and wetting. Experiments showed that crosslinking might or might not improve wetting behavior and might either improve, have no effect, or even worsen bondability of substrates194. These contradictory results seemed to imply that no sure conclusions can be drawn about the wetting of crosslinked surfaces. More recently, however, a phenomenon called "autophobicity" has been found to result when polymers are placed in contact with identical, but crosslinked, surfaces. Autophobicity is a property of certain liquids; it means that they cannot completely wet surfaces of similar chemical compositions. One experimental result was a transition from complete wetting on surfaces without crosslinks, to incomplete wetting after increasing crosslinking, to complete wetting of completely crosslinked surfaces195. These experiments were conducted on very smooth substrates, and the final wetting was thought to result from slight increases in surface roughness caused by the complete crosslinking 196 . These experiments were performed with thermoplastic polymers, and may not be representative of the behaviors of particular thermoset polymers. Also, the results might be somewhat different when the crosslinked substrates are macroscopically rough rather than almost perfectly smooth. In any case, surface crosslinking of thermoset materials may affect the wettability of such surfaces, but the adhesion literature cannot yet describe how. Another aspect to adhesion between successive layers in the SDM process is that all of the interface surfaces are machined. Adhesion to machined polymer surfaces has received no attention in the literature; the ground-surface bonding work by Wool and Raghavan is the only related research. A machined surface would have intermediate surface roughness, being rougher than a surface which was replicated from a polished mold but smoother than a surface abraded by sandpaper or gritblasting. The microscopic effects of the machining process on polymer chain ends at the interface surface are unknown. Additionally, the machining process in SDM may introduce contaminants or oils onto the machined surfaces. Chips from the machining process are one source of contamination, as are debris from previous machining operations which might involve other materials. The high-pressure air used to cool endmills and parts during machining may also contaminate the surface with oil from the shop-air system. Between machining and casting, visible debris are blown off the bonding surfaces, but no solvent wipe operation is conducted during SDM. Therefore, while the machined surface finish should be better than a perfectly smooth one, neither the recommended surface abrasion or solvent wiping practices recommended for optimal adhesion are being performed in SDM.

76

In summary, the strength of interlayer interfaces in Polymer SDM is limited by the adhesion of thermoset polymers to machined surfaces of similar crosslinked polymers. While literature from the adhesion and polymer interfaces community can describe some of the phenomena involved, there is no complete theory which can predict how strong such bonds will be for a given material. As the present test results indicate, there can be substantial differences in relative interface strengths between apparently similar materials. The two polyurethane materials, LUC-4180 and TDT 205-3, provide examples of such differences; possible explanations for this situation are described in the next section.

5.3.2.3. Self-Adhesion Differences between Polyurethanes The previous discussion of adhesion and interfacial phenomena suggests that many competing processes affect the adhesion of successive layers to one another in SDM. A comparison of the tensile strengths of monolithic and multilayer specimens of LUC-4180 polyurethane and TDT 205-3 polyurethane is somewhat surprising. While LUC-4180 has reasonable tensile strength when it is cast monolithically, its double-interface tensile strength is much lower. On the other hand, TDT 205-3 has a lower monolithic tensile strength, but its double-interface tensile strength is nearly the same as its monolithic value. Since both of these materials are polyurethanes, this difference in relative interface strengths is unexpected. A basic question is whether this difference is due to the presence of qualitatively different types of internal bonds or whether the same bonds are involved but some other phenomena limits the strength of the bonds in LUC-4180. The following section addresses this issue. Several types of intermolecular and interatomic bonds are involved in the chemical structure of bulk polyurethanes; polyurethane elastomer chemistry is well described in the literature197. Covalent chemical bonds connect the individual atoms of a polymer chain. Linear polyurethane polymer chains are block copolymers composed of alternating soft and hard segments. The soft segments are made of a polyol material, while the hard segments are made of isocyanates and chain extenders. These segments have different polarities, causing the soft and hard segments to associate together in soft and hard domains. This separation results in the creation of a two-phase structure within the bulk material. Secondary hydrogen bonds can be formed between different hard segments within the hard domains. The hard domains throughout the polyurethane bulk material act as physical crosslinks, restricting the motion of polymer chains at room temperature. There can be

77

high intersegment forces between polymer chains in the hard domains, resulting in the good physical properties of such materials: high modulus and strength198 . While individual covalent bonds are much stronger than individual hydrogen bonds, the greater prevalence of hydrogen bonds makes them a significant source of strength199 . The preceding description of bonds in linear polyurethane elastomers also applies to lightly chemically crosslinked polyurethanes. These materials develop when multifunctional isocyanates or polyols are used, resulting in the formation of a covalently bonded network of polymer chains. As the crosslink density increases, the tighter covalently bonded network prevents the hard and soft segments from arranging themselves for optimal intersegment forces. This limits the strengthening effect from the secondary forces within domains, leading to softer and weaker materials200 . As the degree of crosslinking increases, the materials pass through a minimum in strength, after which the additional covalent bonds make up for the lack of secondary bonding between segments201. At sufficiently high levels of crosslinking, phase separation of the segments into domains becomes impossible, and the polyurethanes have only a single phase202. The previously described bonds determine the bulk properties of polyurethanes; interface properties are determined by the bonds which occur between successive layers of cast material. While bonds in the internal structure of polyurethanes are relatively well understood, there is less of a consensus about the bonds expected to form at interfaces. An area of agreement is that no covalent bonding to previously cured surfaces would be expected203. The reason is that all reactive sites on polymer chains of cured polyurethane are already occupied. However, machining such surfaces might damage the polymer chains, breaking them and possibly making new reactive sites available. Hydrogen bonding across interlayer interfaces is also possible, but experts disagree about whether it is likely204. Besides covalent and hydrogen bonding, other secondary forces are expected to act across interfaces, as in other adhesive phenomena. Dispersion forces (caused by a type of van der Waals bonds) resisting the separation of planar surfaces could theoretically produce interlayer strengths in excess of 100 MPa205. Such an interlayer strength would be significantly higher than any strengths measured in the present study. Overall, experts speculate that the differences in interlayer strength between LUC-4180 and TDT 205-3 are not caused by different types of bonds being present in the two materials206. If this is correct, then other processes must interfere with the creation of strong bonds between successive layers of LUC-4180. This could occur through imperfect wetting or through failure of newly cast material to interpenetrate and establish connectivity

78

with previously cast and machined surfaces. Even though TDT 205-3 has a high relative interface strength, the strength of LUC-4180 interfaces could be limited by these processes since the materials are somewhat different. Although they are both castable polyurethanes, the LUC-4180 and TDT 205-3 materials probably use different polyols and isocyanate prepolymers, and each material also contains a collection of other additives. The two materials' chemistries are relatively complicated and may be quite different. According to Ciba, the TDT 205-3 polyurethane has at least eight different reactive chemicals on the B-side of the material207. The LUC-4180 and TDT 205-3 materials are both two-part castable polyurethanes, but they do have different properties, as shown in Table 5.1. The LUC-4180 material absorbs much more energy than TDT 205-3 during tensile testing, and is ductile enough to neck down before failure. The TDT 205-3 fails in a much more brittle manner. These different properties result from differences in their chemical structures. While both materials are crosslinked thermoset polymers based on MDI polyurethane, their degree of crosslinking and other internal structural properties may be different. These differences in chemical structure could affect their interlayer strengths by affecting wetting or interpenetration of polymer chains. Another difference between the two materials is the different additives involved, including plasticizers, antifoam agents, and fillers. TDT 205-3 includes different aromatic naphtha materials as well as benzene; these components are included for viscosity and exotherm control and also act as plasticizers208. The presence of these plasticizers sometimes becomes obvious during the use of TDT 205-3, especially when the two components are not perfectly mixed. Particularly in that case, a thin coat of liquid is exuded from cured surfaces of TDT 205-3. While this could produce a weak boundary layer209, the top surfaces of SDM layers are almost always machined away; this should prevent such problems. Besides the plasticizers, TDT 205-3 also includes an antifoam agent intended to reduce the surface tension of the mixed resin to allow air bubbles to be released more easily 210 . In comparison, the LUC-4180 material has a somewhat simpler additive package, apparently consisting of a plasticizer (tricresyl phosphate)211,212 and an antifoam agent213. The use of different antifoam additives may affect the air-release properties of the two materials, which are quite different. Trapped air is much more easily released from within TDT 205-3 than from LUC-4180. This results from the lower viscosity and surface tension of TDT 205-3 compared to LUC-4180, and should prevent weak interfaces resulting from air trapped at interfaces.

79

Besides the differences in plasticizers and antifoam agents, the two polyurethane materials also employ different fillers. The TDT 205-3 material has a combination of fillers molecular sieves and talc. The molecular sieves are "moisture scavengers" which are designed to capture water molecules in 3-angstrom holes; the talc component is included to prevent the molecular sieves from packing together and solidifying during storage. Molecular sieves represent about 2.5-3.5% of the polyurethane, and talc is less than 1%. Both fillers are contained in the polyol side. The LUC-4180 polyurethane has up to 2.5% by weight of sodium alumino silicate filler214, in the polyol side of the mixture. This material is a desiccant, intended to prevent moisture contamination of stored polyurethane components215. The differences in fillers may account for the difference in viscosities between the materials. Overall, the differences in additives between the LUC-4180 and TDT 205-3 materials could have a variety of effects on the interface properties of the materials, including the wetting properties of the liquids. One of the effects of plasticizers in thermoset polymers is to improve wetting216. There are often intentionally used to improve wetting of fillers or lamination layers; they may also improve the wetting of previous layers in SDM. While antifoam agents are intended to reduce surface tension to promote air release, such changes in surface tension can also be expected to affect wetting behavior. An additional difference between the two materials is their mixed viscosities - TDT 205-3 is significantly thinner than LUC-4180 (80 cPs vs. 800-900 cPs). According to Kinloch, an ideal adhesive would have a very low viscosity to aid wetting217; TDT 205-3 more nearly meets this criteria. Compounding this viscosity difference is the longer vacuum degassing cycle required for LUC-4180. This material must be mixed, degassed, cast onto the part, and then degassed again. As the process proceeds through the degassing steps, the material's viscosity increases continuously. By the time it is cast onto the part, it is even thicker than it was when originally mixed; this may further hinder surface wetting. As a result, the TDT 205-3 material may be able to wet cured polymer surfaces better than LUC-4180. The various differences in chemical structure and additives between LUC-4180 and TDT 205-3 may cause differences in the wetting and interpenetration behavior of the newly cast liquid materials. This may result in the poor interlayer strength of LUC-4180 even though the interlayer strength of the somewhat similar TDT 205-3 material is reasonable. However, this speculation has not yet been confirmed by testing. Present experiments cannot even rule out the possibility that, contrary to expert opinion, there are different types of bonding between successive layers of LUC-4180 compared to TDT 205-3. However,

80

the value of determining exactly why LUC-4180 does not bond well to itself is questionable. Carnegie Mellon researchers' negative experiences with formulating their own polyurethane materials indicate that choosing different commercial products which satisfy process requirements is more efficient than trying to fix products which do not meet such requirements. LUC-4180 has some good properties, such as medium bulk strength and high elongation to failure, but very poor interlayer strength. TDT 205-3 has relatively low bulk strength, lower elongation to failure, but excellent interlayer strength. An optimal polyurethane material would have both high bulk strength and the excellent interlayer strength of TDT 205-3. Trials of potential materials meeting this specification are probably more useful than attempting to identify the reason for LUC-4180's poor interlayer strength or attempting to improve it.

5.3.2.4. Self-Adhesion of Polyurethane and Epoxy Materials The three materials evaluated for interlayer bonding in Shape Deposition Manufacturing are LUC-4180 polyurethane, TDT 205-3 polyurethane, and 501/530 epoxy. As discussed above, there are significant differences between the relative interface strengths of LUC4180 and TDT 205-3. Even though the bulk strength of TDT 205-3 is lower than that of LUC-4180, the interlayer strength is much higher. A similar observation was made for LUC-4180 and the 501/530 epoxy material; even though the monolithic strength was lower, the better bonding between successive layers of 501/530 ensures a higher interlayer strength than that of LUC-4180. The same properties which affect interlayer bonding of different polyurethane materials also account for the adhesion differences between polyurethane and epoxy materials - chemical structure, its effects on surface forces, and the wetting and interpenetration behavior of the newly cast materials. Of the three materials examined, the 501/530 epoxy has the highest interlayer strength, in spite of the higher relative interface strength of TDT 205-3.

5.4. Effects of Gritblasting on Interlayer Strength As mentioned previously, the adhesion literature recommends the use of surface roughening processes to assure the formation of high-quality adhesive bonds. In an effort to improve the interlayer strength of LUC-4180 polyurethane, gritblasted dual-interface specimens were manufactured and tested. The sides of their interface trenches were

81

gritblasted before material was cast into those trenches. The gritblasting process both roughened and dirtied the interface surfaces, as well as dirtying other exposed polyurethane surfaces. Compressed air was used to blow loose particles out of the trench, but no liquid cleaners were used within the interface trenches. These side interface surfaces were noticeably grayer than freshly machined polyurethane surfaces. The top surface of the original layer of polyurethane was cleaned with a water-soluble cleaner218 to remove particulate debris. This cleaner was rinsed away with water, and the specimens were allowed to air dry before more polyurethane was cast. The gritblasting also eroded the support material surface at the bottom of the trenches, increasing the depth of the trench into which more polyurethane was cast. The resulting specimens had a little bit of extra material on their bottom surfaces between the two interfaces. This material was left on the tensile specimens; no attempts to remove it were made. It was relatively thin, and should not have contributed to the strengths of the interfaces which seem to limit the overall strength of the specimens. After the specimens were completed, they were aged for two weeks prior to testing, to ensure the polyurethane was completely cured. During this time, the specimens warped slightly; they became concave, with bends located at the each of the two interfaces. This had not happened to any other series of specimens, and may have affected the tested strengths since the specimens were not completely parallel to the loading axis. After aging was complete, the gritblasted dual-interface specimens were tensile tested at Stanford, using a test speed of 5.1 mm/min.

5.4.1. Test Results The gritblasted LUC-4180 dual-interface specimens were substantially stronger than the normal LUC-4180 dual-interface specimens had been. Table 5.6 lists the strengths of the specimens, and Table 5.7 shows a comparison between average tensile strengths of the gritblasted dual-interface specimens and normal ones. Figure 5.8 presents the data in a graphical format.

82

Specimen # Strength 1 19.7 2 19.3 3 18.3 4 15.7 5 13.2 6 D Average 17 Engineering stress at break, MPa "D" means destroyed in handling Table 5.6. Tensile Strength of Gritblasted Specimens

Normal LUC-4180

Gritblasted LUC-4180

Dual-Interface Specimens 4.6 17 Average engineering stress at break, MPa Table 5.7. Tensile Strengths - Normal and Gritblasted Specimens

20.0 18.0 Tensile Strength (MPa) 16.0 14.0 12.0 10.0 8.0 6.0 4.0 2.0 0.0 Squares = Normal Specimens, Triangles = Gritblasted Specimens

Figure 5.8. LUC-4180 Tensile Strength - Normal and Gritblasted Specimens The gritblasted specimens were over three times stronger than other specimens, indicating that gritblasting can be used to increase interlayer strengths in the Polymer SDM process.

83

However, the process complexity which this involves could probably not be justified. The TDT 205-3 dual-interface specimens had higher interlayer strengths than even the gritblasted LUC-4180 dual-interface specimens, and TDT 205-3 allows higher process throughput. This comparison indicates that materials with superior properties are more useful for SDM than additional processing steps which can improve the performance of inferior materials.

5.5. Effects of the Support Material Removal Process on Interlayer Strength The final stage in prototype production via SDM is the removal of support material. As explained earlier, the wax support material in the polyurethane part / wax support material system is removed via heating and melting or the use of heated solvents. Following any intermediate heating, the final step is dissolving any remaining wax or residue in heated BIOACT 280 solvent; damage to parts or their material properties must be prevented during this process. In order to determine the effects of the support material removal process, two series of tensile tests were performed. One series of tensile tests examined monolithic LUC-4180 tensile specimens which were soaked in heated BIOACT for a period of time. The objective was to determine whether the solvent or the heat involved in the process had any effect on the bulk material properties of the polyurethane. The second series of tensile tests examined the soaking of dual-interface LUC-4180 specimens to see what additional effects might be expected at interlayer interfaces. Both sets of specimens were made the same way as those for comparing materials had been prepared. The specimens were then soaked in BIOACT on the day the dual-interface specimens were cut out (the day after the monolithic specimens were cut out). Two liters of BIOACT 280 were heated to about 76C on a hotplate. The specimens were placed vertically in this bath; there was sufficient depth to immerse them almost completely. Only the last half centimeter of the grip sections of the specimens remained above the surface of the bath. As soon as the specimens were placed in the bath, they immediately went limp and bent under their own weight since the bath temperature was significantly above the glass-transition temperature (which is between 60-66C 219) of the polyurethane material. The specimens were left in the solvent bath for fifteen minutes, as the temperature slowly increased to about 84C. They were removed from the bath, flattened out, and allowed to

84

dry and cool on a tabletop. The BIOACT residue was wiped clean, the specimens were rinsed in water and allowed to dry overnight, and then the specimens were placed in sealed plastic bags and aged for two weeks. After that standard aging process, the specimens were tensile tested at Stanford, using a test speed of 5.1 mm/min.

5.5.1. Test Results The tensile strengths of the LUC-4180 specimens soaked in BIOACT are reported in Table 5.8. The behavior of these specimens was very similar to those of normal LUC-4180 specimens. All of the monolithic specimens necked down before breaking, while the dualinterface specimens consistently broke at the existing interfaces in a brittle manner. Table 5.9 compares the average tensile strengths of the specimens soaked in BIOACT 280 with those from the earlier, normal tests; Figure 5.9 illustrates this comparison graphically. The monolithic specimens soaked in BIOACT were slightly weaker than the normal monolithic specimens while the dual-interface specimens soaked in BIOACT were significantly stronger than the normal dual-interface specimens. Monolithic Dual-Interface Specimens Specimens 1 54.9 7.4 2 54.4 12.9 3 54.6 12.9 4 54.1 12.6 5 53.3 11.6 6 54.1 7.8 Average 54 11 Monolithic specimens: engineering stress at yield, MPa Dual-Interface specimens: engineering stress at break, MPa Table 5.8. Tensile Strength of Specimens Soaked in BIOACT 280 Specimen #

85

Normal LUC-4180 Monolithic Specimens 58

LUC-4180 Soaked in BIOACT 280 54

Dual-Interface Specimens 4.6 11 Monolithic Specimens: average engineering stress at yield, MPa Dual-Interface Specimens: average engineering stress at break, MPa Table 5.9. Tensile Strengths - Normal and BIOACT-Soaked Specimens

70.0 Tensile Strength (MPa) 60.0 50.0 40.0 30.0 20.0 10.0 0.0

Monolithic Specimens

Dual-Interface Specimens

Squares = Normal Specimens, Triangles = BIOACT-Soaked Specimens

Figure 5.9. Tensile Strength - Normal and BIOACT-Soaked Specimens

5.5.2. Discussion Fractured polymer materials are known to heal above their glass-transition temperatures cracks close, and the original fracture strength of the material is gradually restored. This is called crack healing. The processes involved in crack healing are usually restoration of molecular entanglements and secondary bonding between molecular chains220. These processes are affected by molecular mobility, so they occur significantly faster at temperatures near the glass-transition temperature of materials. Many researchers have examined crack healing in thermoplastic materials, but the process has also been observed in epoxy thermoset materials221. According to Adtech, the glass-transition temperature of LUC-4180 is between 60-66C. The BIOACT etching process occurred between 10-25 C

86

above the glass-transition temperature of the material; this is sufficiently above the glasstransition temperature for some thermoplastic materials to heal222. The BIOACT etching process may act to heal the "cracks" present at the interfaces between successive layers of polyurethane. Additional experiments would be required to ensure that healing was the mechanism responsible for the apparent strength increase, since the healing of unfractured materials has not been demonstrated223. Any possible effects of the BIOACT 280 material on this process should be isolated by performing similar tests on dual-interface tensile specimens in an oven rather than in a solvent bath224. If interface healing occurs during the BIOACT soaking process, then the support material removal process actually increases the strength of the finished parts. This would be unusual within the field of rapid prototyping, but it would be similar to the ultraviolet or thermal postcuring cycles required to achieve full strength parts in processes like Stereolithography.

5.6. Conclusions In order to choose materials to construct parts with improved strength via the Shape Deposition Manufacturing process, several series of tensile tests on current materials were performed. Tensile tests of monolithic tensile specimens were performed to compare the potential maximum strengths of these materials. A new tensile specimen with internal interfaces was designed to evaluate the strength of such materials when processed via layered processes such as SDM. The effect of gritblasting and support material removal processes on tensile strength was studied. Following these tests, the best current materials for monolithic and layered parts were selected. Tensile strengths of Polymer SDM materials were measured; materials tested included LUC-4180 polyurethane, TDT 205-3 polyurethane, and 501/530 epoxy. LUC-4180 polyurethane was the strongest material when cast monolithically; it was also much tougher than the other two materials. New dual-interface tensile specimens were designed to measure interface tensile strengths to evaluate the materials as they would be used in the SDM process. The 501/530 epoxy material had the strongest interface tensile strength, followed by TDT 205-3 polyurethane. LUC-4180 polyurethane was the weakest material in the dual-interface configuration, indicating that high interlayer strength does not necessarily accompany high monolithic strength. The poor interlayer strength of LUC-

87

4180 demonstrates that materials for Polymer SDM cannot be selected by high bulk strength alone. A "relative interface strength" was defined as the ratio of the strength of a dual-interface specimen to the strength of a monolithic specimen of the same material. This relative interface strength is a measure of the anisotropy of tensile strength expected in multilayer parts made via SDM. The TDT 205-3 polyurethane material had the highest relative interface strength (and hence least expected anisotropy) although it did not have the highest absolute interlayer strength. Materials with similar chemical compositions can have very different relative interface strengths. LUC-4180 and TDT 205-3 polyurethanes are an example of this phenomenon. Experts in polyurethanes and polymer interfaces suggest that differences in wetting or polymer chain interpenetration between the LUC-4180 and TDT 205-3 materials may be responsible for the differences in interlayer strength. Wetting properties may differ because of different degrees of crosslinking of the cured surfaces or because of different wetting properties of the uncured liquid resins. Polymer interpenetration may differ because of the crosslink densities of the cured surfaces or the chemical structures of the uncured materials. The effect of gritblasting and support material removal processes on tensile strength was studied. Gritblasting was investigated as a preparation method to improve interlayer strength using LUC-4180 polyurethane material. The procedure increased the strength of the resulting interfaces by a factor of at least three. However, the additional process complexity involved makes this a poor substitute for materials with intrinsically better interlayer bonding. The effects of the support material removal process on the tensile strengths of LUC-4180 polyurethane were investigated. The current support removal process involves soaking parts in heated baths of BIOACT 280 precision cleaner. This was found to decrease slightly the tensile strengths of monolithic specimens. However, it was found to increase the tensile strengths of dual-interface specimens significantly. This may occur through a crack healing mechanism which operates above the glass-transition temperature of the material. TDT 205-3 polyurethane is now the recommended material for Polymer SDM because its cure time is an order of magnitude faster than the 501/530 epoxy material while its interface tensile strength is four times that of LUC-4180 polyurethane. If maximum strength is required, then 501/530 epoxy would be recommended for SDM in spite of its higher viscosity and significantly longer cure time. However, this material is being discontinued

88

by the manufacturer, so supplies are very short. If monolithic parts are being cast, then the higher strength and much greater toughness of LUC-4180 make it the preferred material. The present study indicates that it is better to use materials with good intrinsic interface bonding than to complicate the process by introducing additional steps to fix materials with poor interface bonding. An alternative approach is to avoid interfaces by building monolithic parts; this eliminates potentially weak interfaces and is explored in the Mold SDM process.

89

CHAPTER SIX APPLICATIONS OF POLYMER SDM 6.1. Introduction The Polymer SDM process has applications in a variety of areas, from the rapid prototyping of functional parts to the manufacture of small batches of highly customized parts which cannot be built any other way. Special capabilities of the process include the creation of multimaterial parts, assembled mechanisms, and objects with embedded mechanical devices, sensors, actuators, and electronics. The Polymer SDM process has also been extended through the use of a variant process called "Mold SDM," in which parts are cast into polymer molds made via Shape Deposition Manufacturing. A wide variety of parts have been made using these processes, and a sample are described in this chapter.

