Defect reduction using POU filtration in a new coater/developer
The fundamentals of microbridge defect reduction are not yet completely understood. Previously, it was concluded that microbridge precursors adsorb onto polar moieties within Nylon 6,6, by comparing defect density among polar filters (Nylon 6,6) and non-polar filters (HDPE and PTFE) . The study presented here works toward a greater understanding of the adsorptive filtration-defect reduction relationship by additionally considering the effects of resist solvent polarity. The study also demonstrates how defectivity improvements via enhanced point-of-use (POU) resist filtration can be reproduced on next-generation process equipment .
Toru Umeda, Shuichi Tsuzuki, Toru Numaguchi, Pall Corp. Inashiki-gun, Ibaraki-ken, Japan
The impact of pore size and membrane material polarity on the effectiveness of point-of-use (POU) filtration has been evaluated. Decreased pore size and increased polarity in membrane materials were confirmed to positively influence the effectiveness of microbridge defect removal by a POU filter in the LITHIUS Pro coater/developer system. Comparative analysis of different solvent systems validates a model of competitive adsorption whereby more-hydrophilic solvents and gel-like agglomerates preferentially interact with the Nylon 6,6 membrane surface. This suggests that adsorption is the dominant mechanism for microbridge defect removal via filtration. Therefore, utilizing filtration products built around polar membrane materials (like hydrophilic Nylon 6,6) will result in greater microbridge defect reduction than solely reducing filter pore size.
As lithographic pattern CDs continue to shrink, so does the tolerance for the size of photoresist defects, such as agglomerated microbridge precursors . Microbridge defects are particularly observed in acrylate-based 193nm photoresists . A Tokyo Electron CLEAN TRACK ACT 8 coater/developer system with 200mm wafers was used to demonstrate that POU filtration is effective to reduce microbridge defects, where pore size, membrane polarity, and surface modification were proven to be dominant factors [4,5]. To meet manufacturing requirements, Tokyo Electron developed a CLEAN TRACK LITHIUS Pro coater/developer system that can process 300mm wafers. In this study, the impacts of pore size and membrane material polarity on the effectiveness of POU filtration are evaluated.
The fundamentals of microbridge defect reduction are not yet completely understood. Previously, it was concluded that microbridge precursors adsorb onto polar moieties within Nylon 6,6, by comparing defect density among polar filters (Nylon 6,6) and non-polar filters (HDPE and PTFE) . The present study works toward a greater understanding of the adsorptive filtration-defect reduction relationship by additionally considering the effects of resist solvent polarity.
Effects of membrane polarity and pore size at POU. A standard 193nm resist was dispensed through various point-of-use filters, and was spin-coated onto silicon (Si) wafers using the coater/developer. To study the effect of membrane polarity, Nylon 6,6 and HDPE were used as polar and non-polar filters, respectively. The finest (10nm for Nylon 6,6 and HDPE) and next-finest (20nm for Nylon 6,6, 30nm for HDPE) available removal ratings that were commercially available for each filter type were used in a standard patterning process.
A 90nm half-pitch line/space pattern was printed using a 193nm (dry) exposure tool. After pattern development, defects were quantified using a KLA-Tencor 2360 inspection system. Microbridge defects were then identified using a scanning electron microscope (SEM).
Filtration of microbridge defect precursors in Fe-spiked resist. Microbridge defects do not have a regular morphology, but are likely composed of a gel-like substance. Metal contamination in photoresist increases defectivity by cross-linking acrylate polymer molecules and forming low-molecular-weight gel aggregates . We attempted to create gel aggregates by spiking iron (100ppb, as Fe3+, in diluted nitric acid) into a photoresist polymer + solvent (propylene glycol monomethyl ether acetate [PGMEA]/Ethyl lactate mixture) solution, which was then aged up to 48 hours. Spiked resist of various aging terms were used to challenge 20nm rated Nylon 6,6 filters. Fe concentrations in both the influent and the effluent were measured using inductively coupled plasma-mass spectrometry (ICP-MS). The amount of Fe removed is assumed to correlate to the removal efficiency of gel-like microbridge precursors.
Filtration of microbridge precursors within solvents of varying polarity. Three different solvent systems were evaluated: PGMEA/Ethyl lactate, PGMEA/PGMEA, and PGMEA/Cyclohexanone (each in a mixture). For nuclei metals, we spiked magnesium (Mg), nickel (Ni), and cadmium (Cd) (in diluted nitric acid) into each polymer + solvent system to follow a result of a study that these metals form remarkably greater concentrations of aggregate gels than Fe. The concentration of each spiking metal was 10ppb.
As with Fe-spiked resist, prepared solutions were used to challenge 20nm rated Nylon 6,6 filters. Metal concentrations in both the influent and the effluent were measured using ICP-MS. Again, the amount of removed metal amount is assumed to correlate to the concentration of gel-like microbridge precursors.
Results and discussion
Effects of membrane polarity and pore size. Figure 1 (pg. 19) shows the results of POU filtration in the coater/developer on microbridge defect density. The left y-axis corresponds to microbridge defect density (bar graphs). The right y-axis, "Removal efficiency," is the ratio of microbridge defect density values "with filter" to "without filter."
|Figure 1. Microbridge defect reduction in TEL Lithius Pro using various point-of-use filters.|
In the results, Nylon 6,6 filters showed better microbridge defect removal efficiency (ranging from 98.7 to 99.6%) compared to HDPE filters (77.7 to 91.6% reduction). In HDPE filters, microbridge defect removal efficiency was improved from 77.7 to 91.6% by reducing pore size from 30 to 10nm. In Nylon 6,6 filters, the measured defect density was too small to discern any difference between 20 and 10nm pores.
