Fine Tuning Today's Photoresists

This is an exciting time in the microlithography industry. The promise of immersion lithography extending the 193 nm (ArF) toolset out further than ever imagined has the potential to completely disrupt the position of its once chosen successor, 157 nm (F2) technology. If 193 nm immersion delivers — and there are more people who think it will than think it will not — the ArF scanners could be used through the 45 nm device node, giving the industry time to recoup the investment it made in going from 248 nm (KrF) lithography to 193 nm lithography.
 
"It is clear from the International SEMATECH task force that immersion is not going to require a reinvention of resist platform, but a modification of the basic methacrylate or acrylate platforms to be successful," said Ralph Dammel, director of technology at Clariant Corp., AZ Electronic Materials (Somerville, N.J.). Even so, there is much work ahead to address the practical issues, including outgassing.

"Most of the experimental work in immersion has employed a top coat, where potential interactions between resist and immersion liquid could be masked," said Plamen Tzviatkov, business manager for advanced technology at Arch Microelectronic Materials (North Kingstown, R.I.). "Further studies among supply chain partners are a priority to conclude that existing platforms are capable for commercial lithography."

Rohm and Haas Electronic Materials Microelectronic Technologies (formerly Shipley, Marlborough, Mass.) has an immersion 193 nm lithography resist effort underway. Leo Linehan, research director, agrees that collaboration with partners will be key to solving the materials issues associated with immersion.

In the end, if immersion does prove to be a more cost-effective route than a transition to 157 nm lithography, it will be good news for an industry hard pressed to make its fab investments more profitable. Indeed, Peter Silverman of Intel noted that lithography cost-of-ownership increases are most closely tied to exposure source changes (i-line to 248 nm to 193 nm) because, once these new tool platforms are adopted, they are quickly made more cost-effective through incremental increases in throughput.¹

1. Beyond the traditional function of reflectivity control, spin-on antireflective coatings can also fill vias, provide planarization, or suppress pattern collapse in densely spaced lines. (Source: Brewer Science)

2. Using a 300 mm TEL coat-and-develop track, researchers are able to develop and test products specifically adapted to customers’ substrates. (Source: Clariant AZ)

For photoresist manufacturers, many of the 248 nm processes are mature, but they are continually being extended — in conjunction with high-NA scanners, advanced masks, track systems and antireflective coatings (Fig. 1). The 248 and 193 nm toolsets also happened to coincide with the 300 mm wafer transition (Fig. 2 ). The current limit for 248 nm is somewhere around 90 nm for isolated lines or 120 nm for contact holes. "Due to k₁ factors being pushed so low, you can image 1:1 lines and spaces of 90 nm, but you're using such a specific aerial image that if your mask layout has any variation, you can only print that one pitch and struggle to print, for instance, an isolated space and an isolated line on the same mask level," explained Mark Slezak, technical manager of lithography for JSR Micro (Sunnyvale, Calif.). He defines the ability to manage different pitches of specific CDs across a wafer with overlapping exposure latitude and depth of focus (DOF) as common process latitude.

Slezak estimates that ArF resists are ~80% as mature as KrF resists. But, of course, both are evolving and maturing. Linehan said, "Our customers tell us that they have yet to find a 193 nm resist solution that meets all their needs. For now, leading device manufacturers are making due with what's available." The company is taking a multipronged approach, with acrylates and methacrylates, VEMA and silicon bilayer technologies.

To have reached this stage, the ArF resists, the leading platform of which is acrylate- or methacrylate-based, has reached certain acceptable values for four key criteria: line-edge roughness (LER), defectivity, post-exposure bake (PEB) sensitivity and pattern collapse. Of course, the resist also has to afford adequate etch resistance and transparency to 193 nm light, but these are minimum requirements that must be met in the early stages of resist development. In production, the resist must deliver a good process window (Fig. 3 ) to achieve high yields.

3. In this metal application, 120 nm of advanced ArF resist achieves large exposure latitude (15.8%) and good depth of focus (>0.85 µm), while maintaining low line-edge roughness. (Source: JSR Micro)

LER is a very complicated problem, and there are numerous theories on how to solve LER from a chemical standpoint. An improvement of only 2 nm can be significant, depending on the process (Fig. 4 ). LER problems originate from the ability to control the diffusion of acid, and can be affected by modifying the molecular weight of the polymer, using an advanced photoacid generator (PAG), or trying new additives such as low-molecular-weight compounds, according to Slezak.

4. Cross-sectional SEMs show an improvement in line-edge roughness for 100 nm lines/120 nm spaces — from the AZ AX 1120P resist, with 6.6 nm roughness average, to the AZ EXP T8520 resist, with 4.2 roughness average. (Source: Clariant AZ)

"Line-edge roughness is a persistent problem because we keep pushing our k₁ factors lower, so LER will continue to be an ongoing issue," said Ken Bell, global product manager of 248 and 193 nm resists at Clariant AZ. For resists <2000 Å thick, LER also tends to get worse as the resist is made thinner. To confound matters, the way that LER is measured differs from one customer to the next, according to Bell, making LER even more challenging to address in the industry.