6.2. Rapid Prototyping Polymer Shape Deposition Manufacturing was developed as a rapid prototyping process for the manufacture of polymer parts. Eyerer et al. draw a distinction between three classes of prototype objects and their requirements 225 . These three classes are "geometric" prototypes, "functional" prototypes, and "technical" prototypes. Geometric prototypes are intended mainly as communication tools; their shapes and dimensions are important but their mechanical properties are not. The only exception to this rule is that some geometric prototypes are used as patterns for soft tooling or other properties, where some degree of mechanical robustness may be required. Functional prototypes are intended for actual inservice testing of parts or systems; their mechanical properties are quite important. Not only must their shapes and dimensions be accurate, but their mechanical properties must be sufficient for the planned tests. While this may not require the mechanical properties of the final production parts, it does generally require reasonable strength, stiffness, and toughness. Finally, technical prototypes are prototypes which exactly match production

90

parts. They require accurate shape and size as well as mechanical properties identical to production parts. This implies the use of production materials and production processes such as injection molding. Most commercial rapid prototyping machines target the geometric and functional prototypes markets; Polymer Shape Deposition Manufacturing is intended for the production of functional prototypes. The use of SDM for functional prototypes requires the use of reasonably strong materials with high interlayer strengths. Otherwise, parts will break or delaminate at interlayer interfaces during handling or functional testing. The materials identified during the present research have substantially stronger interlayer strengths than previous Polymer SDM materials, improving the ability of SDM parts to withstand functional tests.

6.3. Multimaterial Structures Multimaterial structures are simply Polymer Shape Deposition Manufacturing parts which use more than one part material. This presents a more complicated set of restrictions on material combinations. The two part materials must be compatible with one another as well as with the support materials, and neither part material may be damaged by the support material removal process. These constraints can be relaxed slightly when one material is completely embedded within the other material. If the interior material is completely shielded within a nonpermeable outer shell, then the interior material won't be exposed to support material removal solvents and need not survive exposure to them. This is similar to the process used for metallic SDM injection molds, where Invar cores are surrounded by stainless steel outer shells226. While Invar would be damaged by direct exposure to the support material removal process, the stainless steel shell protects it from such exposure, allowing its use in the process. An example Polymer SDM multimaterial structure is a stiff outer shell with a lightweight spacefilling material inside. This type of structure was considered for housings for embedded electronic devices. A similar part is pictured in Figure 6.1; this is a half-shell structure with foam on the inside. LUC-4180 polyurethane was used as the material for the outer shell, and X-40 polyurethane foam from TAP Plastics (Dublin, California) was used as the foam filler material. Solder mask was used as the support material for this early part, which caused some surface finish problems on the outer shell. The X-40 polyurethane foam is a very-fast-curing, low-density material (about 32 kg/m3). Its surface can be

91

machined to a relatively smooth finish although such surfaces are porous. Its sole drawback is that it cannot be used as a traditional SDM part material because it cannot be deposited over itself. The material has very small pores when cast over a smooth surface, but if cast over itself, the porosity of the previous surface will create very large pores in the newly-cast material. Therefore, this material can only be used in single-layer applications, where it will then be covered with full-density thermoset materials.

Figure 6.1. Polyurethane Shell with Polyurethane Foam Filler Researchers at Carnegie Mellon University have developed other multimaterial systems for embedded electronic applications. One system used metal-filled polyurethanes as a composite part material for improved heat transfer from electronic devices. Another system combined polyurethane shells with interior phase-change materials. The phase-change materials were designed to act as transient heat sinks to reduce thermal signatures of embedded electronic devices.

6.4. Mechanisms Integrated assemblies and assembled mechanisms are a promising application of Polymer Shape Deposition Manufacturing. The earliest assembled parts made via Polymer SDM were simple parts with embedded axles and shafts. Later assemblies included the sliding disks built inside frames which are shown in Figure 4.3. These were assemblies where the parts were not built in direct contact with one another, but were built with large clearances between the parts. The completed parts allow the sliding disks to move freely within the frames, but do not allow the disks to be removed. More advanced mechanisms made from

92

TDT 205-3 are shown in Figure 4.4. These parts were made in several layers and consist of a cantilevered flap rotating at a vertical hinge. These mechanisms were developed for an aerodynamic testing application where a linear group was required. The combined configuration is shown in Figure 6.2. These mechanisms have relatively small clearances. The radial clearance around the hinge of each flap is fifty microns. The flaps on each end of the assembled "wing" have a slightly different design with endcaps. The axial clearance between the flaps and the end caps is two hundred microns. These parts were made from TDT 205-3 polyurethane using Kindt-Collins 3230A wax support material.

Figure 6.2. Assembled System of Polyurethane Mechanisms227

6.5. Embedded Objects The Polymer Shape Deposition Manufacturing process is particularly suited to the manufacture of parts with embedded objects. The deposition and shaping steps can be halted between layers or parts of layers, objects can be inserted into a part, and the build continued. This allows parts to be made with a variety of different internal components. Purely mechanical components, such as springs, bearings, or metallic inserts, can be placed into parts or between separate pieces of assembled mechanisms. These components can provide secure connections to other objects, or can allow free or restricted motion between pieces of a single part. While this type of embedded component somewhat expands the possible design space for Polymer SDM, the use of "smart" embedded objects provides even more possibilities. Embedded sensors, actuators, and electronics are all

93

possible within Polymer SDM parts. Examples of all three of these classes of objects have been embedded. Figure 6.3 shows a prototype of a rotary selector switch with integrated push-button action. This is an example of a part with purely mechanical embedded objects. It was designed for a series of wearable computers called "Tactical Information Assistants" being developed under DARPA contract. The switch is rotated to move an on-screen selection bar through a number of choices, and is pushed to signal the selection of the current choice. Positive tactile feedback for rotary motion is provided by twelve detent positions around the circumference of the device. The whole knob is springloaded to allow it to be pressed inward to signal a selection. The device was made via Polymer SDM, using Adtech LUC4180 polyurethane with embedded objects. Aremco Crystalbond 555 was used as support material. An embedded wave spring provides the push-button effect for the whole knob. An embedded Teflon bushing with divots provides both a rotary bearing surface for the knob and the detent action for tactile feedback. The outside surfaces were machined to good surface finishes; the bottom of the knob was replicated from a water-soluble wax surface, but its surface finish was poor.

Figure 6.3. Polyurethane Dial Input Device228 Figure 6.4 shows an object with an embedded silicon pressure sensor, made using procedures similar to Polymer SDM by researchers in the Stanford Dextrous Manipulation Laboratory. Researchers in the same laboratory have also built a Polymer SDM part with embedded actuators. The actuators are Nitinol shape-memory-alloy wires whose lengths can be varied under electrical control. The wires are attached to a hinged link, allowing the position of the link to be controlled during operation.

94

Figure 6.4. Part with Embedded Sensor229 The original motivation for the development of Polymer Shape Deposition Manufacturing by researchers at Carnegie Mellon University was the production of wearable computers and other conformal electronic devices. Several generations of these devices have been constructed; an example from the latest series is shown in Figure 6.5. This is "Frogman," a waterproof wearable computer for use by divers. The device was made from LUC-4180 polyurethane and a variety of embedded electronic and mechanical components; it has been successfully tested underwater. As mentioned above, several multimaterial combinations have also been investigated for use in devices like these.

Figure 6.5. Polyurethane Wearable Computer for Divers230

95

6.6. Mold SDM "Mold SDM" is a variant of the Shape Deposition Manufacturing process. It adds additional process steps to allow the manufacture of objects in materials which could not otherwise be used in Shape Deposition Manufacturing. Mold SDM revolves around the use of Shape Deposition Manufacturing to build molds into which other materials are cast monolithically. These molds are eventually removed from around the molded parts, which can then be post-processed. The use of a monolithic casting process both allows the use of part materials which do not bond well to themselves and also eliminates any possibility of interlayer defects which might weaken parts. This process was originally developed by Cooper et al. of the Stanford Rapid Prototyping Laboratory. The solder mask support materials developed earlier for use in Polymer SDM enabled the development of this process. The original motivation for Mold SDM was the production of gelcast ceramic green parts, and this is still the major focus of the process231,232.

6.6.1. The Mold SDM Process In Mold SDM, a mold material (acting as "part" material) and a temporary part material (acting as "support material") are used to make a mold through Shape Deposition Manufacturing (SDM). This mold is built up a layer at a time through material deposition and removal, as in the normal SDM process. At the end of this phase of the process, there is a mold made of mold material, with temporary part material embedded within it. At this point, the temporary part material is removed, using heat or solvents, leaving a mold. The material for the final polymer object is cast into this mold. After the material is cast into the mold, it is allowed to cure. This material can be heated to speed the curing process, as long as the heating does not affect the geometry of the mold. After the polymer material has cured, sprue removal or other secondary machining processes can be performed. The mold can be removed from around the polymer part, using heat, solvents, or other processes. The finished polymer part is then ready for use. Figure 6.6 shows the major process steps for making polymer parts through Mold SDM. Steps one through eight are the construction of a layered mold using SDM. The lightercolored material is mold material, while the darker-colored material is temporary part material. In step nine, the temporary part material is dissolved, leaving only the mold itself. In step ten, final polymer part material is cast into the mold and allowed to cure. At

96

this point, two process paths are possible. In steps 11a and 12a, the mold material is removed and then sprues are machined away. In steps 11b and 12b, secondary machining is performed first, followed by mold removal.

5 11a 12a

10 11b 12b

Mold Material Temporary Part Material Part Material

Figure 6.6. The Mold SDM Process233 Since the final material is cast all at once, some of the restrictions on materials for the general SDM process are lifted, such as the requirement that successive layers adhere to one another. The Mold SDM process therefore expands the family of materials which can be processed by Shape Deposition Manufacturing. While it allows the use of different materials, it does impose new constraints on the overall materials system. The mold material and temporary part material are subject to the same constraints which affect Polymer SDM materials, since molds are made through the "Direct" Polymer SDM process. The final part material must also be compatible with the mold material while the mold material must be removable without harming the final part material. Process planning is also more complicated, since part orientation must be chosen and pouring reservoirs and sprues must be designed to assure complete mold filling to manufacture high-quality parts. Current Mold SDM materials for polymer parts are wax mold materials, solder mask temporary part materials, and castable thermoset final part materials. The molds are made from Kindt-Collins KC3230A wax, which was specially developed for this application. Since this material has a very low viscosity, higher-melting-point waxes are sometimes used to construct dams around areas to be deposited. The current solder mask temporary part material is a 4:1 mixture of ElectroLite ELC4497 and Dymax 9-20311-F solder masks, similar to those used for early Polymer SDM support materials. Current final part materials include LUC-4180 polyurethane, TDT 205-3 polyurethane, and 501/530 epoxy resins.

97

Both single-piece parts and assembled mechanisms have been constructed. Figure 6.7 shows a complicated part made from TDT 205-3 polyurethane. Figure 6.8 shows a mechanism made from colored LUC-4180 polyurethane. This part has geared axles and one-way ratchets intended to allow an "inchworm" motion when the central bow is pressed and released. The individual pieces of the mechanism were cast in separate operations into different cavities of a single mold, producing an assembled mechanism.

Figure 6.7. Mold SDM Polyurethane Part234

98

Figure 6.8. Mold SDM Polyurethane "Inchworm" Toy235

6.6.2. Process Automation To increase the production rate and part quality of this process, an automated Mold SDM machine was developed236. This machine builds molds in an automated fashion. After molds are made by the machine, the temporary part material is removed manually, final part material is cast manually, and mold material removal is performed manually. The machine was built by retrofitting deposition and curing equipment onto a commercial three-axis vertical milling machine. This machine is shown in Figure 6.9. Figure 6.10 shows a schematic view of the retrofitted equipment. All of the equipment shown in the schematic is mounted in or on a three-axis mill from Haas. At the top of the machine are a pair of wax melt units from the Slautterback Corp. They feed wax through hoses to dispensers mounted on retractable mounts on the ram of the milling machine. Below the melt units, inside the housing of the machine, are an ultraviolet light for curing solder mask and an infrared light for preheating wax substrates. The ultraviolet light was made by UVEXS, Inc., and the infrared light was made by Research, Inc. The solder mask dispensing system was made by EFD, operated by pneumatic pressure and controlled 99

by a pneumatic valve. Parts are mounted on the bed of the machine on pallets from Mecatool. A variety of cutting tools are held in the milling machine's toolchanger for use in different situations.

Figure 6.9. Automated Mold SDM Machine

Soldermask Tank Wax Melt Units Ram Tool Changer (20 tools) UV and IR Lamps

Substrate on Pallet Table

Dispensers (2 wax, 1 soldermask)

Figure 6.10. Automated Mold SDM Machine Schematic237 A block diagram of system controls is shown in Figure 6.11. The "Computer" in that schematic is a PC, the "CNC Mill" is the milling machine, and the "I/O Electronics" are control electronics interfaced to the milling machine. The machine-control computer is an IBM-compatible PC running custom control software written in Java under the Windows

100

NT operating system. This computer communicates with CAD and planning software via Ethernet. It is connected to both control electronics on the milling machine and the milling machine itself via serial lines. The control electronics are based on OPTO-22 optoisolated input and output modules. Commands for the milling machine are sent via serial line from the PC, but communication back from the milling machine is only possible indirectly through the control electronics. The control electronics pass synchronization signals to and from the milling machine and also switch power and signal lines which control individual pieces of equipment attached to the milling machine. This system has been used to make Mold SDM molds automatically.

Computer
RS232 Serial RS232 Serial

CNC Mill
Relay Closure

I/O Electronics

Digital and/or Analog

Deposition Equipment and Lights

Figure 6.11. Automated Mold SDM Machine Controls Schematic238

6.6.3. Process Comparison: Direct SDM vs. Mold SDM "Direct" Polymer Shape Deposition Manufacturing and Mold SDM have some similarities and several differences. The similarities arise because both processes involve layered manufacturing and machining, while the differences occur because of the monolithic or layered nature of the final parts produced. Some parts could be made using either process, depending upon the intended applications for the parts. This section will compare the processes, identifying their strengths and weaknesses and showing similar example parts made in each process.

101

There are some significant similarities between these two processes. They are both based upon Shape Deposition Manufacturing, whether for making polymer parts directly or for making polymeric molds for casting polymer parts. As both processes are developments of SDM, each process involves material deposition and machining, and the use of part and support materials. They each require a combination of materials for a complete implementation of the process, and there are many requirements for compatible material combinations. While the material restrictions differ by process, they still allow the use of more than one final part material in each process, adding to the variety of possible applications. Both processes are based upon successive deposition and machining of compacts of material. In order to create geometrical features such as sharp internal corners or flat surfaces crossing multiple layers of material, subtle combinations of material deposition, cutting, redeposition, and recutting must be performed. This high level of complexity makes automated process planning very useful for each process. Although such automation will be difficult to implement for parts of arbitrary geometry, it will be virtually required for commercialization. Direct Polymer SDM is more adaptable to making different classes of objects, but more constrained by the materials from which they can be made. It can easily make mechanisms or parts with embedded objects because there is complete access to the interior of these parts during their construction. The build can simply be stopped at an interior layer, objects placed in their desired locations and glued in place, and the build continued. It is also easier to build precision mechanisms using Direct SDM. The key feature in the construction of precision mechanisms is fine clearances separating regions of final part material, and fine clearances are easier to make in Direct SDM compared to Mold SDM. Direct SDM produces parts whose surface quality is less dependent upon the quality of the casting techniques used. All upward facing surfaces of part material are machined, so porosity at the top of cast layers is simply machined away. On the other hand, it produces multilayer final parts, with interlayer bonds, residual stresses, and possible warpage. The construction of such multilayer parts requires part materials with excellent interlayer strength, preventing the use of many potential materials. The repeated casting of part material also results in very long cycle times if materials with long cure times are used. Mold SDM can make objects from materials which cannot be processed by Direct SDM. Final part materials need not adhere to themselves or to compatible support materials, since they are only cast monolithically at the end of the process. This allows the use of materials like LUC-4180 polyurethane, which has high toughness and reasonable strength but

102

unacceptably low interlayer strength. The process results in monolithic final parts, eliminating interlayer bonds and their weaknesses, and greatly reducing residual stresses. Its principal drawbacks are its greater process complexity, its reduced flexibility for embedded objects and mechanisms, the dependence of cosmetic surfaces on the quality of the casting techniques, and the long and indeterminate waits to remove the current temporary part materials from molds. Compared to Direct SDM, the Mold SDM process is less convenient for the production of parts with embedded objects or embedded regions of different materials. Mold SDM is much more dependent upon high quality casting processes to assure good surface finishes on upward facing surfaces. All external surfaces of a final part are replicated from the mold surfaces in a single casting operation. If any bubbles remain next to the surfaces of the mold, these will cause porosity or voids at the surfaces of the final part. This is typically a problem on flat upward facing surfaces, where air bubbles may collect as they rise through the newly cast material. Figures 6.12, 6.13, and 6.14 show simple objects with moving parts made via Direct SDM and Mold SDM. The Direct SDM parts have visible layer boundaries between the three separate layers in which the parts were made. It is difficult to see this in the pictures, but both parts have delaminated at the layer boundary between the top layer and the middle layer. This occurred because of the poor interlayer bonding which occurs when LUC-4180 polyurethane is used in Direct SDM. Slight warpage in the parts can also be seen; large flat parts often warp when made in this multilayer fashion. The surface finish, especially on the top surfaces, is quite good since it was machined directly into cast polyurethane. The Mold SDM parts show no evidence of interlayer boundaries, since they were cast monolithically. The parts have no interlayer weak points, and are also flatter because less residual stress was developed during construction. The parts were cast upside down, so the casting voids expected on upward facing flat surfaces are actually seen on the bottom of the parts. These flaws indicate that the casting process has not been adequately developed to prevent such voids in this geometry. Also visible on the bottom of the parts are the remains of the casting sprues, which were machined away in manual postprocessing operations. The raised ridge around the upper surface was a machining mistake rather than a processing artifact.

103

Figure 6.12. Direct SDM Assemblies in LUC-4180 Polyurethane

Figure 6.13. Top View of Mold SDM Assembly

104

Figure 6.14. Bottom View of Mold SDM Assembly

6.7. Comparison between Polymer SDM and Commercial RP Systems Polymer SDM is fundamentally different from the rapid prototyping systems which are now commercially available. While commercial systems deposit or fuse material to net shape in plan view, Polymer SDM shapes previously deposited material via peripheral milling. While some of the commercial systems do plane surfaces flat after deposition, none use milling to define side surfaces of parts. The other major difference is that no commercial process uses castable thermoset part materials; these are only used commercially in soft tooling applications where rapid prototyped patterns are replicated. These two basic observations underlie the differences between Polymer SDM and current commercial processes. These differences relate to levels of planning automation, part build rates, and part properties including accuracy, surface finish, strength, and strength anisotropy.

6.7.1. Planning Automation The commercially available rapid prototyping systems generally include automated part planning software. This software can prepare both process plans and low-level machine

105

control instructions although manual intervention may be required for optimal part orientation or the generation of support structures. These part planning systems are simplified by the relative simplicity of commercial layered manufacturing processes, which slice parts into (nearly) independent layers which require only two-dimensional path planning. The use of planar layers greatly simplifies the task of planning three-dimensional arbitrary-geometry parts for these systems. The more complicated requirements of Polymer SDM part planning have prevented it from being developed to this degree of automation. Polymer SDM part planning is inherently three dimensional, requiring arbitrary-geometry parts to be split into non-planar layers made of compacts with slanted or curved surfaces. This more accurate geometrical representation of parts comes at a high cost in planning requirements, currently requiring manual intervention except for 2.5D shapes. The need for special strategies to build sharp interior corners using round milling cutters also complicates the part planning process. It may prove possible to automate this planning process, but manual supervision and possible intervention are expected to be required for the near future. Another approach to this problem is being taken by researchers at the Stanford Center for Design Research, where CAD modeling systems with libraries of easily-planned features are being developed239. The design of parts using such systems could produce parts which can be planned completely automatically.

6.7.2. Part Build Rates Part build rates and their scaling vary from process to process. The various processes fall into four categories based on build rate scaling240: no dependence on part size, dependence on part perimeter, dependence upon part volume, and dependence upon maximum part dimensions. The Cubital Solid Ground Curing process does not depend on part size; an entire vat takes the same amount of time to build whether there are any parts in it or not. In early machines, there was no way to build less than a complete vat, so it was very important to fill the vat with parts stacked in three dimensions throughout the vat. This process is very parallel, however; an entire layer is imaged at the same time rather than drawn line by line. The other processes are serial in nature, sequentially depositing, curing, or cutting material on serpentine paths for each layer. Build times for the Laminated Object Manufacturing and Shape Deposition Manufacturing processes depend on the perimeter of the parts being built, with additional time required to prepare to cut a layer. LOM outlines the desired part with a laser while SDM outlines the part with milling cutters. The setup times preceding cutting do vary greatly between the two processes,

106

however. In LOM, the setup time includes a variable time to crosshatch the non-part area of the previous layer and a fixed time to roll out more part material. In SDM, the setup time includes casting a layer of part or support material and allowing it to cure or cool before cutting can commence; this time normally varies nonlinearly by the volume of the layer being deposited. Several commercial processes' build times depend upon the volume of the parts being built. Stereolithography, Selective Laser Sintering, and Fused Deposition Modeling all require the deposition or fusing of each voxel of part material; bigger parts take longer to build according to their volumes. This duration must be modified to include material handling or other events which occur once every layer, and also to include any environmental warm-up or cool-down times which affect entire parts. Finally, build times for inkjet deposition processes such as Sanders, Genisys, and Actua supposedly depend upon the size of a bounding box containing the part or particular cross-sections of it. Out of all commercial processes, the fastest is the 3D Printing process implemented by Z Corporation. This process is advertised to build a camera-sized object within an hour241; no other commercial process is nearly this fast. Detailed comparisons have not been performed, but Polymer SDM is probably among the slowest of these processes for the production of complicated parts of arbitrary geometry. It is slowed by the need to wait for part and support material to cure or cool between deposition and milling, and by the slow speed of the milling process.

