Photoresists Meet the 193 nm Milestone

For many companies, 2005 will be the year that 90 nm processes are ramped into full-scale production and 193 nm systems become the lithography platform of choice for critical levels (Fig. 1 ). Alongside this significant milestone, immersion lithography is grabbing all the headlines. "Immersion lithography has stolen the spotlight this year, but from a production standpoint, we can say that this is the year that 193 nm lithography moved from development into large-scale production," said Mark Slezak, technical manager of lithography for JSR Micro (Sunnyvale, Calif.).
 
Significant benchmarks have been reached in reduced line edge roughness (LER), defectivity and mask error enhancement factor (MEEF) to allow 193 nm resists to be used on the majority of mask levels at 65 nm. Rick Hemond, director of marketing for the Microelectronic Technologies business of Rohm and Haas Electronic Materials (Marlborough, Mass.), attributes part of 193 nm resist maturity to new levels of quality. "The raw material supply line is strong and capable, and it's ready for high-volume 90 nm and 65 nm production."

At the same time, immersion lithography is taking its first steps, with IBM and TSMC demonstrating prototype chips using immersion scanners. The photoresist suppliers are just beginning to acquire immersion lithography tools. "We've been learning how to tweak the polymer composition and formulation components to produce a resist that works well with immersion lithography," said Hemond.

Pushing the limits of photoresist performance are today's typical etch processes. During etch, two problems stand out as limiters to resist performance: reduced etch selectivity (relative to 248 nm resists) and LER (also known as striations). "For a 45 nm process, we are dealing with 30 nm gates, and these small linewidths can induce many problems," said Meihua Shen, director of process technology for conductor etch at Applied Materials (Santa Clara, Calif.). Based on current results, if the same HF-based chemistry is used on a 193 nm resist as was used on 248 nm resists, the high-selectivity process would produce too much polymer buildup and worse striations. "So we do different things to pattern the gate. For instance, we open the hard mask and rely on the hard mask to open the gate rather than relying on the photoresist. And for the resist trim, a 70 nm feature might be trimmed to 30 nm, which requires an oxygen-based etch, and perhaps an additive like HBr to help control LER. The choice of etch chemistry becomes very important." Shen said the challenge of achieving high selectivity while controlling LER is met through a combination of hardware modifications as well as process modifications.

1. AZ Electronic Materials has recently upgraded its 300 mm capabilities for generating data on its 193 nm photoresists and antireflective coatings with a 0.85 NA Nikon S306D scanner and TEL Clean Track ACT 12 coater/developer track.

In many cases, bilayer photoresist approaches prove to be simpler and more cost-effective than hard mask approaches. "A bilayer solution utilizes a thick underlayer to control etch resistance, reflection and topography, leaving the role of the top layer to what it was designed for: imaging performance," said Plamen Tzviatkov, business development manager, advanced technology, for Fujifilm Electronic Materials (North Kingstown, R.I.). "The bilayer approach is becoming a mainstream technology, and is no longer perceived only as a backup technology if the single-layer options fail. Bilayer stands on its own merits, offering a solution to etch, CD control and advanced imaging performance, rivaling that of traditional single-layer approaches." Freescale Semiconductor, for one, has committed to bilayer for its 90 nm BEOL processes.

By far, the greatest change to occur in the lithography arena in the past five years has been the rapid rise of immersion lithography, which in two years went from a laboratory curiosity to a critical technology on the ITRS, and eventually put 157 nm projects on hold indefinitely. By placing water between the lens element and the photoresist on the wafer, immersion lithography roughly doubles the depth of focus at a given wavelength because of water's higher refractive index (1.43) than air (1.0). While immersion is currently being applied to 193 nm (ArF) scanners, there is some talk of adding immersion to 248 nm scanners to extend their lifetime and lower cost of ownership.

