Resists Join the Sub-l Revolution

But resist advances do not occur in a vacuum. Today, more than ever, integrated development efforts - taking simultaneous advantage of high-NA scanners, OPC, PSMs, anti-reflective coatings (ARCs) and advanced resists chemistries - are critical. In fact, concurrent optimization of each lithographic component and 'image process integration,'¹ may provide the only path into deep sub-wavelength regions. Resist development also requires the latest tools. 'We are eager to gain access to the latest 0.70 NA scanners and get a feel for how much lithographic improvement the new tools deliver versus how much additional technology development with the resists is required,' said Murrae Bowden, director of R&D at Arch Chemicals (Norwalk, Conn.).

Resists extend life of i-line

The vast majority of i-line processes use positive-tone single layer resists based on the same diazonaphthoquinone (DNQ) novolac-based chemistry used since the 1980s. Exposure to 365 nm light converts the DNQ dissolution inhibitor into a base-soluble acidic photoproduct that increases the dissolution rate of the novolac matrix in exposed regions. Though i-line chemistry is robust and mature, resist chemists continually tweak formulations to attain incremental improvements in resolution, exposure latitude and depth-of-focus, three of lithography's most crucial performance parameters. Resist requirements include thermal stability, good adhesion to substrates, high purity, aqueous base development and compatibility with aggressive plasma etching processes.

Photoresist for both deep UV and i-line lithography can be spun onto 300 mm wafers by tools such as this Polaris 3500 from FSI International (Chaska, Minn.)

Success in extending i-line is so great that surprisingly, some IC manufacturers have converted some DUV processes back to i-line due to recent advances in resist technology and RETs -- taking advantage of i-line's lower cost of ownership while freeing DUV scanner capacity, a frequent bottleneck in leading-edge fabs. 'There's often a long discussion as to which technology is best suited for a particular line,' said Mark Thirsk, manager of lithographic products at Shipley (Marlborough, Mass.). Depending on the pattern and device requirements, the limit of i-line in production appears to be in the 0.25-0.20 µm range (Fig. 1).

Fig. 1 OIR 88 resist and a 0.57 NA I-line stepper with 2/3 annular illumination produced 0.25 mm dense lines with 1.4 mm DOF. (Source: Arch Chemicals)

248 nm maturing

The resolution limit for 248 nm resists with enhancements continues to decrease. Most experts indicate 0.15 µm devices are possible with a strong push to 0.13 µm. 'Internal results with a digital signal processor (DSP) vehicle indicate we can fabricate 120 nm devices with DUV,' said Om Nalamasu, technical manager for optical lithography and imaging materials research at Lucent Technologies/Bell Labs (Murray Hill, N.J.). 'Positive resists for 248 can image down to 100 nm in logic applications, 130 nm in dense arrays,' said Rick Hemond, technical development manager for new products at Shipley.

The CA resists based on polyhydroxystyrene (PHOST) polymer used at 248 nm represent a radical departure from i-line's novolac-based resist chemistry. The HOST polymer backbone has protecting groups that become deprotected when a photoacid generator (PAG) decomposes upon exposure to 248 nm wavelength light, beginning a catalytic chemical reaction described as chemical amplification. The deprotection mechanism causes a polarity change in the resist polymer from lipophilic to hydrophilic, making exposed regions soluble in developer, typically TMAH (tetramethyl-ammonium hydroxide). Imaging in CA resists occurs in two steps: acid generation during exposure and acid-catalyzed reactions during the post-exposure baking (PEB) step.

Because acid diffusion greatly influences resist characteristics during post-exposure delay and PEB, control of amine groups that neutralize the acid-catalyzed reactions is essential. This requires molecular contamination monitoring of the environment in the track system and stepper, which, if uncontrolled causes resist 'T-topping.' Similarly, some processed SiON ARCs cause 'footing.' As a result all DUV production lithography cells have airborne molecular contamination filtration and monitors provided by Extraction Systems (Franklin, Mass) and other firms. Though environmental control requirements have not changed, newer 248 nm resists are more resistant to amine contamination due to incorporation of low activation energy protecting groups or annealing that reduces the resist's free volume.

Fig. 2 10 nm lines and 225 nm spaces imaged in 400 nm of UV110 positive DUV resist and AR3 ARC on an ASML PAS 5500/300 scanner with 0.60 NA lens and conventional illumination. (Source: Shipley)

With manufacturers supplying fifth and sixth generations of 248 nm resist, deep UV lithography is maturing rapidly. Commercial DUV resists have undergone tremendous improvement, now offering imaging performance to the 150 nm range (Fig. 2) with etch resistance comparable to i-line resists. DUV resists also are being optimized for specific challenges, resulting in PSM resists, resists for isolation layers, contact holes, metal lines, etc. Though such specialization introduces complexity to the fab line, it is absolutely necessary to sub-wavelength lithography. In addition, certain device designs may emphasize a given requirement. 'Memory devices typically have more demanding etch requirements than, say, logic devices,' explained Arch Chemicals' Bowden.