6.7.3. Part Properties Many different properties of rapid prototyped parts are important to their users. Some of these are accuracy, surface quality, and mechanical properties such as tensile strength. Some users are concerned with one of these properties while others have requirements on most or all of them. Standardized methods for evaluating such properties have not been developed and promulgated, and finding independent test data on commercial processes can be very difficult. Since Polymer SDM has not been completely characterized either, a direct comparison is impossible. However, some general comments about these properties can be made. Part accuracy is an important property of rapid prototyped parts, but no standardized accuracy tests have been developed. The ASTM has not yet published a standard for the evaluation of accuracy of rapid prototyping systems. Commercial rapid prototyping systems have a variety of process-specific accuracy concerns, which are only beginning to

107

be explored. These include part distortion during build, damage during support material removal, and the effects of manual finishing operations. While some system manufacturers have published statistical accuracy studies, this is relatively uncommon, and independent academic verification of such claims is unusual. Similarly, the accuracy of Polymer SDM parts has not yet been analyzed. The use of direct machining and replication of surfaces should lead to high accuracy, but material shrinkage, layered material deposition, and the thermal cycling which occurs during the process may compromise this capability. Shrinkage of both part and support materials contributes to Christmas-tree effects, overall part distortion, and the evolution of residual stresses in parts. Thermally harsh support material removal processes also contribute to part distortion. Besides the accuracy achievable through build processes, another aspect of the part accuracy issue in rapid prototyping is the disconnect between a solid model and the simplified part a system attempts to build. Commercial layered manufacturing systems cannot build smooth slanted surfaces and cannot make planar features thinner than a layer. Polymer SDM can build smooth slanted or curved surfaces and can make arbitrarily thin features. However, some pathological combinations of interior sharp corners and rounded surfaces cannot be produced because of cutter gouging. This aspect of not being able to plan a part's exact geometry through either commercial rapid prototyping systems or Polymer SDM is rarely examined. In summary, no direct comparison between real-world accuracy of Polymer SDM and commercial rapid prototyping systems has yet been made. Surface finish is another part property which is important for rapid prototyped parts. It affects both a part's fidelity to the original computer model and the part's appearance. The commercial rapid prototyping processes produce parts with stairsteps on all slanted surfaces, where the edges of successive layers approximate the desired smooth surfaces. Rapid prototypes intended for marketing studies or other "appearance" applications require the smooth cosmetic surfaces typically expected on consumer products. The stairsteps present on parts produced on commercial rapid prototyping equipment must be manually finished to achieve this surface finish. This manual finishing may also damage or smooth out desired features on a part, representing additional deviation from the original computer model. The Polymer SDM process can avoid most of these problems. All part surfaces are either machined directly or else replicated from machined surfaces. This can produce an excellent surface finish, with no need for hand-finishing; this preserves the desired geometry from the original solid model. Polymer SDM operation sequences, however, tend to produce Christmas-tree surface finish defects which can detract from the potential surface finish of such parts. Depending on the part and support materials involved, this

108

tendency can be reduced by enforcing a certain operation sequence. Features should be machined into the higher-shrinkage material, rather than casting high-shrinkage material around geometry already machined into low-shrinkage material. In the polyurethane and wax material combination, features should be machined into wax, rather than casting wax around machined polyurethane features. Overall, Polymer SDM parts should have surface finish superior to those directly achievable in commercial rapid prototyping processes, especially for parts with slanted surfaces. Mechanical properties are another important issue for rapid prototyped parts; commonly measured values include tensile, flexural, and impact strengths as well as tensile and flexural moduli. The only parameters yet measured for Polymer SDM are tensile strengths, so only these properties can be compared. Table 2.1 listed tensile strengths quoted by manufacturers for commercial rapid prototyping materials; these can be considered baseline strengths. However, none of the specifications includes the effect of anisotropy, which reduces the strength of multilayer parts. Nor is it clear how representative these values are; some independent researchers have measured half these values for tensile strengths. Considering most of these issues, Polymer SDM tensile strengths are intermediate between those of different commercial rapid prototyping processes. Compared to Polymer SDM, several processes make weaker parts, Fused Deposition Modeling makes parts almost as strong, Selective Laser Sintering and Solid Ground Curing make roughly comparable parts, and Stereolithography produces stronger parts. This comparison recognizes that the dual-interface tensile strengths measured for Polymer SDM materials represent upper bounds on the likely strengths of Polymer SDM materials. Compared to Mold SDM using materials like LUC-4180, all processes except Stereolithography produce weaker parts. Stereolithography is an impressive process, since the tensile strengths reported for current materials are significantly stronger than any current Polymer SDM material. They are also as strong as the strongest available polyurethane prototyping resins. Finally, Stereolithography results in the production of parts with isotropic tensile strengths; that is, uniformly high strengths in all directions. This means that even if the high-strength polyurethane resins were found compatible for use in Polymer SDM, and they happened to have 100% relative interface strengths, they would still only match the strength of commercially available Stereolithography parts. Similarly, even if Mold SDM is used to build monolithic final parts, and the strongest commercially available castable polyurethanes were used, such parts might only match the strength of current Stereolithography parts. However, Mold SDM using LUC-4180 can produce parts which are substantially tougher than any of the strong Stereolithography materials.

109

6.8. Preferred Applications of Polymer Shape Deposition Manufacturing There are three categories of parts for which Polymer SDM can be used most advantageously: parts which can be made faster via Polymer SDM, parts which can be made better via Polymer SDM, and parts which can only be made via Polymer SDM. There is a class of parts which can be built faster via Polymer SDM than through other rapid prototyping techniques. These parts tend to have moderate or larger sizes with simple geometry and large section thicknesses; this would be a prototyping use of the Polymer SDM process. There is another class of parts which can be made better through Polymer SDM than through any other rapid prototyping method; part surface finishes can exceed those achievable in commercial rapid prototyping processes. Using Polymer SDM to build such parts would also be considered a prototyping use of the process. Finally, there are a variety of parts which can be made via Polymer SDM but which are difficult or impossible to build any other way. Such parts contain small mechanisms, embedded objects, or multimaterial structures; building such parts would make Polymer SDM a limited production process. This is the optimal use of Polymer Shape Deposition Manufacturing.

6.8.1. Parts Made Faster through Polymer SDM Although Polymer SDM is among the slowest processes for the production of arbitrary geometry parts, there are some parts it could build most rapidly. This occurs because it scales with perimeter not volume, making bigger parts comparatively faster. Additionally, the every-layer setup times largely depend on casting and cooling processes. If a large part with relatively few coarse features surrounded by limited support material is being produced, then even the casting times can benefit from the size of the part. Simple geometry implies that there are few overcut to undercut transitions, requiring only a small number of layers. Additionally, the current thermoset part materials cure exothermically and are relatively poor conductors of heat. This means that thicker sections grow hotter and cure faster. While this does raise part production speed, there are some accompanying drawbacks. First, high cure exotherms may melt support material or distort machined features; second, faster cure speeds cause additional shrinkage, possibly increasing part distortion. The other drawback of pursuing maximum production speed via Shape Deposition Manufacturing is that materials with the fastest cure speeds are not usually the materials with the best mechanical properties. Nevertheless, TDT 205-3 polyurethane part material allows thick sections of part material to be machined an hour after curing and

110

bonds to itself very strongly. This allows relatively rapid production of medium or large parts with simple, large-radius features. While this is an interesting application of the process, continued progress in materials for the very fast Z Corporation system will probably take over this niche.

6.8.2. Parts with Better Surface Finish through Polymer SDM Another advantageous application for Polymer SDM is making parts with better surface finishes than achievable in any commercial rapid prototyping process. This eliminates a manual finishing step which may distort the desired geometry of a prototype. Skipping the manual finishing step reduces the cost of these parts, and preventing damage to part features greatly improves prototype quality. While such slight changes in part shape are not an important issue for appearance models, they are important for tooling masters. These parts are intended for use with soft-tooling or other downstream processes which replicate small series of parts from rapid prototyped master patterns. This is clearly a prototyping application, where the final parts will be produced by other means. Since improved accuracy of initial prototype parts can speed design and testing cycles, the more accurate parts achievable this way can justify the use of the relatively costly Polymer SDM process. If very small numbers of even more accurate functional parts are required, the Mold SDM process can be used. This process can use monolithic castable materials which have higher strength and less anisotropy than achievable in direct Polymer SDM. Accuracy is improved compared to traditional soft-tooling practices because one fewer part replication step is required. Rather than making a positive pattern, casting a negative mold in silicone, and then casting a positive part in polyurethane, a negative wax mold can be produced initially, and polyurethane can be cast into it directly. This may produce parts with closer tolerances than those possible after a second replication. The drawback of this sequence is that the mold is fugitive; only one part is produced per mold manufactured. This is much less efficient than standard soft tooling methods, which produce several parts per mold and several molds per pattern. If the improved accuracy justifies the high cost of this process, then this could be a successful application of Mold SDM for prototyping polymer parts.

111

6.8.3. Parts Manufacturable only through Polymer SDM The last and best of these advantageous uses of Polymer Shape Deposition Manufacturing is the construction of parts which would be almost impossible to build any other way. Since the parts cannot be made by other production processes, this is not really a prototyping application, but instead a limited production application. The principal types of parts in this category are those that exploit process capabilities like small assembled mechanisms, embedded objects, and multimaterial structures. The use of Mold SDM to gelcast green ceramic or metallic parts into polymer molds also falls into this category. Small mechanisms are a category of parts for which Polymer SDM is particularly well suited. While conventional machining and assembly practices can be used instead, these processes become more difficult as part dimensions shrink. Fasteners and other packaging and attachment methods also become more cumbersome at smaller sizes, and can be larger and more cumbersome than the parts being attached. Commercial rapid prototyping systems can almost build such parts, but there are serious limitations. Stereolithography produces very strong, finely detailed parts, but its non-soluble support structures prevent the construction of mechanisms. Laminated Object Manufacturing and the Actua and Genisys processes are also eliminated by their non-soluble support structures. The Z Corporation and Sanders processes both produce parts with very weak mechanical properties. Selective Laser Sintering suffers from poor surface finish and the difficulty of blowing supporting powder out of small cavities. The Cubital process has produced some medium-sized assembled structures, but users complain about the difficulty of support material removal from small features242. The new soluble material for Fused Deposition Modeling should allow the construction of mechanisms, but the bead width will limit achievable feature sizes. The solvent-soluble waxes used as current Polymer SDM support materials make this process superior to all commercial rapid prototyping systems for the construction of small mechanisms. Parts with embedded objects are another category of parts possible in Polymer SDM but difficult to build otherwise. So-called "smart structures" are one application - parts with embedded sensors or actuators accompanied by embedded electronics can be built using Polymer SDM. Competing processes for this application include Stereolithography, simple electronic encapsulation, and conventional housings. Stereolithography has been used to embed sensors but the process is very cumbersome. The normal build process must be interrupted for the introduction of sensors, and then the process must be restarted, with the

112

hope that the recoating hardware will not disturb the position of the sensors. The embedding of thick sensors would be even more involved in such a process. The Polymer SDM process easily permits the installation of embedded components, which can have any regular or irregular shape since matching cavities can be planned and machined. If desirable, embedded objects can be installed and later shaped to conform to curved surfaces; this has been done with LED lenses for computer indicators. This allows significantly more flexibility than possible with Stereolithography. Traditional electronic encapsulation normally requires an exterior form into which electronic components have been mounted; it is usually performed in a single step. This process is not adaptable to the arbitrary part geometries processable through Polymer SDM. The use of conventional housings and fasteners is another alternative for the construction of this type of device. With this arrangement, sensors, actuators, or electronics are contained within a separate multipiece structure. This may be acceptable, or it may not, depending upon the application. If sensors are to measure stresses or other properties inside a monolithic structure, then a poorly-coupled, multipiece housing would represent a different situation. Electronics can be mounted within such housings, but there are tradeoffs between the two alternatives. Polymer SDM embedded electronics cannot be repaired. On the other hand, Polymer SDM embedded electronics can be made tamperproof or tamperevident in ways that conventional housings with fasteners cannot. Thermal characteristics of Polymer SDM devices can be varied greatly, either by insulating electronics from the environment or closely coupling the two; great thermal design flexibility is possible using Polymer SDM. Polymer material cast in-place around electronic devices can also improve heat transfer at interfaces by ensuring intimate contact through complete wetting. Overall, a great deal of design flexibility is provided by the use of Polymer SDM for the production of parts with embedded sensors, actuators, and electronics. Parts of multimaterial construction are another advantage of the Polymer Shape Deposition Manufacturing process. They are particularly useful for building embedded electronic systems with varying properties. Both thermal and structural characteristics of such devices can be varied in this way. Thermal characteristics can be modified by using materials with different thermal conductivities or unusual phase-change transitions. Polymer foams can be used to insulate components from one another or from the environment, while polymers filled with aluminum, boron nitride, or other materials can be used to increase heat transfer between two regions. Phase-change materials can be used to absorb transient heat input, later releasing the heat to the environment or back into the

113

device. Structural characteristics can be varied widely within a part. An extreme example might be stiff internal frames, surrounded by tough impact-absorbent shells, containing low-density space-filling foam which protects embedded electronics. The use of different materials in the Polymer SDM process allows part properties to be tailored to match a given application. Structures like this would be difficult to build using any process except Polymer SDM. None of these applications could be easily accomplished without Polymer SDM. This makes Polymer SDM a limited production process, rather than a prototyping process for parts intended to be made some other way. Using the process for production does raise some issues. First, the current lack of fully automated part planning tools means that each new design requires substantial human process development. This means that one-off parts will be comparatively expensive. If small or medium batches of identical parts are needed, this process planning investment can be amortized over a reasonable number of parts. The second issue is cost; Polymer SDM parts will be relatively expensive because of the extensive machining and long curing and cooling cycles required by the process. This limits the realistic use of the process to short-run production of costly parts, at least initially.

6.9. Conclusions Several of these applications of Polymer Shape Deposition Manufacturing were enabled or improved by the current material identification and material strength research. The greatest impact was made by the identification of the class of solder mask materials as potential support materials for Polymer SDM. These materials allowed the development of the Mold SDM process, which greatly expands the choices of materials from which parts can be made. The other major advance was the development of solvents which allow complete support material removability for wax support materials. This allows castable thermoset part materials to be used to make parts with fine features, assembled mechanisms, and small clearances. The best example of this class of applications is the construction of aerodynamic flap assemblies made with polyurethane part material. While the material identification phase led to the development of these new applications, the material strength research allowed improved materials to be selected for use in functional prototypes. These newer materials, TDT 205-3 polyurethane and 501/530 epoxy have interlayer strengths several times those of the previous material, LUC-4180 polyurethane. This allows the

114

construction of parts with smaller interlayer contact areas with greatly reduced risk of delamination during use or handling. The Mold SDM variant process presents new opportunities for the production of polymer parts. Final parts are cast monolithically, avoiding interlayer interfaces, potential interlayer defects, and residual stresses from layered construction. This allows the use of materials such as LUC-4180 which have good bulk properties but poor interface properties, expanding the choices of potential final part materials.

115

CHAPTER SEVEN CONCLUSIONS 7.1. Contributions Several new aspects of Polymer Shape Deposition Manufacturing were investigated. The principal contributions were in the area of material identification, material characterization, and application development. These efforts resulted in improved materials for the direct Polymer SDM process, and allowed the development of the new Mold SDM process. The constraints on part and support materials and their combinations for use in Polymer Shape Deposition Manufacturing were identified and categorized. Many potential part and support materials were identified and tested, alone and in combination with each other and with previously identified materials. Part materials examined included acrylate and epoxy Stereolithography resins, castable filled epoxies, castable filled and unfilled polyurethanes, fast-curing castable polyurethanes, and epoxy adhesives. Support materials examined included photoresists, solder masks, water-soluble waxes, water- and solvent-soluble fixturing compounds, and machinable waxes. Several of these part and support materials were found to have useful properties. Acrylate Stereolithography part material and solder mask support material were identified as a possible Polymer SDM material combination, and a model of an injection molding tool was built using these materials. The use of different part materials within a single object was demonstrated using parts with polyurethane foams inside solid polyurethane shells. Fastercuring polyurethane part materials were identified, and comparison parts were made from one of them. Very-fast-curing polyurethane part materials were tested, but other process restrictions prevented their successful use. A new solvent (Petroferm BIOACT 280) was identified for the removal of wax support materials. This solvent now allows the complete removal of support material from the previous thermoset part material and wax support

116

material combination, without requiring any manual material removal. This allows the manufacture of parts with intricate internal mechanisms and small clearances. A new dual-interface tensile strength specimen was designed to measure interface tensile strengths of materials as they would be used in Polymer SDM. Using such specimens, process-specific material strengths were tested for three part materials. The relationship between interface tensile strength and bulk material strength was found to vary greatly among the different materials. New materials with improved interface tensile strengths were identified: Ciba TDT 205-3 (a fast-curing castable polyurethane) and Adtech 501/530 (a slow-curing castable epoxy). Ciba TDT 205-3 is a part material with a dual-interface tensile strength four times higher than the original polyurethane material; it also cures six times faster, allowing substantial cycle time reduction. Ultraviolet-curable, water-soluble thermoset materials called "solder masks" were identified as useful support materials. Their identification allowed the development of the Mold SDM process by Cooper et al. In this process, fugitive wax molds are made via Polymer Shape Deposition Manufacturing; other castable materials are then cast into these rapid prototyped molds. This indirect process allows the creation of parts using a wider variety of materials than would be possible through Direct SDM. Mold SDM produces monolithic final parts whose strengths are not limited by interlayer adhesion. This process was automated through the addition of material dispensing and curing hardware to a 3-axis vertical milling machine. Similar parts were manufactured through Direct Polymer SDM and through Mold SDM as a means of process comparison. A variety of applications for the Polymer SDM process were identified, and several test parts were constructed. Direct Polymer SDM was compared to current commercial rapid prototyping processes. Polymer SDM's greatest strength is the ability to make parts which cannot otherwise be made - parts with internal mechanisms, embedded objects, or multimaterial structures. Much of this work was published previously in a shorter form 243,244,245,246.

117

7.2. Future Work This research has improved the state of the Polymer Shape Deposition Manufacturing process and allowed the development of Mold SDM. Several issues which arose during this work deserve additional consideration. These opportunities include the development of stronger and easier to use materials for Polymer SDM, the development of quantitative measurements or tests to simplify the material selection process, improvement of the Polymer SDM process, and the development of improved materials for Mold SDM.

7.2.1. Polymer SDM Materials Improvements Further research into several aspects of Polymer SDM materials is needed. Materials which produce strong parts yet have short cycle times and are easy to use must be identified or developed. To assure valid comparisons, the properties of the currently recommended material, Ciba TDT 205-3, should be verified. If strong parts are made with materials which have excellent bulk strengths but only moderate interlayer strengths, then research into postprocessing techniques may allow the production of even stronger parts. Finally, knowledge of the underlying mechanisms which determine interface tensile strength of materials would be very valuable for efficient material selection. Besides material strength, both material shrinkage issues and thermal issues affecting support material removal require more research for materials improvement. The 501/530 epoxy part material is an example of a relatively high-strength material. However, this material is poorly suited to the SDM process. It is highly loaded with fillers and requires twenty-four hours between casting and machining; additionally, the manufacturer has discontinued production. Its slow cure time would make this material a poor candidate for a commercial process because of the long cycle times which would be required for part production. The TDT 205-3 material is much better suited to the SDM process; it has a very low mixed viscosity, vacuum deairs very well, and can be machined two hours after casting, even in thin sections. However, this material has an interlayer strength of 20 MPa at best. There is a need for improved materials which combine the high relative interface strength and ease-of-use of the TDT 205-3 material and the higher strength of the 501/530 epoxy material.

118

The currently recommended material is Ciba TDT 205-3 polyurethane, which was tested in the present study. However, only a single round of testing of this material was performed. These results should be verified to ensure that batch-to-batch variations will not produce poorer than expected interlayer strengths. Additionally, the effects of the support material process on TDT 205-3 should be checked. The experience with LUC-4180 indicated that this process will probably improve interlayer strength, but this assumption should be verified. Since the relative interface strength of TDT 205-3 is so high, the support material removal process cannot significantly improve the interlayer strength, but testing will ensure that interfaces are not weakened by the process. The most important current requirement is materials with higher interlayer strengths. Since the 501/530 epoxy material is being discontinued by its manufacturer, a replacement material for applications requiring maximum interlayer strength will need to be identified. Materials which improve upon TDT 205-3 would also be especially useful - a combination of low viscosity, rapid cure time, and high interlayer strength is needed. Several materials are worth investigating for this role. One possibility is materials with high bulk strengths and moderate relative interface strengths; such materials may still have higher interlayer strengths than the TDT 205-3 material. A potential material is TUFFalloy 4270, from Hapco, Inc.; this material's bulk strength is 72 MPa247. Another possibility is materials of moderate strength and excellent relative interface strength. Such materials might combine the strength of LUC-4180 with the relative interface strength of TDT 205-3, resulting in materials with much higher interlayer strengths. A candidate material is Por-A-Kast Mark 15248, from Synair; this material's bulk strength is 55 MPa. A similar possibility would be materials of average strength with very good relative interface strength. For example, some laminating resins are engineered with secondary bonding mechanisms to assure that later layers will adhere to earlier layers. The existence of a secondary cure may provide improved interlayer bonding compared to typical thermoset polymers; an example of a material like this is TDT 232-1 epoxy249 from Ciba Specialty Chemicals. If high interlayer strength is developed through the use of strong materials with only moderate relative interface strength, then additional processes for interlayer strength improvement may be worth developing. If the support material removal process can be optimized to improve interlayer strength through crack healing, this would be a useful side benefit. Before doing so, additional testing to confirm crack healing during the removal process should be performed since crack healing of unfractured materials has not yet been verified. This means confirming that the strength improvement is the result of crack

119

healing alone, rather than side-effects of solvents or additional curing due to elevated temperatures. Therefore, the observed improvements in tensile strength should be measurable after heating within an oven rather than in a bath of BIOACT. Crack healing should also be a progressive process, so additional exposure to elevated temperatures should produce strengths progressively approaching those reached in monolithically cast specimens. While improved materials are the most urgent need, a better understanding of the processes leading to high and low interlayer strength would be very beneficial. The current theory is that poor wetting or poor polymer chain interpenetration is responsible for inferior interlayer strength, and that different types of interlayer bonds are not responsible for such problems. This hypothesis could be tested, but would require a very involved research program. The only subpart of this hypothesis which could be tested easily is the question of wetting. If wetting is the reason for poor interlayer strength of LUC-4180, then improving the wetting of this material should improve its bond strength. This can be tested by raising the temperature of cured surfaces before adding newly cast material. The hotter surfaces would raise the temperature of the uncured liquid polyurethane, reducing its viscosity and improving its wetting performance. If this change produces improved interlayer strength, then poor wetting is a responsible for the low strength of LUC-4180 interfaces; if there is no effect, then no conclusions can be drawn. Aside from material strength, other current concerns include material shrinkage and thermal issues affecting support material removal. While high-shrinkage part and support materials can cause warpage and interlayer delamination, upper limits on acceptable shrinkage have never been determined. The 2% linear shrinkage of Somos 3110 part material and 6% linear shrinkage of Kindt-Collins File-a-wax support material are high enough to cause delamination in multilayer parts. The 0.2% linear shrinkage of the LUC-4180 and TDT 205-3 polyurethane part materials seems to be low enough not to cause Christmas-tree effects when cast in multiple layers without support material. Somewhere between these values of linear shrinkage is a maximum acceptable shrinkage level, above which parts will be visibly distorted by residual stresses or even destroyed by delamination. Part warpage during support material removal is another area which requires more research. This is especially an issue with part materials with low glass-transition temperatures and support materials which require heated solvents for removal. During the removal process, the parts soften and may warp under their own weight. While they return to roughly the desired final shapes, some warpage can occur. The identification of solvents with improved

120

dissolving abilities at low temperatures could reduce distortion by reducing the heating and consequent softening of part materials. Alternatively, part materials with higher glasstransition temperatures could be identified and used to accomplish the same effect.

7.2.2. Polymer SDM Material Selection Process Improvements Selecting materials for use in Polymer Shape Deposition Manufacturing is currently an imprecise process. The development of a small number of quantitative criteria or standardized tests for materials would expedite this process in the future. Complete wetting of previously machined surfaces by newly deposited material is believed to be necessary for high interlayer strength. If this link can be confirmed, then simple wetting tests, even if qualitative, should precede any extensive tensile testing of a potential part material. This would be useful since wetting tests are easier to perform than building and breaking complete tensile specimens. Wetting tests could either directly check whether liquid polymer can completely wet similar cured polymer surfaces, or they could check contact angles formed by water or other liquids over cured polymer surfaces. Only if wetting tests predict good interlayer bonding should further testing of such interfaces be conducted. If maximum acceptable volumetric shrinkage values were determined, future material selection efforts could use this limit as another initial check on the suitability of a part or support material. Next, the deposition or curing temperatures and the glass-transition temperatures of potential part and support materials could be checked. If specifications indicate that materials can be deposited without greatly exceeding the glass-transition temperature of complementary materials, then the materials may be physically compatible. After this indication, deposition tests should be performed to confirm chemical and physical deposition compatibility. Only after wetting, shrinkage, glass-transition-temperature, and deposition tests were passed could it be justified to build and test tensile specimens while searching for new materials. The development of such a series of criteria and tests would allow efficient consideration of a wider variety of potential part and support materials for Polymer Shape Deposition Manufacturing.

7.2.3. Polymer SDM Process Improvements Small mechanisms are an interesting application of Polymer SDM. As mechanisms grow smaller, assembly becomes progressively more difficult, so the advantages of pre-

121

assembled mechanisms become greater. The Polymer SDM process has already been used to make relatively small clearances, and even smaller ones are possible. The achievable clearances have not been characterized; possible axial and radial clearances depend upon the materials in use. Specific test part geometries should be developed to determine the finest clearances possible using a given material. This will allow materials optimized for the creation of fine mechanisms to be selected. Special techniques may be required to leave thin films of support material in places where they are required. Additional research into the reduction of part distortion during the SDM process is also required. For small parts, such as the flap mechanisms shown in Figure 4.4, part distortion is not a major issue. For larger parts, especially with thin-walled features, wide flat areas, or large variations in section thickness, part distortion can be a problem. Both Christmas-tree surface finish defects and overall part warpage can reduce the quality of completed parts. As mentioned in the discussion of the Christmas-tree effect, support materials with high shrinkage can cause major distortion of thin-walled features during their construction. Such defects are quite apparent at interlayer boundaries, where geometrical features are visibly misaligned at layer boundaries. This is an important issue since the current material systems include wax support materials with relatively high shrinkage. This problem can be approached from several directions. The best solution is to develop lowershrinkage support materials; filled waxes may be usable in this role although relatively low viscosities must be maintained to assure void-free support material deposition. Another approach is to modify part planning so that critical features are always replicated from support material rather than directly machined. The reason is that replicated features will not be surrounded later by support material. Since current support materials shrink much more than part materials, the step of surrounding features with part material causes most of the Christmas-tree effect. Re-ordering process sequences to omit this step can improve surface quality; the flap parts (Figure 6.22) were generally constructed using this method. If this approach is unsuccessful, then limits on minimum feature thicknesses for part material and limits on maximum section thicknesses for support materials could limit the distortion caused by support material shrinkage. Alternatively, a compensation strategy could be developed, where compensated geometries are milled with the expectation that later shrinkage will produce the desired final geometry. While possible, this approach would require both a very detailed understanding of material solidification and shrinkage effects and a large software development effort.

