Optical lithography to 200 and beyong
02/01/1999
Optical lithography to 2000 and beyond
Pieter "Pete" Burggraaf, Senior Technical Editor
Despite recent predictions of its demise, optical lithography is now poised to sustain semiconductor manufacturing beyond the turn of the millennium. Indeed, there are indications that it can be extended to the application of 157-nm wavelength, two generations beyond today`s leading edge.
Plotting the role of lithography in the evolution toward 256-Gbit memory ICs and 180-million-transistor microprocessors - the 2012 "50-nm technology node" and the limit of SIA`s 1997 National Technology Roadmap for Semiconductors - is "ludicrous," jabs one industry spokesperson. The danger of forecasting a fast-moving target is nowhere more evident than in lithography.
Just 14 months ago, the revised Roadmap presented a scenario in which DUV would play out in 2001, at the 150-nm node. This prompted the hyping of the lack of readiness of "next-generation lithography" (NGL) methods; some scenarios put viable NGL methods out to 2010.
Now, as DUV lithography rounds the corner into the new millennium, it is being re-engineered as "sub-wavelength DUV," where lithography is more intimately tied to IC design. Indeed, DUV has its own roadmap (see Table 1 on p. 32), albeit "hypothetical," which is a product of International SEMATECH`s Project DELPHI. DELPHI has concluded that incremental risks and costs associated with extending optical lithography are seemingly more tolerable than NGL costs, which are not completely defined.
John Petersen of Petersen Advanced Lithography [2], formerly the leader of the project, notes that the DELPHI roadmap outlines 248-nm lithography with the potential to be used through the 130-nm technology node, and 193-nm lithography through the 100-nm node and the beginning of 70 nm, where contact holes will require NGL technology. "Our roadmap was for 1:2 features that have pitches larger than the minimum requirement. If you believe that contacts are the lithography driver (a lot of us think this may not be right) then NGL, either optical or otherwise, will come in at contacts. Production contact-hole resolution is limited to a pitch of wavelength/NA. This means that it will be very difficult to do 1:1 contacts below 180 nm unless we can break this barrier," he says.
One key to the DELPHI scenario is "imaging process integration" - proper selections of mask types, exposure tools, illuminator designs, and resists. "The goal is to make each component of the imaging system work to the best benefit of other components to produce focus-exposure process windows large enough to use in manufacturing," says Petersen. "In addition, photoresist design and the exposure tool has to be used to simplify reticle design as much as possible."
 |
Industry momentum
DUV lithography in production is unfolding in 1999 as the first milestone on the DELPHI roadmap indicates: 180-nm features with chrome on glass reticles having some optical proximity correction (OPC), various uses of conventional illumination or off-axis illumination (OAI), and single-layer resist processing.
Adolph Hunter, product manager at ASML, notes, "US and Asian customers are using our 0.63-NA, 248-nm, 4:1 step-and-scan system in production at 180 nm, achieving depth of focus (DOF) up to 1.5 ?m and critical dimension (CD) uniformities of 13 nm."
Similar process results have been achieved with SVGL`s 248-nm, 0.60-NA Micrascan III at SEMATECH. John Shamaly, VP of corporate marketing, says, "SEMATECH performance results with the Micrascan III is similar to what our customers are seeing in advanced fabs around the world, accelerating their 180-nm production transition."
Indeed, fabs already doing 180-nm lithography with ~0.60 NA, 248-nm commercial tools (see Table 2 on p. 34), as reported or extrapolated from the literature, include TI, Micron, NEC, Samsung, IBM, AMD, and Lucent Technologies.
OPC
This application at 180 nm is the beginning of the re-engineered "integrated" approach to sub-wavelength optical lithography. For example, straightforward OPC is part of the 180-nm node, and even had its roots in late 250-nm applications. Officials at OPC Technology tell Solid State Technology, "We`ve seen many customers adapting existing design rules to address most of their OPC needs at these nodes."
"While a few companies are using a rigorous data set that puts OPC everywhere," notes John Wiesner, senior VP of engineering at Nikon Precision, "everyone assumes that some form of straightforward OPC, such as decorating line ends, is required with binary masks." Ken Rygler, vice president of strategic business at DuPont Photomasks, adds, "We have been doing this for three to four years. OPC assist bar features are ~250 nm at the reticle, which is driving the need and application of dry etch in maskmaking."
A key advance in maskmaking technology, dry etch means reticle quality is much less dependent on chrome thickness variation, for example; subresolution assist features need dimensions with good edge fidelity. While its application is still ramping with the need to solve plasma-induced CD uniformity variations, pinhole defects, etc., technologists at ASML have observed 40-50% reductions in reticle CD uniformity on reticles fabricated with dry etching for 180 nm and below.
OAI
On the difficult side of the 180-nm node, for example isolated lines and contacts, 248-nm OAI is also required. For leading-edge lithography systems, capabilities such as independently selectable NA and sigma, and annular and quadrupole illumination, have been available even on advanced i-line steppers. "Users are familiar with these capabilities to achieve increased process latitudes without loss of intensity and therefore no production penalty, and are increasingly using them routinely in production," says Hunter.
John Nistler at AMD says, "The use of various optical techniques such as OPC, variable NA, variable sigma, OAI, and even phase-shift masks are considered part of the toolbox available for an integrated process. The decision to use these tools is based on technology and equipment availability at the time of insertion. As one example, AMD used PSM with i-line at 365 nm, but used conventional reticles with OPC and 248 nm at 300 nm. Now, reinsertion of PSM with 248 nm is being considered."
 |
PSM
A key point on the DELPHI roadmap is the addition of phase shift masks (PSMs) to extend 248-nm lithography further. Table 1 anticipates progressive use of attenuated PSM (attPSM) with OAI beginning with isolated lines at the 150-nm node, then alternating PSM (altPSM) with conventional illumination, and finally attPSM with OAI possibly stretching 248-nm DUV to 130-nm contact-hole applications. These forms of PSM also play a significant role with the application of 193-nm lithography.
