The effect of sub-0.1µm filtration on 248nm photoresist performance
07/01/2000
Barry Gotlinsky, Michael Mesawich, Pall Corp., Port Washington, New York
James Beach, International Sematech, Austin, Texas
OVERVIEW
The implementation of resist filtration below 0.1µm within existing dispense systems raises concern as the removal rating of the filter approaches the size of large molecular weight components of the photoresist. A comprehensive study has shown, however, that as the filter removal rating became finer, the resist performance in terms of photospeed, process window, and thermal stability did not change. This indicates that, using existing dispense systems, photoresists can be filtered to as fine as 0.03mm without significant polymer shearing or the unintentional removal of important materials from the resist. Based on these data, appropriate protection in terms of particle removal is possible as linewidths necessitate the use of finer filters in resist dispense pumps.
Each step in wafer processing introduces the possibility of deleterious particulate contamination, microbubble void defects, and metallic contamination onto the wafer surface. For photoresists, the removal of particles larger than half the smallest feature size is imperative to prevent circuit failure.
 |
Photoresists designed for use at 248nm-wavelength exposure operate by a mechanism of chemical amplification [1] that can be extremely sensitive to external influences. Complicating matters further, these photoresists contain different resin systems and operate by an entirely different mechanism than traditional positive i-line and g-line photoresists. Studies performed on previous-generation photoresists are therefore not applicable to deep ultraviolet (DUV) photoresists. Accordingly, we set out to explore the ability to remove deleterious particles at and below the feature size without compromising photoresist performance.
Current filtration is typically 0.1µm. With the need for finer filtration, the major concern is whether sub-0.1µm-rated filters could have deleterious effects on resist performance. In addition, the role of membrane materials themselves and what effect they could have on photoresist constituents is of concern. In particular, we wanted to know if filtration at these levels would reduce process latitude and/or thermal stability, or completely render the photoresist useless by removing critical components. The selection of newer materials for optimizing photoresist dispense processing complicates this situation further.
The 1999 International Technology Roadmap for Semiconductors cites the removal of particles <0.1µm as critical for features at £180nm [2]. Reduction of feature sizes to <180nm has led to the offering of tighter membranes to ensure the removal of yield-reducing particles.
Delivery mechanism
The method of delivering photoresist to the wafer surface is best accomplished by a precision dispense system. Point-of-use (POU) filtration is an integral part of any dispense system to reduce particles that might be suspended in the fluid. The pump, however, cannot be considered a passive component in the proper functioning of the filter.
To determine if the photochemical delivery method has any effect on filter performance, we used two distinct types of delivery systems:
- Nitrogen (N2) pressurization. A nitrogen-based pump applies N2 pressure directly on the resist surface in the pump body; the pressure pushes the photoresist through the filter into the dispense delivery tube. Specific delivery pressures and rates are established by adjusting N2 pressure.
- Motor-based pressurization. A stepper motor elongates a diaphragm, with the displacement rate and amount determining the delivery pressure and rate. The resist passes through the filter and into the delivery tube.
Both pump systems are typical on advanced track systems. With both systems, we controlled delivery pressure and dispense rate to match dispense conditions for all the filters tested, for both types of pumps.
Filter material
 |
The variety of microlithographic chemicals allows the selection of the appropriate membrane material and micron ratings to best accomplish the removal of particles. For this study, we used membranes constructed of polytetrafluoroethylene (PTFE), high-density polyethylene (HDPE), and Nylon 6,6, (N66); each has a different surface characteristic. PTFE and HDPE are hydrophobic. N66 is hydrophilic. We selected membranes for POU filtration rated from 0.1µm to 0.03mm. In all three cases, the removal of particles occurs throughout the depth of the membrane.
 Figure 1. CDs as dispensed with IDI Model 450 pump: a) CD vs. focus at 29mJ; and b) CD vs. focus at 31mJ. |
The effects of the polymer makeup and the morphology of the pores of each membrane have not been carefully studied in POU photoresist filtration. Literature cites concerns of fractionating resins and removal of high-molecular-weight components when going to finer filters [3]. One additional concern is that the push toward finer filtration will increase the operating pressure required to dispense photoresists. Although all the filters selected for testing use a high-surface-area pleated design to minimize differential pressure, any increased pressure could damage sensitive DUV photoresist or cause microbubbles. Due to this concern, we monitored differential pressure to correlate any observed effects.
Testing methodology
We set up each dispense pump to work on 200mm wafers in a track system physically linked to a DUV scanner. Because the filters and pumps were being changed during the test, the pumps were external to the track and the dispense lines being hand-held above the wafers. We timed the dispense so that the puddle was formed immediately prior to the initiation of spinning. A standard recipe was followed in the track, including application of bottom ARC, temperature stabilization, and soft bake after dispense. The coated 200mm wafers were exposed at various exposure doses and focuses on the scanner to observe any change in the process window of the 180nm dense features. The wafers were also analyzed for changes in thermal stability and resist profiles.
 Figure 2. CDs when resist was dispensed with the Cybor stepper motor pump. |
We programmed the N2 pump via a PC interface to yield a dispense of 5.1g in 2.5 sec. The N2 pressure was adjusted by the pump controller to maintain these parameters; it also adjusted for backpressure from differential pressure across the filter. The last step in the program was set to eliminate resist suck-back effects.
