Silicon-rich-methacrylate bilayer resist for 193-nm lithography
06/01/1998
Cover Article
Silicon-rich-methacrylate bilayer resist for 193-nm lithography
Andrew Blakeney, Allen Gabor, Daniela White, Thomas Steinh?usler, Olin Microelectronic Materials, Providence, Rhode Island, William Deady, John Jarmalowicz, Roderick Kunz, MIT Lincoln Laboratory, Lexington, Massachusetts, Kim Dean, Georgia Rich, David Stark, Sematech, Austin, Texas
A prototype bilayer resist system, based on a silicon-containing methacrylate imageable layer and a crosslinked styrenic copolymer undercoat, shows sub-0.15-?m resolution after 193-nm exposure and O2 etch. The resistance to the substrate-etch plasma is better than current DUV resists. Making bilayer resists feasible for sub-0.18-?m manufacturing requires solutions for some recentlyidentified etch process and materials issues.
ArF excimer laser lithography at 193 nm is the prime candidate for sub-0.18-?m patterning [1-14]. The leading resist technology candidates for practical 193-nm lithography are top surface imaging (TSI) [12, 13], bilayer (BL) [7-11] and single-layer (SL) resists [1-6], with a customer preference for the latter. The underlying technologies and materials of each approach have characteristic advantages and disadvantages. The numerous challenges for the ArF resist chemists (e.g., transparency, photospeed, adhesion, sensitivity, various process time delay latitudes, plasma etch resistance, etc.) have somewhat different weights for each technology.
Lithographic aspect ratios and other issues require that resist films be thinner (about 0.5-0.6 ?m) for sub-0.18-?m devices. This, in turn, requires either greatly improved etch processes, or improved etch resistance for the resist, or both. Thus, excellent plasma etch resistance (preferably even better than earlier formulations) is a critical feature for realistic ArF resist formulations. Most SL g-line, i-line, and DUV resists are based on hydroxyaromatic polymers such as novolac and polyhydroxystyrene (Fig. 1).
 |
Figure 1. Hydroxyaromatic polymers used in typical previous photo-resists. Both the etch-resisting group (red) and alkali-solubilizing group (blue) had to be replaced for 193-nm transparency.
These highly aromatic materials have excellent resistance to the plasma used to etch the substrate and excellent aqueous alkali solubility for image development. Extremely high light absorption, however, precludes the use of hydroxyaromatic resins at 193 nm (except for TSI applications). This presents a materials problem to the resist chemist because both the group providing plasma-etch resistance and the group providing the alkali solubility required for the image development must be replaced to make the resist sufficiently transparent at 193 nm. Thus we must find new materials (or groups of materials) with high transparency, etch resistance, and a different alkali-solubilizing group.
 |
Figure 2. Generic alicyclic polymers considered for 193-nm photoresist resins. The moiety R may be H, COOH or COOR`, where R` is an acid-sensitive group.
Polymers with alicyclic groups (as in Fig. 2) are sufficiently transparent at 193 nm and have reasonable plasma-etch resistance. Acid-sensitive carboxylic acid derivatives of such systems become alkali-soluble after deprotection [2, 6, 14, 15, 16], thus appearing suitable as SL photo-acid-catalyzed resist candidates. However, the alicyclic resins in their "pure state" suffer from high hydrophobicity and adhesion problems. Modifications (e.g., blending with other resins, copolymerization, or functionalization) to improve these and other properties tend to decrease the plasma etch resistance significantly below that of novolac-based resists. SL alicyclic polymer resists with acceptable adhesion and development parameters may not show etch resistance superior to that of the aromatic resins [15]. Bilayer resists, however, have different materials options and offer some other potential advantages.
 |
Table 1 rates the competing resist technologies in terms of some key advantages and disadvantages. Reflectivity issues at 193 nm are the same as or worse than those at 248 nm [17, 18], so the SL resists will probably require bottom antireflective coatings (BARCs) for many layers. Therefore, Table 1 shows that option as SL+BARC.
 |
Figure 3. Bilayer resist processing scheme with a silicon-containing photosensitive top layer.
Bilayer resists can be competitive with SL+BARC processes and may offer advantages (e.g., resolution, DOF) as a result of thin-film imaging in the top layer or the ability to use different materials. The basic process steps shown in Fig. 3 for a bilayer resist are the same as for a SL plus BARC. Both technology approaches have two layers (with the corresponding apply and bake steps), exposure, post-exposure bake, development, and oxygen-plasma-etch steps.
