Advancing the standards for ultrahigh-purity fluoropolymer components
11/01/2004
New wet processes for semiconductor manufacturing require higher standards for purity and improved design and materials for fluid system components. Semiconductor equipment makers and tool integrators are calling for longer life cycles, easier maintenance, resistance to permeation, and reduced particle shedding in valves and other system components. The four-year-old Semi F57-0301 standard provides a baseline for purity, but third-party tests underscore the need for more stringent industry standards to address new wet processes and materials requirements. Methodologies and conclusions discussed in this article are applicable not only to valves but also to many other liquid system components, such as flow meters, filter housings, and regulators.
Increased use of liquid processing, along with new processes and materials, such as abrasive slurries for chemical mechanical planarization (CMP), are pushing the requirements higher for improved performance and reliability in fluid system components. As the industry moves toward larger wafer sizes as well as narrower and denser linewidths, greater demands are placed on the purity of fluid system components. Productivity, profitability, and process yield are directly related to overall system-component performance.
Such trends led to the development and approval of Semi Standard F57-0301 in October 2000. The purpose of the standard is to specify the minimum performance requirements for ultrahigh-purity (UHP) polymer components used throughout semiconductor ultrapure water and chemical distribution systems. The scope of the standard encompasses polymer component purity and mechanical specifications, as well as references for qualification test methods. Certification, traceability, and packaging requirements are also included. The standard focuses primarily on purity requirements, however, leaving mechanical requirements such as dimensional tolerances, flow characteristics, leak integrity, and mechanical strength to the supplier's discretion.
This article will discuss and provide evidence supporting Semi F57 and more stringent industry standards that should be incorporated into the Semi standard. It will also propose some guidelines concerning the design, material selection, and manufacturing of fluid system components that enable component suppliers to meet the industry's highest standards.
Overriding design criteria
When developing components for UHP applications, the fluid system designer must be in close touch with the present and future requirements of leading semiconductor equipment makers and integrators. Communication must be ongoing and thorough, covering the following points in designing next-generation valves.
- Designs should exceed, where possible and practical, the current purity requirements of Semi F57.
- Design efforts should focus on the most chal-
- lenging process requirements, such as those involving CMP and acids. Next-generation valves should demonstrate superior performance in any liquid-chemical distribution system.
- Designs should address long life, reliability, and ease of maintenance. These issues should be given equal weight compared to initial performance issues, such as pressure/temperature ratings and footprint.
- Designs should be statistically and data driven, given the availability of thresholds required by the industry.
Material selection and manufacturing
Given the listed design criteria, particularly the second item, material selection is critical. Traditionally, suppliers of high-purity valves for the semiconductor industry have chosen perfluoroalkoxy (PFA) for wetted components. PFA is produced by the copolymerization of trifluoroethanol (TFE) and perfluoroalkyl monomers. These are produced by reacting fluoroepoxide with a metal fluoride to obtain an acid fluoride, which is then pyrolyzed over calcium carbonate to obtain propylvinyl ether (PVE). In recent years, Dupont, Dyneon, and other fluoropolymer resin manufacturers have improved the purity of PFA.
In certain fluid components, such as valves, there is good reason to call PFA into question, given the ever-increasing demands for higher-purity wetted surfaces, as evidenced in Semi F57 and higher performance requirements for acid and CMP processes. The alternative to PFA is PTFE, which is inherently less reactive than PFA and has superior chemical resistance and qualities for purity. It may well be the material of choice for next-generation wetted components [1].
PTFE, the original fluoropolymer resin, is simply polymerized TFE. It has a less complex structure than PFA and requires less processing to reach a usable state. DuPont's Teflon NXT grade of PTFE (modified PTFE) provides the following benefits: high purity; ease of machining; and resistance to acids, abrasives, and other operating stresses. It contains <0.01% PPVE, so it is still considered PTFE.
 Figure 1. Comparison of modified PTFE and PFA processing steps for wetted components. |
Historically, most high-purity valves have been manufactured using injection-molding manufacturing technology, but this technology carries with it certain risks from a purity standpoint. PFA, the material of choice for injection molding, must pass through several processes in close contact with metal (Fig. 1). Melt extrusion of PFA into molding pellets occurs in an extruder with a metal barrel and screw that conveys the highly corrosive PFA melt at temperatures in excess of 300°C through a metal-pelletizing die. These PFA molding pellets are then remelted at temperatures in excess of 300°C in the injection-molding process. Again, the molten PFA is conveyed in a metal barrel and screw, and then into a high-temperature metal injection-mold die to form the final valve shape.