Immersion lithography is scheduled to enter production around mid-2006. The first full-field tools have yet to be delivered. After introduction, extension of the immersion technology will involve the manufacture of lenses with increasingly higher numerical aperture (NA) from 0.85 to 0.93 to >1.0, increasing the refractive index of the photoresist (~1.7 today), and increasing the refractive index (n) of the immersion fluid (to 1.6 or higher). In this manner, the 193 nm immersion approach could be extended to the 22 nm node, though much remains to be proven about the new lens systems, high-n fluids and immersion scanner systems in production.

2. 45 nm lines and spaces patterned using immersion lithography and a barrier topcoat. (Source: Rohm and Haas Electronic Materials)

Though it is unclear how far 193 nm lithography will be extended, most agree it may well cover the needs of the 45 and 32 nm device generations (Fig. 2 ). Meanwhile, several hurdles have yet to be overcome, including the control of microbubbles in the water and any effects on imaging; control of outgassing of materials from the resist to the water and/or lens element; and addressing defectivity issues that might arise from the immersion approach.

Prototype devices by IBM and TSMC were likely fabricated with a commercial 193 nm resist, with a protective topcoat between the photoresist and immersion liquid (water). The role of the topcoat is primarily to protect the lens element from outgassed materials from the photoresist. There is debate over whether a topcoat is ultimately desired due to cost. "Our strategy is to develop a non-topcoat approach, and studies to date indicate it is technically feasible. We know the major leaching components are PAC and photoacid, both inherently important for imaging," said Tzviatkov. "However, the leaching does not impact the imaging to the extent we previously thought because it is limited to the very surface of the resist, and we believe we have established means to compensate for that and maintain lithographic performance." Tzviatkov emphasized the desire by customers to use the same 193 nm resists for dry imaging as for immersion.

"We do know that when low molecular weight components leach into the water, there are extractables," said Slezak. "But, since there are still questions about the dispensing system that will be used for the water, it's unclear whether the extractables will affect the imaging or the tool. And since nobody is willing to take the risk, the first generation of immersion lithography will likely be done with an immersion topcoat."

George Barclay, 193 nm program leader for Rohm and Haas Electronic Materials' Microelectronic Technologies business, said that through their development, they have gotten to the point where the levels of extractables are approaching detection limits, but that still does not preclude the use of a topcoat in production. "We're letting our customers decide; they will tell us what level of leaching is allowable for 193 nm immersion."

Ralph Dammel, director of technology at AZ Electronic Materials (Somerville, N.J.), added, "Detailed studies need to be performed to look at immersion's impact on CDs, for example, because the puddles that are being moved around on the wafer are bigger than the exposure field, so we need to know if the resist has a history dependence, but this requires a full-field production immersion tool, which very few facilities have right now."

Regarding the chemistry of the topcoats that have been developed, the suppliers are being understandably cautious about disclosing proprietary information. However, Dammel said, "The trend is toward developer-soluble topcoats, because the additional process complexity associated with a solvent-removed topcoat is too high."

One advantage to the presence of a topcoat is that it can also be engineered to have antireflective properties; however, this will not preclude the use of an underlying antireflective coating. "For immersion, we believe the industry will have to adapt a dual-layer BARC approach to have good reflectivity control," said Shree Deshpande, global marketing manager, ARC products, Brewer Science Inc. (Rolla, Mo.). "In addition, it is too early to say, but if the resist suppliers have to change the polymer backbone to achieve a certain refractive index for immersion resists, then BARCs may need to change to maintain compatibility with photoresists to achieve good profiles."

Antireflective coatings serve to control substrate reflectivity for better CD control and a wider process window. In dual-damascene applications, they also provide via filling and/or planarization. Antireflectants can also play a role in reducing the potential for pattern collapse. These films are continually being optimized for higher etch rate and/or better compatibility with different photoresists. "We've realized the importance of addressing these issues, so we are developing materials that now allow our antireflectants to be made thinner and to etch faster, if necessary, to allow thin 193 nm photoresists to be used above them," said Hemond.

