 |
 |
 |
 |
|
Comments? Suggestions? Send us your feedback.
MicroMagazine.com |
SURFACE CHEMISTRIESEvaluating the performance of polysulfone membrane filters in SC-1 chemistriesDarryl Weddington, Todd Ridley, and Vincent Ladigo, Motorola APRDL; and James Campbell and Mark Gogol, Memtec Electronics Even as cleaning technologies are becoming more critical to achieving high yields in submicron semiconductor fabrication, particle problems in these wet processes have almost disappeared. This accomplishment can be traced to the use of automated wet benches, higher-grade starting chemicals, filtration systems with increasingly smaller pore sizes, and other advanced techniques that reduce the sources of contamination in the manufacturing process. In wet processes, polytetrafluoroethylene (PTFE) membrane filter cartridges have been the industry standard for many years because the membrane material can withstand elevated temperatures in acetic and base chemical cleaning solutions while providing excellent particle capture efficiency. Despite these capabilities, PTFE cartridges have their drawbacks: they exhibit limited flow characteristics, require prewetting, and are generally more expensive than non-PTFE membranes.
The study reported in this article evaluated the use of a microporous polysulfone membrane material with a highly asymmetric pore structure as an alternative to the conventional PTFE offerings.1 The goal of this evaluation was to reduce process operating costs. It was anticipated that the test filter could lower cost of ownership as a result of its lower unit cost while providing equivalent or better filtration performance. According to the manufacturer, the unique structure of the chemically resistant polysulfone membrane maximizes flow capability and allows for a double-layer filter configuration that enhances particle capture at a lower flow resistance than is possible with conventional single-layer PTFE filters. The membrane material is offered in Experimental Equipment and Methods The investigation of the polysulfone filter focused on its performance in a standard clean 1 (SC-1) immersion bath that is part of a standard RCA clean used in front-end processing during the manufacture of 0.35-µm logic technology at Motorola's Advanced Products Research and Development Laboratory (APRDL) in Austin, TX. The SC-1 solution used is a standard mixture of 5:1:1 deionized (DI) water, ammonium hydroxide, and hydrogen peroxide. All chemicals are gigabit grade, and the DI water has bubbled ozone. During the test period, this solution was heated to 50°C, and changeout frequency was every 12 hours.
Figure 1: Diagram of the automated surface preparation system used in the evaluation. SC-1 Process Equipment. The evaluation was carried out on an automated surface preparation system (a wet bench) enclosed in a Class 1 minienvironment. As seen in Figure 1, there are six process modules on this bench: a piranha (H2SO4/O3) clean, a buffered oxide etch (BOE), a hydrofluoric acid (HF) bath, an SC-1, an SC-2, and an isopropyl alcohol (IPA) dry. Each process module except the IPA dry is followed by a DI-water rinse. The SC-1 process tank is a megasonic (300-W) unit constructed of PVDF, with an inner tank and outer weir constructed of PFA. The tank has a capacity of approximately 7 gal, while the module's recirculation loop and filters may hold up to 4 gal. As the schematic in Figure 2 shows, the recirculation unit is composed of a pump, two filters, plumbing lines that circulate the cleaning chemistry, and several infrared (IR) heaters, which warm the solution to the process temperature set point.
Figure 2: Diagram of the SC-1 tank and recirculation unit. After the wafers are processed through the SC-1 solution, a robot moves the cassette into a quick-dump rinse (QDR) module. Five seconds later, heated DI water is sprayed onto the wafers and fed into the bottom of the tank at a rate of 14 gal/min. The DI water then ramps down to ambient temperature and is quickly removed. The dryer module uses liquid IPA that is heated to a vapor state, then forced into the chamber by flowing nitrogen (N2). The system is purged with N2, followed by a final dry. The wet bench's robot system uses a quartz end effector to pick up and move the cassettes in a module. The system can accommodate either quartz or PFA Teflon carriers; however, only Teflon carriers were used because quartz is known to etch and shed particles while in an etchant solution.2 Test Methods. The polysulfone filters, which had been factory flushed with ultrapure water, were installed on the tool without prewetting or flushing. After the SC-1 tank was filled with the cleaning chemistry, the following tests were performed over the 5-month study period to assess the performance of the polysulfone filters. Surface Particle Measurements. As part of the standard manufacturing qualification, particle monitors were processed through the tool using the aforementioned modified RCA clean every 24 hours for 130 days. The particle monitors were 200-mm bare-silicon P-type test wafers (W-545). Pre- and postprocess light point defect measurements for particles >=0.2 µm were taken on the monitors using a wafer surface analysis system with a 5-mm edge exclusion. Two particle monitors were processed for each product run, and resulting data were averaged over the entire study period. In Situ Particle