One example of such specialization is Mitsubishi's process for shrinking KrF-imaged via holes from 0.2 µm to 0.1 µm using a coat, bake and rinse sequence following resist patterning. Clariant AZ licensed the material and commercializes it as RELACS (resist enhancement lithography assisted by chemical shrink) using R200 coating and R2 developer (Fig. 3), with solutions also available for i-line.

Once the performance of single-layer PHOST-based resists is exhausted at 248 nm, bilayer resists can become useful. 'Bilayer schemes represent a key way of extending resist technology to push a given wavelength capability,' said Allen. Bilayer processes decouple requirements for imaging, performed by the thin (200 nm or less) top resist, from needs for etch resistance, managed by the thick (400-1000 nm) bottom resist. The top layer uses standard single-layer resist imaging, and the image is transferred down anisotropically using O₂/SO₂ plasma etching. The bottom layer acts as ARC, while reducing pattern-width degradation from underlying topography. Generally, a higher carbon:oxygen ratio in the polymer structure promotes etch resistance. Importantly, bilayer image performance tends to be limited more by the capability of the scanner, optimized in ultrathin resist layers, compared with thicker, single-layer resist (SLR) approaches.

Fig. 3 The RELACS process shrinks these KrF imaged 0.22 mm contact holes (left) to 0.11 mm (right) using a coat, diffusion bake and rinse step after wafer patterning. (Source: Clariant)

At least one bilayer approach has been implemented in manufacturing. The CARL (chemically amplified resist lines) process developed by Infineon Technologies (Erlangen, Germany), is being evaluated at 248 nm and 193 nm. Siemens has used the CARL process in i-line lithography in DRAM and logic IC production fabs since 1995 to extend i-line capability to 0.22 µm lines and spaces without optical enhancements. At 248 nm exposure (0.6 NA), it provides 0.175 µm contacts with a remarkable, 1 µm estimated depth-of-focus.³ The process especially lends itself to creation of high-aspect-ratio trenches or contact holes (Fig. 4).

In the CARL process, the top resist is first patterned and wet developed. Then a puddle silylation step swells the resist by 30-100 nm, providing lateral line widening or shrinking of contact holes, while increasing silicon content for etch resistance. This offers the option of overexposing a positive-tone resist to assure contact hole punch-through and returning to the target CD using the silylation 'chemical bias.'³ The novolac-based bottom resist becomes insoluble in solvents after hard baking at elevated temperatures. Finally, a dry develop is followed by O₂/SO₂ etching of the bottom resist. The CARL process increases focus latitude dramatically because fine structures can be exposed at the isofocal point.

Unlike the CARL process, conventional bilayer approaches do not use silylation to add silicon content to the top resist layer but instead incorporate silicon in the resist matrix itself. The silicon converts to a silicon oxide hard mask/etch barrier.

Fig. 4 The CARL process uses a chemical biasing step and high-density plasma etching to define 30 nm contact holes with >20 aspect ratio using a 193nm, 0.6 NA scanner and standard binary mask. (Sources: Infineon, Clariant, Trikon Technologies, International SEMATECH)

Bilayer drawbacks include an unconventional etch process, stripping issues with the resist and added process steps (though process time compares favorably to SLR processes with ARCs). A variation of standard bilayer approaches, top surface imaging (TSI), appears to have line-edge roughness issues. TSI images the resist, then silylates it to improve pattern transfer to underlying resist.

193 nm timing

The insertion point for 193 nm lithography probably will be second-generation 0.13 µm or even 0.10 µm devices. Successful extension of 248 nm resists is stalling the 193 nm transition, but acceleration of industry technology roadmaps also contributes to this decision. 'The compressed timeline is the biggest problem,' explained Kurt Romse, head of the lithography group at IMEC (Leuven, Belgium). 'If there were four years to develop 0.13 micron technology, then 193 would have a very good chance of being accepted into manufacturing from the beginning, but there's only two or less, causing people to put more emphasis on pushing 248 nm.'

