The evolution of semiconductor materials management in the last decade
Resin inspection, materials and production management are changing to keep pace with the semiconductor industry`s stringent contamination control requirements
By Penny Van Sickle, Fluoroware, Inc.
Prior to October 1987, the Dow Jones Industrial Average was at 2,000 and linewidths on semiconductor devices were approximately 1.0 micron. This year, the Dow Jones is up to 8,000 and 0.18-micron devices are currently being developed in laboratories. In addition, working transistors for 300 mm wafers have been manufactured1 but 300 mm wafers are not yet available in production volumes. As these milestones indicate, numerous changes have taken place in industry, and within semiconductor manufacturing specifically, during the last decade.
To achieve this linewidth reduction, the requirements for manufacturing materials have become tighter and tighter. Because of this, the acceptable contamination levels of the materials used to manufacture devices have been reduced significantly. For 0.18-micron linewidth products, specifications on chemicals are approaching 100 ppt for each metal while particle specifications are now less than 10 particles/ml 0.1 micron. To be safe, actual metal and particle concentrations generally need to be 5 to 10 times lower than the specification.2
DI water, ultrapure chemicals and silicon wafers used by device manufacturers are stored and transported using polymeric-based products such as piping, tubing, fittings, valves, drums, day tanks, wafer shippers, wafer carriers, storage boxes and chip trays. As the cleanliness specifications tighten on DI water, chemicals and silicon wafers, the performance expectations of the polymeric products with which they come in contact also have risen.
Many polymeric resins are used to manufacture products for the semiconductor industry. Perfluoroalkoxy (PFA) and other fluoropolymers are used in products such as piping, tubing, drums and process wafer carriers that contact strong chemicals. Polypropylene, polycarbonate, polybutyleneterphthalate (PBT), polyetheretherkeretone (PEEK) and polystyrene are used for wafer shippers, wafer carriers, storage boxes, mask packs, and chip trays.
In order to ensure that the final products meet specification, it is important that the resins used in these products are tightly controlled. As the semiconductor industry has evolved over the past 10 years, so have inspection requirements on resins. A comparison of resin inspection testing for 1987 and 1997 is shown in Table 1. Some tests have not changed in 10 years while others have become far more sophisticated.
One test that has not changed is the visual inspection of the resin. For this test, a sample is collected from the resin container and spread out on a white tray. Using bright white light and a magnifying glass, the resin is examined pellet by pellet. Pellets that are the wrong shape, the wrong color or visibly contaminated with specks of impurities are identified. The overall sample is also examined for evidence of gross contamination. During the last 10 years, gross contamination in the form of metal shavings due to wear on resin extruders or wood chips due to carpentry work performed at the supplier have led to rejection. This test, while low tech, is surprisingly effective.
One of the tests that has changed significantly in the last 10 years is the inspection for metal contamination. In 1987, this testing was performed by spreading the resin out on a table and then slowly running a very strong magnet above the resin. The number of pellets that moved or stuck to the magnet were counted to ensure that this number was not above specification. Of course, some metals such as aluminum do not respond to magnets and this test did not quantify how much or what type of metal was present.
Early in the 1990s, the use of inductively coupled plasma-mass spectroscopy (ICP-MS) became more and more commonplace. With ICP-MS, the bulk amount of metal in a resin can be quantified by dissolving or ashing the resin. Extractable metal concentrations can be determined by immersing the resin in a liquid such as nitric acid and then testing the liquid. ICP-MS can differentiate metals so that ppb/ppt extraction values of 68 different cations can be determined. The use of ICP-MS for incoming resin inspection ensures that material lots with high metal levels are identified and rejected before being used to manufacture products.