122

7.2.4. Mold SDM Materials Improvements The newer Mold SDM process requires several materials improvements for optimal utility. The current solder mask temporary part materials only cure in very shallow thicknesses, limiting the build rate and contributing to defects when solder mask regions do not cure. For faster production rates, solder masks with shorter cure-times and greater depths of cure are required. Improved solder mask machinability is desirable although greater cure depth is more important. Wax mold materials with better machinability and reduced shrinkage would improve mold quality. Thermoset part materials with higher strengths should be identified. These materials should have higher heat-distortion temperatures to maintain desired geometry during mold removal. They can however have long cure times. While such cure times would contribute to very long part production times in Direct Polymer SDM, the single casting process required for Mold SDM means that process speed is not affected much by final part material cure time. New high strength castable polyurethane final part materials could be stronger than any commercial rapid prototyping materials except the newest Stereolithography resins.

123

APPENDIX A MATERIAL COMBINATIONS EXAMINED A.1. Support Materials for Thermoplastic Part Materials Several types of materials were examined in an effort to find support materials usable with high-temperature injection-molded thermoplastic part materials. These materials included ultraviolet-curable Stereolithography resins, ultraviolet-curable adhesives, integrated-circuit photoresists, printed-circuit-board plating resists, and printed-circuit-board solder masks. Ultraviolet-curing Stereolithography resins were examined as potential support materials for thermoplastics because of their ease of deposition and curing. Early Stereolithography resins had glass-transition temperatures of as high as 100-150C250 , meaning that geometry machined into them would be somewhat temperature stable. Almost all current Stereolithography resins, however, have different base chemistries with lower glasstransition temperatures, usually from 40-85C251. This represents significantly lower tolerance of high temperatures and implies correspondingly greater distortion of previously machined geometry following deposition of heated thermoplastic material. Ultraviolet-curable adhesives were investigated because of their better high-temperature performance. Some materials with maximum use temperatures of 200C were found, and at least one material with a maximum temperature of 260C was identified. These materials, however, were silicones with good solvent resistance, complicating their removal. Loctite Nuva-Sil 76 is an ultraviolet-curing silicone sealant which can operate to 260C; it can be cured to a depth of 9 mm within a minute252. One drawback of Nuva-Sil 5076 is its low hardness, Shore A 30. It can be shaped with milling cutters, but its machinability is very poor. This material is formulated to be more easily removed than other Loctite silicones, but this may cause adhesion problems in multilayer parts. A solvent which should be able to remove this material is Amtex CCR253. This solvent should not harm Lexan or Ultem part materials, but has only limited removal abilities for these support materials. It will gel silicones after they are soaked in it, but can only completely dissolve

124

thin layers of the material. Based on the poor support material removal techniques available, no additional work was performed with ultraviolet-curing silicone adhesives. The next materials investigated were photoresists intended for the fabrication of integrated circuits. These materials, unlike the previous silicones, are designed to be removed after they are patterned. Two issues with the use of these photoresists are their usable layer thicknesses and their performance under high-temperature conditions. Most photoresists are spincoated in very thin layers; both negative and positive photoresists are normally used in thicknesses of five microns at most254. Researchers have developed special methods to use photoresists in thicknesses as great as forty microns to form molds for electroplating of small parts255. Some experimental materials have been developed with glass-transition temperatures of 270-310C256, but this is much higher than materials normally used. One commercial material for use at 220C was found257, but the limited layer thicknesses achievable using photoresists intended for integrated-circuit fabrication led to an end of this line of research. Photoresist systems intended for use with printed circuit board manufacturing were investigated next. These materials are used in significantly thicker layers than similar materials intended for integrated circuit fabrication - fifteen to one hundred microns for printed circuit boards258. Researchers at Armstrong World Industries even developed a plating resist which could be used at thicknesses of 2 mm259. These materials are used as plating resists during the generation of copper conductor geometries for printed circuit boards. As with the resists for integrated circuit production, these are intended for complete removal after the plating process is complete. However, these materials are not formulated to withstand high temperatures since the plating processes are often performed near room temperature. This drawback ended research into these materials, but led to the knowledge of solder masks for printed circuit boards. Solder masks for printed circuit board production are used to protect regions of a board from an area-soldering process260. Several classes of solder mask are available: permanent solder masks, temporary peelable solder masks, and temporary soluble solder masks. Permanent solder masks are intended to remain on the printed circuit boards forever, as a protective environmental barrier. Temporary solder masks are intended to be removed after a soldering process is complete. Some of these are designed to remain intact during manual removal via peeling. Other temporary solder masks are intended to be dissolved through washing with solvents or water. These soluble temporary solder masks are generally

125

solidified either through curing under ultraviolet light or solvent evaporation at room temperature or in an oven. Ultraviolet-curing solder masks solidify in seconds or tens of seconds, while solvent-evaporation solder masks solidify in tens of minutes. Table A.1 lists a variety of commercially available solder masks and their removal methods. Besides these materials, a much larger variety of solvent-evaporation solder masks is available261. Both Dymax262 and Tech Spray263 have patents on particular solder mask formulations or processes using them. Current Materials Material Tech Spray Inc. Wondermask W / WA / WS Tech Spray Inc. Wondermask W-UVA Dymax Corp. Ultra Light-Weld 9-20311-F Electro-Lite Corp. ELC4497 Cure Method solv. evap. UV UV UV Viscosity Varies 28,00035,000 cPs 20,000 cPs 28,00035,000 cPs Normal Thickness 10-15 mils 30-40 mils Removal Method water wash water wash Reference 264 265 266 267

20 mils typ. water wash 30-40 mils water wash

Discontinued Materials Material Tech Spray Inc. Wondermask S-UVA Dymax Corp. 948 Series Lackwerke Peters SD 2956 UV-HAL Cure Method UV UV UV Viscosity 7,000-10,000 cPs Varies 19,00027,000 mPas Normal Thickness 30-40 mils Removal Method solvent wash Reference 268 269 270

20 mils typ. water wash ~1 mil caustic soda wash

Table A.1. Commercially Available Solder Masks Of these materials, those most useful for the Shape Deposition Manufacturing process are those that are ultraviolet-curable and water-soluble: Dymax 9-20311-F, ElectroLite ELC4497 and Tech Spray Wondermask W-UVA. These materials can be deposited and solidified in a short period of time, and can be later removed through the use of nontoxic materials. They are highly-branched, lightly-crosslinked polymers which have hightemperature stability for short time periods, allowing them to withstand deposition of hot thermoplastic material. The materials photopolymerize during exposure to UV-A or fullrange ultraviolet light, solidifying into solids.

126

The Dymax solder mask has a gummy consistency when cured, while the ElectroLite and Tech Spray products are hard and brittle after curing. The Dymax material is too soft for machining, while the others are quite brittle and prone to chipping. Blends of the materials were found to have hybrid properties with improved machinability. The ElectroLite solder mask seemed to be somewhat less brittle than the Tech Spray material, so Dymax and ElectroLite solder masks were mixed to make solder mask blends. Originally, the ratio of 3.5:1 ElectroLite ELC4497:Dymax 948 was used, but this was later changed to 4:1 ElectroLite ELC4497:Dymax 9-20311-F. Combining materials into mixed solder masks is straightforward. The materials are manually mixed and allowed to stand overnight for air bubbles to be released from the mixture. These mixtures, however, may not be stable for long-term storage. The Dymax and ElectroLite materials have different solidification mechanisms, and their components slowly react, resulting in viscosity increasing with time. After days or weeks, significant viscosity changes can be detected, and the material will become unusably thick after a period of months. This means that deposition systems may require frequent recalibration to maintain constant deposition rates. Material deposition can be accomplished in a variety of ways. The individual solder masks have relatively high viscosities (20,000 - 40,000 cPs), but will eventually flow and selflevel. Mixed solder masks have similar characteristics. They are often dispensed using syringes or air-over-plunger positive displacement systems. For manual deposition, spatulas have been used to place solder mask exactly where it is needed. One automated testbed system used hardware from EFD271 including a cartridge reservoir with a pressurized plunger and a pneumatically controlled 725DA-SS dispense valve, depositing solder mask through a removable needle. With such a system, flow rates can be varied by needle size, reservoir pressure, and valve opening setting. The reservoirs are designed to accommodate 180 cc or 360 cc Semco cartridges, which are relatively standard packaging for solder masks. Larger reservoirs are available to allow less frequent container changes; one-liter and one-gallon systems are available from EFD. Air bubbles must be avoided inside the reservoirs, or gaps in deposition patterns will occur. After material has been deposited, it will flow and self-level if allowed to stand for a few minutes before curing. The materials' viscosities change significantly with temperature; heated material flows much more easily. To cure these materials, a Dymax 2000-EC ultraviolet floodlight system has been used272. This system provides a peak intensity of 70 mW/cm2 over a 200 mm-square footprint. This light uses a 400 watt metal halide bulb in a line-source configuration, filtered to

127

provide mainly illumination from 300-500 nm. This system includes an added shutter to block the output of the lamp when it is not needed, since the lamp should not be power cycled often. Mixed solder mask of 1 mm thickness can be cured completely in three fortyfive-second exposures to this light system. Parts are normally rotated ninety degrees between exposures to ensure a complete cure under the shadows of shutter panels. Thinner layers can be cured much more rapidly, but insufficient cure time produces unmachinably gummy material. If layers thicker than 1 mm are desired, multiple thin layers of solder mask must be separately deposited and cured. The curing process tends to generate large amounts of heat, both from absorption of ultraviolet radiation and from the exothermic curing reaction of the solder masks. This can damage the geometry of part material or previously shaped support material. When liquid solder mask is deposited onto warm, previously cured solder mask, its lower viscosity will ensure that it flows more easily, spreading into small features and producing a smooth surface finish. After the solder masks are cured, they immediately begin to absorb moisture from the atmosphere. This results in a gummy consistency after several hours, limiting the machinability of the materials. Some experiments were performed where cured solder mask was placed in desiccators to extend the time between curing and machining, but these experiments were unsuccessful. The solder mask surface cracks when exposed to very dry air, ruining the surface finish. Therefore, solder masks should be machined immediately after curing and then covered by another material to protect them from atmospheric moisture. These materials can be dissolved in water, given sufficient exposure time. Physical agitation and ultrasonic energy can be used to accelerate the process, as can replacement of water after it has dissolved a substantial amount of solder mask. A dissolution test was made to determine the time required to dissolve the material. A 5 x 30 x 30 mm slab of 4:1 mixed solder mask dissolved completely in 400 mL of agitated room-temperature tap water in about seventy minutes. The time to dissolve solder mask support material from parts depends greatly on the geometry of the parts. If solder mask must be removed from small internal passages or cavities, then the dissolving process may require several days, frequent manual reorientation, and water changes. These solder mask materials are intended to remain intact during area-soldering processes involving exposure to high temperatures. The Tech Spray and ElectroLite solder masks are rated to withstand 343C, and the Dymax solder mask is rated to withstand wave-solder

128

processes. While the materials may remain intact when exposed to those temperatures, they do not retain the geometry of machined features when so heated - sharp features will soften or begin to flow. An earlier 3.5:1 ElectroLite ELC4497:Dymax 948 mixed solder mask was tested for the effect of high bulk temperatures on geometrical features. After an hour's exposure to a 90C oven, there was no measurable effect on geometry. After an hour in a 95C oven, the material had flowed enough for a square to change by about 1% in length; after exposure to 100C, a square changed by 2.5% in length. These tests indicate that if deposited part material is hot enough to heat the solder mask support material to temperatures above 90C, then machined features will be affected. This has been verified experimentally. Wax heated to 125C was deposited onto mixed solder mask, and the geometry of machined solder mask features was distorted by exposure to heat. As a result of this high-temperature distortion, these materials were not selected for use as support materials for injection-moldable thermoplastic part materials. These solder masks, however, were later used as support material for ultraviolet-curable part materials, and as support material for thermoplastic waxes to make molds for the Mold SDM process; both of these uses are described later. Researchers at Advanced Ceramics Research in Tucson, Arizona, developed support materials called ACR200 and ACR50 for use with injection moldable thermoplastic part materials. ACR200 is a proprietary non-ionic, water-soluble, machinable thermoplastic material, which is deposited via extrusion at high temperatures and high pressures 273. ACR50 is a similar material with a lower melt viscosity. These materials have been used as support materials for extruded polycarbonate part materials by researchers at Carnegie Mellon University274. This material combination was used to make a polycarbonate impeller part via Polymer Shape Deposition Manufacturing. Unfortunately, the high temperatures associated with the extrusion of polycarbonate tend to cause significant softening of ACR200, producing replicated geometry with poor surface quality.

A.2. UV-Curable Part and Support Materials for Fast Cycle Times Since good support materials could not be found for use with injection-molding thermoplastic materials, other applications of Shape Deposition Manufacturing using polymers were investigated. The earliest successful application was the development of a prototyping environment for the manufacture of metallic parts via SDM. This required the

129

identification of part and support materials with short cycle times. Different ultravioletcuring thermoset resins were considered as part materials to complement the solder mask materials identified as potential support materials. The first material examined as an ultraviolet-curing part material was Du Pont Somos 3110 Stereolithography resin. This is an acrylate resin which is hard and relatively brittle. Acrylate Stereolithography resins are some of the earliest materials developed for use with 3D Systems' Stereolithography process. These materials are designed to cure upon exposure to ultraviolet light from a laser. In that application, a layer of material is not completely cured in a single exposure, but is partially cured at first, and then cured more as a second layer of material is deposited and illuminated. This provides good bonding between successive layers of material. After removal from the SLA machine, the curing process is completed in a postcuring apparatus with ultraviolet lights. Acrylate Stereolithography resins are oxygen inhibited, meaning that surface layers exposed to atmospheric oxygen will not cure275. This is considered a drawback for Stereolithography purposes since it might result in surfaces which were not fully cured or tacky even after the ultraviolet curing and postcuring processes. For Shape Deposition Manufacturing, however, this should not cause problems and may even have a beneficial effect. One reason is that exposed surfaces of part material are always machined in SDM any uncured top portion of a layer will be removed during this machining process. Part surfaces built via replication of support material surfaces are not exposed to oxygen at those surfaces, so they should also cure completely. Therefore, the main drawback present in Stereolithography should not apply to SDM. A more beneficial possibility is that successive layers of material may bond better to one another since the earlier layers' surfaces may not have completely cured at first. This allows thick sections to be built monolithically even though only thin layers can be cured under a given ultraviolet lighting system. For use in Polymer SDM, the Somos 3110 resin was deposited by an EFD 740V dispense valve attached to a cartridge reservoir. It was cured with the same Dymax 2000-EC ultraviolet flood light used for curing solder mask. The 3110 can be cured into a hard, rigid, machinable material and bonds well to solder mask materials. It is not affected by the solder mask support removal technique of soaking the part in water. However, when layers of Somos 3110 are deposited and cured over solder mask, the combined layers

130

delaminate from their substrates. This is believed to occur because of the high shrinkage of the 3110 material. Since the 3110 material's shrinkage characteristics seemed too extreme, a different Stereolithography resin with much lower shrinkage was selected: Du Pont Somos 2100276. Manufacturer's specifications for both materials are shown in Table A.2. These properties were measured after ultraviolet (3110) or ultraviolet and thermal (2100) postcuring was performed. Tensile Tensile Notched Volumetric Strength Modulus Elongation Izod Shrinkage (%) (MPa) (MPa) (J/m) (%) 21 7.1 810 37 9.2 46 15 150 6.2 3

Material Somos 3110 Somos 2100

Reference 277 278

Table A.2. Acrylate Stereolithography Resin Part Materials The Somos 2100 material cures into a relatively hard, machinable material although it was much softer than cured Somos 3110 resin. Somos 2100 also bonds well to solder mask support material, but causes much less delamination of parts from their substrates. Its lower shrinkage should limit this tendency and reduce residual stresses in parts compared to those made of Somos 3110. Somos 2100 is also unaffected by the solder mask support material removal process of soaking parts in water. After support material removal, the Somos 2100 resin can be thermally postcured to increase its tensile strength, impact strength, and ductility. The recommended postcure is thirty to sixty minutes at a temperature of 130C. The pairing of Somos 2100 part material and mixed solder mask support material is a relatively successful material combination for Polymer Shape Deposition Manufacturing. An object made from Somos 2100 and solder mask is shown in Figure 4.1. This part duplicates an injection-molding tool made from stainless steel via SDM; a serpentine cooling channel is located in the middle of the part. The stainless steel tool is shown in Figure 4.2. The Somos part is roughly 75 x 50 x 20 mm. It was constructed on a 2 mmthick release layer of support material. Somos 2100 was used as the part material, and mixed ElectroLite ELC4497 / Dymax 948 solder mask was used as the support material. Material was deposited via EFD dispensers mounted on an Adept SCARA robot although some additional manual deposition was used to correct errors. To begin building the part, a

131

release layer of solder mask was deposited, cured, and milled. The part itself was divided into 1 mm-thick layers for construction . The first several layers were made by depositing, curing, and milling part material, and then depositing and curing support material around it. The liquid part material had too low a viscosity to retain its shape before curing, so it was restrained by additional retaining walls made of solder mask. Each cured layer was completed by planing the top flat before depositing the next layer. This process was continued until the layers with cooling channels were finished. The remaining layers of the part were deposited sequentially and then machined all at once. No support material was deposited around this upper segment of the part. The part was then sawn off its substrate and immersed in water to remove the support material. This process was accelerated via agitation of the water bath and some use of an ultrasonic bath. Deposition and cutting required approximately twenty hours, and the etching process required about seventy hours. More aggressive support material removal techniques could have reduced the latter time. The polymer part duplicated the features of the metallic part extremely well, including the same "Christmas-tree" stairsteps which were present at layer boundaries. The construction of a larger rectangular test part pointed out some possible problems with the material combination of Somos 2100 and solder mask. The most important is that the Stereolithography resin shrinks too much when an entire layer is deposited and cured during Shape Deposition Manufacturing. In the laser-based Stereolithography processes, special build styles can be used to work around the high bulk shrinkage of such materials. Different hatching or outlining schemes are used to reduce part shrinkage to small fractions of the bulk material shrinkage of the resins279. However, these techniques cannot be implemented using the flood ultraviolet light sources used in Shape Deposition Manufacturing. While path-based curing with lasers could be implemented for Polymer SDM, this represents such a complication of the process that it has never been seriously considered. One alternative might be to use a moving mask with a slit to produce curing along a moving wave front280; this might allow uncured material to compensate for the shrinkage of newly cured resin. Another issue with the combination of Somos 2100 and solder mask is the strength of the Somos 2100 material. While its impact performance is good, its strength is quite low, even after thermal postcuring. As a result of this poor strength, other ultraviolet-curable resins were investigated to replace Somos 2100 as a part material to build stronger fast-cycle-time parts. The first alternative part material investigated was a mixture of acrylate

132

Stereolithography resins; later materials investigated included epoxy Stereolithography resins, and epoxy ultraviolet-curable adhesives. The first part material examined was a mixture of the previously described Somos 2100 and Somos 3110 materials. It was hoped that the combination of the stronger 3110 with the lower-shrinkage 2100 material would result in an improved mixture. However, this mixture still had a significantly higher shrinkage than the 2100 resin alone did. Since even the shrinkage of Somos 2100 was only marginal for the SDM process, this part material was not investigated any further. The next material was an epoxy Stereolithography resin. Epoxy resins were introduced after the original acrylate Stereolithography resins, and had some advantages 281. The first is that they tend to shrink less, especially when processed via new optimized build styles in the Stereolithography process. The second is that they cure via a cationic mechanism rather than the free-radical curing process used by the acrylate materials. This different curing mechanism both eliminates the oxygen-inhibition problem which led to tacky surfaces and also provides a continuing cure mechanism even after ultraviolet illumination has ceased. These curing differences usually mean that epoxy resins have better interlayer adhesion than acrylate Stereolithography resins. An epoxy resin tested as a part material was Ciba 5180282. This material could be deposited and cured, but was not compatible with solder mask support material. When Ciba 5180 was deposited over cured, machined solder mask, the bulk of the 5180 would cure, but the interface between the two materials would remain liquid. Epoxy Stereolithography resins can be contaminated by exposure to even small amounts of acrylate resins283. 3D Systems reports that contamination causes the epoxy resins to lose photo-sensitivity if contaminated by 0.1% of acrylate resin 284 . Stereolithography resin vendors prescribe complete equipment cleaning procedures to prevent contamination when a machine is being switched from an acrylate resin to an epoxy resin. After verifying this incompatibility, research into epoxy Stereolithography resin part materials was discontinued. The only other category of commercially available Stereolithography resins is based on vinylether chemistry. This type of material also reacts with acrylate resins285, and was therefore not investigated as a part material for Polymer SDM. The last class of ultraviolet-curable part materials examined was ultraviolet-curing adhesives. There are many manufacturers of ultraviolet-curing adhesives, including Dymax (Torrington, Connecticut), UVEXS (Sunnyvale, California), Grace Specialty Polymer

133

(Lexington, Massachusetts), Loctite North America (Newington, Connecticut), Electronic Materials Inc. (Breckenridge, Colorado), and Epoxy Technology, Inc. (Billerica, Massachusetts). The main material tested was Electronic Materials' Emcast 1722 Gen II UV-Curable Epoxy286. This material is chemically similar to epoxy Stereolithography resins; it can be cured by exposure to ultraviolet light, but also has a secondary "dark-cure" mechanism which continues the curing process even after exposure to ultraviolet light. This material cures to a hard consistency, but not immediately. Immediately after exposure to ultraviolet light, the surface of the material is still flexible, but it becomes much harder over a period of tens of minutes. However, the material cannot be cured in contact with solder mask. A liquid layer remains between cured epoxy and previously cured and machined solder mask; this seems to be the same cure inhibition which affected epoxy Stereolithography resins. After this result, experimentation with ultraviolet-curing part materials was discontinued, in favor of work on castable thermoset part materials described in the next section. If ultraviolet curing adhesives again become an active area of study, a material that warrants examination is the 518 series of adhesives from UVEXS287. This is a urethane acrylate resin which cures hard and moderately strong and may be curable in reasonable thicknesses. Its base chemistry should make it more compatible with the solder mask materials than any epoxy or vinylether resin.

A.3. Completely Removable Support Materials for Castable Part Materials As mentioned in Chapter Two, researchers at Carnegie Mellon University identified a castable polyurethane part material called LUC-4180; this was selected for use in wearable computers for divers. The support material used with it is a machinable investment casting pattern wax called "Master Protowax" from the Kindt-Collins company. This wax is normally removed by heating a part and its support material in an oven. Much of the wax melts, and most of the remaining wax can be removed manually. Unfortunately, a waxy residue usually remains even after this removal process. This residue and the manual support material removal required are the principal drawbacks to this material. Completely removable support materials were needed, especially for the production of parts with fine features or assembled mechanisms. This was the next research avenue pursued at Stanford University. Several potential support materials and support material removal methods were identified and evaluated. These included solder masks dissolved by water, water-soluble

134

waxes dissolved by water, epoxies dissolved by solvents, and waxes dissolved by organic solvents. Several characteristics of solder mask would make it a good support material for polyurethane part material. It can be deposited and machined, is not affected by exposure to the deposition temperatures of the polyurethane material, and can be dissolved easily by water. Therefore, extensive tests with this material combination were conducted, with inconclusive results. Some high-quality polyurethane material could be deposited over solder mask, but polyurethane was often damaged by this contact. Bubbles and other surface defects often occurred at interfaces between polyurethane and solder mask. The hygroscopic nature of solder mask probably resulted in a high absorbed moisture content at the surface, which interfered with the curing process of polyurethane. Polyurethanes are very sensitive to the presence of moisture, both during storage of raw materials and during the curing process. Special housekeeping procedures must be followed to maintain polyurethane components free from moisture. These include keeping containers closed as well as blanketing their contents with dry gas to displace moist air. This extreme sensitivity to moisture renders polyurethane incompatible with solder mask support materials. An example of the poor surface finish often encountered with this combination is shown in Figure A.1. Although some high-quality surfaces were made this way, the process could not be made reliable or robust.