The major differences between SL+BARC and BL processes are the film thicknesses used for the individual layers and the distribution of lithographic functions between the layers. In both the bilayer and SL plus BARC processes, the top layer is assigned the imaging function. However, the bilayer uses a thin (~0.25 ?m) imaging layer while SL+BARC uses a thicker (~0.6 ?m) film, with a different protection function. The bilayer imaging film protects the underlayer from oxygen etch and usually contains silicon to provide etch resistance. As the top resist region is converted to refractory silicon oxides, the underlying resist and undercoat becomes better protected, losing additional film only through sputtering. Thus, a moderately thin silicon-containing film will suffice for bilayers. In contrast, the SL+BARC top layer must protect the substrate from etching by aggressive fluorine and chlorine plasmas. Because the substrate-etch plasma will continually erode the film, this layer must be significantly thicker than the upper layer of the bilayer, consuming more of the lithographic focus budget and increasing the danger of pattern collapse during wet development.
Both the undercoat in the bilayer and the BARC in the SL+BARC process reduce substrate reflections and swing-curve effects. The BARC must have a significantly higher O2 etch rate than the imaging layer to avoid diminishing the imaging layer thickness during BARC-etch, thus reducing the protection provided for the substrate etch process. In contrast, the undercoat in the bilayer resist needs only a reasonable O2 etch rate combined with a good substrate-plasma etch-resistance. Because the underlayer in a bilayer resist stack provides much of the substrate plasma etch resistance, it will be significantly thicker than the imaging layer and approximately the same thickness as the SL resist.
The redistribution of the lithographic functions for a bilayer resist means fewer restrictions in the choice of materials. An element capable of forming refractory oxide, such as silicon, is required in the imaging layer of the bilayer resist to obtain good O2 etch resistance. Beyond this requirement there is wide latitude in the choice of materials for the imaging layer. Aromatic moieties can be used in the undercoat for their plasma etch resistance since the exposing radiation need not penetrate the underlayer. Completely aromatic systems, however, will degrade the BARC properties of the undercoat as well as slow the O2 etch rate [17]. Overall, the resist chemist has a wider selection of materials for bilayer resist/undercoat development.
Results and discussion
The Prototype Bilayer Resist System.The bilayer resist results described in this paper were obtained using a methacrylate terpolymer synthesized from the monomers using conventional radical polymerization. Each of the three different monomer units shown in Fig. 4 is responsible for one primary function.
 |
Figure 4. Imaging layer terpolymer before and after exposure.
The silicon-containing monomer at right provides the O2 etch resistance. The monomer with the "property enhancing group" R1 at left is used to optimize the polymer hydrophilicity and coating properties. The third monomer contains an acid sensitive, alpha-alkoxy ester and is the key monomer for the dissolution-switch required for imaging. As the result of 193-nm exposure, a photoacid generator (PAG) in the resist produces an acid strong enough to catalyze a reaction of the alpha-alkoxy ester during the post-exposure bake (PEB) step. The previously insoluble ester, after the reaction, becomes a highly alkali-soluble methacrylic acid moiety. The undercoat used is a styrenic copolymer with a crosslinker. DUV curing crosslinks the mixture to prevent intermixing with the resist top layer.
Bilayer Resist Processing. The data described below was obtained using a 0.25-?m thick imaging layer over a DUV cured undercoat layer 0.45 ?m thick. The resist processing used a 100?C-60 sec softbake and a 100?C-60 PEB on an FSI Polaris Cluster using unlinked proximity bake plates. Unless otherwise specified, the resist was developed by immersion in OPD 262 (a standard 0.262 N TMAH developer) for 30 sec. The 193-nm exposure tool was a 0.6-NA ISI Microstepper. A Trikon (formerly PMT) Pinnacle 8000 HDP etcher performed the O2 etch. The etching parameters were 1375 W (source), 350 W (bias), -30?C and 60 sccm O2 flow at 1.875 mtorr.
Lithographic results.
 |
Figure 5. a) Dense and b) isolated lines patterned in bilayer resist with 27 mJ/cm2 exposure and developed for 30 sec in OPD 262.
Figure 5 shows SEMs of dense line/space patterns with assorted dimensions and corresponding isolated lines obtained using an exposure optimized for 0.15-?m dense lines. The 0.13-?m L/S pairs are almost resolved at this exposure. Higher exposures can resolve down to 0.12-?m L/S pairs although at the cost of noticeably overexposing larger geometries. The DOF of the 0.15-?m L/S features is about 0.6 ?m.
This bilayer resist is very robust in terms of its tolerance for changes in PEB temperature and thus does not require extremely tight temperature controls to obtain reproducible linewidths. The linewidth change for a large PEB temperature change (20?C) is very small (<0.5 nm/?C), resulting in significant latitude for hot plate control and calibration.
