Resist Developers Try to Beat the Clock
Laura Peters, Senior Editor -- Semiconductor International, 2/1/2001
 | At a Glance | | |
| | ||
| |||
The 193 nm (ArF) photoresists must be selected this year to target the 100 nm node. Such pressure might not be so daunting were there not three seemingly viable resist polymer platforms to choose from at 193 nm, each with benefits but none that can be applied across production lines. "What's going to happen is folks are going to choose different chemistry platforms for different applications," said Will Conley of Motorola (Austin, Texas). Device makers will stick with single-layer resists whenever possible, he added, but bilayer approaches are being considered, especially for dual-damascene patterning.
Even more daunting is the conversion to 157 nm lithography for patterning critical levels of the 70 nm device generation, which could come as soon as 2003. But many believe all lithography pieces will not be in place until the 50 nm node, especially since the two chemical platforms for resists were just identified. "We used to ask, 'Can we do it?' Now the question is, 'How fast can we get it done?'," said Grant Willson, professor of chemistry and chemical engineering at the University of Texas at Austin. Willson is confident the industry will develop workable resists for 157 nm lithography, but is less sure the 2003 deadline can be met.
248 nm imaging capability
With enhancements, engineers currently estimate the limit of 248 nm imaging at 100 nm for isolated lines and 110 nm dense structures using the latest high-NA (0.8) scanners. Part of the difference is due to the structure the light is illuminating. Scattering bar and OPC technology are also more readily applied to isolated structures. At these critical dimension (CD) limits, it is very hard to maintain adequate process windows in a production environment and process optimization is engineering-intensive. "There will be some applications where 248 nm will be pushed down to 100 nm, but this is going to be very difficult and the application of that will be limited," said Kurt Ronse of IMEC (Leuven, Belgium).
|
|
Though people often think of the gate as the most challenging imaging layer, printing isolated lines is easier than printing nested features. "Although logic chips continue to lead the industry in the introduction race, gate layers may lose their role as the flagship layer for new lithographies since isolated features are reasonably easy to print," explained Ralph Dammel of Clariant Corp., AZ Electronic Materials (Somerville, N.J.). "Contact layers have rapidly become critical, and may be the first to transfer to 193 nm."
 |  2. Ion implantation applications require a thicker resist to withstand ion bombardment and provide adequate stopping power. (Source: Shipley) |
In going from i-line exposure to 248 nm, resist chemistry and the mechanism of dissolution inhibition switching changed from the novolac-based chemistry with DNQ dissolution inhibitor to chemical amplification using photo-acid generators (PAGs) and acid-catalyzed deprotection mechanisms. Chemically amplified (CA) KrF resists are typically based on hydroxystyrene polymers, alone or in combination with other monomers such as t-butylacrylate. Resist designers incorporate different PAGs, acid-labile deprotecting groups and additives to increase performance or tailor the resist for a specific application.
Resist patterning with all CA resists occurs in two steps: acid generation during exposure and acid-catalyzed reactions during the post-exposure bake (PEB) step. Chemical amplification refers to a catalytic reaction that begins when a photoacid generator decomposes upon exposure to DUV light. Acid formed on decomposition of the PAG cleaves the protecting groups on the polymer backbone, causing a polarity change in the polymer from lipophilic to hydrophilic, making exposed regions soluble in basic developers like tetra-methyl ammonium hydroxide (TMAH).
With chemical amplification comes a need to control amine-based contaminants throughout the lithography system and, importantly, to control the PEB and pre-PEB environment. Through 248 nm lithography work, the industry has learned that resist contamination must be approached from all angles, including interaction with neighboring films. For instance, a silicon-rich nitride hard mask in direct contact with resist causes feature footing, requiring an intermediate oxide passivation layer formed by removing ammonia and nitrogen near the end of the CVD process.
Depending on the baking temperatures, CA resists are either high-activation energy — baked at 130°C (PAB) and 140°C (PEB) — or low-activation energy — baked at 90°C (PAB) and 110°C (PEB). Low-activation energy means that, shortly after the light hits the resist, it can undergo transformation near room temperature. Because the reaction goes to completion more quickly, low-activation energy resists are less susceptible to time delays after exposure and before PEB. With high-activation energy resists, engineers must better control delays in the track system (see "Track Requirements for Chemically Amplified Resists"), which can influence CD variation within a lot, Shipley's Thirsk said.