Figure A.1. Polyurethane Surface Cured in Contact with Solder Mask Following the experimentation with solder mask support material, other easily removable support materials were investigated for use with polyurethane part materials. Most of these

135

materials were waxes intended for different applications: investment casting pattern cores, temporary adhesives, or temporary fixturing compounds. Investment casting pattern cores act as place holders for desired cavities inside investment casting patterns; these cores are normally cast or shaped from water-soluble waxes. After cores are shaped, insoluble investment casting waxes are cast or injected around them to form investment casting patterns. These patterns are soaked in water to remove the core materials, and then the finished patterns are invested and cast. Investment casting core materials are available from the Kindt-Collins company (Cleveland, Ohio), M. Argueso & Co., Inc. (Mamaroneck, New York), and Dussek-Campbell-Yates (Chicago, Illinois). Temporary adhesives and temporary fixturing compounds are used to hold objects during secondary operations such as sawing or machining. Semiconductor wafers, single crystals, ceramic parts, and thinwalled or honeycombed metal parts are often held this way during such operations. Aremco Products, Inc. (Ossining, New York) makes water- and solvent-soluble temporary adhesives; the J. H. Young Company, Inc. (Rochester, New York) makes solvent-soluble adhesives and waxes; and the Rigidax division of M. Argueso & Co., Inc. (Mamaroneck, New York) makes temporary fixturing waxes. Most of the water-soluble waxes are made of polyethylene glycol polymer, and many are highly filled. A variety of these materials were tested, including Yates 14A4, Argueso Cerita 28-8, Argueso Cerita 28-8Bl, Kindt-Collins KC2277A, Aremco Crystalbond 555, Crystalbond 509, and Crystalbond 590. Out of these materials, Kindt-Collins KC2277A water-soluble wax and Aremco Crystalbond 555 water-soluble temporary adhesive288 were tested the most extensively. Both materials experienced persistent problems with casting porosity; void-free castings could not be consistently obtained. Both materials are machinable, but careful choice of machining speeds and feeds must be made. Some of the same interface quality problems encountered with solder mask support materials occurred with these water-soluble materials as well. Both waxes are less hygroscopic than solder mask; they do not noticeably absorb moisture from the atmosphere. Crystalbond 555 was preferred over KC2277A but the combination of wax surface voids and possible moisture contamination of polyurethanes results in inconsistent polyurethane surface quality using either material. Despite these issues, an embedded mechanism was constructed using Adtech LUC-4180 polyurethane part material and Crystalbond 555 support material. This was a prototype of an input device for a wearable computer: a rotary selector switch with integrated push-button action (Figure 6.3).

136

The surface interactions between polyurethane and water-soluble support materials led to consideration of castable thermoset polymer support materials for use with castable thermoset polymer part materials. This requires differential solvent resistance and careful choice of solvents and support materials. Thermoset resins often swell in response to solvents; this type of swelling may damage part material features with internal support material and must therefore be avoided. Certain strippable epoxies were evaluated for use as support materials. Standard polyurethane and epoxy polymers were also considered for use as support materials removable by special organic solvents. Emerson & Cuming advertises a pair of epoxy encapsulant materials as strippable. These are STYCAST 3020 and 3050 epoxy encapsulants, intended for potting or encapsulating electrical and electronic components. Both are described as "readily machinable" and easily removed by soaking overnight in white vinegar at 70C289,290. This removal procedure was tested on STYCAST 3050 and found not to work. While a color change did occur during such a soak, no softening or swelling was observed. There were reports of past successful use of this material for optical fixturing, where a tenth of a millimeter thickness may have been soluble in vinegar. After these poor test results using vinegar and special epoxies, standard thermoset resins and more aggressive solvents were investigated. A variety of solvents are commercially available for dissolving polyurethane and epoxy resins. A full line of these are available from Dynaloy, Inc. (Hanover, New Jersey), and some similar materials may be available from W. R. Grace and BASF. Materials are available which can dissolve cured polyurethane or anhydride-curing epoxy materials without harming amide-curing epoxies291. However, these solvents have significant drawbacks. Most of them are toxic or flammable, requiring special handling and heating equipment. Some of the strongest solvents are acutely toxic and highly flammable, requiring explosion-proof equipment. The multiple hazards of these solvents ended further research into their use for support material removal. The difficulty of finding safe solvents ended research into the use of castable thermoset polymers as support materials. Since these materials were so difficult to dissolve, research shifted to support materials known to be much easier to dissolve: machinable waxes. Machinable waxes were the original support materials used by researchers at Carnegie Mellon University; their latest one is Kindt-Collins "Master Protowax." It is easy to deposit, relatively machinable, and chemically compatible with polyurethane part materials; its primary drawback is the difficulty of complete removal. In an effort to improve this

137

situation, solvent systems were investigated. The requirement was that the wax support material be removed without swelling or otherwise damaging polyurethane part material. This is a similar situation to that found in the Sanders process, where a wax must be dissolved without harming another material. The Sanders process uses a product called BIOACT VSO from Petroferm (Fernandina Beach, Florida) to accomplish this task292. Petroferm sells a variety of materials for dissolving wax and pitch materials in various applications. A material called "BIOACT 280"293 was evaluated as a solvent for the Protowax support material. This product is relatively non-toxic and has a low evaporation rate. As it dissolves wax, the solvent thickens over time; its ability to dissolve wax is reduced, and it must be replaced. Industrially, the used solvent and dissolved wax is normally burned for its fuel content. BIOACT 280 dissolves Protowax completely when heated slightly above the melting point of the wax, without leaving a residue on parts. This support material removal method results in a materials combination allowing complete removability of support material. This allows the production of objects with moving parts or small internal cavities; some example parts are shown in Figure 4.3. These objects were built already assembled, using LUC-4180 polyurethane part material and Protowax support material which was removed by heating in an oven and soaking in heated BIOACT 280. This LUC-4180, Protowax, and BIOACT 280 system has been used to make several parts, but there are some limitations. One difficulty is the relatively poor adhesion between LUC4180 and Protowax. When broad, flat polyurethane layers are attached only at their bases to wax surfaces, they are easily delaminated from the support material. In the middle of a build, this is not a problem since previous layers are surrounded by part material or anchored to previous part material. At the first layer of a part, however, the attachment between polyurethane parts and the support material underneath them is relatively weak. This generally requires the use of temporary anchors, where holes or trenches are machined into the support material before part material is cast. These features will support milling loads much better than flat interfaces would, preventing delamination during machining of the first layer. These features are then machined off the part later, either during the build or after part completion. Another difficulty with the LUC-4180 and Protowax combination occurs during support material removal. Polyurethane parts soften during the removal process, returning to their normal hardnesses only after cooling to room temperature. Since they are so soft and weak, they cannot support their own weights and warp greatly during the removal process. This occurs because of the polyurethane's very low glass-transition temperature of 60-

138

66C, which is below the solvent bath temperature. While they return to roughly their original shapes after cooling, some warpage does typically occur. Warpage is especially noticeable in multilayer or planar parts; most Polymer SDM parts fall into these categories. In an effort to reduce the temperature of the support material removal process, several approaches were examined. These included both using a lower temperature BIOACT 280 bath and replacing BIOACT 280 with other solvents which could be used at a lower temperature. BIOACT 280 will dissolve Protowax at temperatures below the 60C melting point of the wax, although this process takes a long time and is greatly aided by some manual wax removal. For example, an LUC-4180 assembly was soaked in BIOACT 280 at temperatures of up to 55C for a week before the support material was completely removed. After such long exposure times, the LUC-4180 polyurethane material slightly softens; it is rigid but not as hard as it would be after a short exposure to the solvent. Another way to increase the speed of the removal process would be to provide ultrasonic agitation of the solvent bath. This could be as simple as placing a beaker of BIOACT within a water bath in an ultrasonic cleaner. A superior solution would be an ultrasonic system optimized for BIOACT as a working fluid; such systems are available as custom models from Sonicor Instrument Corporation (Copiague, New York). Either long soaking times or the use of more aggressive bath agitation would permit the use of BIOACT 280 at lower temperatures, which should reduce part warpage. The other alternative to BIOACT 280 is the use of another solvent which can dissolve wax at a lower temperature. Several other Petroferm products were tested by that company for use with this materials combination294. These products included BIOACT 105, 120, 121, 144, 200, and 286, Solvating Agent 70, AXAREL 2200, and AXAREL 4100. These materials were tested at room temperature, and the most promising were BIOACT 120 and 121; AXAREL 4100 was also quite good. These three materials dissolved Protowax faster than BIOACT 280 at equivalent temperatures, and did not have any obvious ill effects on LUC-4180 polyurethane. While BIOACT 120 and 121 cannot be heated during use because of low flashpoints, AXAREL 4100 could probably be heated somewhat for improved performance. Another product tested for use as a solvent was EnSolv Precision Vapor Degreasing and Cleaning Solvent from Enviro Tech International (Melrose Park, Illinois)295. This product is designed for use in vapor degreasers, and dissolves Protowax very rapidly in this application. Unfortunately, it also reduces the surface hardness of polyurethane parts, and tends to warp them. At lower temperatures, it dissolves wax much faster than BIOACT 280 at equivalent temperatures, but always warps polyurethane parts

139

and softens their surfaces. It also evaporates rapidly and must only be used under a fume hood. Overall, most of these alternative solvents are probably inferior to BIOACT 280 for support material removal. The only material which warrants further study is AXAREL 4100, which should be usable at lower temperatures which would lessen part warpage.

A.4. Castable Part Materials for Short Cycle Times While LUC-4180 polyurethane and Protowax support material are a useful material combination, this combination is not perfect. A major drawback is the long cure time required between casting and machining the polyurethane part material. To shorten the production cycle for Polymer SDM parts, a variety of faster-curing thermoset resins were investigated. Many polyurethane resins with short cure times are commercially available. Many of these fast-curing materials are intended for use with heated molds for high-volume, short cycletime production applications. Some materials, however, are intended for the prototyping market for use with silicone molds and soft tooling. These materials are designed to cure rapidly at room temperature so that small series of prototypes may be molded in a short time. Some fast-curing epoxy materials exist, but a much greater variety of polyurethane prototyping materials are available. One such series of materials is Ciba's "Parts in Minutes" line of fast-curing polyurethanes296. These materials gel in one or two minutes, and can be demolded after 15-90 minutes, depending on the materials. The materials are designed for use with automated dispensing systems rather than manual mixing. Two of these materials were tested for use in Polymer SDM: RP 6450 and RP 6455. RP 6450 is the fastest-curing of these materials - it gels in 40-50 seconds and can be machined after 30-60 minutes. The high cure speed has a couple of drawbacks. The material gels so fast that vacuum degassing cannot be performed. While equipment is available to prevent air entrainment during stirring, air bubbles cannot be removed from part cavities; small pores are typically found at interior corners after casting. The material's fast-curing chemistry results in very high cure exotherms; the center of a twelve-millimeter-thick sample reached a temperature of 125C a minute after beginning to gel. This high temperature can melt wax support materials, leading to geometrical distortion. Thicker castings of this material would be expected to reach even higher peak temperatures. The other material tested, RP 6455,

140

cures slightly more slowly. It gels in about 100 seconds, and can be machined in 90-120 minutes. While somewhat slower, it still gels too fast to allow vacuum degassing. Since void-free surfaces are required for high-quality prototype parts, the inability to vacuum degas these materials forced an end to research into such fast-curing materials. In early Polymer SDM research at Carnegie Mellon University, part materials from Adtech and Ciba were used. These companies offer polyurethanes with intermediate cure times, faster than LUC-4180 but slower than the Ciba Parts in Minutes materials. Two such materials were tested at Stanford for use in Polymer SDM. The first material was Adtech LUC-4102297, which had been used for a short period of time at CMU. This material can be machined about an hour after curing, and gels slowly enough to allow vacuum degassing. However, it is very highly filled with an abrasive filler. This makes mixing difficult, and leads to rapid wear of dispensing equipment, cutting tools, and machines. The fillers also settle over time, requiring agitation to maintain them in suspension. The other material tested was Ciba Pro-Cast 20298. This material can be machined about ninety minutes after curing, gels just slowly enough to degas, and has a moderate cure exotherm. An eleven-millimeter-thick section reached a peak temperature of 40C, so reasonable thicknesses of this material should not remelt support materials while curing. As with LUC-4102, however, this material is highly filled, and the high filler loading is inconvenient for storage, mixing, deposition, and milling operations. The previous difficulties with filled materials provoked an investigation into unfilled or lightly filled resins with intermediate cure times. This search led to Ciba TDT 205-3, another polyurethane material299. It had been used as a base material for early filled resins formulated at CMU, and was recommended by Ciba as an unfilled version of Pro-Cast 20. It does have a small particulate component, but both components have low viscosities and are non-abrasive. The material gels in about eight minutes, which is slow enough to allow vacuum degassing. The uncured material has very good air-release properties, and can be vacuum degassed very well in a short time. Its intermediate cure time is accompanied by an intermediate cure exotherm. A six-millimeter-thick sample reached 48C during gellation, while a nineteen-millimeter-thick sample reached 75C after gelling. The material can be machined in less than two hours after casting. The material combination of TDT 205-3 part material, Protowax support material, and BIOACT 280 solvent for support material was used to make the assembly shown in Figure A.2.

141

Figure A.2. TDT 205-3 Polyurethane Assembly However, TDT 205-3's high exotherm promotes melting of or chemical interaction with the Protowax support material. This interaction leads to surface color changes in cured polyurethane as well as texture changes. Fine machined details in wax surfaces seem to be smoothed or blurred slightly during replication onto polyurethane surfaces. This problem was solved through the use of a different support material with a higher melting point. This new support material is Kindt-Collins KC3230A, which is a custom blend of 25% Master File-a-wax and 75% Master Protowax. This wax has a melting point of about 80C and machines significantly better than Master Protowax. Its higher melting point reduces melting and other interactions between polyurethane and wax, leading to higher surface quality. This support material can still be removed by BIOACT 280 although higher removal temperatures are required. The material combination of TDT 205-3, KC3230A, and BIOACT 280 is the current, preferred Polymer SDM materials combination. It was used to make the mechanisms which are shown in Figure 4.4. This materials combination resulted from a search for castable resins to allow short cycle times for Polymer SDM parts. This was successful since the part material cure time was reduced from the twelve hours of LUC-4180 to two hours for TDT 205-3. This allows the production of multiple layers within a single shift, instead of less than one layer per shift. A side benefit of the new part material is increased interface tensile strength. While the LUC-4180 has a high bulk strength, its interlayer interfaces are relatively weak. TDT 2053 has a lower bulk strength than LUC-4180, but its interlayer interfaces are substantially stronger than those of the other material. Tests of these parameters are described in Chapter Five.

142

A.5. Part Materials for Higher Interlayer Strength To overcome the problem of LUC-4180 polyurethane parts delaminating during handling, materials with higher interlayer strength were investigated. One material identified for this application is Adtech EE-501/530 epoxy300. It had been used in early experiments at CMU, but was examined for this application because it had exhibited much better adhesion than LUC-4180 when used in multimaterial testing at Stanford. To learn more about achievable interlayer strengths of this and previous materials, several rounds of tensile testing were conducted. These experiments are described in detail in Chapter Five. The conclusion was that LUC-4180 has very weak interlayer interfaces, and that those of both TDT 205-3 and 501/530 are much stronger. Adtech 501/530 has a bulk tensile strength intermediate between those of LUC-4180 and TDT 205-3, but its interlayer strength is stronger than that of TDT 205-3 by a factor of two. This makes the 501/530 epoxy material the strongest material identified for Polymer Shape Deposition Manufacturing of multilayer structures. It can be used with either the original Kindt-Collins Protowax support material or the newer KC3230A support material identified for TDT 205-3. Both materials can be removed with BIOACT 280 without damage to the 501/530 part material. However, the 501/530 epoxy's high strength is accompanied by a few drawbacks in ease of use. It is highly filled; these fillers settle during storage, requiring stirring to put them back into suspension for use. The fillers are described as nonabrasive, and no resulting tool wear problems have been noted. Compared to the slightly filled TDT 205-3 material, this heavily filled resin has a much higher mixed viscosity. This makes filling complicated cavities more difficult, and lengthens the degassing process. The epoxy takes much longer to cure than the TDT 205-3 material; it cannot be machined for roughly 24 hours after casting. While this slow speed greatly lengthens production cycles, it does allow degassing operations to be completed before the material begins to gel. These various issues have kept 501/530 from being used much for Polymer Shape Deposition Manufacturing although it has the highest tensile strength of any material in current use.

143

APPENDIX B TENSILE TESTING NOTES B.1. Insufficient Cure Time before Tensile Testing All of the tensile test data presented in the text describe specimens which were aged for two weeks after casting. This delay was allowed to assure that the specimens were completely cured. One other series of monolithic LUC-4180 specimens was tested almost immediately after casting, but was found to have widely varying and lower tensile strengths, probably indicating that the specimens had not fully cured. The resulting tensile test values were not included in the summary data in the text, but are presented for reference in Table B.1. These specimens were machined about 18 hours after the initial casting, and were tensile tested about 25 hours after machining. The first specimen was tested at a test speed of 51 mm/min., while the remaining specimens were tested at test speeds of 5.1 mm/min. All of the specimens were tested at Stanford University, and plots of load versus crosshead position were generated by a chart recorder. Specimen # Strength 1 33.1 2 37.5 3 19.9 4 38.3 5 36.4 6 25.6 Average 32 Engineering stress at yield, MPa Table B.1. Tensile Strength of Incompletely Cured LUC-4180 One possible problem with these test results was that the first four specimens were tested with an incorrect set of grips on the tensile machine. The grips were not aligned parallel with one another, so the specimens were being twisted as well as pulled axially. This error was corrected before testing the last two specimens, but its effect on the strengths of the initial specimens is unknown.

144

B.2. Material Inconsistencies during Tensile Tests As mentioned in the text, LUC-4180 polyurethane of varying age was used for the tensile test specimens. This choice was made to duplicate the conditions experienced during the manufacture of actual prototype parts. Some polyurethane used for specimens was taken from newly opened containers, while other polyurethane was from containers which had previously been opened. This may have an effect on the quality of the resulting specimens since polyurethane components are very sensitive to moisture and each container opening represents exposure to atmospheric moisture. Each time a container was opened, it was gas-blanketed with dry argon gas before being closed and stored. This precaution should have extended the life of the containers of polyurethane, but there still may have been an eventual effect on the quality of the resulting cast material. One additional aspect of the LUC-4180 material used was that some material arrived crystallized from the factory. This happened to some batches of Part B of the material when they were shipped during the winter and not heated in transit. After arrival, they were heated on hot plates or in ovens until the crystals melted (a temperature of 52C is recommended by Adtech). After the crystals melted, the material was stirred and then allowed to cool slowly. Adtech claims that this should not affect the properties of the resulting cast polyurethane. The amount of time containers had been open before being used to cast tensile specimens is tabulated in Table B.2. Series I through IV of the LUC-4180 polyurethane, and the TDT 205-3 and 501/530 materials represent the tensile specimens used in the material comparison section. The "Grit Blast" LUC-4180 column represents the series of specimens whose trenches were gritblasted before performing the second casting operation for dual-interface specimens. The "BIOACT" column of tensile specimens was used to investigate the effects of the support material removal process. Finally, the "Not Cured" LUC-4180 column represents the tensile test series described in section B.1 where insufficient time was allowed for the specimens to cure completely. An "X" in a column of the table indicates that none of the indicated type of specimens were made in the series. In the row marked "Crystallization," "CRX" indicates that the batch of polyurethane had crystallized during shipping and required heating for decrystallization.

145

Ser. Ser. Ser. Ser. Grit- BIO- Not I II III IV TDT 501/ Blast ACT Cured LUC- LUC- LUC- LUC- 205-3 530 LUC- LUC- LUC4180 4180 4180 4180 4180 4180 4180 Monolithic Time Before 15 Casting wks. Crystallized? Dual-Interface Time Before 15 1st Casting wk. Time Before 20 2nd Casting wks. Crystallized? 2 days 1 hr. 2 hr. 11 days 10 wks. CRX 6 mos.

X X

1 hr. 1 day

1 day 2 days CRX

2 hr. 6 hr.

11 days 12 days

1 10 day wks. 2 10 day wks. CRX CRX

X X

Table B.2. Container Open-Times Before Casting Specimens Of all LUC-4180 test specimens, those in Series III were cast from the newest polyurethane. Monolithic specimens in this series had slightly higher tensile strengths than monolithic specimens of other series, but only by about 4%. Strength differences in dualinterface specimens between this and other series, however, are less easy to measure. While the strengths of this series of specimens were lower than those of other series, all of them did survive handling. Dual-interface specimens from other series had higher measured strengths, but only one third of the specimens in the other series survived to be measured. Overall, the effect of newer or older polyurethane material is unclear; there may have been a difference in strength, but it was probably a small one. There was a possible problem with the first casting of the Series IV LUC-4180 dualinterface tensile specimens. An aluminum foil cover meant to shield the newly cast material from debris fell into partial contact with the polyurethane soon after it was cast. Several square centimeters of material cured abnormally at the surface. Instead of having a hard, shiny coating, the polyurethane surface was duller and much softer. During the machining process, most of this abnormal material should have been removed, and all of the completed tensile specimens did appear normal and seemed to have normal hardness. However, this phenomenon may have had some effect on the tensile strength values reported for this series of specimens.

146

B.3. Timing of Tensile Testing Activities Besides the incompletely cured specimens described in section B.1, most tensile testing operations were performed on similar timetables, but there were some variations. The timing variations probably did not have a significant effect on the results of the tensile testing. The intervals between casting and cutting out monolithic specimens, between casting and trenching dual-interface specimens, between casting material into the trenches and cutting out dual-interface specimens, and between cutting out specimens and tensile testing specimens are listed in Table B.3. Ser. Ser. Ser. Ser. Grit- BIO- Not I II III IV TDT 501/ Blast ACT Cured LUC- LUC- LUC- LUC- 205-3 530 LUC- LUC- LUC 4180 4180 4180 4180 4180 4180 4180 Monolithic Delay before ~26 cutting out hrs. Delay before 15 testing days Dual-Interface Delay before ~25 cutting trench hrs. Delay before 27 cutting out hrs. Delay before 14 testing days 4 days 14 days 18 hrs. 14 days 6.5 hrs. 17 days 46 hrs. 16 days 19 hrs. 15 days 18 hrs. ~25 hrs.

X X

X X

X X X

18 hrs. 24 hrs. 13 days

22 hrs. ~19 hrs. 14 days

4 hrs. 4 hrs. 14 days

24 hrs. 39 hrs. 15 days

22 hrs. ~19 hrs. 14 days

21 hrs. 18.5 hrs. 14 days

X X X

Table B.3. Interoperation Delays during Tensile Testing

B.4. Experimental Uncertainty Experimental uncertainty can be described in terms of precision and bias. Precision errors in data can be estimated statistically, but bias errors can only be estimated by examining the experimental techniques involved301. Table B.4 summarizes estimates of the precision of the tensile testing data. The estimates of uncertainty were calculated by determining a 95% confidence interval around the sample mean tensile strengths for each group of data, assuming a t-distribution302. As shown in the table, some groups of data were much tighter than others. One interesting aspect was that some materials had more scatter in the

147

test data from monolithic tensile specimens while others had more scatter in data from dualinterface specimens. This was surprising since the statistical nature of fracture failures was expected to cause increased variability in the measured strengths of dual-interface tensile specimens (which already had "cracks" at the interfaces). Conf. Interval # of Average Low High Samples Strength Uncertainty End End 58.7 6.2 27.3 20.2 49.3 45.0 20.6 54.8 13.6

Test Series

LUC-4180 17 57.9 0.8 57.1 Monolithic LUC-4180 10 4.6 1.7 2.9 Dual-Interface TDT 205-3 6 22.0 5.3 16.7 Monolithic TDT 205-3 6 19.8 0.4 19.3 Dual-Interface 501/530 6 48.6 0.7 47.9 Monolithic 501/530 6 39.9 5.1 34.8 Dual-Interface LUC-4180 5 17.2 3.4 13.8 Gritblasted Dual-Interface LUC-4180 6 54.2 0.6 53.7 BIOACT Monolithic LUC-4180 6 10.9 2.7 8.2 BIOACT Dual-Interface Engineering Stress at Break or Yield, MPa Table B.4. Statistical Summary of Experimental Results

Estimating the size of possible bias errors in the data was more difficult. The largest component of uncertainty was the lack of recent full-load calibration data for the tensile test machine itself. A LabVIEW data acquisition system was used for the Series IV LUC4180, Gritblasted LUC-4180, BIOACT-soaked LUC-4180, TDT 205-3, and the 501/530 tests. A chart-recorder system was used for the Series II and III LUC-4180 tests, as well as the incompletely-cured LUC-4180 tests. Neither of these systems was calibrated up to the load levels present during monolithic tensile tests of any material. An estimate of the possible bias errors is presented in Table B.5.