Traditional organic bottom antireflective coatings provide superior reflectivity control (to <0.5% @ 193 nm, most cases) for improved photolithography process windows and to extend the limits of KrF patterning, for instance. ARCs have also been used between the resist and hard mask to prevent resist poisoning from SiN, SiON and SiCN hard masks. These BARCs are tuned in their etch selectivities not only to photoresists but also to the hard masks. There are also wet-developable BARCs, which do not require an extra dry etch step, but can more importantly save on resist budget by ~20-30%, according to Deshpande. "Wet-developable BARCs, gap fill and underlayers are being implemented by our customers in manufacturing for <130 nm technology nodes for ion implants as well as dual-damascene integration," he said.

Another solution for dual-damascene integration for logic and deep trench patterning for DRAMs is a hybrid BARC (inorganic/organic), ensemble ARC (Fig. 3 ), which offers tuned reflectivity, etch selectivities and ease of integration. In via-first dual-damascene, it enables seamless etch integration and a wide lithography process window.

3. Ensemble ARC performance with a 90 nm L/S resist pattern (left) and filling 0.22 µm dense vias (middle) and isolated via (right). (Source: Brewer Science)

Developing compatible antireflection strategies will become more complicated for next-generation immersion lithography (NA >1.0). "We have a very large range of angle of incidence from near vertical for the very large features to 45° for an NA of 1.2 for small features. You can no longer control reflectivity with a single-layer bottom antireflective coating due to the different optical path lengths," said Dammel. This will give rise to dual bottom antireflectants, either both organic; one organic, one inorganic; or both inorganic. Both spin-on and CVD inorganic layers are currently available, and often an inorganic layer can also act as hard mask.

In the early stages of development, 193 nm resist platforms included cyclic olefins, acrylates, methacrylates, vinylethers, and hybrids and derivatives of these various forms. As the platforms matured, it became evident that methacrylates and acrylates were the most robust from the standpoint of flexibility of design, ability to integrate with the manufacturing flow and in terms of shelf life. Today, methacrylates and acrylates make up ~90% of the 193 nm resists sold. Other formulations have found their place in specialized applications.

"We focus mostly on methacrylates because they are more mature, we can apply chemical shrinks or reflow processes due to the higher glass transition temperatures, and the initial problem of etch resistance has been solved," Slezak said. He added that methacrylates are simpler to tailor for a specific property, such as reduced MEEF, for instance, because certain monomers or protecting groups can be substituted to deliver a given property. "The flexibility of these polymers allows us to do that."

Barclay agreed, saying, "Methacrylates are made by free radical polymerization that allows us to combine functionalities that we would like to have for resists, and polymers made by different polymerization methods make it more difficult to incorporate those functionalities."

Many of the early problems for 193 nm photoresists — including post-exposure bake sensitivity, MEEF and defectivity — have largely been solved. "The biggest issue that the industry has not found a comprehensive answer for is line edge roughness, and it is still lagging behind the desired values that the industry would like to see," said Dammel. "There is not comprehensive theory that if you do A, B and C, the LER will be good." Though may people point to the fact that the CD budget (2-3 nm) is approaching the size of the polymer, Dammel said the molecular weight of the polymer does not have a dominant influence on LER, based on their experiments, and that there are other factors that are more significant. "It's a very complex field, and it's not always the intuitive answers that lead to improvements in LER," he said.

While the exact causes of LER are unknown, some mechanisms have been identified. "One is definitely high-energy bombardment from the etch, and another is uneven polymer deposition," explained Peter Hsieh, senior manager of technology development for dielectric etch at Applied Materials. He suggested that by going to an etch process with lower-energy bombardment (lower bias), better LER performance can be achieved. Changes in etch chemistry can help reduce polymer deposition on feature sidewalls. Hsieh emphasized, however, that there is often pattern sensitivity to LER across the wafer, and that striations also depend on the type of resist used and the structure of the resist.