At 0.13 µm, Lucent Technologies' Nalamasu emphasized a benefit to transferring to 193 nm using standard binary reticles rather than enhancements using 248 nm technology. 'Most of the development work for 130 nm will be done using 248 nm, and the decision going into volume production will depend on the maturity of 193 nm resist materials and performance of exposure tools with very high NA, say, above 0.68, at that time,' he said. Nalamasu added that the necessity for optical enhancements at the insertion point of 193 nm appears high, probably requiring some type of PSM and OPC. 'To an extent this depends on whether you're a DRAM supplier or ASIC supplier,' he said. Shipley's Thirsk added: 'Ingenuity using the current resist and exposure tools is eroding the economic requirement to move to 193. If you're faced with an investment in engineering time versus a new tool, resists and masks, the 248 nm work will always take precedence.' Thirsk noted that some 193 nm R&D has helped further optimize 248 nm resists. 'Efforts in controlling dissolution, influencing the etch rate and increasing the transparency of 193 resists have positively influenced 248 nm developments,' said Thirsk.

Scanner availability also may contribute to 193 nm delay, since production units only became available early this year. From the resist side, transition to an entirely new chemistry is required, introducing numerous challenges. The now-familiar hydroxystyrene-based polymers used at 248 nm absorb 193 nm light too strongly, making them unsuitable for use at 193.

Progress in 193 nm resists

The most interesting aspect of 193 nm resist development is that nearly a decade of R&D has not produced a clearly superior polymer platform. Though one might expect the field to be narrowing this close to the technology insertion point, IBM's Allen contends that almost the opposite is happening. 'Many companies are working full tilt in two or three completely different polymer families, whereas two years ago each company actively involved in 193 resists had its own platform.' He added, 'It's going to be very interesting because no single platform is going to provide the best process for every application at 193.' Companies also are vying for intellectual property positions. 'The 193 nm field is crowded with IP and patents, much more so than at 248,' explained Thirsk.

Fig. 5 Reportedly the world's smallest (0.09 mm2) conventional non-volatile memory device was fabricated using 193nm exposure, a strong PSM and multi-layer ARC. (Source: Lucent Technologies/Bell Labs)

Another example of the broad-spectrum work involves a lithography consortium headed by IMEC whose charter includes evaluating various 193 nm resist chemistries from Arch Chemicals, Clariant, JSR (Sunny vale, Calif.) and Shipley on a common platform (an ASML 5500/900 scanner with integrated TEL track). This will simplify comparisons of resists whose performance often is measured under different conditions.

The fortunate result of industry-wide activity in 193 is the emergence of several very promising single-layer resists, though most remain 1-2 years from production-worthiness. Resolution results to date are impressive. For instance, Lucent Technologies/Bell Labs fabricated an 80 nm floating gate flash memory device using optimized 193 nm resist, multilayer ARC and alternating PSM (Fig. 5). Using single-layer resist, 0.6 NA scanner and inorganic ARC, 0.14 µm lines and spaces with 0.9 µm DOF are possible (Fig. 6). Also using conventional illumination, 100 nm lines were imaged with an experimental resist (Fig. 7).

Improvements in 193 nm are most needed in the areas of etch resistance and reduction of line-edge roughness. 'Line-edge roughness is much more pronounced at 193 than at 248, indicating the resists are not at the level of maturity they need for production,' explained Romse, of IMEC. 'There's a complex interaction between lithographic performance, line-edge roughness and dry etch resistance. Resist chemists need to find the right compromise -- a viable but difficult task.' Nalamasu claimed line-edge roughness can be reduced to the 5-7 nm range, 'but trying to get it below that requires a better understanding of the whole process.' Frank Houlihan, technical staff member at Lucent, added, 'It's even more complicated than that because there's some question as to the effect line-edge roughness has on device performance, something we are currently exploring.'

Fig. 6 9010-ES-B resist on inorganic ARC and 193 mm scanner produced 0.13 mm L/S with illustrated DOF. (Source: Arch Chemicals)

The majority of 193 resist development work is performed empirically, partially due to complexity of reactions in the resist. Simulation is used, but mainly to gain understanding of the physical and chemical interactions and resist behavior rather than to optimize resist chemistries. 'Modeling is more difficult because we're dealing with completely different classes of materials and additives than we were working with at 248. We model interactions and performance of materials in order to couple that understanding with processing of the resist,' said Elsa Reichmanis, head of polymer and organics research at Lucent. 'Once we extract the appropriate parameters and model dissolution behavior as a function of components in the resist, we can build a suitable model.' Shipley's Hemond added: 'The simulation problem is very complex, and it is complicated to extract the chemical values the simulators require to make the necessary calculations.'