148

Tensile Load Bias Errors Load Measurement Dimensional Bias Errors Width Measurement Thickness Measurement Width Estimation Thickness Estimation Overall Bias Error Worst Case Pythagorean Sum

Magnitude ~5% 0.4% 1.6% ~0.4% ~2.4% 10% 6%

Table B.5. Estimated Bias Errors One of the components of the estimated error in tensile load measurement was the apparent nonlinearity in loads reported by the LabVIEW data acquisition system. Estimates of linearity were based upon calibrations of the testing machine, its electronics, and the data acquisition hardware at the lower end of the load range of the system. Known weights were hung from the testing machine, and the system load output was recorded. The linearity errors from those reported loads were considered estimates of the linearity error for the whole system, although the entire load range was not tested in this way. This procedure was only performed over the range of 0-400 N, while some specimens did not break until loads of over 2400 N were applied. Over all of the LabVIEW data acquisition sessions, 1.7% linearity error was the greatest value measured. Since the system was only calibrated over such a small part of the range used during actual tests, 3% was guessed as an estimated linearity error. The chart-recorder system could also only be indirectly calibrated. The low-load scale used for dual-interface specimens and all TDT 205-3 specimens could be calibrated directly, although only at low load levels. This calibration estimated that the load scale was accurate within 1%, but the higher load levels of that scale were not calibrated. The high-load scale was not directly calibrated, it was simply checked against the calibration of the low-load scale; this check could only assure linearity of 5% over the tested range. Combining all of these factors, the measured loads can only be guessed to be accurate within 5%, with the low-failure-load LUC-4180 dual-interface specimens even less accurate. Width and thickness measurement errors result from measuring the dimensions of the specimens with calipers. Variation in measurements between users and measuring instruments resulted in uncertainty of 0.4% in width and 1.6% in thickness. Width and thickness estimation errors occurred because two series of specimens were not measured at

149

the time of their tensile tests. The preceding tensile specimens had been sufficiently uniform in size that their dimensions were assumed to be the same. Later specimens varied more widely in size, indicating that the assumed dimensions may have been incorrect. The two series of specimens whose dimensions were estimated were the Series III LUC-4180 specimens and the incompletely cured LUC-4180 specimens. The width and thickness estimation errors were calculated by assuming that the unmeasured specimens were at the outer size limits of all measured specimens. The width and thickness measurement and estimation uncertainties were added before determining overall uncertainties of the stress measurements. The uncertainties in tensile load and specimen width and length were combined to determine the uncertainty in reported tensile stresses. A worst case estimate was prepared by adding magnitudes of uncertainties using Taylor's theorem. A Pythagorean sum was prepared by evaluating the square root of the sum of the squares of such terms303. These summary bias error estimates were recorded in Table B.5. In conclusion, the reported tensile strengths may be somewhat in error in magnitude, but the trends revealed by comparisons between different materials and different processing conditions should be accurate.

150

NOTES Chapter One - Introduction J. W. Kietzman, A. G. Cooper, L. E. Weiss, L. Schultz, J. L. Lombardi, and F. B. Prinz, "Layered Manufacturing Material Issues for SDM of Polymers and Ceramics", Proceedings of the 1997 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1997, pp. 133-140.
1

Chapter Two - Background

2

The figure was published in Lee E. Weiss, "Process Overviews," in Friedrich B. Prinz, Chair, Rapid Prototyping in Europe and Japan, Vol. I, Analytical Chapters, Baltimore: International Technology Research Institute, Loyola College in Maryland, March 1997, p. 5. Used by permission.
3

J. G. Conley and H. L. Marcus, "Rapid Prototyping and Solid Freeform Fabrication,", ASME Transactions - Journal of Manufacturing Science and Engineering, vol. 119, no. 4B, November 1997, pp. 811-816.
4 J. P. Kruth, "Material Incress Manufacturing by Rapid Prototyping Techniques," CIRP Annals, vol. 40, no. 2, 1991, pp. 603-614.

Stacey Au and Paul K. Wright, "A Comparative Study of Rapid Prototyping Technology," Intelligent Concurrent Design: Fundamentals, Methodology, Modeling, and Practice, DE-vol. 66, ASME, New Orleans, Louisiana, November / December 1993, pp. 73-82. Marshall Burns, Automated Fabrication: Improving Productivity in Manufacturing, Englewood Cliffs, New Jersey: PTR Prentice Hall, 1993.
7 6

D. Kochan, ed., Solid Freeform Manufacturing, Manufacturing Research and Technology Series, vol. 19, Amsterdam: Elsevier, 1993. Paul F. Jacobs, Rapid Prototyping & Manufacturing: Fundamentals of StereoLithography, Dearborn, Michigan: Society of Manufacturing Engineers, 1992.
9 8

Paul F. Jacobs, Stereolithography and other RP&M Technologies: From Rapid Prototyping to Rapid Tooling, Dearborn, Michigan: Society of Manufacturing Engineers, 1996.

151

J. J. Beaman, J. W. Barlow, D. L. Bourell, R. H. Crawford, H. L. Marcus, and K. P. McAlea, Solid Freeform Fabrication - A New Direction in Manufacturing, Dordrecht: Kluwer Academic Publishers, 1997.
11

10

Friedrich B. Prinz, Chair, Rapid Prototyping in Europe and Japan, Vol. I, Analytical Chapters, Baltimore: International Technology Research Institute, Loyola College in Maryland, March 1997.

12 Rapid Prototyping Report - The Newsletter of the Desktop Manufacturing Industry, San Diego: CAD/CAM Publishing, Inc.

Terry Wohlers, Rapid Prototyping and Tooling State of the Industry: 1998 Worldwide Progress Report, Fort Collins, Colorado: Wohlers Associates, Inc., 1998.
14 15

13

http://www.patents.ibm.com or http://patent.womplex.ibm.com

Terry Wohlers, Rapid Prototyping and Tooling State of the Industry: 1998 Worldwide Progress Report, Fort Collins, Colorado: Wohlers Associates, Inc., 1998.
16 Paul F. Jacobs, Rapid Prototyping & Manufacturing: Fundamentals of StereoLithography, Dearborn, Michigan: Society of Manufacturing Engineers, 1992.

Paul F. Jacobs, Stereolithography and other RP&M Technologies: From Rapid Prototyping to Rapid Tooling, Dearborn, Michigan: Society of Manufacturing Engineers, 1996. C. R. Deckard, "Selective Laser Sintering," PhD Thesis, The University of Texas at Austin, Austin, Texas, 1988.
19 18

17

Rapid Prototyping Report - The Newsletter of the Desktop Manufacturing Industry, March 1997, p. 6.

20

Stratasys, Inc., "Fast, Precise, Safe Prototypes with FDM," Proceedings of the 1991 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1991, pp. 115-122. "Elastomers for Rapid Prototyping," Mechanical Engineering, vol. 120, no. 1, January 1998, p. 36. M. K. Agarwala, R. van Weeren, R. Vaidyanathan, A. Bandyopadhyay, G. Carrasquillo, V. Jamalabad, N. Langrana, A. Safari, S. H. Garofalini, and S. C. Danforth, "Structural Ceramics by Fused Deposition of Ceramics," Proceedings of the 1995 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1995, pp. 1-8. M. Feygin and B. Hsieh, "Laminated Object Manufacturing (LOM): A Simpler Process," Proceedings of the 1991 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1991, pp. 123-130.
24 23 22 21

Sung S. Pak, "Fabrication of Plastic LOM Objects," Prototyping Technology International, Issue 1, July/September 1997, pp. 33-35.

152

25

Don Klosterman, Brian Priore, and Richard Chartoff, "Laminated Object Manufacturing of Polymer Matrix Composites," Proceedings of the Seventh International Conference on Rapid Prototyping, San Francisco, California, April 1997, pp. 283-292.
26

James M. Ogg, "Composite - Successful for Tooling," Prototyping Technology International '98, pp. 40-43.
27

H. Levi, "Accurate Rapid Prototyping by the Solid Ground Curing Technology," Proceedings of the 1991 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1991, pp. 110-114.

28

Steven Ashley, "Rapid Prototyping is Coming of Age," Mechanical Engineering, vol. 117, no. 7, July 1995, p. 66. Marshall Burns, Automated Fabrication: Improving Productivity in Manufacturing, Englewood Cliffs, New Jersey: PTR Prentice Hall, 1993, p. 66.

29

30

E. Sachs, P. Williams, D. Brancazio, M. Cima, and K. Kremmin, "Three Dimensional Printing: Rapid Prototypes and Tooling Directly from a CAD Model," Proceedings of Manufacturing International '90, Atlanta, Georgia, March 1990, pp. 131-136.
31 32 33

Tom Clay, "Ideas Proofed," Prototyping Technology International '98, pp. 26-29. Marina Hatsopoulos (Z Corporation), private communication, April 1998.

Robert D. Stoddart, "Drop-on-Demand Yields Precision Tooling Patterns," Prototyping Technology International '97, pp. 224-227.
34

"Rapid Concept Modelers," Mechanical Engineering, vol. 118, no. 1, January 1996, pp. 64-66. "A New Dimension for Office Printers," Mechanical Engineering, vol. 118, no. 3, March 1996, pp. 112-114.
36 35

W. E. Masters, "The ballistic particle manufacturing process," National Conference on Rapid Prototyping, 1990, pp. 39-48.

37

K. E. Richardson, "The production of wax models by the ballistic particle manufacturing process," Second Conference on Rapid Prototyping, 1991, pp. 15-22. "Rapid Prototypes," Mechanical Engineering, vol. 117, no. 5, May 1995, p. 20.

38 39

D. Kochan, ed., Solid Freeform Manufacturing, Manufacturing Research and Technology Series, vol. 19, Amsterdam: Elsevier, 1993, pp. 87-91.

40

T. B. Heller, "Three-Dimensional Lithography: Laser Modeling Using Photopolymers," Beam Deflection and Scanning Technologies, San Jose, California, February / March 1991, SPIE Proceedings, vol. 1454, pp. 272-282.

41

Steven Ashley, "Rapid Prototyping Systems," Mechanical Engineering, vol. 113, no. 4, April 1991, pp. 34-43.

153

J. J. Beaman, J. W. Barlow, D. L. Bourell, R. H. Crawford, H. L. Marcus, and K. P. McAlea, Solid Freeform Fabrication - A New Direction in Manufacturing, Dordrecht: Kluwer Academic Publishers, 1997, p. 27.
43

42

D. Kochan, ed., Solid Freeform Manufacturing, Manufacturing Research and Technology Series, vol. 19, Amsterdam: Elsevier, 1993, pp. 63-68. D. Kochan, ed., Solid Freeform Manufacturing, Manufacturing Research and Technology Series, vol. 19, Amsterdam: Elsevier, 1993, pp. 69-73.

44

45

Marshall Burns, Automated Fabrication: Improving Productivity in Manufacturing, Englewood Cliffs, New Jersey: PTR Prentice Hall, 1993, pp. 60-62. D. Kochan, ed., Solid Freeform Manufacturing, Manufacturing Research and Technology Series, vol. 19, Amsterdam: Elsevier, 1993, pp. 74-76.

46

47

Lee E. Weiss, "Process Overviews," in Friedrich B. Prinz, Chair, Rapid Prototyping in Europe and Japan, Vol. I, Analytical Chapters, Baltimore: International Technology Research Institute, Loyola College in Maryland, March 1997, pp. 7-10.
48

Chua Chee Kai and Leong Kah Fai, Rapid Prototyping: Principles and Applications in Manufacturing, New York: John Wiley & Sons, Inc., 1997, pp. 66-71.

49

K. Ikuta and K. Hirowatari, "Real Three Dimensional Micro Fabrication using Stereo Lithography and Metal Molding," IEEE Micro Electro Mechanical Systems, Fort Lauderdale, Florida, February 1993, pp. 42-47. T. Takagi and N. Nakajima, "Photoforming Applied to Fine Machining," IEEE Micro Electro Mechanical Systems, Fort Lauderdale, Florida, February 1993, pp. 173-178.
51 50

D. Kochan, ed., Solid Freeform Manufacturing, Manufacturing Research and Technology Series, vol. 19, Amsterdam: Elsevier, 1993, pp. 53-62.

52

Friedrich B. Prinz, Chair, Rapid Prototyping in Europe and Japan, Vol. II, Site Reports, Baltimore: International Technology Research Institute, Loyola College in Maryland, March 1997, pp. 14-17.

53 "A Different Way to SLA," Prototyping Technology International, Issue 4, April/June 1998, pp. 58-60.

James W. Comb, William R. Priedeman, and Patrick W. Turley, "FDM Technology Process Improvements," Proceedings of the 1994 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1994, pp. 42-49.
55 56 57 58

54

Stratasys, Inc., "Stratasys Materials," December 1997. 3D Systems, "SL 5510," P/N 70332, January 22, 1998. 3D Systems, "SL 5410 Materials Datasheet," P/N 70292, June 27, 1997. www.dupont.com/somos/product/select.html, before April 27, 1998.

154

59

Du Pont, "Preliminary Product Data: Somos 8120 Epoxy Photopolymer," Version 1.1, received May 1998.
60 61 62

Du Pont, "Somos Product Data Sheet - Somos 7100," Version 3.3, received May 1998. www.dupont.com/somos/product/select.html, before April 27, 1998.

Sandra Seitz, Simon van de Crommert, Klaus J. Esser, and Kevin McAlea, "Routes to Plastic Models and Prototypes using the SLS Selective Laser Sintering Process," Rapid Prototyping and Flexible Manufacturing, Munich, Germany, June 1997, SPIE Proceedings, vol. 3102, pp. 106-111.
63

3D Systems, "SL 5510," P/N 70332, January 22, 1998. [This data sheet includes data for DTM materials for comparison.]
64 65 66

Stratasys, Inc., "Stratasys Materials," December 1997. Cubital, "Solimer XA7501 Data Sheet," received May 1998.

Paul S. Fussell, "Materials - Overview, Plastics, and Resins," in Friedrich B. Prinz, Chair, Rapid Prototyping in Europe and Japan, Vol. I, Analytical Chapters, Baltimore: International Technology Research Institute, Loyola College in Maryland, March 1997, pp. 41-51.
67

"Materials on the Market," Rapid Prototyping (newsletter of the Rapid Prototyping Association of the Society of Manufacturing Engineers), vol. 4, no. 2, Second Quarter 1998, p. 6. Stratasys, Inc., "Stratasys Materials," December 1997.

68 69

Paul S. Fussell, "Materials - Overview, Plastics, and Resins," in Friedrich B. Prinz, Chair, Rapid Prototyping in Europe and Japan, Vol. I, Analytical Chapters, Baltimore: International Technology Research Institute, Loyola College in Maryland, March 1997, pp. 41-51.
70

Wilbrand Woebcken, Saechtling International Plastics Handbook for the Technologist, Engineer, and User, 3rd ed., Munich: Hanser Publishers, 1995, Tables 4.13, 4.14, 4.37, 4.41. The exact materials are Nylon 6-6 (conditioned to atmospheric moisture), mediumimpact ABS, "Standard type" polycarbonate, and "general-purpose" polystyrene.
71

Eric Fodran, Martin Koch, and Unny Menon, "Mechanical and Dimensional Characteristics of Fused Deposition Modeling Build Styles," Proceedings of the 1996 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1996, pp. 419-442.
72

Michael Ebenhoch (Institut fuer Maschinenelemente, Stuttgart), private communication, May 1998.

155

73

M. Bertoldi, M. A. Yardimci, C. M. Pistor, S. I. Guceri, and G. Sala, "Mechanical Characterization of Parts Processed via Fused Deposition," Proceedings of the 1998 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1998, pp. 557-565. P. Kamatchi Subramanian, N. K. Vail, J. W. Barlow, and H. L. Marcus, "Anisotropy in Alumina Processed by SLS," Proceedings of the 1994 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1994, pp. 330338.

74

75

B. Badrinarayan and J. W. Barlow, "Effect of Processing Parameters In SLS Of MetalPolymer Powders," Proceedings of the 1995 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1995, pp. 55-63. The only anisotropy discussed in this paper is green strength after scanning in the transverse rather than the longitudinal direction.
76

T. D. Stewart, K. W. Dalgarno, T. H. C. Childs, and J. Perkins, "Strength of the DTM RapidSteel 1.0 Material," Proceedings of the 1998 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1998, pp. 443-9. This paper describes property variation as a function of distance away from infiltration points in Rapid Steel parts.

David C. Thompson, "The Optimization of Part Orientation for Solid Freeform Manufacture", Master of Science Thesis, The University of Texas at Austin, Austin, Texas, December 1995. The experimental data is summarized on p. 24. Ian Gibson and Dongping Shi, "Material Properties and Fabrication Parameters in Selective Laser Sintering Process," Rapid Prototyping Journal, vol. 3, no. 4, 1997, pp. 129-136.
79 78

77

Sun Pak, Gregory Nisnevich, Jennifer Maher, and Michael Feygin, "Laminated Object Manufacturing," in Paul F. Jacobs, Stereolithography and other RP&M Technologies: From Rapid Prototyping to Rapid Tooling, Dearborn, Michigan: Society of Manufacturing Engineers, 1996, p. 375. Thomas Pang (3D Systems), private communication, June 1998. Dr. Pang mentioned unpublished internal 3D Systems tests from 1992 showing relatively isotropic tensile strength for green and postcured SLA parts.

80

81

D. G. Dahl and B. Cleary, "Constructing Multiple Prototype Parts from ComputerGenerated Master Models," In Rapid Prototyping Systems: Fast Track to Product Realization [Compilation of Papers from Rapid Prototyping and Manufacturing 1993], Dearborn, Michigan: Society of Manufacturing Engineers, 1994, pp. 35-54. I found two how-to pamphlets describing soft-tool casting techniques:

82

David G. Dahl and Bill Cleary, "Construction of Multiple Prototype Parts from ComputerGenerated Master Models," Ciba-Geigy Corporation, East Lansing, Michigan, 1993 [this is a reprint of the article above]. "Conap Urethane and Epoxy Tooling Resins Mold Making Guide," CON 9978, Conap, Inc., Olean, New York, November 1994.

156

83

"Tooling Options: ACES Injection Molding (AIM)," Rapid Prototyping Report - The Newsletter of the Desktop Manufacturing Industry, January 1997, pp. 5-6.

84

Uday Hejmadi and Kevin McAlea, "Selective Sintering of Metal Molds: The RapidTool Process," Proceedings of the 1996 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1996, pp. 97-104.

Steven Ashley, "Rapid Prototyping is Coming of Age," Mechanical Engineering, vol. 117, no. 7, July 1995, p. 65.
86

85

Emanuel Sachs, Samuel Allen, Honglin Guo, Javier Banos, Michael Cima, James Serdy, and David Brancazio, "Progress on Tooling by 3D Printing; Conformal Cooling, Dimensional Control, Surface Finish and Hardness," Proceedings of the 1997 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1997, pp. 115-123.

Steven Ashley, "From CAD Art to Rapid Metal Tools," Mechanical Engineering, vol. 119, no. 3, March 1997, pp. 82-87. R. Merz, F. B. Prinz, K. Ramaswami, M. Terk, and L. E. Weiss, "Shape Deposition Manufacturing," Proceedings of the 1994 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1994, pp. 1-8.
89 88

87

F. B. Prinz and L. E. Weiss, "Method for Fabrication of Three-Dimensional Articles," U. S. Patent No. 5,301,415, 1994.
90

This figure was drawn by Alexander Cooper of the Stanford University Rapid Prototyping Laboratory. This figure was drawn by Alexander Cooper of the Stanford University Rapid Prototyping Laboratory.

91

Krishnan Ramaswami, "Process Planning for Shape Deposition Manufacturing," PhD Thesis, Stanford University, Palo Alto, California, January 1997.
93

92

Krishnan Ramaswami, Yasushi Yamaguchi, and Fritz Prinz, "Spatial Partitioning of Solids for Solid Freeform Fabrication," Proceedings of the Fourth ACM/SIGGRAPH Symposium on Solid Modeling and Applications, Association for Computing Machinery, Atlanta, Georgia, May 1997, pp. 346-353.
94

A. Sudhalkar, R. Bhargava, R. Mattikalli, and J. M. Chen, An Introduction to the NOODLES Geometric Modeling System, Technical Report, EDRC, Carnegie Mellon University, Pittsburgh, Pennsylvania, October, 1990.

Spatial Technology, Inc., ACIS Geometric Modeler, Boulder, Colorado: Spatial Technology, Inc., 1993.
96

95

This figure was assembled by Ju-Hsien Kao of the Stanford University Rapid Prototyping Laboratory.

157

97

Krishnan Ramaswami and Fritz Prinz, "CNC Cutter Path Generation in Shape Deposition Manufacturing," Concurrent Product and Process Engineering, New York: American Society of Mechanical Engineers, Design Engineering Division, 1995, pp. 213220. Ju-Hsien Kao and Fritz Prinz, "Optimal Motion Planning for Deposition in Layered Manufacturing," ASME DETC '98 Computers in Engineering Conference, Atlanta, Georgia, September 1998.

98

Robert Merz, "Shape Deposition Manufacturing," PhD Thesis, Department of Electrical Engineering, Technical University of Vienna, Vienna, Austria, 1994, pp. 49-56. R. Merz, F. B. Prinz, and L. E. Weiss, "Method and Apparatus for Depositing Molten Metal," U. S. Patent No. 5,281,789, 1994.
101 100

99

J. R. Fessler, R. Merz, A. H Nickel, F. B. Prinz, and L. E. Weiss, "Laser Deposition of Metals for Shape Deposition Manufacturing," Proceedings of the 1996 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1996, pp. 117-124. J. Fessler, A. Nickel, G. Link, F. Prinz, and P. Fussell, "Functional Gradient Metallic Prototypes through Shape Deposition Manufacturing," Proceedings of the 1997 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1997, pp. 521-528.

102

F. B. Prinz and L. E. Weiss, "Automated System for Forming Objects by Incremental Buildup of Layers," U. S. Patent No. 5,301,863, 1994. L. E. Weiss, F. B. Prinz, and D. P. Siewiorek, "A Framework for Thermal Spray Deposition: The MD* System," Proceedings of the 1991 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1991, pp. 178186. L. E. Weiss, F. B. Prinz, and L. E. Gursoz, "Method and Apparatus for Fabrication of Three-Dimensional Articles by Thermal Spray Deposition," U. S. Patent No. 5,126,529, 1992.
106 F. B. Prinz, L. E. Weiss and D. A. Adams, "Method and Apparatus for Fabrication of Three-Dimensional Articles by Spray Deposition using Masks as Support Structures," U. S. Patent No. 5,203,944, 1993. 107 105 104

103

Robert Merz, "Shape Deposition Manufacturing," PhD Thesis, Department of Electrical Engineering, Technical University of Vienna, Vienna, Austria, 1994, p. 6.

F. B. Prinz and L. E. Weiss, "Method and Apparatus for Fabrication of ThreeDimensional Metal Articles by Weld Deposition," U. S. Patent No. 5,207,371, 1993. Some researchers have developed variable layer thickness construction and adaptive slicing software techniques to optimize surface finish and build time in commercial rapid prototyping machines. See the following references:
109

108

158

Prashant Kulkarni and Debasish Dutta, "Adaptive Slicing of Parametrizable Algebraic Surfaces for Layered Manufacturing," in Advances in Design Automation 1995, vol. 1, Proceedings of the 1995 Design Engineering Technical Conferences, Boston, Massachusetts, September 17-20, 1995, pp. 211-217. Justin Tyberg and Jan Helge Bohn, "FDM Systems and Local Adaptive Slicing," Proceedings of the 1998 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1998 pp. 375-82. Fritz Klocke, H. Wirtz, and W. Meiners, "Direct Manufacturing of Metal Prototypes and Prototype Tools," Proceedings of the 1996 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1996, pp. 141-148. Kendrick E. Richardson, "Multiple Material Machining: An Innovative RP Technology," Rapid Prototyping (newsletter of the Rapid Prototyping Association of the Society of Manufacturing Engineers), vol. 4, no. 2, Second Quarter 1998, pp. 1-4.
112 111 110

Alexander Cooper (Stanford University), private communication, September 1998.