Antireflective coatings (ARCs), both spin-on organic materials and CVD inorganic films, play a crucial role in imaging most sub-200 nm features. ARCs improve resolution and CD control by minimizing reflectivity from the underlying substrate. The more common bottom ARCs (BARCs) decouple the imaging process from the substrate properties and the periodic part of the swing curve. Top ARCs reduce the swing curve and lead to improved yields, especially on contact layers where they resolve the missing contact problem, according to Clariant's Dammel. Though many process engineers prefer CVD-deposited ARCs, there are applications in which organic films are preferred. For example, Shipley's line of ARCs has grown to address needs of conformal coating, planarization, different etch rates, etc., Thirsk explained. At the same time, ARCs should be as broadly compatible with resists to minimize complexity in manufacturing.
|
|
|
193 nm solutions
 5. 80 nm lines and spaces (1:3) imaged in AZ EXP AX 2010P 193 nm resist using conventional illumination, a binary mask, NA=0.6 and s =0.7. (Source: Clariant AZ Electronic Materials) | |
As was the case with initial 248 nm formulations, the most pronounced issues with 193 nm resist maturity include process latitude (exposure latitude and DOF) and line-edge roughness. With lithography's close link to etching capability, it is also important to note that advances in etcher design and high-density plasma systems are helping to improve 193 nm resist performance. In cases of insufficient dry etch resistance, use of a hard mask can work because the top resist layer only has to work well enough to pattern the hard mask. "The concept of hard mask is getting more and more accepted in the industry, on the one hand, for 193, but also in preparation for 157," Ronse said. This is especially true as resist thickness scales rapidly with illumination wavelength, around 400-600 nm at 193 nm and perhaps 200 nm at 157 nm.
The likely insertion point for 193 nm lithography is the 100 nm device generation. Ronse estimates possibly 10 critical mask levels will require the new systems. Critical masks will include gate, active area, contact, first via and metal levels.
Six months into the availability of full-field 193 nm scanners, resist suppliers began commercially introducing 193 nm products. Despite the fact that device engineers prefer one resist/one supplier, at 193 nm there are three useful polymer platforms: acrylates, pure cyclic olefins and cyclic olefin/maleic anhydrides (COMAs). Each platform has its advantages. "We have found that the acrylates, particularly the 2-adamantane methacrylate chemistry developed by Fujitsu, generally provides the best dense line performance and good contact hole performance," said Clariant's Dammel. In addition to lithographic performance, each chemistry platform offers different etch resistance depending on the substrate, and each offers different cost structures. Thus, resist suppliers are pursuing multiple platforms to determine the optimum results for each device layer.
Since resist suppliers do not have access to the latest exposure tools, they often form partnerships with leading customers and share data on resist performance. "We have a joint development program with Samsung where we licensed VEMA, a variation on COMA chemistry, and they supply us with feedback on its performance in production," Shipley's Thirsk said.
Fortunately, 193 nm development can help 248 nm processes today. "One of the things that spins out of 193 nm resist development is the 193 nm resists are very much more transparent at 248 nm than 248 nm resists," Willson explained. "So, in principle, you can back-integrate." The same phenomenon is occurring with the 157 and 193 nm resists, he added.
One issue regarding 193 nm imaging is the shrinking of photoresist features when exposed to the electron beam in SEMs. Resist companies are devising formulations that will be less sensitive to SEM analysis, while researchers fine-tune their analysis settings to prevent this problem.
157 nm lithography will happen
Recent breakthroughs in resist formulation are making 157 nm lithography technology possible. Resists were long believed to be the toughest parts of the overall solution because all previous materials used in photoresists absorb 157 nm light too strongly. Willson noted that the simplest polymer — polyethylene — and even Teflon are opaque to 157 nm.
Fortunately, the industry recently identified two promising platforms: highly fluorinated hydrocarbons and silicon-based siloxanes. "The main challenge in developing a siloxane resist is achievement of a sufficiently high glass transition temperature," Dammel explained. Willson said they currently favor silsesquioxanes and are also examining fluorinated alcohols. Part of the problem lies in the fact that every component of the resist must be highly transparent to 157 nm radiation with absorbance in the 2.0/µm or lower range.