According to Applied's Shen, LER can be expressed in many different forms. In the worst cases, LER and the stresses induced during etching or resist trimming can lead to line breakage, line twisting or line movement. "We see a lot of pattern dependency; if we have big pads connected to thin lines, there is a lot of stress at the point where the thin line meets the big pad," she said. "The Advanced Patterning Film has become a major player when flexibility in thermal budget is needed as well as etch selectivity and LER control."

Another issue is pattern collapse — where resist features literally collapse via capillary forces during the spin develop process. Pattern collapse usually shows up in dense lines and spaces, when the aspect ratio is 3:1 or higher. The issue is serious at the 65 nm node and will become increasing critical with continued scaling. "As features become smaller, the collapsing force scales with one over the space between the structures, so it's a hyperbolically increasing force. We find that 45 nm features, which we are currently testing, cannot maintain an aspect ratio of 3:1. We are closer to 2:1, and that puts further constraints on the film thickness we can use for immersion lithography for very small features," said Dammel.

Slezak said three main factors play a role in pattern collapse: polymer selection, profile control and surface characteristics. Collapsing force is directly proportional to surface tension, which is the reason why surfactant rinses, with their low contact angle, help prevent pattern collapse during the spin develop process. Changing the chemistry of the underlying antireflectant can also modify the pattern collapse margin. "There are two areas for scientific improvement in pattern collapse. One is to improve the interface between the resist and antireflective coating, to make the resist 'stickier,' so that it adheres better; the second is to modify the polymer or the rinse step to control the capillary forces that cause the resist to collapse," said Barclay.

Barclay went on to explain the larger, ongoing issue with pattern collapse: "From a chemistry perspective, the challenge we have is building in a chemical switch to match the aerial image of contrast. As we go down in k1 factor, the aerial image is getting to be more shades of gray rather than black and white, and we need to develop chemistry that can respond to those shades of gray, and we can do that. So we think we have enough contrast to image the 45 nm node and the 32 nm node features, but we have to make sure the patterns stand up."

Multilayer as needed

4. Spin-on hard mask approaches can help address specific patterning issues while offering a cost advantage over CVD hard mask approaches.

Multilayer techniques (Fig. 4 and Table ) involving CVD layers, spin-on organic or inorganic layers help facilitate CD shrinking as well as control over a number of processing problems including LER. "When you implement these spin-on hard mask (bilayer) and dual spin-on hard mask (trilayer) processes, you start checking off many of the problems that have given you headaches, such as poisoning from the substrate, topography control, which helps depth of focus and relieves pressure from CMP, and line edge roughness, which can improve during the etch transfer into the underlayer," Slezak said. Multilayer approaches require tighter coupling between the litho and etch groups to ensure as little line edge roughness and linewidth roughness as possible post-etch.

"Multilayer techniques are being aggressively looked at for the 45 nm node," Slezak said, adding, "In the case of spin-on hard masks, in some cases it is possible to push out the purchase of another CVD tool or a higher-NA litho tool; it can also take pressure off the CMP tool because of the planarization that organic underlayers provide."

For contact hole applications, technologies used to shrink an already established resist feature, using thermal means (reflow) or chemicals (chemical shrink), are finding increasing use. "You have to have good aspect ratio control, meaning that if you have an ellipse-shaped hole, it has to stay an ellipse-shaped hole, and the shrink cannot have much of a temperature dependence," said Slezak. He added that, interestingly, negative-based photoresists, once thought to be necessary for applications like contact holes and isolated trenches, have not taken off. "About two years ago, there was a big push for negative photoresists, but as it turns out, the demand really isn't there."

Many of the early problems with 193 nm resists, such as defectivity and post-exposure bake sensitivity, have been addressed. However, issues of LER and pattern collapse are ongoing issues that must be solved for current and future generations of devices. The good news is that 193 nm resists are ready for the production line, and their modification for immersion lithography, at present, appears to be manageable. Time will tell whether a topcoat for immersion will be required and how serious defectivity issues will be.