113 L. E. Weiss, R. Merz, F. B. Prinz, G. Neplotnik, P. Padmanabhan, L. Schultz, and K. Ramaswami, "Shape Deposition Manufacturing of Heterogeneous Structures," Journal of Manufacturing Systems, vol. 16, no. 4, June 1997, pp. 239-248. 114

Marshall Burns and Charles L. Thomas, "Sheet Autofab by Offset Fabrication," Rapid Prototyping (newsletter of the Rapid Prototyping Association of the Society of Manufacturing Engineers), vol. 3, no. 2, Second Quarter 1997, pp. 1-6.

Hugh Denham, Graeme George, Llew Rintoul, and Paul Calvert, "Fabrication of Polymers and Composites Containing Embedded Sensors," Third International Conference on Intelligent Materials / Third European Conference on Smart Structures and Materials, Lyon, France, June 1996, SPIE Proceedings, vol. 2779, pp. 742-747. This paper describes embedded fiber optic sensors and piezoelectric film. W. Nau, R. L. Dooley, A. A. Ogale, C. C. Jara-Almonte, and A. Bachi, "Embedding Sensors using Stereolithography," Clemson University Intelligent Design and Rapid Prototyping Laboratory; this paper was presented at the Fourth International Conference on Rapid Prototyping, The University of Dayton, Dayton, Ohio, June 1993, but was not published in the proceedings of that conference. The authors embedded strain gages within SLA tensile and bending specimens and compared their outputs to external strain gages during testing. W. H. Nau, "Embedding Sensors using Stereolithography," Master of Science Thesis, Clemson University, Clemson, South Carolina, 1991.
118 119 117 116

115

Lee Weiss (Carnegie Mellon University), private communication, May 4, 1998.

F. B. Prinz, L. E. Weiss, and D. P. Siewiorek, "Electronic Packages and Smart Structures Formed by Thermal Spray Deposition," U. S. Patent No. 5,278,442, 1994.

120 Asim Smailagic and Daniel P. Siewiorek, "Matching Interface Design with User Tasks: Modalities of Interaction with CMU Wearable Computers," IEEE Personal Communications, vol. 3, no. 1, February 1996, pp. 14-25.

159

Susan Finger, John Stivoric, Cristina Amon, Levant Gursoz, Fritz Prinz, Daniel Siewiorek, Asim Smailagic, and Lee Weiss, "Reflections on a Concurrent Design Methodology: A Case Study in Wearable Computer Design," Computer-Aided Design, vol. 28, no. 5, 1996, pp. 393-404.
122 Asim Smailagic and Daniel P. Siewiorek, "A Case Study in Embedded-System Design: The VuMan 2 Wearable Computer," IEEE Design and Test of Computers, vol. 10, no. 3, September 1993, pp. 56-67.

121

Thomas Martin and Daniel P. Siewiorek, "Wearable Computers: Paradigm Shift in Computing," IEEE Potentials, August / September 1994, pp. 36-38. Page 37 describes the planned fabrication of a follow-on version of VuMan 2 via MD*.
124

123

J. E. Beck, Fritz B. Prinz, D. P. Siewiorek, and L. E. Weiss, "Manufacturing Mechatronics Using Thermal Spray Shape Deposition," Proceedings of the 1992 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1992, pp. 272-279.

Susan Finger, John Stivoric, Cristina Amon, Levant Gursoz, Fritz Prinz, Daniel Siewiorek, Asim Smailagic, and Lee Weiss, "Reflections on a Concurrent Design Methodology: A Case Study in Wearable Computer Design," Computer-Aided Design, vol. 28, no. 5, 1996, pp. 393-404.
126

125

Krishnan Ramaswami, "Process Planning for Shape Deposition Manufacturing," PhD Thesis, Stanford University, Palo Alto, California, January 1997, p. 89.

L. E. Weiss, R. Merz, F. B. Prinz, G. Neplotnik, P. Padmanabhan, L. Schultz, and K. Ramaswami, "Shape Deposition Manufacturing of Heterogeneous Structures," Journal of Manufacturing Systems, vol. 16, no. 4, June 1997, pp. 239-248.
128

127

L. Weiss, F. Prinz, G. Neplotnik, K. Padmanabhan, L. Schultz, and R. Merz, "Shape Deposition Manufacturing of Wearable Computers," Proceedings of the 1996 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1996, pp. 34-35. L. Weiss, F. Prinz, G. Neplotnik, K. Padmanabhan, L. Schultz, and R. Merz, "Shape Deposition Manufacturing of Wearable Computers," Proceedings of the 1996 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1996, pp. 31-38.

129

Cristina Amon, Kyongsoo Chae, Eric Egan, Chris Kasabach, Daniel Siewiorek, Asim Smailagic, and John Stivoric, "System-level Thermal Management and Concurrent System Design of a Wearable Multicomputer System," I-THERM V: InterSociety Conference on Thermal Phenomena in Electronic Systems, Orlando, Florida, May 29 - June 1, 1996, pp. 46-55.
131

130

C. H. Amon, J. S. Nigen, D. P. Siewiorek, A. Smailagic, and J. M. Stivoric, "Concurrent Design and Analysis of the Navigator Wearable Computer System: The Thermal Perspective," IEEE Transactions on Components, Packaging, and Manufacturing Technology, vol. 18, no. 3, 1995, pp. 567-577.

160

E. Egan and C. H. Amon, "Thermal Design of Wearable Computers: Application to the Navigator2, Thermal Management Devices, and Embedded Electronics," EDRC Report no. 24-123-95, Carnegie Mellon University, Pittsburgh, Pennsylvania, 1995.
133

132

Eric Egan and Cristina H. Amon, "Cooling Strategies for Embedded Electronic Components of Wearable Computers Fabricated by Shape Deposition Manufacturing," ITHERM V: InterSociety Conference on Thermal Phenomena in Electronic Systems, Orlando, Florida, May 29 - June 1, 1996, pp. 13-20.

Chapter Three - Polymer Shape Deposition Manufacturing These flap mechanisms were designed and built by Byongho Park of the Stanford University Rapid Prototyping Laboratory. This sequence of drawings was prepared by Byongho Park of the Stanford University Rapid Prototyping Laboratory. J. M. Pinilla, J. Kao, and F. B. Prinz, "Process Planning and Automation for Additive-Subtractive Solid Freeform Fabrication," Proceedings of the 1998 Solid Freeform Fabrication Symposium, The University of Texas at Austin, August 1998, pp. 247-57.
137 136 135 134

This drawing was prepared by Scott Crawford of the Stanford University Rapid Prototyping Laboratory.

Chapter Four - Materials for Polymer SDM For a discussion of residual stresses resulting from solidification of metal droplets in SDM, see F. B. Prinz, L. E. Weiss, C. H. Amon, and J. L. Beuth, "Processing, Thermal and Mechanical Issues in Shape Deposition Manufacturing," Proceedings of the 1995 Solid Freeform Fabrication Symposium, The University of Texas at Austin, August 1995, pp. 118-129.
139 138

There is a longer discussion of polymer shrinkage issues in J. J. Beaman, J. W. Barlow, D. L. Bourell, R. H. Crawford, H. L. Marcus, and K. P. McAlea, Solid Freeform Fabrication: A New Direction in Manufacturing, Dordrecht: Kluwer Academic Publishers, 1997, p. 94.

This part was made by John Fessler and Alexander Nickel of the Stanford University Rapid Prototyping Laboratory.

140

161

141

If LUC-4180 polyurethane parts are built by a facility operating two shifts (08001600, 1600-0000), then reasonable assumptions lead to a maximum of three layers of a single part every two days. If twelve hours are required for a layer to cure, and an hour and a half of machining and preparation are required before a second casting, then a pattern soon develops of three castings every two days. A layer is machined at the start of the first day, followed by a casting at 0930 and a second casting at 2300. The second day, a casting cannot be performed until 1230, and there isn't sufficient time for a second casting. Therefore, machining and casting cannot begin until the third day, beginning a new cycle of the same pattern. If only a single shift is used, then only a single layer can be made per day. While multiple parts can be built on an interleaved schedule, and multiple machines can be used to increase throughput, the end-to-end time for LUC-4180 polyurethane parts is still quite long.

142 These flap mechanisms were built by Byongho Park of the Stanford University Rapid Prototyping Laboratory.

L. Weiss and F. Prinz, "Novel Applications and Implementations of Shape Deposition Manufacturing," Third Pacific Rim International Conference on Advanced Materials and Processing, Honolulu, Hawaii, July 1998.
144

143

This green ceramic impeller was made by Lee Weiss and the staff of the Shape Deposition Laboratory at Carnegie Mellon University.

This wax mold was made by Alexander Cooper of the Stanford University Rapid Prototyping Laboratory.
146

145

Robert Merz, "Shape Deposition Manufacturing," PhD Thesis, Department of Electrical Engineering, Technical University of Vienna, Vienna, Austria, 1994, pp. 134135.

147

A discussion of the problems of residual stresses in metallic SDM parts can be found in Robert Merz, "Shape Deposition Manufacturing," PhD Thesis, Department of Electrical Engineering, Technical University of Vienna, Vienna, Austria, 1994, pp. 132-134.

Values from non-wax materials were published by the respective manufacturers. The linear shrinkage value for Somos 2100 was calculated from the reported volumetric shrinkage value. Wax shrinkage values were calculated from volumetric shrinkage data measured by Alexander Cooper of the Stanford Rapid Prototyping Laboratory. Wax shrinkage was measured between typical molten casting temperatures and room temperature. The KC3230A material's shrinkage was estimated from measurements of the shrinkages of the two components from which it is made.

148

Chapter Five - Material Strength in Polymer SDM Edward R. Salmon, Encapsulation of Electronic Devices and Components, New York: Marcel Dekker, Inc., 1987, pp. 147-148.
149

162

150

Armand F. Lewis, "Epoxy Resin Adhesives," in Clayton A. May, ed., Epoxy Resins: Chemistry and Technology, 2nd ed., revised and expanded, New York: Marcel Dekker, Inc., 1988, pp. 664-665. This reference discusses different types of adhesive joints and loading conditions and the types of adhesives preferred for each.

151

L. Weiss, F. Prinz, G. Neplotnik, K. Padmanabhan, L. Schultz, and R. Merz, "Shape Deposition Manufacturing of Wearable Computers," Proceedings of the 1996 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1996, pp. 31-38.

"LUC-4180 Lowex Casting System," Adtech Plastic Systems Corp., Charlotte, Michigan. The shrinkage specification was reported for a seven-day cure cycle. "TDT 205-3 Resin and Hardener / DT 081 / DT 082 Three Component Rapid-Curing Polyurethane Casting System," Ciba Specialty Chemicals Corp., East Lansing, Michigan, October 1996. The shrinkage specification was reported in accordance with ASTM D 2566, using Mold #0.
154 "System5 Epoxy Encapsulation Systems," Adtech Plastic Systems Corp., Charlotte, Michigan. The shrinkage specification was reported in accordance with ASTM standard D 2566-86. 153

152

Mark Schlack, "Performance Testing of Plastics Products," Chapter 27 in Michael L. Berins, ed., SPI Plastics Engineering Handbook of the Society of the Plastics Industry, Inc., New York: Van Nostrand Reinhold, 1991, p. 806.
156

155

S. Turner, Mechanical Testing of Plastics, 2nd ed., New York: Longman, 1983, pp. 134-5. Floyd Roberts (Rapid Prototyping Subcommittee of the ASTM Mechanical Testing Committee), private communication, July 1998. "Standard Test Methods for Flexural Properties of Unreinforced and Reinforced Plastics and Electrical Insulating Materials," ASTM D 790-96a, 1997 Annual Book of ASTM Standards, vol. 08.01, West Conshohocken, Pennsylvania: ASTM, 1997, pp. 141151.

157

158

159

S. Turner, Mechanical Testing of Plastics, 2nd ed., New York: Longman, 1983, pp. 134-5. S. I. Krishnamachari, Applied Stress Analysis of Plastics: A Mechanical Engineering Approach, New York: Van Nostrand Reinhold, 1993, pp. 76-80.

160

Thomas H. Pang, "Advances in Stereolithography Photopolymer Systems," Chapter 2 in Paul F. Jacobs, Stereolithography and other RP&M Technologies: From Rapid Prototyping to Rapid Tooling, Dearborn, Michigan: Society of Manufacturing Engineers, 1996, p. 63. S. Turner, Mechanical Testing of Plastics, 2nd ed., New York: Longman, 1983, p. 139.
162

161

163

163

Ralf Michael Criens and Huter-Georg Mosle, "Knit-Lines in Injection Molding and Mechanical Behavior," in Witold Brostow and Roger D. Corneliussen, ed., Failure of Plastics, Munich: Hanser Publishers, 1986, pp. 415-429. This reference describes the greater sensitivity of impact tests instead of tensile tests for observing the effects of knitlines in injection molded parts.

164

"Standard Test Method for Tensile Properties of Plastics," ASTM D 638-96, 1997 Annual Book of ASTM Standards, vol. 08.01, West Conshohocken, Pennsylvania: ASTM, 1997, pp. 44-56.

165 V. K. Stokes, "Vibration Welding of Thermoplastics. Part III: Strength of Polycarbonate Butt Welds," Polymer Engineering and Science, vol. 28, no. 15, MidAugust 1988, pp. 989-997.

These representative plots of load vs. crosshead position are from the load histories of the following specimens: The LUC-4180 polyurethane load history was recorded during the testing of monolithic BIOACT-soaked specimen #4 (see section 5.5). The TDT 205-3 polyurethane load history was recorded during the testing of monolithic specimen #5. The 501/530 epoxy load history was recorded during the testing of monolithic specimen #5. These representative plots of load vs. crosshead position are from the load histories of the following specimens: The LUC-4180 polyurethane load history was recorded during the testing of dual-interface BIOACT-soaked specimen #1 (see section 5.5). The TDT 2053 polyurethane load history was recorded during the testing of dual-interface specimen #4. The 501/530 epoxy load history was recorded during the testing of dual-interface specimen #4. Mark Schlack, "Performance Testing of Plastic Products," Chapter 27 in Michael L. Berins, ed., SPI Plastics Engineering Handbook of the Society of the Plastics Industry, Inc., 5th ed., New York: Van Nostrand Reinhold, 1991, p. 807.
169 Turner has defined an "anisotropy ratio" that works in the opposite manner - the ratio of the maximum to minimum strength or modulus within a part. S. Turner, Mechanical Testing of Plastics, 2nd ed., New York: Longman, 1983, p. 93. 170 168 167

166

A. J. Kinloch, "Review: The Science of Adhesion - Part 1. Surface and Interfacial Aspects," Journal of Materials Science, vol. 15, 1980, p. 2156. A. J. Kinloch, "Review: The Science of Adhesion - Part 1. Surface and Interfacial Aspects," Journal of Materials Science, vol. 15, 1980, pp. 2141-2166. Georges Fourche, "An Overview of the Basic Aspects of Polymer Adhesion. Part I: The Fundamentals," Polymer Engineering and Science, vol. 35, no. 12, June 1995, pp. 957-967.

171

172

173

K. W. Allen, "Theories of Adhesion," in D. E. Packham, ed., Handbook of Adhesion, Polymer Science and Technology Series, D. M. Brewis and D. Briggs, ed., Essex, England: Longman Scientific & Technical, 1992, pp. 473-475. J. W. McBain and D. G. Hopkins, "On Adhesives and Adhesive Action," Journal of Physical Chemistry, vol. 29, 1925, pp. 188-204.

174

164

K. W. Allen, "Mechanical Theory of Adhesion," in D. E. Packham, ed., Handbook of Adhesion, Polymer Science and Technology Series, D. M. Brewis and D. Briggs, ed., Essex, England: Longman Scientific & Technical, 1992, pp. 273-275.
176

175

B. V. Deryaguin, Research, vol. 8, 1955, p. 70ff. This obscure first reference to the electronic theory of adhesion was provided in Kinloch's review article.

177

Georges Fourche, "An Overview of the Basic Aspects of Polymer Adhesion. Part I: The Fundamentals," Polymer Engineering and Science, vol. 35, no. 12, June 1995, p. 958. S. S. Voyutskii, Autohesion and Adhesion of High Polymers, New York: John Wiley and Sons (Interscience), 1963.

178

179

A. J. Kinloch, "Review: The Science of Adhesion - Part 1. Surface and Interfacial Aspects," Journal of Material Science, vol. 15, 1980, p. 2154. I think there was an error in the original, which said that diffusion did not occur above the glass-transition temperature. After comparison with Wu and Comyn , I think the proper statement must be that diffusion is unlikely below the glass-transition temperature.

180 J. J. Bikerman, The Science of Adhesive Joints, New York: Academic Press, 1961 and 1968. 181

Georges Fourche, "An Overview of the Basic Aspects of Polymer Adhesion. Part I: The Fundamentals," Polymer Engineering and Science, vol. 35, no. 12, June 1995, pp. 961-962.

182

A. J. Kinloch, "Review: The Science of Adhesion - Part 1. Surface and Interfacial Aspects," Journal of Material Science, vol. 15, 1980, p. 2156.

"Theory Gives Direction to Adhesion Work," Chemical Engineering News, vol. 41, no. 15, 1963, p. 67ff. This article describes work done by L. H. Sharpe and H. Schonhorn which was described in the Colloid and Surface Chemistry Division at the 144th American Chemical Society's National Meeting.
184 Vijay K. Stokes, "Joining Methods for Plastics and Plastic Composites: An Overview," Polymer Engineering and Science, vol. 29, no. 19, Mid-October, 1989, pp. 1310-1324.

183

J. Raghavan and Richard P. Wool, "Interfaces in Repair, Recycling, Joining and Manufacturing of Polymers and Polymer Composites," Journal of Applied Polymer Science, to appear. This paper mentions the requirement that chains must cross and recross interfaces for optimal interface strength; this is not a citation to the original work on that concept. R.-J. Chang and A. N. Gent, "Effect of Interfacial Bonding on the Strength of Adhesion of Elastomers. I. Self-Adhesion," Journal of Polymer Science: Polymer Physics Edition, vol. 19, 1981, pp. 1619-1633. The authors describe bonding polybutadiene and ethylene-propylene copolymers of varying degrees of previous partial crosslinking, and then allowing the materials to continue crosslinking in contact with one another.
186

185

165

Edward M. Petrie, "Joining of Plastics, Elastomers, and Composites," in Charles A. Harper, ed., Handbook of Plastics, Elastomers, and Composites, 3rd ed., New York: McGraw-Hill, 1996, pp. 7.43-7.49.
188 Edward M. Petrie, "Joining of Plastics, Elastomers, and Composites," in Charles A. Harper, ed., Handbook of Plastics, Elastomers, and Composites, 3rd ed., New York: McGraw-Hill, 1996, p. 7.87.

187

"LUC-4180 Lowex Casting System," Adtech Plastic Systems Corp., Charlotte, Michigan, p. 1. "TDT 205-3 Resin and Hardener / DT 081 / DT 082 Three Component Rapid-Curing Polyurethane Casting System," Ciba Specialty Chemicals Corp., East Lansing, Michigan, October 1996, p. 1.
191 190

189

J. Raghavan and Richard P. Wool, "Interfaces in Manufacturing, Repair and Recycling of Polymer Composites," presented at the 1996 ASME International Mechanical Engineering Congress and Exhibition, November 17-22, 1996, Atlanta, Georgia, Advanced Materials: Development, Characterization, Processing, and Mechanical Behavior (Book of Abstracts), MD-vol. 74, New York: ASME, 1996, pp. 79-80. This is only an abstract of the initial work by Raghavan and Wool.

192 J. Raghavan and Richard P. Wool, "Interfaces in Repair, Recycling, Joining and Manufacturing of Polymers and Polymer Composites," Journal of Applied Polymer Science, to appear. This is a complete description of the authors' current work, including casting, grinding, and recasting thermoset material. 193 The authors raise the issue of whether homopolymerization to form linear polystyrene rather than crosslinked vinyl ester might have resulted in regions of material with different network structures than expected. 194

Souheng Wu, Polymer Interface and Adhesion, New York: Marcel Dekker, Inc., 1982, pp. 316-317.

T. Kerle, R. Yerushalmi-Rozen, and J. Klein, "Cross-Link-Induced Autophobicity in Polymer Melts: A Re-Entrant Wetting Transition," Europhysics Letters, vol. 38, no. 3, 1997, pp. 207-212. This article describes the transition from complete wetting, to partial wetting, and back to complete wetting as a surface's degree of crosslinking was increased from none, to some, to complete. Tobias Kerle, Rachel Yerushalmi-Rozen, and Jacob Klein, "Wetting and Autophobicity of a Polymer Melt on a Network of Itself," Macromolecules, vol. 31, 1998, pp. 422-429. This article more completely explains the experiments of the previous article, including details about the surface-roughness effects on the return to complete wetting of the substrate.
197 Zoran S. Petrovic and James Ferguson, "Polyurethane Elastomers," Progress in Polymer Science, vol. 16, 1991, pp. 695-836. There is a discussion of soft and hard segments and their domains in polyurethane on pp. 696-7. 196

195

166

198

Wilson F. Gum, Wolfram Riese, and Henri Ulrich, ed., Reaction Polymers Polyurethanes, Epoxies, Unsaturated Polyesters, Phenolics, Special Monomers, and Additives: Chemistry, Technology, Applications, Markets, Munich: Hanser Publishers, 1992, p. 269.

199 Christian Maier, Andrew N. Hrymak, and Philip E. Wood, "Knit Lines in Reaction Injection Moldings," Advances in Polymer Technology, vol. 10, no. 1, 1990, pp. 61-74. On p. 63, the authors reference a German paper which said hydrogen bonding may be a significant source of strength across knit lines in reaction injection molding. The article says that hydrogen bonding between molecules may be 100 times more common than covalent bonds within individual molecules in a polyurethane material. Although covalent bonds are ten times stronger than hydrogen bonds, the large collection of hydrogen bonds would represent significant bond strength. The quoted article is H. V. Boenig, Kunststoffe, vol. 71, no. 1, 1981, p. 41ff. (in German). 200

Dimitrios V. Dounis and Garth L. Wilkes, "Effect of Toluene Diisocyanate Index on Morphology and Physical Properties of Flexible Slabstock Polyurethane Foams," Journal of Applied Polymer Science, vol. 66, 1997, pp. 2395-2408. On p. 2396, the authors quote a reference which claims that increased crosslinking leads to reduced properties in polyester-MDI polyurethane elastomers. The referenced paper is Z. H. Ophir and G. L. Wilkes, "Time Dependence of Mechanical Properties and Domain Formation of Linear and Crosslinked Segmented Polyurethanes," Stuart Cooper and Gerald Estes, ed., Multiphase Polymers, Advances in Chemistry Series, vol. 176, 1979, pp. 53-67.

Zygmunt Wirpsza, Polyurethanes: Chemistry, Technology, and Applications, Ellis Horwood PTR Prentice Hall Polymer Science and Technology Series, New York: Ellis Horwood, 1993, pp. 114-118.
202

201

Wilson F. Gum, Wolfram Riese, and Henri Ulrich, ed., Reaction Polymers Polyurethanes, Epoxies, Unsaturated Polyesters, Phenolics, Special Monomers, and Additives: Chemistry, Technology, Applications, Markets, Munich: Hanser Publishers, 1992, p. 269.

203 Roger Kaminski (Adtech Plastic Systems Corp.), Kevin Schroeder (Ciba Specialty Chemicals), Peter Markusch (Bayer Corporation), and Lloyd Colgrove (Dow Chemical), private communications, 1998. All of these chemists agreed that little or no crosslinking would be expected across successively cast layers of LUC-4180 and TDT 205-3. However, some researchers have found evidence of covalent bonding between a polyurethane adhesive and an epoxy primer; see I. E. Klein, J. Sharon, A. E. Yaniv, H. Dodiuk, and D. Katz, "Chemical Interactions in the System Anodized Aluminum-PrimerAdhesive," International Journal of Adhesion and Adhesives, vol. 3, 1983, p. 159ff.

