Chemically Amplified resists for advanced lithography: Road to success or Detour?
06/01/1997
Chemically amplified resists for advanced lithography: Road to success or detour?
David Seeger, IBM Research Division, Yorktown, New York
Chemically amplified (CA) resists play an important role in the semiconductor industry, primarily for 248-nm lithographic applications. Great advances have been made in the development of these materials, though some issues remain. This paper explores the development of CA resists and demonstrates some of their limitations for more advanced applications. The CD budget crunch may limit the utility of CA resists. Recent efforts to design resists in fundamentally different ways face daunting challenges.
The bulk of all new photoresists are CA. In these resists, the lithographically important chemistry is not driven directly by photons absorbed in the exposure step. Instead, an acid formed during exposure of a photo-acid generator (PAG) catalyzes chemical changes in the resist in a bake step immediately following exposure (Fig. 1). Since the reaction is catalytic, the acid is regenerated after each chemical reaction; the same acid molecule can participate in further reactions. Many chemical reactions result from a single photon-absorption event, leading to very fast resist materials.
This technology was pioneered by IBM [1] in the 1980s in order to accommodate early deep-ultraviolet light sources. At the time, the choices were mercury arc lamps (like those found in the widely used Perkin Elmer Micralign systems) or a then relatively new source, excimer lasers. While lasers held the promise of large flux, at the time they were in a very early development phase and there was concern over their ability to operate reliably. Mercury arc lamps, already proven in manufacturing, were deemed a more reasonable approach. These lamps required very fast resists, leading to the development of CA resists.
Nearly all new photoresists under development today are CA resists. All positive and negative tone 248-nm resists are chemically amplified, as are new single-layer resists for 193-nm applications [2-4]. CA resists are even showing up in some i-line applications [5] where exposure speed is critical.
 |
Figure 1. Description of how a chemically amplified resist works, where `PAG` represents a photo-acid generator and `insol` and `sol` represent fragments of the polymer that are insoluble and soluble, respectively, in developer.
To understand the prevalence of CA systems, one merely needs to look at the problem from a chemical perspective. The number of photochemical reactions in organic chemistry is small compared to the number of known acid and base catalyzed chemistries. Catalysts have been widely used in synthesis reactions using "protecting groups." Selective incorporation of protecting groups can shield a functionality from chemical attack while the chemist is busy working to build up a remote part of the molecule. Later, mild catalytic reactions remove the protecting groups to regenerate the original organic function [6].
In photoresists, protecting groups mask acidic functionalities on the resist polymer. When the groups are catalytically removed, the unmasked acidic moiety drastically changes the solubility of the resist polymer in (basic) developer (Fig. 1). Efficiently creating an acid or base for the catalytic de-protection reaction, in only the exposed parts of the resist, was a key challenge for this approach. Photo-acid generators served this function well and now they are commonplace [7-8]. Since merely changing a photo-acid generator or protecting group produces a "new" resist system, we have seen an unprecedented explosion of new resists.
Advantages and issues with CA resists
The biggest advantage of CA systems is speed. These materials also show very high contrast and typically have very high resolution, with gamma [9] values of 5-10, vs. 2-3 for novolak resists (Fig. 2). The catalytic nature of the CA process makes reproducibility an important concern. Should the catalytic chain be interrupted unexpectedly, many lithographically important reactions will not take place, leading to catastrophic resist failure. The most common interruption occurs when an airborne base penetrates into the photoresist and quenches the photo-generated acid. The resulting severe "skins" on the surface of the resist prevent clean development. Use of a protective top coat as a barrier [10] and careful filtration of the air to prevent airborne contamination [11] can overcome this problem. Other solutions include additives to the resist formulation [12] and reducing the rate of absorption of contaminants by "densifying" the resist through annealing [13-14]. Contamination from the substrate can also take place, particularly if previous deposition steps required a resist-poisoning gas. Some solutions to airborne contamination problems have shown promise in minimizing substrate contamination.
 |
Figure 2. High-resolution imaging capability of APEX (CA system) with x-ray exposures
The kinetics of the catalytic reaction are also important. Careful control of temperature uniformity and bake time must be maintained. Any nonuniformities will change the critical dimension (CD) on the wafer. The change in CD due to variations in bake temperature depends on the temperature uniformity of the bake plate and the bake latitude of the resist. Figure 3 estimates line width control for different hypothetical resist systems assuming different baking uniformities. The post-exposure bake latitudes (5-20nm/?C) are representative of commercially available resist systems. Even the best of these systems will require hot plate uniformity on the order of 0.1?C in order to meet the desired line width control for 0.1-?m lithography. Since this control is beyond current specifications for hot plate uniformities, more advanced lithographies will require improvements to either baking or resist technology. This behavior is a resist limitation and is independent of the exposure technology.