Monitoring. Samples of the SC-1 chemistry were taken every 60 seconds for 11 days and analyzed with an in situ particle counter. The fluid chemical sampler was connected to the SC-1 module's recirculating wet tank to draw samples from the outer weir. Each fluid sample passed through a burette into an overflow chamber. After this overflow chamber was drained to force any bubbles back into the solution, the sample was compressed at about 60 psi, then forced through the spectrometer-type particle counter at a known flow rate and pressure. When the analysis was completed, the pressurized lines were vented and the sample was returned to the recirculation unit of the SC-1 bath.3 Metallic Contamination Measurements. Total x-ray reflection fluorescence (TXRF) was used to determine the presence on the monitor wafer surface of high-molecular-weight metal contamination that may have leached out of the chemical solution during processing. Fourteen TXRF monitors--bare-silicon P-type prime wafers (W-103) that had not undergone any previous chemical processing--were processed through the SC-1 tank during the 5-month study period. A control wafer was measured to determine the level of metallics on the test wafers prior to processing. Inductively coupled plasma mass spectrometry (ICP-MS) was used to determine trace quantities of metal elements in the chemical solution. Five samples were taken during the study period for analysis by this technique. Product Monitoring. In addition, an electrical data analysis system was used to determine the effects on the cache and unit circuit areas of 0.35-µm logic technology being manufactured. The cache circuits are process development vehicles consisting of multiple-bit cells used in the actual product; the unit circuits are actual product die. Several testing methods were used, including cross-sectioning and applying voltage to the gate, to determine the reliability of these circuits. Baseline data for the PTFE membrane filters that had been in use in the tool were obtained in similar fashion before the polysulfone membrane filters were installed on the tank. Wafer surface particle data were collected from the lab's SPC system by manufacturing personnel during their daily qualification of the tool. SC-1 chemistry samples were taken every 60 seconds for 11 days and analyzed with the in situ particle counter. Six TXRF monitor wafers were used to determine contamination levels of high-molecular-weight metals that may have leached out of the chemical solution during wafer processing. ICP-MS was used to determine trace quantities of metal elements in the chemical solution. Finally, baseline cache and unit data from 12 product lots were extracted from the APRDL data storage system.
Table I: Comparative particle data for polysulfone and PTFE filters. Results As described above, particle data were collected for monitor wafers and the SC-1 chemistry using different methods. Two particle monitors were processed with each run, and an in situ particle monitor analyzed fluid samples taken every 60 seconds. The collected data were averaged, yielding the results shown in Table I. The specification limit for this process is a delta of 50 at 0.2-µm adders. The data in Table I indicate that there was no significant variance in the performance of the two filter materials. The particle data also revealed that there was no significant variance from run to run. Figure 3 shows the day-to-day trends for particle tests performed using bare-silicon monitors, and Figure 4 shows the day-to-day trends for the in situ particle monitor.
Figure 3: Average particle delta per sample run for the two filter materials.
Figure 4: Average particles in solution per sample run for the two filter materials.
Table II: Polysulfone membrane filter TXRF data. The TXRF data collected during the 5-month study with the polysulfone membrane filters installed are given in Table II, while Table III presents the baseline data collected with PTFE membrane filters installed over 15 months of standard product operation. The metal specification limit for this process is 1010 atoms/cm2 for all elements listed in these tables. The TXRF measurements were taken at three locations on the monitor wafers: off center (20:0); at the upper edge below notch (0:80); and at lower left (40:40).
Table III: PTFE membrane filter TXRF data. With regard to the TXRF data, it should be noted that heavy-metal impurities (iron, zinc, and aluminum) are known to be a problem in alkaline processes such as an SC-1 clean even when the chemicals meet stringent purity specifications.4,5 As the temperature of the solution increases, so does the metallic contamination on the wafer surface. The mechanisms for increased deposition of iron at higher temperatures are not clearly understood, although it is possible that the increase in temperature may shift the reaction rate constants such that dissolution onto the silicon wafer surface is promoted.6 Table IV presents ICP-MS data for both the PTFE and polysulfone membrane filters. Columns dated 10/31 through 11/15 were generated with the PTFE membrane filters installed, and those dated between 11/21 and 5/25 were generated with the polysulfone membrane filters. The process specification limit for the combined metallic levels in the SC-1 solution is 50 ppb. These data show that the polysulfone membrane filters are stable for metallic contaminants.