The stepper-motor pump uses a dedicated controller. We set this positive displacement pump to dispense 5.1g in 2.5 sec.
 |
In all cases, we did not change filter hardware, but filter removal ratings and materials were varied, from 0.1-0.03µm, and from PTFE to N66 to HDPE, respectively. Upon installation of each filter, the pump was purged to ensure sufficient wetting of the filter and minimization of microbubbles. Using Shipley UV6 0.6 — a DUV photoresist — we kept dispense rates, dispense volumes, and wafer processing constant throughout testing.
 |
Some wafers were tested for thermal stability by subjecting them to 155°C for 3 min. The photoresist used reportedly had a thermal stability of 150°C. Cross-section scanning electron micrograph images (SEMs) were then taken after the thermal exposure for comparison.
The dispensed photoresist was also analyzed by gel permeation chromatography (GPC) to determine if any molecular weight distribution changes occurred after filtration and dispense. This was done by collecting photoresist in light-protected vials from the dispense tube.
Results
Three wafers for each pump-membrane set were coated with 6000Å of photoresist, as summarized in Table 1. The first round of tests with the N2 pump compared a standard 0.1µm PTFE filter to a finer 0.05µm version and to equivalent rated N66. A second round of tests, using the motor pump system, carried this sequence further by including an HDPE filter with a membrane rated at 0.03µm.
 Figure 3. Top-down SEMs from a wafer coated with photoresist dispensed through a) a 0.05mm PTFE membrane and b) a 0.04mm N66 membrane. |
Analysis of the data revealed no changes in DUV resist performance when switching among PTFE, N66, and HDPE filters. For the wafers coated using the N2 pump, at dosage levels of 29 and 31MJ, the critical dimensions (CDs) were within experimental error for all the filters tested. Therefore, when going from PTFE to N66 membranes, and using filters finer than the standard 0.1µm filter, we did not see a performance shift. Figures 1a and 1b give CDs and depth of focus (DOF) for the N2 pump.
Similarly, for the motor pump, filter materials and ratings did not affect the measured performance of the photoresist. For example, at a dose of 29mJ, filters using PTFE, N66, and HDPE, with ratings from 0.1mm to 0.03µm, all yielded CDs of 170±5nm, with a DOF of 0.6-0.8µm (Fig. 2 and Table 2).
SEMs revealed sharp features (Fig. 3). Our examination of SEM cross sections before and after thermal treatment did not reveal any significant differences in the profiles of the finely filtered resists (Fig. 4). Therefore, these materials did not have any observable deleterious effects on the resist itself.
We also used GPC to analyze photoresist collected from the pump dispense. This was done using 0.1µm PTFE, 0.05µm PTFE, 0.04µm N66, and 0.03µm HDPE membranes to determine that no significant polymer shearing or unintentional removal of important materials from the resist occurred.
The GPC data revealed that the average molecular weight and weight fraction distribution did not change after being dispensed through the filters (Table 3). Based on this data, we concluded that dispensing photoresist through finer filters did not have an effect on the polymer chain length through shear, removal, or other mechanisms. The filter membrane materials also did not have an effect upon the photoresist. The lack of change in properties of the photoresist itself is in accordance with the performance data obtained.
Conclusion
 Figure 4. SEM cross sections of resist filtered with a 0.04mum N66 membrane a) before and b) after thermal stability testing. |
Filtering a DUV photoresist at that sub-0.1µm level did not significantly affect resist performance in terms of photospeed, process window, or thermal stability. Our study has indicated that, using existing dispense systems, photoresists can be filtered as fine as 0.03µm without polymer shearing or the unintentional removal of important materials from the resist. Based on these data, appropriate protection in terms of particle removal is possible as linewidths necessitate the use of finer filters in resist dispense pumps.
Acknowledgments
The authors thank Ken Reichman and David Gonzales of Cybor Corp. and Jay Foersterling of Integrated Designs L.P. for the use of the dispense pumps in this study. Nylon N66 is a registered trademark of Pall Corp. Shipley and UV6 are registered trademarks owned by Shipley Company L.L.C., Marlborough, MA.
References
- L.F. Thompson, C.G. Willson, M.J. Bowden, Introduction to Microlithography, American Chemical Society, pp. 212-232, 1994.
- The International Technology Roadmap for Semiconductors, Semiconductor Industry Association, pp. 272-273, 1999.
- W. Conley, R. Allen, R. Kunz, "Deep UV Resist Technology," Future Fab International, 4, 1, pp. 145-149, 1998.
Barry Gotlinsky received his PhD in chemistry from the City University of New York. He has more than 15 years of experience in technical support of the microelectronics market. Gotlinsky is VP of microelectronics support in the scientific and laboratory services department of Pall Corp., Port Washington, NY 11050; ph 516/484-3600, fax 516/484-3628, e-mail [email protected].
Michael Mesawich received his BS in physics from the State University of New York at Stony Brook. He has more than seven years of experience in market and product development for microlithography applications. Mesawich is currently senior marketing manager of the Pall Microelectronics Group, East Hills, NY.
James Beach has a PhD in polymer chemistry and has worked in resist processing at International Sematech ATDF for the past four years.
Robust Pump Systems
 |
Pressure-on-demand and direct-drive pump systems have provided a reliable, repeatable, and extremely accurate method of delivering a precision volume of photochemicals in lithography. These systems have proven extremely robust in large-scale semiconductor manufacturing environments for many years. It is clear that most pump systems can be properly configured to meet the demands of a DUV 248nm coater system and this is an important step in reducing defects to meet the SIA Roadmap at linewidths below 100nm. Ultimately, however, the best way to minimize the potential for particle generation is by placing filtration downstream of the pump, as this study demonstrates. Significantly, these researchers have found that important components of the photoresist were not damaged or removed when filtered as fine as 0.04mm and 0.03mm. This is contrary to the perception that a point-of-use filtration model causes chemical stripping.
Jay Foersterling, president, Integrated Designs, L.P., Carrollton, Texas