The best 193 nm polymers today include acrylic polymers (acrylate-based platforms with attached alicyclic structures) and cyclic olefin platforms. Acrylic platforms were the original formulations for 193 nm. They traditionally provide high imaging performance yet significantly lack etch resistance, though incorporation of cycloaliphatic side groups in the polymer backbone helps slow the etch rate. Etch resistance is a critical issue, particularly at the thinner resist thicknesses (400-600 nm) required for 0.13 µm processes and beyond. For all 193 nm resist platforms, performance is substantially below that of 248 nm systems, due to short resist development times and difficulties in making the 193 nm carboxylic acid chemistry work with TMAH developers. 'If 193 nm resists were as good as today's 248 nm resists, 193 nm exposure would be able to operate at the 100 nm node without optical enhancements,' said Ralph Dammel, head of Clariant AZ's 193 nm program.

'The best resolution performance for dense lines is obtained with acrylic systems, particularly with the adamantane methacrylate chemistry developed by Fujitsu,' said Dammel. He bases his statement on a recent SEMATECH comparison of six resists from leading suppliers, among other sources. Clariant AZ plans to further develop this chemistry.

Cyclic olefin polymers were developed over the last five years and appear to be a most promising approach for 193 nm. IBM, and Lucent Technologies/Bell Labs in collaboration with Arch Chemicals, are strongly pursuing this group of materials that 'have intrinsically higher robustness toward the fairly harsh etch steps that are being performed,' explained Allen. IBM and BF Goodrich co-developed a norbornene copolymer that uses a transition-metal catalyst to polymerize monomers that were not readily synthesized before. For the first time this year these resists demonstrated imaging performance at the sub-100 nm level with etch resistance roughly equivalent to today's resists, according to Allen. The Bell Labs approach, being developed jointly with Arch Chemicals, uses alternating copolymers of norbornene and maleic anhydride, iodonium-based PAG and cholate-based dissolution inhibition. 'It is our belief that this chemistry's free radical polymerization is easier to control than some other approaches,' said Reichmanis. Alternatively, a free radical copolymer of functionalized cycloolefins and maleic anhydride that Clariant AZ licensed from Hyundai 'promises to be very cost-effective and may migrate back into 248 nm to reduce resist cost,' said Dammel.

Fig. 7 100 nm lines with 200 nm spaces were patterned using experimental 193 nm resist and AR10 ARC on an ISI 193 nm, 0.6 NA scanner with conventional illumination. (Source: Shipley)

Beyond selecting the most robust polymer platform, resists chemists are challenged to determine the PAG best able to deliver strong dissolution inhibition, high photospeeds, thermal stability and tailorable photoacid size, the last of which correlates strongly with imaging performance.

Another critical issue at 193 is resist outgassing of organic compounds that build up on scanner optics - addressed by PAG selection or using various additives.⁴ Aromatic fragments such as the iodobenzene liberated by iodonium salt sensitizers are among the most critical outgassing species. 'We've explored structural property relationships between outgassing and chemical structure of the resist, leading to identification of systems that can substantially decrease deleterious outgassing or even completely eliminate it,' said Houlihan. While the outgassing problem is evident with 248 nm systems as well, 'the working distance between the lenses and resist at 193 nm is shorter, and the absorption of vapors is greater at shorter wavelengths,' he said, so the outgassing issue will only be exacerbated for 157 nm systems.

Bilayer approaches continue to be evaluated at each lithography wavelength. At the 0.13 µm node, resist thickness will be scaled back even more, so the combination of high-NA lenses and ultrathin resists may become necessary. However, experts agree that a substantial roadblock in single-layer resist technology must occur before bilayer approaches will gain wide acceptance.

Conclusions

Advanced resist chemistries and ARCs enable penetration into the deep sub-wavelength regime at i-line, DUV and 193 nm exposure. The amazing advances in stretching 248 nm capabilities via resists, ARCs, high NA scanners and optical enhancements will stall 193 nm lithography insertion, but production-worthy resists should be prepared for the 0.13 µm device generation. Acid-catalyzed deprotection is the imaging mechanism of choice, and novel polymer structures, PAGs, additives and resist structures are paving the path to subsequent generations of sub-wavelength lithography.

References

1. R. DeJule, 'Revolutionizing Process Integration,' Semiconductor International, Jan. 1999, p.62.
2. P. Singer, 'Anti-reflective Coatings: A Story of Interfaces,' Semiconductor International, Mar. 1999, p.55.
3. W-D Domke, et al., 'Chemical Amplification of Resist Lines: The CARL Process,' Microlithography World, Spring 1999, p.2.
4. F. Houlihan, et al., 'A Study of Resist Outgassing as a Function of Differing Photoadditives,' SPIE Advances in Resist Technology and Processing XVI, March 1999, p. 264.
5. R.A. Cirelli, et al., 'Chemically Amplified Negative-Tone Resist Using Novel Acryl Polymer for 193 nm Lithography,' SPIE Advances in Resist Technology and Processing XVI, March 1999, p. 295.