Thirsk estimates the cost of developing a 157 nm resist at $50-60M. Since such an undertaking is not manageable for any one company, International SEMATECH formed a university-based group to perform fundamental research on 157 nm resists. It includes researchers from Caltech, Clemson, Cornell, UC Berkeley and UT Austin. Lessons learned through this study are open to the public. Just as significantly, IMEC and ASML are cooperating on 157 nm lithography development with funding from many top device manufacturers, including AMD, Infineon, Intel, Micron, Motorola, Philips, Samsung, ST Microelectronics and UMC.
|  6. At 157 nm exposure wavelength, top surface imaging was used to pattern 80 nm lines and spaces (1:2). Lines are 300 nm tall. (Source: G. Willson, UT Austin) |
Beyond needing to meet the performance requirements for reduced line-edge roughness, fast photospeed, dry etch resistance, solubility in aqueous base developer, etc., 157 nm resists have a strong tendency to outgas materials, which settle on the lens. At 157 nm, even a monolayer of oxygen or water can significantly change the transmission of light through the lens and even attenuate light. "Fluorocarbons are less of a concern than silicon dioxide," Willson explained, because the user can bleed oxygen into the system with the laser on, making fluorocarbon residues volatile. With siloxanes, however, SiO2 deposits are more difficult to remove.
Beyond creating a suitable environment for processing these resists, engineers will have to make necessary changes to the track systems in terms of handling priorities, Conley said. "Variations in time between post-apply bake and exposure will cause repeatability problems." Dammel claims a further problem at an absorbance of 1.5/µm (a generous value), because the wall angle of the resist will be lower than the standard 85°C requirement.
| Track Requirements for Chemically Amplified Resists Rob Crowell, TEL Clean Track Marketing Manager, Austin, Texas | |
| Photoresist track design is dictated by needs to tightly control resist processing together with the demands of running integrated track/scanner systems in a production environment. Sub-180 nm imaging demands tight specifications on coating uniformity, adhesion, baking temperature uniformity, short and consistent delay times, and rigorous environmental control. For advanced 248 and 193 nm lithography, many photoresist processes today are layer-customized. Gone are the days when engineers used one or two resist processes for all device layers. Additionally, track manufacturers must deliver innovative hardware and software solutions to address key components of process non-uniformity. As feature sizes approach 100 nm, issues such as pattern collapse and adhesion become more critical. New 193 nm photoresists are especially vulnerable to lifting since the formulations are not yet optimized for adhesion as are i-line and 248 nm materials. Adhesion can be optimized with respect to wafer temperature and HMDS delivery for each resist type. Lifting and pattern collapse can also be addressed by reducing resist thickness to improve structural strength. Multilayer resist processes offer another potential remedy. During resist coating, film thickness uniformity is essential. Across-wafer and wafer-to-wafer tolerances of 50 Å were acceptable for i-line processes but sub-180 nm processes require <20 Å ranges. To achieve tighter film thickness tolerances, the engineer must rigorously control cup humidity and temperature, resist temperature, resist dispense rate and wafer temperature. Some newer resist materials have exhibited sensitivity to temperature and humidity not only in coater bowls but in other modules as well. Some sub-130 nm processes require ±0.1°C temperature control and humidity within ±0.2%. Post-apply bake and post-exposure bake processes play a large role in determining final critical dimensions. While i-line resists require hotplate uniformities <1.0°C, bake modules for 248 nm resists need to be <0.5°C. The ideal develop process renders defect-free and uniform photoresist features, requiring a rapid, uniform application of developer. Defects can be minimized by optimizing developer nozzle designs and the post-develop rinse steps. The control of airborne amine contaminates such as NMP and ammonia is crucial in chemically amplified resist processes. HMDS application is an ammonia-producing process that requires ample exhaustion to prevent leakage into the track environment. As feature sizes approach 100 nm, engineers must precisely control wafer transfer times so that all wafers experience similar thermal cycles and time intervals between exposure and develop. Track hardware can place wafers in "safe holding" locations after bake processes, while software can manage wafer flows through the track and stepper to ensure wafer transfer consistency. Throughput is a key factor, especially with the cost of track/stepper cells approaching $10M. Steppers and tracks must operate at 100 wafers/hr or better — while maintaining very tight CD and defect control. Engineers should scrutinize all process steps in a track to ensure they do not adversely impact stepper throughput. |
Clariant, AZ Electronic Materials
DuPont Central Research and Development