In 1987, DuPont PFA 340 J, DuPont 340 EG and Daikin Neoflon AP210 were the PFAs available worldwide for manufacturing products for strong chemical contact, including piping, tubing, drums and process wafer carriers. However, these PFAs increased fluoride concentrations in liquids to unacceptable levels and etched silico wafers stored in PFA carriers.3 To address this issue, DuPont and Fluoroware, Inc. began a joint effort to develop a fully fluorinated material having extractable fluoride concentrations that were 20 to 100 times lower than other PFAs on the market, as shown in Figure 1. This fully fluorinated PFA was commercialized in 1990 and given the trade grade of PFA 440HP. Since 1990, other PFAs have been developed and released with similar properties.
In 1987, the polypropylenes used to manufacture wafer handling products were standard polypropylenes available on the commodity market. In 1990, it became apparent that the outgassing qualities of the commodity polypropylenes were not meeting the needs of the semiconductor industry. Device manufacturers began noticing that the surfaces of wafers stored in wafer shippers became hydrophobic. At the same time, research showed that hydrophobic wafers were more difficult to clean than hydrophilic wafers and that particles in wet baths were attracted to and redeposited on hydrophobic wafers when pulled through the liquid/air interface.4 The plasticizers, heat stabilizers and other additives in the commercially available polypropylene were then evaluated to determine which additives affected the wafer surface. Once non-contaminating additives were identified, new polypropylenes were developed and introduced for use in wafer shippers, transport wafer carriers and storage boxes.
By the late 1980s, the emphasis was changing from manual handling to robotic handling to address contamination and ergonomic issues in the fab. The standard mechanical interface (SMIF) pod was introduced. Roll transfer of wafers was eliminated in favor of robotic and slide transfer, especially as the use of 8-inch wafers increased. Higher particle increases were seen with manual handling compared to robotic handling. The “top wafer effect” was also seen on manually handled wafers but not on wafers handled robotically.5,6 However, robots only work well if the wafers are always in the same place. This means that the wafer carrier had to have tighter dimensional tolerances initially and during use in the fab.
In 1987, carbon powder filled polypropylene was one of the more commonly used static-protective wafer carriers. However, polypropylene carriers can warp when exposed to force or high temperature.7 Thus, a material that was more resistant to dimensional changes due to stresses or temperature was needed. This led to the development of the first carbon-fiber loaded PEEK carrier, which combined the thermal and dimensional properties of PEEK with the improved electrostatic discharge performance of carbon fiber. As shown in Figure 2, this product is more resistant to dimensional changes than polypropylene-based carriers and also has the benefit of much lower particle generation. Carriers made from this material have been in use for six years at some fabs without requiring replacement.8
The drive for quick response on new products and the need for tighter tolerances and fewer adjustments also led to a revolution in mold design and product manufacturing, as shown in Table 2.
Mold design has been more of an art than a science. In 1987, the mold designer would obtain general information from the resin supplier for recommended mold pressures, gating positions, gating thickness and mold temperatures to design the mold. After manufacturing the mold, which can cost from $50,000 to $250,000, the first round of testing would be performed to find areas requiring modification. The mold would then be modified, retested and modified again until all dimensions were correct.
Tight dimensional tolerances coupled with the fact that carbon-fiber filled resins have different shrink rates depending on fiber orientation required a more precise method of mold design. Several finite element analysis programs now exist for evaluating mold designs. Using a software package, such as C-Mold or MoldFlow, the location and size of gates can now be evaluated numerically to determine how well the mold will fill and whether the part will warp during cooling. By varying gate locations, gate size, cooling line placement, mold temperature, molding pressure, etc., the mold design can be evaluated and modified before any steel is cut.
See Figures 3 and 4 for examples of MoldFlow analysis for a shipper product. In Figure 3, about half of the end wall of the original shipper design would have warped inward 0.008 inch (depicted by the dark blue areas). When numerical analysis showed that changes in mold temperature and gate location would not improve the warp, a rib feature was added to the exterior of the endwall. As shown in Figure 4, the warpage has been significantly reduced.
There has been an additional revolution in the actual manufacturing of the products in order to create a cleaner and more reproducible product. In 1987, the gates used to fill the mold became a part of the product when the product was removed from the mold. These gates then had to be removed (degated) and discarded before the product could be used, generally by an operator wielding a pair of clippers and an Exacto knife.