Table IV: ICP-MS data in parts per billion. Columns dated 10/31, 11/7, and 11/15 contain results for the PTFE filters; the remaining columns contain results for the polysulfone filters. The cache probe data from 24 product lots (12 lots per filter type) were compared using JMP Version 2 analysis software (SAS Institute, Cary, NC) to determine the statistical variance between the two filters. Comparable circuit yields were observed for all of the lots, and no statistical difference in cache probe results was observed between the filter types. Similar results were observed for the unit probe data. (These product data are confidential.) Conclusion Polysulfone membrane filters evaluated in an SC-1 process module performed comparably to conventional PTFE membrane filters. Thus, this study demonstrated that either filter is acceptable. Using the polysulfone membrane filters in SC-1 solutions could yield several benefits. Because the material has better flow characteristics than PTFE, less stress will be placed on system pumps, minimizing pump failures. In addition, no prewetting is needed with polysulfone filters, and they cost approximately 60% less than PTFE filters. The total decrease in cost of ownership will be even greater than this unit replacement savings, since it will include the cost savings associated with the other two benefits as well. Future research will address the particle capture and recovery rates of the polysulfone filters and their applicability in other process steps, such as BOE and HF baths. Acknowledgments The authors wish to recognize the contributions of Lisa Clark, Jeanne Rannells, and Evelyn Ferrero, who assisted with ICP-MS chemical analyses, and Susan Backer, who assisted with the TXRF measurements. They also wish to thank Jeff Cope and Brad Smith for participating in several discussions on how to evaluate the product data; the members of the APRDL pilot line; and John Alvis, Roc Blumenthal, and Lou Parrillo, who provided managerial support. References 1. Shucosky AC, "Filtrative Enhancements Due to Graded Microporous Membrane Structure," TAPPI Journal, 79(11):203206, 1996. 2. Weddington D, and Pena R, "Comparison of Teflon and Quartz as a Material for Wafer Carriers in an Automated Acid Bench," Internal Report 97-22, Austin, TX, Motorola APRDL, 1997. 3. Mitchell JR, and Knollenberg BA, "New Techniques Move In Situ Particle Monitoring Closer to the Wafer," Semiconductor International, 19(10):145154, 1996. 4. Berry M, Depinto G, and Steinberg L, "A Methodology for Continuous Defect Reduction in a High Volume Sub-Micron CMOS Factory," in Electrochemical Society Manual, Schmidt DN (ed), PV92-21, Pennington, NJ, Electrochemical Society, pp 208222, 1992. 5. Helms CR, and Park H, "Electrochemical Equilibrium of Fe in Acid/Base/Peroxide Solutions Related to Si Wafer Cleaning," in Electrochemical Society Manual, Ruzyullo J, and Novak R (eds), PV94-7, Pennington, NJ, Electrochemical Society, 1992. 6. Hall M, Rosato J, Jarvis T, et al., "Effect of SC-1 Process Parameters on Particle Removal and Surface Metallic Contamination," presented to the SCP Cleans Symposium, Boise, ID, April 1997. Darryl Weddington has been with Motorola for five years, the past three as a cleans process technician in the Advanced Products Research and Development Laboratory (APRDL) in Austin, TX. In this position, he is involved in seeking, developing, and maintaining wet clean processes for 0.25-µm logic technologies. Weddington attended State Technician Institute in Memphis and St. Edwards University. He has coauthored more than 10 research papers in the area of diffusion and wet etch processes. (Weddington can be reached at 512/933-7527.) Todd Ridley is a maintenance technician for cleans tools at APRDL, where he is responsible for all equipment maintenance issues. He has been with Motorola 14 years. Ridley holds an OJT degree in diffusion equipment maintenance and is currently working on a cleans process and equipment maintenance technician degree. Vincent Ladigo is a senior process and equipment technician in Motorola's APRDL, where he is involved in maintaining existing tools and processes as well as developing and evaluating new tools and processes that align with the goals of APRDL's device development groups. Prior to joining Motorola six years ago, Ladigo was at Texas Instruments for nine years. He has an associate degree in electronic engineering from ITT Technical Institute in Indianapolis. James Campbell is a sales and marketing manager for the Memtec Electronics Group of Memtec/Filterite (Timonium, MD), responsible for providing applications support and customer assistance to the semiconductor industry. He has more than 20 years of experience with both equipment and materials suppliers in wafer fab applications with a concentration in fluid processing and filtration. Campbell has a degree in business administration from the University of Texas at Arlington. (Campbell can be reached at 972/271-5032 or jpcampbell@memtec.com) Mark Gogol is the engineering/marketing manager for Memtec Electronics. Associated with Memtec applications engineering for six years, he has been active in the semiconductor manufacturing industry for more than 10 years. Gogol holds a BS in engineering and physics from West Virginia Wesleyan College.
MicroHome |
Search | Current Issue | MicroArchives Questions/comments about MICRO Magazine? E-mail us at cheynman@gmail.com. © 2007 Tom Cheyney |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