Defectivity is key, of course, because it is tied to device yield. This doesn't refer simply to particle defects in the dispensed resist or even the known interaction between airborne amines and the acid components in photoresist that cause footing or T-topping (resist poisoning). Defectivity in this context includes chemical- or process-related defects resulting from interactions between the substrate, resist, antireflective coating (if present), and hard mask (if present). "With our joint-venture partner, FujiFilm-Arch, we designed the 193 nm chemistry to be homogenous and inherently capable of delivering low defectivity along with excellent lithographic performance," Tzviatkov explained. Process tweaking of an established platform, which may have worked to reduce defectivity on previous-generation resists, has not proved effective with the current 193 nm materials, he added.

Tzviatkov went on to describe the shrinking degrees of freedom available when designing resists. "These performance criteria — of process window, pattern collapse, etch resistance, defectivity, etc. — were known and solved issues at 248 nm. Though the requirements remain the same, the tolerance level is more demanding, with defectivity driven to 0.5 defects/cm²."

Likewise, PEB sensitivity is extremely important for 193 nm resists, whereas it was important, but had a less dramatic effect on critical dimensions for 248 nm resists. PEB non-uniformities affect CDs across the wafer, and wafer to wafer. "Very good lithographically performing resists come in at below 1 nm/°C, and device manufacturers typically reduce PEB non-uniformities by cherry-picking the most uniform hot plate to go into the post-exposure bake station on the track," Bell explained. "At the same time, you can have a total range of 0.75°C across three hot plates, which doesn't sound like much, but it causes a considerable CD variation, so the user has to adjust the exposure or bake temperature to get uniform CDs."

Another important challenge with 193 nm resists is pattern collapse, where the capillary effect causes high-aspect-ratio (>3:1) structures to fall over during the development process. It can be a problem for isolated or dense lines. Improving the adhesion of the resist to the underlying antireflective coating can alleviate pattern collapse issues. Alternatively, the photoresist itself can be modified so that it has more structural integrity. Clariant, Tokyo Electron Ltd. (TEL, Tokyo) and Tokyo Electron America (TEA, Austin, Texas) recently developed a surfactant rinse solution and process for 90 nm production that is designed to eliminate pattern collapse and may be extendible to 55 nm patterns.

Antireflective coatings traditionally control substrate reflectivity for better CD control and a wider process window. They can also be used to improve the adhesion between resists and substrate. In dual-damascene applications, bottom antireflective coatings additionally provide via filling (Fig. 5 ) and also planarization. Their ability to be both conformal and planarizing distinguishes them from CVD antireflective coatings, making them compatible with a wider range of features, including vias, contact holes, lines and spaces, etc.

5. Bottom antireflective coating over isolated (top) and dense (bottom) vias shows uniform filling independent of bias following development (right). (Source: Brewer Science)

"Organic BARCs not only work by destructive interference ( 'n' dependent) but also by absorbing exposure radiation ( 'k' imaginary refractive index), and thus the process has a much wider thickness latitude and process window compared to inorganic BARCs, which are very sensitive to thickness variations since their refractive index is a function of thickness," said Shree Deshpande, global marketing manager of the ARC business at Brewer Science Inc. (Rolla, Mo.). A further advantage is the ability to rework the film by simply removing it with a standard develop process, or removing it after exposure by either dry plasma etch or resist ashing process.

Rohm and Haas recently introduced a tunable, fast-etching 193 nm antireflectant that is gaining use in manufacturing. "Based on a polyester technology, this ARC offers tunable optical properties and significantly faster etch rate than the well established acrylate-based ARCs," said Rick Hemond, marketing director for lithography.

ARCs are being applied, essentially, whenever the resist budget for pattern transfer is limited, according to Deshpande. For instance, ion implant layers, often considered non-critical in terms of CD, are requiring greater control of reflectivity for today's KrF resists. In this application, the BARC is removed in the same develop step as the resist. A final, relatively new application for antireflective coatings is as a barrier between low-k dielectric films and the photoresist, to prevent resist poisoning.

Because of their shape and the poor aerial image created in projection lithography, contact holes by nature are difficult to resolve. Companies have developed various chemical approaches to resolve sub-100 nm contacts. For instance, Clariant's RELACS process uses an overcoat that reacts with the resist to shrink the contact hole.