The prototype bilayer resist is more sensitive to post-exposure time delay (PED) than might be desired. A 20-min delay between exposure and baking (in a 1.5-ppb NH3 ambient) produces a slight thickening of the linewidth of the top of the line
("t-topping"), but no bridging. Fortunately, charcoal filters for contamination control and an exposure system linked to the PEB station minimize problems with PED.
Pattern transfer to the underlayer. Initial pattern-transfer experiments indicated that results of the O2 pattern-transfer step can vary greatly depending on the etch conditions. Table 2 shows the undercoat
esist etch selectivity ratio for three different processes, which emphasize different physical and chemical mechanisms.
 |
Contributions to the etching process can be broken down into three main components; sputtering, ion assisted etching, and chemical etching [13]. In a sputtering-controlled process (Process A in Table 2), the bond strength, atomic mass, and the density of the materials being sputtered control the relative etch rate. In our case, this means that the methacrylate system is sputtered faster than the aromatic undercoat, which could again lead to some CD loss.
In chemical etching (Process B), the surface of the silicon-containing polymer is transformed into a pseudo-silica-like surface by the oxygen plasma, while the organic polymer is converted into carbonaceous gases and water. Chemical etching gives the highest etch selectivity relative to the undercoat but exhibits poor isotropy, leading to an undercut image profile. Because the chemical etching is not a directional etch, the undercoat is prone to lose linewidth.
The best process for the pattern transfer is probably one with a significant quantity of ion assisted etching (Process C). Ion assisted etching occurs when an oxygen neutral or ion adsorbs to the polymer surface followed by an oxygen ion hitting the adsorbed species, causing the polymer to be etched. This type of etching is directional, much like sputtering, because the trajectories of the ions that assist the adsorbed species are perpendicular to the wafer plane. Figures 6 and 7 were obtained using such a process.
 |
Figure 6. Binary resist pattern-transfer: 0.15- and 0.14-?m dense L/S patterns exposed at 27 mJ/cm2 a) before and b) after O2 etch, and c) after-etch patterns for 25 mJ/cm2 exposure.
Figure 6 shows before- and after-etch SEMs of 0.14- and 0.15-?m dense L/S patterns printed using a 27 mJ/cm2 exposure. The after-O2-etch linewidth has shrunk about 0.02 ?m (lateral etch) and the profiles are somewhat re-entrant. To obtain equal line and space CDs at 0.15 ?m after etch, the resist must be exposed at 25 mJ/cm2 (Figure 6c). The DOF of the 0.15-?m after-etch L/S pattern at 25 mJ/cm2 exposure dose is about 0.6 ?m, the same as the DOF before etch.
 |
Figure 7. Binary resist pattern-transfer: 0.16- and 0.15-?m isolated lines exposed at 27 mJ/cm2 a) before and b) after O2 etch, and c) after-etch patterns for 25-mJ/cm2 exposure.
Linewidth loss for isolated lines is significantly larger than for the dense lines (Figure 7). As a result, the 0.15-?m isolated line at the 27-mJ/cm2 exposure is no longer present after etch. However, the isolated line at the lower 25-mJ/cm2 dose shows the extent of the loss (about 0.07 ?m). Thus, the imaging-layer undercoat transfer process currently employed has an iso/dense bias that is separate from any resist or optical proximity effect.
The etch results are highly dependent on the etcher, etch process, and etch gas employed, a fact which offers significant potential for improving results. Since iso/dense biases have been observed in the O2 etch process, it is reasonable to assume that additional biases are possible in the substrate etch steps. To obtain proper geometry sizes in the final device, the effects of all the etch processes must be considered.
Recent experiments at IMEC and Sematech have shown that use of an O2/SO2 etch gas mixture can lead to a lower lateral etch rate and less etch-derived iso-dense bias. As a bonus, using the O2/SO2 etch gas mixture for etching the undercoat yields significantly less line-edge roughness, which has been identified as an important issue at small dimensions.
Substrate etching. Protecting well-defined regions of the substrate from etching is the primary purpose of photoresist, and thus the resist etch rate is a key parameter. The variety of substrate materials requires a number of different etch chemistries, too many to test on all developmental resists. Typically, the screening criterion for a new resist system is a low etch rate in a harsh Cl2-based metal etch plasma. In bilayer resists, the undercoat layer has the function of plasma etch resistance. Our cross-linked styrenic copolymer underlayer etched at a rate of 296 nm/min in a Cl2 plasma [19] (better than current typical DUV resists), while polyhydroxystyrene (MXP-7) and Novolac resist (OiR-620) showed rates of 290 nm/min and 278 nm/min respectively. Thus we anticipate that the resist/undercoat stack will show excellent results when substrate etching experiments currently underway are completed.