By contrast, the manufacturing of wetted components with PTFE is a cleaner process that helps to eliminate metallic contamination. Sintering of PTFE is accomplished in block form. Machining of the block into a valve occurs without lubricants, using special cutting tools at room temperature. Further, precision machining is performed at closer tolerances compared to injection molding, and machining enables faster and more economical turnaround time for product and manufacturing-process development. It allows for great flexibility in changeovers between valve configurations.
Design and flow path
Table 1 translates desirable valve attributes into design guidelines. The attributes are based on Semi F57 and other industry requirements, whereas the design guidelines are based on principles of physics and valve geometry. In large part, Table 1 underscores the following axiom: The valve must not adversely affect critical fluids (such as slurries), and the critical fluids must not adversely affect the valve.
 |
Six of the eight guidelines in Table 1 are in part, if not wholly, related to flow path within the valve. Flow path is a critical consideration, not only from the standpoint of Semi F57 purity requirements for particle contribution but also from the standpoint of recent advances in semiconductor manufacturing, such as copper CMP, which requires gentle treatment of slurries to minimize agglomeration. Many valves employed in the semiconductor industry today were not designed for advanced processes. In fact, their design sometimes predates the invention and widespread use of these processes.
Computational fluid dynamics (CFD), the leading computer-aided modeling tool for fluid systems, enables engineers to model different geometries (with different critical system fluids) to determine how those geometries will perform and whether they meet the design guidelines in Table 1. CFD predicts the flow coefficient for any given valve design, and illustrates and calculates fluid velocity, shear forces, and incremental pressure drops throughout the valve.
In terms of fluid flow, the body and diaphragm are the most important sections of the valve (Fig. 2). CFD enables the designer to hold certain design elements constant while experimenting with others for optimized performance of the flow path. For example, a designer may experiment with the size of the inlet and outlet orifices or their angle in relation to one another. Or the designer may focus on the shape of the bowl or cavity itself, in combination with the diaphragm geometry.
Besides flow path, Table 1 addresses issues of material integrity and wear, particle shedding, and permeation, all of which may be tested at the design stage through finite element analysis (FEA). FEA provides the fluid component engineer with a time-stepped analysis of the stresses and deformation of each material or component section, such as the diaphragm and seat seal. FEA is particularly valuable in the design of extended life-cycle fluoropolymer diaphragms, which must resist permeation and attack by acids, slurries, and other critical chemicals used in the semiconductor industry. While extended life cycle requires a flexible diaphragm, resistance to acids requires a thick diaphragm. FEA assists the engineer in determining the best balance between these two competing requirements.
In employing FEA software, designers specify material properties; boundary conditions, such as temperature and pressure; stress limitations; and diaphragm geometry. The program then models and tests designs in 3D space. Results enable engineers to experiment with diaphragm contours, thickness, and functionality until the desired or optimal design is reached.
High-purity testing
As noted earlier, Semi F57 focuses on purity, particularly static leach tests utilizing DI water. The standard is specific in identifying levels of ionic metallic and total organic carbon contamination that could lead to complications in semiconductor manufacturing. Given the increasing demands of OEMs, integrators, and toolmakers, Semi F57 performance levels should be regarded as baseline or minimum requirements.
Unfortunately, there are some in the industry who question whether Semi F57 is realistic and attainable, or too exacting. The Semi standard is, first and foremost, a means of communicating between those who use and those who make fluid system components. If OEMs or toolmakers desire a particular standard, fluid system manufacturers must determine if the standard is attainable and repeatable under independent testing conditions. Tables 2–4 contain results for surface-extractable contamination tests performed on Swagelok's DRP series of UHP fluoropolymer diaphragm valves. The independent tests were conducted by CT Associates Inc., in accord with Semi Standard F40 testing procedures. Semi F57 contamination limits are provided in the tables for reference and comparison.
 |
While static leach tests required by Semi F57 provide solid baseline-performance guidelines, most field applications are dynamic, not static. Dynamic leach tests, not only for DI water but also for HCl and HF, as required by leading OEMs in the semiconductor industry, should find their place in the Semi standard. One such test is BOC Edwards' Dycon Ex dynamic extraction test, which utilizes 37% HCl. HCl is an appropriate medium for aggressive testing because it is one of the most effective chemicals in permeating fluoropolymers and in extracting metallic particles. Dycon Ex employs minimal chemical volume over time, enabling detection of extractables at very low levels. The industry expectation is <20ng/cm2 for bulk chemical extraction for the area normalized surface contamination from 37 extracted elements. The requirement for the area normalized extraction rate is <0.5ng/cm2/day at 7 days. With proper component design and manufacturing procedures, these standards are attainable. CT Associates conducted the Dycon Ex dynamic extraction test with the DRP series valve and results were well under industry limits (Table 5). The area-normalized surface contamination from 37 extracted elements was 9.36ng/cm2. The area normalized extraction rate was 0.06ng/cm2/day at 7 days.