One of the reasons acrylate resist platforms have been the most successful to date is the combination of great resolving capabilities and the flexibility to manipulate the resist components to attain application-specific properties. Patterning dense contact holes requires a different formulation than that used to pattern isolated gates or dense line arrays or high-AR contacts. For more advanced applications, bilayer resist schemes are gaining favor.

The complexity of thin-film processing has led to more customized photoresist solutions, Slezak said. "Pretty much every customer has their own back-end integration scheme. So we have to know what hard mask or sacrificial layer they are using and what the substrate is in order to know, for instance, if we need a slightly more acidic photoacid generator or a slightly more hydrophobic quencher, so that we can get the best resist-to-substrate interaction, using the best material for that application."

One example of a customized solution is the development of an organosiloxane underlayer that adds process latitude to the etch process to make a dual-damascene process more production worthy.² In this case, a dyed organosiloxane film was used as the underlayer to improve etch selectivity and also as antireflectant.

An approach being used in 248 and 193 nm dual-damascene applications is the thin imaging system (TIS) offered by Arch. This two-layer system was designed to mimic a single-layer resist on antireflective coating in terms of process complexity. However, like other bilayer approaches, the top layer contains silicon and exclusively performs the patterning in an ultrathin layer, while the organic underlayer provides antireflective control, high etch selectivity and profile integrity. This TIS process (Fig. 6 ) is also being developed for 157 nm lithography. "One of the unique advantages of this combination is the very low PEB sensitivity of 0.5 nm/°C relative to a single thick layer resist," Tzviatkov said. "We have demonstrated 50 nm contact holes using this technology."

6. 50 nm contact holes were patterned using a thin imaging layer (IL) over an etch-resistant underlayer (UL). Reflow facilitates a final aspect ratio of 9:1 after oxide etch. (Source: Arch Microelectronic Materials)

In the transition from 248 to 193 nm resists, ARCs changed considerably to achieve greater etch selectivity to the resist, superior resist compatibility and better CD control, Deshpande explained. With the hyper-NA (>0.85) scanner systems, the materials will have to accommodate for polarization effects caused by differences in refractive indices of air and resist. "Also, etch rates and defectivity will be key issues for organic BARC designs for high-NA ArF lithography."

"The polarization effects are going to become more severe as we move to scanners with NA greater than 0.85," Clariant's Dammel explained. "These hyper-NA tools will require extremely good suppression of substrate reflectivity, so will see more elaborate antireflectivity schemes in the future such as graded ARCs, dual ARCs, etc."

Hard masks are playing an equally important role. Resist films must get thinner to achieve better resolution at a given optical transparency, so sufficient etch resistance is more and more difficult to attain without a hard mask. "As technology goes to smaller linewidths, for 157 nm or whatever comes after 193 nm dry, the aspect ratio of the resist is always going to be such that there will be a need for some type of hard mask scheme to transfer patterns, simply because the resist is getting thinner and thinner," said Andy Romano, global product manager of advanced products at Clariant.

The greatest challenges in developing 157 nm photoresists currently include increasing the transparency and etch resistance of the formulations, followed by tackling the same challenges of pattern collapse, LER, defectivity and PEB sensitivity. Fluorocarbon polymers (a single-layer solution) are performing slightly better than SSQ-based 157 nm resists (bilayer solution) — offering absorbance of 1/µm vs. 2/µm, respectively.

A major concern with silicon-containing resists is silicon outgassing and deposition of SiO₂ on the optics. The measurement and subsequent elimination of silicon outgassing through concurrent resist materials design is the subject of intense research today.³

Amine contamination issues with 157 nm are an order of magnitude greater than for 193 nm materials, according to Dammel. The primary concern is ammonia, Romano said, but "we're not sure that amines are the only thing that could have an impact on 157 nm chemically amplified system at present." Other concerns may be organics or oxides, which might need to be kept out of the light path.

The struggle to find organic molecules transparent enough to begin to make good photoresists suggests that it should not be difficult to find absorbing films for the underlying antireflective coating. "Contrary to common belief, it is very challenging to design 157 nm BARCs with the right balance of optical properties with reasonable film thickness, etch properties, adhesion properties and photoresist compatibility," Deshpande said. The BARCs for 157 nm will have to be thinner than those used for 193 nm (~100 vs. ~200 nm) and provide faster etching. With SSQ-based resists, organic BARCs as well as spin-on hard masks will be needed to achieve compatibility with the silicon-containing resist.

The emergence of immersion lithography for 193 nm could throw off the timeline for 157 nm by three or four device generations. "For a while there, DRAM manufacturers wanted to quickly go to 157 nm, while logic manufacturers were pressing for 193 nm immersion. Now DRAM manufacturers seem to be seriously considering immersion, so that makes one technology for the industry," Dammel said. "Certainly immersion has moved to the forefront everywhere." He warned, however, that the industry does not yet know enough about 193 nm immersion technology to know whether there will be serious showstoppers.