We have completed preliminary substrate plasma etching tests on 2300 ? of polysilicon using an Applied Materials 5000 MxP etcher. An aggressive etch process utilized first CF4 and then an HBr gas mixture as the etch gases.
 |
Figure 8. 0.15-?m nominal lines a) before and b) after etching into230-nm thick polysilicon.
Figure 8 shows the results for isolated and dense 0.15-?m lines after the polysilicon etch. Even with over-etch, approximately 50% of the undercoat thickness remains.
As noted in the previous section, the resist/undercoat lines were narrower after the O2 etch process. In contrast, after the polysilicon etch, the polysilicon lines were wider than the starting resist/undercoat lines. In addition, the linewidth of the isolated lines were increased more than the linewidth of the dense lines, so the iso/dense bias was opposite to that of the O2 etch. The linewidth loss from the lateral etch in the O2 etch process is more than counteracted by the bias in the polysilicon etch process, which results from the generation of sidewall-polymer during the substrate etch. The substrate etch conditions have not yet been optimized to reduce - or better completely avoid - sidewall-polymer build-up.
Resist/undercoat stripping. The stripping process required for the bilayer resist may differ depending on the etch chemistry and on whether or not the stripping is for rework or for the next step in device manufacture. Several cleaning steps may be required. For instance, in the polysilicon etch process, the silicon-containing resist is completely etched away by the plasma, seemingly leaving only the undercoat to be removed by oxygen ashing. However, sidewall polymer formed in the substrate-etching process remains behind, requiring subsequent treatment for removal. Recent polysilicon etch/strip work at IMEC [20] using a three-step strip process - oxygen ashing, sulfuric acid/hydrogen peroxide immersion, and ammonia/hydrogen peroxide rinse - yielded near vertical polysilicon profiles with no apparent remaining resist, undercoat, or sidewall polymer. With some engineering effort, suitable stripping/cleaning processes can be developed for other etch chemistries as well.
Conclusion
Preliminary results with our bilayer resist demonstrate significant promise for practical sub 0.18-?m lithography using ArF excimer laser exposures. Certain bilayer resist performance features require some improvement. To make manufacturing with 193-nm bilayer resists practical, however, the critical parameters must involve processing, notably the optimization of the O2 etch chemistry and process and the optimization of the substrate etch processes to control lateral etch and iso/dense biases.n
Acknowledgments
The Lincoln Laboratory portion of this work was performed under a Cooperative Research and Development Agreement with Olin Microelectronic Materials. The authors of this paper would like to acknowledge the support of many colleagues at OMM, MIT/LL, IMEC, and Sematech for their contributions to this work. Opinions, interpretations, conclusions, and recommendations are those of the authors and not necessarily endorsed by the US government or its employees.
References
1. R.D. Allen, et al., J. Photopolymer Science and Technology, Vol. 10, 503, 1997 and references therein.
2. F.H. Houlihan, et al., J. Photopolymer Science and Technology, Vol.
3. T. Hattori, et al., J. Photopolymer Science and Technology, Vol. 10, 535, 1997 and references therein.
4. K. Nozaki and E. Yano, J. Photopolymer Science and Technology, Vol. 10, 545, 1997 and references therein.
5. T. Ohfuji, et al., J. Photopolymer Science and Technology, Vol. 10, 551, 1997 and references therein.
6. K. Nakano, et al., J. Photopolymer Science and Technology, Vol. 10, 561, 1997 and references therein.
7. T. Morisawa, et al., J. Photopolymer Science and Technology, Vol. 10, 589, 1997 and references therein.
8. Y.-J Kang, et al., J. Photopolymer Science and Technology, Vol. 10, 585, 1997 and references therein.
9. U. Schaedeli, et al., J. Photopolymer Science and Technology, Vol. 9, 435, 1996 and references therein.
10. U. Schaedeli, et al., SPIE, 2724, 344, 1996.
11. B.W. Smith, et al., SPIE, 2438, 504, 1995.
12. S.C. Palmateer, et al., SPIE, 2438, 455, 1995.
13. S.C. Palmateer, et al., J. Vacuum Science and Technology, A, 14, 1132, 1996.
14. S. Takechi, et al., J. Photopolymer Science and Technology, 9, 475, 1996.
15. R. R. Kunz, et al., SPIE, 2724, 365, 1996.
16. U. Okoroanyanwu, et al, SPIE, 3049, 92, 1997.
17. R. R. Kunz, R.D. Allen, SPIE, 2195, 447, 1995.
18. J. Sturtevant, B. Roman, Microlithography World, p. 13 Autumn, 1995.
19. Lucas Helicon Etcher; 2 mtorr Cl2; 80 sccm, 2500-W Source, 75-W chuck.
20. Personal communication from Veerle Van Driessche and Mieke Goethals.