 |
Besides tests for ionic, metallic, and total organic carbon contamination, Semi F57 requires a test for particle contribution. This calls for a rinse-timed test with DI water, followed by cycle evaluations. The objective is to measure particle shedding resulting from valve actuation, as well as subsequent cleanup over time. This is an important test, but unfortunately, specific limits for particle contribution have not been designated in the standard because no consensus has been reached. While some individuals (or companies) doubt that particle contribution can be measured reliably with repeatable results, there is growing acknowledgment that it can. Leading OEMs have issued standards pertaining to particle shedding. Table 6 represents particle contribution test results for the DRP, performed by Air Liquide-Balazs Analytical Services and corroborated by CT Associates.
 |
In field applications, semiconductor components process acidic chemicals, which in some cases may undermine component integrity and increase susceptibility to particle shedding. Future iterations of Semi F57 should look toward additional media, such as HCl and HF. Testing protocols for acids, slurries, and other critical chemicals should be an integral part of any UHP fluoropolymer valve design, manufacturing, and testing process.
Semi F57 should also look toward longer, more rigorous testing for particle shedding. While life-cycle testing is not the focus of Semi F57, it is germane to issues of purity and particle shedding. How does extensive life-cycle testing in HF or HCl affect a valve's particle shedding? How many liters of DI water are required to flush the valve clean after testing in HF or HCl? What is the overall projected COO given the amount of DI water required to bring the particle count down to acceptable levels?
One standard issued by a major equipment manufacturer requires accelerated life-cycle testing in fluids appropriate to the valve's intended use, such as HCl or HF. Accelerated life-cycle testing evaluates the statistical number of cycles, designated as the B10 life, where 10% of the valves would be expected to fail. In accord with the industry standard, CT Associates tested the 1/2-in. DRP in 49% HF, checking the test valves for cracking pressure and port-to-port pressure every 150,000 cycles. Because the valves were not cycled to failure, the Weibayes method [2] was used to calculate the B10 life. No valve failures were recorded throughout the 1.05 million cycles run for the test; for comparison, the industry standard is 1 million B10 life. The Weibull mean time to failure was 1.9 million cycles, and the lower bound on the mean-time-between-failures confidence interval was 3.6 million cycles (at a 90% confidence level).
 |
The same industry standard identifies specifications for particle shedding tests. First, a passive flush test measures the amount of DI water required to bring the particle contribution down to <0.1 particles/ml, with a designated particle size of <0.10µm. The standard calls for ≤300ml. The 1/2-in. DRP met the standard in just 97 liters before life-cycle testing in HF and 45 liters after HF testing. Second, a cycle test records the number of on-and-off cycles required for the valves to produce <100 particles/actuation, with a designated particle size of <0.10µm. Table 7 contains results for the 1/2-in. DRP. Both tests were conducted by CT Associates.
Conclusion
Semi F57 is an attainable, appropriate, and useful standard for measuring the purity of fluoropolymer components. With appropriate selection of materials (modified PTFE for all wetted surfaces), advanced manufacturing practices (machining), computer-aided modeling techniques, and rigorous third-party testing, it is possible to design and manufacture components that meet or exceed this Semi standard. If Semi F57 is to remain a viable and relevant standard, additional tests and standards already in use within the industry must be incorporated. These new standards, while stringent, are attainable and appropriate, given current processes and practices in the field. They include dynamic leach tests, life-cycle and particle-contribution testing, and tests utilizing specific media used in the semiconductor manufacturing process, including slurries, HF, and HCl.
Acknowledgments
Teflon NXT is a registered trademark of DuPont. DRP is a trademark of Swagelok Co. Dycon Ex is a servicemark of BOC Edwards.
References
- J. Hanson, "Innovations in Semiconductor Manufacturing," Semicon West Technical Symposium, 2003.
- R.B. Abernethy, The New Weibull Handbook, 2000.
For more information, contact James M. Hanson at Swagelok Co., 31400 Aurora Rd., Solon, OH 44139; ph 440/349-5934, fax 440/349-5806, e-mail [email protected].