Figure 2. Fe removal efficiency of Nylon 6,6, filter vs. aging term of spiked resist.
Results showing enhanced microbridge defect removal via decreased pore size and increased membrane polarity in POU filters have been reproduced within a next-generation coater/developer system.
Microbridge precursor reduction
Filtration of microbridge defect precursors in Fe-spiked resist. Figure 2 shows the iron removal efficiency of Nylon 6,6 filters in Fe-spiked resist that is aged to various extents. The removal efficiency of Nylon 6,6 filters appeared to increase with aging time. An interpretation of this result is illustrated in Fig. 3. Fe is introduced (spiked) into the resist solution in an ionic state. As time elapses, Fe ions serve as nucleation sites and begin to coordinate with aggregates of cross-linked acrylate gels. With additional aging, a greater amount of Fe ions becomes complexed with gel defect precursors. Given the demonstrated ability of polar Nylon 6,6, membrane to remove aggregated gel contaminants, it follows that a greater amount of Fe will be removed with resist aging time. Even after 48 hours of aging, however, the Nylon 6,6 filter does not completely remove all of the Fe. This is because all of the Fe has not complexed with aggregate gels.
|Figure 3. Interpretation model of the results shown in Fig. 2.|
Filtration of microbridge precursors within solvents of varying polarity. It is known that solvent molecules compete with solute molecules for available adsorption sites on a solid surface . The interpretation of the aging result suggested that the gel-like aggregates were polar in nature, and thus, hydrophilic. To study this hypothesis, the effect of solvent polarity (hydrophilicity) within a spiked resist on metal removal efficiency was evaluated.
|Figure 4. Water solubility in the solvent vs. metal removal efficiency in Nylon 6,6 filtration.|
Filtration results are given in Fig. 4. An inverse relationship is observed between metal removal efficiency and solvent hydrophilicity. In correlation with earlier results on Fe-spiked resist, a schematic model used to interpret this observation is given in Fig. 5. In hydrophobic solvent systems, a greater number of adsorption sites on the Nylon 6,6 surface will be available to retain gel aggregates. Conversely, if the solvent is hydrophilic, the polar Nylon 6,6 adsorption sites are partly covered by solvent molecules and may have some effect on gel adsorption to the membrane surface. This suggests that competitive adsorption between gel aggregates and more-hydrophilic solvents can affect microbridge defect removal filtration. These results further validate surface adsorption as the dominant mechanism for microbridge defect removal via filtration.
|Figure 5. Schematic of competitive adsorption between hydrophilic aggregated gel and solvent molecules.|
Decreased pore size and increased polarity in membrane materials were confirmed to positively influence the effectiveness of microbridge defect removal by a POU filter in the LITHIUS Pro coater/developer system. Also, comparative analysis of different solvent systems supports a model of competitive adsorption onto the Nylon 6,6 membrane surface between more-hydrophilic solvents and gel-like agglomerates. Results further validate membrane surface adsorption via polar (electrostatic) interactions as the dominant mechanism for microbridge defect removal via filtration. With an increased understanding of contaminant-surface interactions, it is clear that filtration products built around hydrophilic Nylon 6,6 membranes will continue to be a strategic component within next-generation litho process equipment to drive enhanced fluid cleanliness, defect reduction, and increased product yield.
The authors thank Tokyo Electron Kyushu Ltd. for providing POU filter evaluation data on microbridge defect reduction in the TEL Lithius Pro coater/developer system. The authors also thank JSR Corporation for providing test solution used in the investigation of the fundamentals of microbridge reduction. CLEAN TRACK ACT and LITHIUS Pro are trademarks of Tokyo Electron Limited.
1. D. Hall et al., "The Effectiveness of sub 50nm Filtration on Reduced Defectivity in Advanced Lithography Applications," ARCH Interface, Sept. (2003).
2. T. Umeda, S. Tsuzuki, T. Numaguchi, "Defect Reduction by Using Point-of-Use Filtration in a New Coater/Developer," in Advances in Resist Materials and Processing Technology XXVI, edited by Clifford L. Henderson, Proceedings of SPIE Vol. 7273 (SPIE, Bellingham, WA, 2009), 72734B.
3. International Technology Roadmap for Semiconductors, 2007 edition, http://www.itrs.net.
4. T. Umeda et al., "Research of Appropriate Filter Membrane for Reducing Defects of ArF Lithography," FUJIFILM Interface, Oct. (2005) San Diego.
5. T. Umeda et al., "Study on Effective Property of Point of Use Filter for Defectivity Reduction in 75nm ArF Lithography and 120nm KrF Lithography," FUJIFILM Interface, Oct. (2006) San Diego.
6. Y. Nagata et al., Gel Handbook, NTS Inc., pp.13 (1997).
7. T. Keii, Adsorption, Kyoritsu Shuppan, pp.148 (1983).
Toru Umeda received his BE in chemistry from Yamagata U., Japan, and is a senior staff scientist in the Scientific Laboratory Services Division at Pall Corp, Nihon Pall Ltd., 46 Kasuminosato, Ami-machi, Inashiki-gun, Ibaraki-ken, Japan 300-0315; ph. +81 29 889-1951; fax 81 29 889-1957; email@example.com; www.pall.com.
Shuichi Tsuzuki received his BS in chemistry from Tokyo U. of Science, Japan, and is ME group manager in the Scientific Laboratory Services Division at Pall Corp, Nihon Pall Ltd.
Toru Numaguchi received his PhD in catalytic science from Kyoto U., Japan, and is director of PI technology-Japan at Pall Corp, Nihon Pall Ltd.