Finally, CA resists are becoming more and more complex. Early resists (e.g. PMMA, PBS) were very simple homopolymer single-component systems; i-line novolak resists typically are composed of a polymer resin and a photo-active compound. Deep ultraviolet resists currently in use consist of four major components: copolymer resin (counting as two components because control of the copolymer ratio is critical for consistent resist performance), photo-acid generator, and an additive for environmental stability. Some proposed 193-nm resists have as many as six major components [15]. Clearly, such systems will increase cost and complicate resist manufacturing.
 |
Figure 3. Dependence of linewidth on hotplate uniformity and resist bake latitude.
Alternatives to CA resists
While there may be viable solutions to these issues, other resist technologies could be considered. In CA resist processing, more stable processes generally use lower post-exposure bake conditions, thereby minimizing the importance of the amplification step [16]. Nonamplified resist chemistries may therefore offer an alternate solution. Nonamplified resists have many advantages, including simplicity, stability with respect to environmental contamination, insensitivity to baking conditions, etc. The high exposure doses that are typically required are a major drawback. A nonamplified resist chemistry that combines these advantages with a photo-speed that approaches amplified resists is desired.
 |
Figure 4. Modification of diazoquinone structure to allow exposures at 248-nm: a) conventional diazoquinone, b) modified diazoquinone, c) UV absorption of modified diazoquinone, and d) UV absorption of conventional diazoquinone.
 |
Figure 5. Contrast of high-sensitivity non-CA resist with broadband deep-UV exposures compared to PMMA.
Several years ago, IBM researchers reported an effort to improve the sensitivity of diazoquinone novolak resists for x-ray lithography [17-18]. X-ray resists typically absorb only 10% of the incident radiation /?m of resist thickness. Incorporating highly absorbing moieties may increase absorption, and thus improve sensitivity. In this study, 2? improvements in sensitivity were achieved by doping the resin with highly absorbing halogen atoms.
Furthermore, placing these highly absorbing moieties at a location close to where the chemistry will happen will increase the efficiency of the reaction. Incorporating halogen atoms close to the diazoquinone functionality improved the conversion efficiency of the diazoquinone during exposure. Speed roughly doubled with only a 10% increase in bulk absorption [17-18]. Comparison of this new resist, ESX (enhanced sensitivity x-ray resist), with conventional novolak resist showed that process latitude improved by a factor of two at the same time. Thus there is nothing fundamental about the perceived trade-off between sensitivity and process latitude. One can improve both at the same time.
Diazoquinone-type resists for 248-nm applications have recently been developed [19]. At this wavelength, conventional novolak resists absorb so much radiation that light never penetrates the full thickness of the resist. Modifying the photoactive compound (PAC) structure from diazonaphthoquinone (Fig. 4a) to diazo-4-hydroxy coumarin (Fig. 4b) decreased the absorption; the PAC efficiently bleaches during exposure at 248 nm (Figs. 4c, d). The exposing light can reach throughout the thickness of the resist, allowing high-resolution lithography. This technology must next be combined with a resin that is also less opaque to the exposing radiation.
 |
Figure 6. High-resolution imaging of non-CA resist by x-ray exposure.
In a third example, MIT has been using non-CA top surface imaging resists for 193-nm lithography [20]. In this process, pure resist resin materials crosslink at the surface upon exposure. Then, exposure to a silylating compound selectively silylates unexposed areas. Dry development by reactive ion etching then uses the silicon oxide as an etch mask. This process has been used with high-resolution imaging, but low sensitivity has been a problem. Some work has shown promise in improving speed [21]. More research is needed.
IBM is investigating a different chemical approach than the novolaks described above [22]. This new resist, called SPAR, is also non-CA and is aqueous base developable. The average contrast (gamma) (Fig. 5) is >6, rivaling CA resists. The material shows the stability with respect to changes in processing conditions that would be expected from a non-CA system. Environmental stability with respect to amine contaminants is also demonstrated since these materials are cast from n-methyl pyrrolidone (a known contaminant for CA resists) solvent. The resist is capable of high-resolution imaging (Fig. 6). We have demonstrated resolution to 0.125 ?m (limited by the mask) using x-ray exposures. The truly surprising attribute of this system is that the speed is similar to CA systems, as shown in the table.