On the question of whether hydrogen bonding would be expected across multilayer polyurethane interfaces, experts had differing opinions. Dr. Peter Markusch of Bayer Corporation thought that no hydrogen bonding would occur at room temperature. Dr. Lloyd Colgrove of Dow Chemical thought that hydrogen bonding could occur across polyurethane interfaces.
205

204

A. J. Kinloch, Adhesion and Adhesives: Science and Technology, London: Chapman and Hall, 1987, pp. 79-80.

167

Lloyd Colgrove (Dow Chemical), Peter Markusch (Bayer Corporation), and Richard Wool (University of Delaware), private communications, September 1998. None of these chemists thought it likely that the differences between the interface strengths of LUC-4180 and TDT 205-3 were due to qualitatively different types of bonds being present in the two situations.
207 208

206

Jim Hoge (Ciba Specialty Chemicals), private communication, June 1998.

Teruko Miyazaki (Ciba Specialty Chemicals) and Jim Hoge (Ciba Specialty Chemicals), private communications, June 1998. Ciba chemists identified these components as plasticizers; the functions they ascribed to the materials match those described by the literature. See J. K. Sears and N. W. Touchette, "Plasticizers," in Herman F. Mark, Norbert M. Bikales, Charles G. Overberger, and Georg Menges, ed., Encyclopedia of Polymer Science and Engineering, Supplement Volume, New York: Wiley-Interscience, 1985, p. 614 and p. 628. These sections describe the uses of plasticizers in thermoset materials in general and in polyurethane in particular.

D. M. Brewis, "Weak Boundary Layers," in D. E. Packham, ed., Handbook of Adhesion, Polymer Science and Technology Series, D. M. Brewis and D. Briggs, ed., Essex, England: Longman Scientific & Technical, 1992, pp. 501-502. The author notes that plasticizers often migrate to polymer surfaces. However, this only causes a weak boundary layer if the applied adhesive cannot absorb this material before solidifying.
210 211

209

Teruko Miyazaki (Ciba Specialty Chemicals), private communication, June 1998.

"Material Safety Data Sheet - LUC-4180 Poly A" and "Material Safety Data Sheet LUC-4180 B (Iso)," Adtech Plastic Systems Corp., Charlotte, Michigan. The LUC-4180 B-side data sheet identifies tricresyl phosphate as less than 10% of the B-side mixture of the polyurethane.

212

J. K. Sears and N. W. Touchette, "Plasticizers," in Herman F. Mark, Norbert M. Bikales, Charles G. Overberger, and Georg Menges, ed., Encyclopedia of Polymer Science and Engineering, Supplement Volume, New York: Wiley-Interscience, 1985, p. 569. This article identifies tricresyl phosphate as a plasticizer.

Steve Haggerty (Adtech Plastic Systems), private communication, June 1998. An (unspecified) antifoam agent is used in the Adtech LUC-4180 polyurethane.
214

213

"Material Safety Data Sheet - LUC-4180 Poly A" and "Material Safety Data Sheet LUC-4180 B (Iso)," Adtech Plastic Systems Corp., Charlotte, Michigan. The LUC-4180 A-side data sheet identifies sodium alumino silicate as less than 10% of the A-side mixture of the polyurethane. According to Adtech, this is a particulate filler material. Jim Hoge (Ciba Specialty Chemicals), private communication, June 1998.

215 216

J. K. Sears and N. W. Touchette, "Plasticizers," in Herman F. Mark, Norbert M. Bikales, Charles G. Overberger, and Georg Menges, ed., Encyclopedia of Polymer Science and Engineering, Supplement Volume, New York: Wiley-Interscience, 1985, p. 614.

168

217

A. J. Kinloch, "Review: The Science of Adhesion - Part 1. Surface and Interfacial Aspects," Journal of Material Science, vol. 15, 1980, p. 2142. This is a reference to the qualities of an ideal adhesive, mentioning that a viscosity of a few centipoise would be preferred.

218 "Precision Clean Concentrate," Part Number 02701, LPS Laboratories, Tucker, Georgia.

Richard Bullock (Adtech Plastic Systems Corp.), private communication, June 1998. Mr. Bullock stated the glass transition temperature for LUC-4180 as 140-150F; no value appears in the data sheets for the material. Richard P. Wool, Polymer Interfaces: Structure and Strength, Cincinnati: Hanser/Gardner Publications, Inc., 1995, pp. 420-22. Richard P. Wool, Polymer Interfaces: Structure and Strength, Cincinnati: Hanser/Gardner Publications, Inc., 1995, pp. 463-464. The author describes a variety of investigations into crack healing of thermoplastic materials, and mentions experiments which he and other authors conducted to examine the process for epoxy thermoset materials. Richard P. Wool, Polymer Interfaces: Structure and Strength, Cincinnati: Hanser/Gardner Publications, Inc., 1995, p. 446. Initial crack-healing experiments by Jud, Kausch, and Williams on crack healing in PMMA and other materials are described; the healing process occurred between 5-17 C above the glass-transition temperature of the materials. The referenced experiments were reported in K. Jud, H. H. Kausch, and J. G. Williams, "Fracture-Mechanics Studies of Crack Healing and Welding of Polymers," Journal of Materials Science, vol. 16, no. 1, 1981, pp. 204-210. Richard P. Wool, Polymer Interfaces: Structure and Strength, Cincinnati: Hanser/Gardner Publications, Inc., 1995, p. 464. The author poses the question of whether unfractured epoxy interfaces can heal. This idea was suggested by Alex Nickel of the Stanford University Rapid Prototyping Laboratory.
224 223 222 221 220

219

Chapter Six - Applications of Polymer SDM

225

P. Eyerer, B. Wiedemann, K.-H. Dusel, and B. Keller, "Materials for Solid Freeform Manufacturing Processes," Computers in Industry, vol. 28, 1995, pp. 35-45. J. Fessler, A. Nickel, G. Link, F. Prinz, and P. Fussell, "Functional Gradient Metallic Prototypes through Shape Deposition Manufacturing," Proceedings of the 1997 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1997, pp. 521-528.

226

This system of aerodynamic flaps was designed and built by Byongho Park of the Stanford University Rapid Prototyping Laboratory.

227

169

This prototype input device was designed and built by Eric Chamness of the Stanford University Rapid Prototyping Laboratory.
229 This part with an embedded pressure sensor was made by Jorge Cham and Yanjie Sun of the Stanford University Dextrous Manipulation Laboratory. 230

228

This wearable computer was made by researchers at Carnegie Mellon University. Its construction is described in L. Weiss, F. Prinz, G. Neplotnik, K. Padmanabhan, L. Schultz, and R. Merz, "Shape Deposition Manufacturing of Wearable Computers," Proceedings of the 1996 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1996, pp. 31-38.

J. W. Kietzman, A. G. Cooper, L. E. Weiss, L. Schultz, J. L. Lombardi, and F. B. Prinz, "Layered Manufacturing Material Issues for SDM of Polymers and Ceramics", Proceedings of the 1997 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1997, pp. 133-140. A. G. Cooper, S. Kang, J. W. Kietzman, F. B. Prinz, J. L. Lombardi, and L. Weiss, "Automated Fabrication of Complex Molded Parts Using Mold SDM," Proceedings of the 1998 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1998, pp. 721-8.
233 232

231

This figure was published in A. G. Cooper, S. Kang, J. W. Kietzman, F. B. Prinz, J. L. Lombardi, and L. Weiss, "Automated Fabrication of Complex Molded Parts Using Mold SDM," Proceedings of the 1998 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1998, p. 722. Used by permission of the authors.

234

This part was made by Alexander Cooper of the Stanford Rapid Prototyping Laboratory.

This part was made by David Miller and Brad Levin of the Stanford Rapid Prototyping Laboratory. The development team included Alexander Cooper, Sangkyun Kang, Miguel Pinilla, Falk Kaehny, Tom Hasler, and John Kietzman.
237 236

235

This figure was published in A. G. Cooper, S. Kang, J. W. Kietzman, F. B. Prinz, J. L. Lombardi, and L. Weiss, "Automated Fabrication of Complex Molded Parts Using Mold SDM," Proceedings of the 1998 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1998, p. 725. Used by permission of the authors. This figure was published in A. G. Cooper, S. Kang, J. W. Kietzman, F. B. Prinz, J. L. Lombardi, and L. Weiss, "Automated Fabrication of Complex Molded Parts Using Mold SDM," Proceedings of the 1998 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1998, p. 726. Used by permission of the authors.

238

M. Binnard and M. R. Cutkosky, "Building Block Design for Layered Shape Manufacturing," Proceedings of the 1998 ASME DETC/DFM Conference, Atlanta, Georgia, September 13-16, 1998.

239

170

Cf. Charles L. Thomas and Kenneth J. Hayworth, "Automating Sheet-Based Fabrication: The Conveyed-Adherent Process," Proceedings of the 1996 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1996, pp. 281-290. These authors present a different classification of part build times based on layer creation method.
241

240

"Starch Used in New Rapid Prototyping Technology," Medical Device and Diagnostic Industry, September 1997, p. 32. Marshall Burns, Automated Fabrication: Improving Productivity in Manufacturing, Englewood Cliffs, New Jersey: PTR Prentice Hall, 1993, pp. 66-67.

242

Chapter Seven - Conclusions J. W. Kietzman, A. G. Cooper, L. E. Weiss, L. Schultz, J. L. Lombardi, and F. B. Prinz, "Layered Manufacturing Material Issues for SDM of Polymers and Ceramics", Proceedings of the 1997 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1997, pp. 133-140. A. Cooper, J. Kietzman, F. Prinz, and J. Lombardi , "Fabrication of Complex Ceramic Parts using Mold SDM," conference presentation at the Solid Freeform Fabrication Symposium of the American Ceramics Society Annual Meeting, Cincinnati, Ohio, May 4-6, 1998. A. G. Cooper, S. Kang, J. W. Kietzman, F. B. Prinz, J. L. Lombardi, and L. Weiss, "Automated Fabrication of Complex Molded Parts Using Mold SDM," Proceedings of the 1998 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1998, pp. 721-8.
246 245 244 243

John Kietzman and Fritz Prinz, "Material Strength in Polymer SDM," Proceedings of the 1998 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1998, pp. 567-74. "TUFFalloy liquid molding compounds with thermoplastic properties," data sheet, Hapco, Inc., Hanover, Massachusetts, June 1997.

247

"The Por-A-Kast Products," data sheet, Synair, Chattanooga, Tennessee, Spring 1998.
249

248

TDT 232-1 Epoxy Room Temperature Laminating Material, Ciba Specialty Chemicals Corporation, East Lansing, Michigan. The existence of a secondary curing mechanism was mentioned by Jim Hoge (Ciba Specialty Chemicals Corp.), private communication, June 1998.

171

Appendix A - Material Combinations Examined Early materials with high glass-transition temperatures include Ciba-Geigy resins Cibatool SL XB 5131 (Tg=150C) and XB 5139 (Tg=100-120C) and DSM DeSolite 4112-143-1 (Tg=100C). The Ciba resins were designed for the 3D Systems Stereolithography process, and the DSM resin was designed for the Cubital Solid Ground Curing process. These values were reported in Marshall Burns, Automated Fabrication: Improving Productivity in Manufacturing, Englewood Cliffs, New Jersey: PTR Prentice Hall, 1993, pp. 262-263.
251 250

Almost all Stereolithography materials advertised on the World Wide Web in July, 1998 by 3D Systems (Ciba) and Du Pont (SOMOS) have glass-transition temperatures below 85C. The exceptions are Ciba SL 5170, which has a broad glass transition between 65C and 90C, and SL 5410, which can have a glass-transition temperature of 88C if thermally postcured. A very new entrant in this category is Ciba SL 5530HT, which has a glass-transition temperature of 122C [advertised on the World Wide Web by January 1999]. "Nuva-Sil 76 Technical Data Sheet," Newington, Connecticut: Loctite Corporation, Industrial Group, November 1987. Amtex-CCR Silicone Solvent is sold by Amtex Chemical Company, West Chester, Pennsylvania.

252

253

Stuart M. Lee, "Film Deposition Techniques," Martin Grayson, ed., Encyclopedia of Semiconductor Technology, New York: John Wiley & Sons, 1984, p. 308. D. R. Ciarlo and N. M. Ceglio, "Ultrathick Photoresist Processing," Semiconductor Microlithography V, 1980, SPIE Proceedings, vol. 221, pp. 64-73. This paper describes the work of researchers at LLNL who are using photoresists in thicknesses of forty microns as molds for electroplating tiny Fresnel lenses.
256 255

254

Jack L. R. Williams and Michel F. Molaire, "Photoreactive Polymers," in Martin Grayson, ed., Encyclopedia of Semiconductor Technology, New York: John Wiley & Sons, 1984, p. 621. This reference describes the chemical structure of two polyester photoresists with glass-transition temperatures of 271C and 312C. It implies but doesn't say that these are extreme values.

John Grunwald, Edwin J. Turner, Allen C. Spencer, David A. Sawoska, and Giyora Ben-Shushan, "An ultra high temperature positive photoresist," Advances in Resist Processing and Technology II, Santa Clara, California, March 11-12, 1985, SPIE Proceedings, vol. 539, pp. 275-283. This paper describes Macdermid ULTRAMAC PR78T, a commercial photoresist for IC fabrication which can survive baking at 220C without significant dimensional change. Jurgen Diekmann, "Lithography and Photofabrication," Charles A. Harper and Ronald N. Sampson, ed., Electronic Materials and Processes Handbook, 2nd ed., New York: McGraw-Hill, Inc., 1994, p. 10.22. This section claims that PCB resists are used in thicknesses of 15-100 microns.
258

257

172

Robert T. Currier and Stephen G. Mearig, "Thick Photoresist Technology," PC Fab, July 1985, pp. 38-47. This article describes PCB plating resists which could be used at thicknesses of .080".
260 Charles A. Harper and Ronald N. Sampson, ed., Electronic Materials and Processes Handbook, 2nd ed., New York: McGraw-Hill, Inc., pp. 2.52-2.61. This section outlines the uses and characteristics of solder masks for printed circuit boards although it describes only permanent solder masks. 261

259

Solder masks which solidify via the evaporation of a carrier solvent are available from 1) Aim Products, Smithfield, Rhode Island; 2) Chemtronics, Kennesaw, Georgia; 3) Kester Solder Company (division of Litton), Des Plaines, Illinois; 4) Metron Optics, Solana Beach, California; and 5) Multicore Solders Inc., Richardson, Texas.

Gary V. Grosclaude and Andrew G. Bachmann, "Method utilizing water soluble masking formulation for producing an article," US Patent No. 5,388,754, February 14, 1995. This patent describes the use of a water-soluble, ultraviolet-curable solder mask with a particulate which bubbles upon exposure to water, for ease of removal of cured material. This patent is assigned to the Dymax Corporation.
263

262

Greg R. Unruh, "UV curable temporary solder mask," US Patent No. 5,420,171, May 30, 1995. This patent describes two temporary solder masks claimed to withstand solder at a temperature of 266C. One is intended to be soluble in an "aqueous nonalkaline solution," while the other is intended to be soluble in an organic solvent.

"Tech Spray Product Catalog - Revised February 1994," Tech Spray Inc., Amarillo, Texas, p. 34. This section describes Tech Spray's temporary soluble solder mask products. The W, WA, and WS products differ by intended application method and have different viscosities. The S-UVA and W-UVA products are rated to withstand solder deposition at up to 343C; no temperature specification is provided for the W, WA, and WS products. "Tech Spray Product Catalog - Revised February 1994," Tech Spray Inc., Amarillo, Texas, p. 34.
266 "Ultra Light-Weld 9-20311-F UV/Visible Curing Water Soluble Mask: Preliminary Product Data Sheet 9-20311-F," Dymax Corporation, Torrington, Connecticut, February 1996. This product superseded Dymax 948. 267 265

264

"ELC4415 and ELC4497 UV Curable Solder Masks," Electro-Lite Corporation, Danbury, Connecticut, September 1992. Note: There are persistent rumors that ElectroLite does not formulate materials but that Tech Spray does. Tech Spray personnel claim that Electro-Lite resells the Tech Spray Wondermask W-UVA product under an Electro-Lite label. Therefore, Electo-Lite may either modify the Wondermask W-UVA solder mask into ELC4497 or simply resell it as ELC4497 solder mask.

"Tech Spray Product Catalog - Revised February 1994," Tech Spray Inc., Amarillo, Texas, p. 34. "Dymax UV Curable Water Soluble Masks: Preliminary Product Data Sheet 948 Series", Dymax Corporation, Torrington, Connecticut, May 1995.
269

268

173

"Preliminary Technical Information: Strippable Solder Mask SD 2956 UV-HAL, blue, UV-curing," Lackwerke Peters GmbH, Kempen, Germany. Lackwerke Peters products are distributed in the North America by Paradign, Inc., Ashland, Massachusetts.
271 272

270

"Dispense Valve Systems," EFD Inc., East Providence, Rhode Island.

"UV/Visible & Visible Light Curing Equipment / Equipment Selector Guide," Dymax Corporation, Torrington, Connecticut.

The ACR200 material is Poly 2-ethyl-2-oxazoline [John Lombardi (Advanced Ceramics Research), private communication, September 1998]. Following its use as a support material for SDM, researchers at Advanced Ceramics Research also modified it for use as a support material in the Fused Deposition Modeling process. John L. Lombardi, Gregory J. Artz, Dragan Popovich, Ranji Vaidyanathan, and Sajiv Boggavarapu, "Issues Associated with the Development of a Water Soluble Support Material for use in Extrusion Freeforming & Fused Deposition Modeling," Proceedings of the 1998 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1998, pp. 511-7. J. W. Kietzman, A. G. Cooper, L. E. Weiss, L. Schultz, J. L. Lombardi, and F. B. Prinz, "Layered Manufacturing Material Issues for SDM of Polymers and Ceramics", Proceedings of the 1997 Solid Freeform Fabrication Symposium, The University of Texas at Austin, Austin, Texas, August 1997, pp. 133-140.
275 Thomas H. Pang, "Advances in Stereolithography Photopolymer Systems," Chapter 2 in Paul F. Jacobs, Stereolithography and other RP&M Technologies: From Rapid Prototyping to Rapid Tooling, Dearborn, Michigan: Society of Manufacturing Engineers, 1996, p. 32. This is a reference to the oxygen inhibition process as it affects acrylate stereolithography resins. Further information is available in Graham Plews and Roger Phillips, "Some Factors Affecting Cure of UV Curing Inks and Varnishes," Journal of Coating Technology, vol. 51, no. 648, January 1979, pp. 69-77. 274

273

I am indebted to Paul Fussell of ALCOA (Pittsburgh, Pennsylvania) for providing samples of Du Pont Somos 2100 resin. "Du Pont Solid Imaging Materials Group Product Data - Somos 3100 and 3110 Photopolymers," Du Pont Somos Solid Imaging Materials Group, New Castle, Delaware, March 1994. The properties reported in Table A.2 are those measured following postcuring under fluorescent or ultraviolet light.
278 277

276

"SOMOS 2100 Family Product Data," Specification Ver. 1.26.93, Du Pont SOMOS Group, New Castle, Delaware, January 1993. The properties reported are those measured following thermal postcuring.

279

Raw volumetric shrinkage of acrylate materials of 6-7% can be reduced to an effective linear shrinkage of 0.7-0.9% through build-style choices. Raw shrinkage of epoxies of around 7% can be reduced as low as 0.05% using Ciba SL 5170 and other optimized build styles. Reported in Thomas H. Pang, "Advances in Stereolithography Photopolymer Systems," Chapter 2 in Paul F. Jacobs, Stereolithography and other RP&M Technologies: From Rapid Prototyping to Rapid Tooling, Dearborn, Michigan: Society of Manufacturing Engineers, 1996, p. 40.

174

280

This idea was suggested to me by Paul Fussell of ALCOA.

281 Thomas H. Pang, "Advances in Stereolithography Photopolymer Systems," Chapter 2 in Paul F. Jacobs, Stereolithography and other RP&M Technologies: From Rapid Prototyping to Rapid Tooling, Dearborn, Michigan: Society of Manufacturing Engineers, 1996, pp. 31-37. This section describes the chemistry of stereolithography resins and their cure mechanisms, including those of acrylate, epoxy, and vinylether materials.

I am indebted to Gayle Link, who procured a sample of Ciba 5180 resin from General Motors. "Resin Data Supplement: Cibatool SL 5180," P/N 19140-S10-00 Rev. A, Ciba-Geigy Corporation, Formulated Systems Group, East Lansing, Michigan, p. 2. This applications note mentions that small amounts of acrylate resins will contaminate epoxy SLA resins, although it does not describe the negative effects which may result. "Tips & Info Technical Bulletin: Polymer Change and Vat Cleaning Procedures for CIBATOOL Resins," P/N 19093-S08-02 Rev. A, 3D Systems, Valencia, California.
285 284 283

282

Vinylether resins cure immediately upon exposure to acrylate resins although this might still allow their use with solder mask part materials. The deposition process would have to ensure that material was initially placed at its final location, rather than needing to flow into corners or cavities before being cured. The chemical incompatibility was described by Paul Fussell (ALCOA), private communication, February 1996. "Technical Data Sheet: Emcast 1720, 1721, 1722, 1723 Generation 2 and H. V. Generation 2 U. V. Epoxy Adhesives," Rev. 3/96, Electronic Materials Inc., Breckenridge, Colorado.

286

"518 Series UV Curable Adhesives," UVEXS Incorporated, Sunnyvale, California. This is an ultraviolet-curable urethane acrylate resin with 80 Shore D hardness and 31 MPa tensile strength. The data sheet advertises it as usable for potting applications, so thick layers may be curable in a single step. "Technical Data: Washaway Mounting Adhesives," Bulletin No. M10, Aremco Products, Inc., Ossining, New York, March 1995. "Technical Data - STYCAST 3020 Epoxy Encapsulant," Emerson & Cuming, Inc. Woburn, Massachusetts, 1993.
290 "Technical Data - STYCAST 3050 Epoxy Encapsulant," Emerson & Cuming, Inc. Woburn, Massachusetts, 1993. 291 289 288

287

"Data Sheet: Dynasolve 700 Series," Dynaloy, Inc., Hanover, New Jersey, June 1986. This is a material for dissolving cured urethanes and anhydride-curing epoxies without harming amine-curing epoxies. "Model-Maker Application Note #1 - Build and Support Material Characteristics," Sanders Prototype, Inc., Wilton, New Hampshire. This article describes the characteristics of the Sanders material combination and describes the removal of support material via soaking in Petroferm BIOACT VSO slightly below the melting point of the support material.

292

175

293

"BIOACT 280 Precision Cleaner Technical Data Sheet," Petroferm, Inc., Fernandina Beach, Florida. Another useful document is "BIOACT 280 Precision Cleaner Wax Removal Applications," Petroferm, Inc., Fernandina Beach, Florida.

I am indebted to Tricey H. Godfrey of Petroferm, Inc. (Fernandina Beach. Florida), who performed this series of tests for me.
295 I am indebted to Larry Clark of Enviro Tech International (in the Alameda, California, office) for providing samples of EnSolv and procuring a test site with equipment to test it on polyurethane samples. 296 "Typical Properties of Ciba Parts in Minutes Polyurethanes," Ciba-Geigy Corporation, Formulated Materials Group, East Lansing, Michigan. 297

294

"Product Bulletin: LUC-4102 Super Lowex Casting System - Set Fast Reproduction Urethane," ADTECH Plastic Systems Corp., Charlotte, Michigan, May 1995.

"Product Data: Pro-Cast 20 Reproduction Urethane," Ciba-Geigy Corporation, Formulated Materials Group, East Lansing, Michigan, May 1996. "Product Data - TDT 205-3 Resin and Hardener / DT 081 / DT 082," Ciba Specialty Chemicals Corporation, East Lansing, Michigan, October 1996. "System5 Epoxy Encapsulation Systems," Adtech Plastic Systems Corp., Charlotte, Michigan.
300 299

298

Appendix B - Tensile Testing Notes

301

Thomas G. Beckwith and Roy D. Marangoni, Mechanical Measurements, 4th ed., Reading, Massachusetts: Addison-Wesley Publishing Company, 1990, pp. 37-66.

302

Robert V. Hogg and Elliot A. Tanis, Probability and Statistical Inference, 3rd ed., New York: Macmillan Publishing Company, 1988, pp. 348-353. Thomas G. Beckwith and Roy D. Marangoni, Mechanical Measurements, 4th ed., Reading, Massachusetts: Addison-Wesley Publishing Company, 1990, pp. 43-44.

303

176