Since 1987, there have been several improvements in degating technology. First, pneumatically operated degating robots were built to remove the gates consistently. Molds are also manufactured with valve gates, which are placed at the end of the gate, just before the melted material enters the product. After the mold is filled, the valve gate closes and the product is removed from the mold. Since the valve is placed at the entrance to the mold, there is no gate to discard and particle generation is eliminated. True cleanroom manufacturing can now be performed when needed, which leads to a much cleaner final product. Once the molding is complete, it is important that the bags used for packaging have low particle, extractable anion and outgasssing properties so that the bag does not recontaminate the cleanroom manufactured product.
The last 10 years have seen great strides in integrated circuit manufacturing that have led to a revolution in the manufacturing of plastic materials management products for the semiconductor industry. Resin inspection has become far more sophisticated, finite element analysis is used to design molds, manual handling of the products after molding has been reduced significantly and products can truly be cleanroom manufactured. As the semiconductor industry transitions to 300-mm wafers and into the 21st century, product manufacturing will continue to evolve. CR
Acknowledgments. Thanks to Barb Cygan for the MoldFlow figures, Dave Nyseth and Bernie Shambour for helpful discussions on mold improvements, and Chuck Extrand and Kerry Kiser for their insightful comments on this article.
Penny M. Van Sickle is a senior contamination control engineer in the technology research group at Fluoroware (Chaska, MN) and serves as one of the program directors for the contamination control section of the Institute of Environmental Sciences 1997 and 1998 annual technical meetings. She holds a BS in material sciences and engineering from the University of Minnesota and is the author and co-author of over 10 technical articles. She joined Fluoroware in 1989.
1. C. Haris, M. Hiatt, et. al., “Fabrication of the First Transistors on 300 mm Wafers,” Semiconductor International, August 1997, pp. 67-74.
2. L. Hall, J. Sees, and A. Misra, “Chemical Management for UHP Wafer Fab Applications,” Proceedings of the 1996 Ultra Clean Processing of Silicon Surfaces Conference, pp. 13-15.
3. J. Goodman and S. Andrews, “Fluoride Contamination from Fluoropolymers in Semiconductor Manufacture,” Solid State Technology, July 1990.
4. D. Riley and R. Carbonell, “The Deposition of Liquid-Based Contaminants onto Silicon Surfaces,” 1990 Proceedings of the Institute of Environmental Sciences Conference, pp. 224-228.
5. P. Muller, S. Silverman, et. al. “The Norcross Project: Comparing Manual and Robotic Wafer Handling Using Minienvironments in a Degraded Cleanroom,” Microcontamination, September 1994, pp. 25-30 and 68.
6. P. Muller, S. Silverman, et. al., “The Norcross Project: Investigating the Relationship Between Ergonomic Factors and Particle Addition in Manual-Access Minienvironments,” Microcontamination, October 1994, pp. 39-43.
7. J. Prestidge and P. Van Sickle, Fluoroware Lab Report 9101-13 “Effect of Proper and Improper Stacking on Wafer Carrier Dimensions,” April 25, 1991.
8. P. Jones, I. Emami, J. Goodman, K. Mikkelsen, “Evaluating PEEK Based Transport and Storage Wafer Carriers,” Microcontamination, January 1993, pp. 29-38.
As part of a joint effort, DuPont and Fluoroware, Inc. developed a fully fluorinated material with extractable fluoride concentrations that were 20 to 100 times lower than other PFAs on the market. The result was DuPont`s 440 HP, shown in Figure 1. Carbon-fiber loaded PEEK carriers (shown in Figure 2) are more resistant to dimensional changes than polypropylene-based carriers and also have the benefit of much lower particle generation.
In Figure 3, about half of the end wall of the original shipper design would have warped inward 0.008 of an inch (depicted by the dark blue areas). In Figure 4, the warpage has been significantly reduced by adding a rib feature to the exterior of the end wall. Without MoldFlow analysis, this warpage would not have come to light until after the mold was completed.