Conclusion
The importance of CA resists in 248-nm lithography is undeniable. These materials can meet aggressive 0.25-?m ground rules with impressive CD control, once the proper manufacturing controls are in place (e.g. charcoal filtration, integrated photosectors, high precision hot plates etc). Notable issues associated with CA resist systems include resist complexity, resist poisoning by contaminants from the air or substrate, and sensitivity to post-exposure bake conditions.
Though impressive progress has been made in addressing these issues, alternatives to chemical amplification may also merit investigation. This paper highlights several recent studies of non-amplified systems. Process stability and lithographic performance can be achieved with these systems; however, the most daunting challenge is to maintain these properties and match the speed of CA systems.
References
1. H. Ito, C.G. Willson, "Polymers in Electronics," T. Davidson, ed., Symposium Series 242; American Chemical Society; Washington, D.C., p. 11, 1984.
2. Wallraff, et al., "Single layer chemically amplified photoresists for 193-nm lithography," J. Vac. Sci. Technol., B11(6), 2783, 1993.
3. R. Allen, et al., "Single layer resists with enhanced etch resistance for 193nm lithography," J. Photopolym. Sci. Technol., 7, 507, 1994.
4. S. Takechi, et al., "Alicyclic polymer for ArF and KrF excimer resist based on chemically amplification," J. Photopolym. Sci., Technol., 3, 439, 1992.
5. L. Linehan, et al., U.S. Patent No. 5296332, 1994.
6. "Protective Groups in Organic Chemistry," J. McOmic Ed; Plenum Press, NY, 1973.
7. H. Ito, "Deep UV Resist Systems, Radiation Curing in Polymer Science and Technology," J. Fouassier and J. Rabek, eds., Elsevier, London, Chapter 11 Vol 4, 1993.
8. G. Pawlowski, et al., "Novel Photacid Generators: Key Components for the Progress of Chemically Amplified Photoresist Systems," J. Photopolym. Sci. Technol. 4, pp. 389-402, 1991.
9. L. Tompson, M. Bowden, "Introduction to Microlithography," ACS symposium series 219, p. 169, 1983.
10. T. Kumada, et al., "Study on the Over-Top Coating Suppressing Surface Insoluble Layer Generation for Chemical Amplification Resist," Proc. SPIE, 1925, pp. 31-2, 1993.
11. S. McDonald, et al., "Airborne Chemical Contamination of a Chemically Amplified Resist," Proc. SPIE, 1466, p. 13, 1991.
12. D. Funhoff, H. Binder, R. Schwalm, "Deep-UV resists with Improved Delay Capabilities," Proc. SPIE, 1672, p. 46, 1992.
13. H. Ito, "Approach Toward Environmental Stabilization of Chemical Amplification Resists," J. Photopolym. Sci. Technol., 6, 547-562, 1993.
14. H. Ito, et al., "Environmentally Stable Chemical Amplification Positive Resist: Principle, Chemistry, Contamination Resistance, and Lithographic Feasibility," J. Photopolym. Sci. Technol., 7, pp. 433-448, 1994.
15. R. Allen, et al. "193nm Single Layer Resists: Building Etch resistance Into a High Resolution Imaging System," SPIE, 2438, pp. 474-485, 1995.
16. K. Petrillo, "Process Optimization of APEX-E," SPIE, 1926, pp. 176-187, 1993.
17. D. Bucca, "Nonchemically Amplified Positive Photoresist for Synchrotron Radiation X-ray Lithography," Proc. SPIE, 2438, p. 403, 1995.
18. A. Aviram, D. Bucca, D. Seeger, "A diazoquinone positive photoresist for x-ray lithography," J. Vac. Sci. Technol. B, 13(6), 2968, 1995.
19. C.G. Willson, et al., "New NonChemically Amplified 248-nm Resist," SPIE`s 22nd International Symposium on Microlithography, Santa Clara, CA, March 1997.
20. M. Hartney, et al., Vac. Sci. Technol., B 8, 1476, 1990.
21. M. Hartney, D. Johnson, A. Spencer, Proc. SPIE,1466, p. 238, 1991.
22. A. Aviram, K. Petrillo, A. Pomerene, D. Seeger, to be published.
DAVID SEEGER joined IBM in 1983 after receiving his PhD degree in organic chemistry from Yale University. He spent the majority of his career working on resist processes for x-ray applications, including those used in the first all x-ray CMOS devices through 11 levels of lithography in the mid-1980s. He currently manages a group involved in materials and process development for advanced optical
applications, as well as for the fabrication of optical and x-ray masks. IBM Research Division, T.J. Watson Research Center, Rte. 134, Yorktown, NY 10598; ph 914/945-3838, fax 914/945-4015, e-mail [email protected].