Petersen notes, "There are fabrication issues with PSM, starting with etch selectivity with today`s bilevel attenuator material, i.e., two layers of Mo-Si-N on quartz." The new solutions needed in maskmaking are perhaps the biggest challenge to the extension of 248-nm DUV and the development of 193 nm. "With respect to reticles, we need an attenuator with better etch selectivity to quartz, an ability to make high transmission ternary (HTT) attPSMs, and zero-bias fabrication techniques with <60-nm mask dimension overlay for the different levels. In addition, we need accurate SEM, phase and transmission metrology of live features, hierarchical inspection using the actinic wavelength and phase-and-transmission repair capabilities."
AMD`s Nistler adds that while altPSMs are difficult, the real issue is design. "Unwanted phase transition areas have always been the concern, but proper design of transition areas leads to production quality reticles as AMD has demonstrated with DuPont Photomasks on numerous occasions," he says.
Status of PSM
Maskmakers are correct when they tell us that PSMs are difficult to manufacture and costly, but this technology is increasingly being used in production. "Along with OPC, attPSM and altPSM are very much a reality," says Nistler. Indeed, two years ago at the annual SPIE Microlithography conference, AMD reported its application for the manufacture of fifth- and sixth-generation microprocessors using i-line lithography in full-scale production. While altPSMs may not be widespread, they are also used in production by several DRAM manufacturers.
Rygler clearly expresses maskmakers` arguments on the "costly" side of the equation. "It is the variety of strategies that can be pursued and the accompanying strain on photomask makers. Without the industry adhering to a defined roadmap, different customers will take different approaches as a function of size, technical ability, strategy, product line, available capacity, and technical needs," he says.
With greater application of OPC and particularly PSM, it seems inevitable that the cost of masks will increase as a percentage of IC revenue, perhaps increasing from the current level near 1% to as high as 5%. However, the pay-off for IC manufacturers is increased yield.
One route for PSM
The message is not the feasibility of PSMs, but the route from preliminary use in i-line applications to its ultimate enhancement of DUV lithography, possibly for the 70-nm node. Part of the solution seems to be emerging from companies like Numerical Technologies and the recent acquisition of OPC Technologies by Mentor Graphics. Atul Sharan, VP of marketing at Numerical Technologies, says, "These new solutions lie in judicious combination of OPC and PSM with the entire IC design-to-manufacturing flow, not just applying PSM or OPC to mask-level engineering, even though PSMs are the enabling technology."
Indeed, this is an important part of the "integration" in the DELPHI roadmap. Petersen says, "Our goals mean starting at circuit design and layout to limit geometries so there are no critical features at angles and then using strongly grid-based designs for easy scaling and OPC adjustment. Also, using designs that are altPSM friendly, with no phase conflicts and all critical features can be phase-shifted."
Briefly described, this new design capability of lithography brings software solutions that link IC physical design with OPC and PSM; in essence the software applies phase shift to the design in addition to the mask level (Fig. 1).
Numerical Technologies` efforts are typical for this emerging niche of subwavelength design-to-manufacturing solution suppliers. (IBM has recently revealed similar work and is offering such design tools to its foundry partners.) Working jointly with reticle suppliers DuPont Photomasks and Photronics, physical design house Cadence, and IC library supplier Duet Technologies, Numerical Technologies is developing fully characterized PSM-enhanced physical libraries that bury some of the complexity of phase shifting within these designs. Numerical also enables the intelligent inspection of masks for these phase-shifted and OPC designs by embedding its technology in inspection equipment from KLA-Tencor, Applied Materials, and Zygo.
Hints can be seen in planned presentations for next month`s pivotal SPIE 1999 Microlithography conference. Several papers will report on 248-nm performance using OAI and attPSM to print features down to 100 nm, for example. Another paper will report an innovative application of existing novel technology, extending the resolution of a NA 0.6, 248-nm lens to 120 nm. Phil Ware, director of technical marketing for Canon, says, "Widespread use of such methods could delay the need for 193 nm for several years, with its insertion closer to the 100-nm node. Then, these same exposure techniques can be used with both 193 nm and 157 nm, potentially pushing optical lithography to 60 nm or 70 nm."
Status of 193-nm lithography
It is interesting how, with focused investigation such as that delivered by Project DELPHI, the industry`s argument about lithography has switched so dramatically. Now, for the 130-nm technology node, instead of debating the entry of NGL, the emphasis is on where the industry will switch from 248- to 193-nm DUV; milestones are identified as "work required" or "hard work required," indicating no "show stoppers."
In other words, as the DELPHI roadmap indicates, many experts think that the 150-nm technology node and even the 130- and some applications at 100-nm nodes are still the territory of 248-nm DUV with full application of OPC, PSM, and OAI technologies. But achieving optical lithography at this level will clearly be an economic issue, and economic comparisons will be measured against the costs of 193-nm DUV (Fig. 2), not NGL. This is not to say that the DUV route will be straightforward, but the path seems brighter today than a year ago.
As of 1Q99, first-generation, 193-nm production systems are being installed on development lines. In addition to the reticle challenges discussed above, the systems have a full slate of issues to address:
The cost of ownership (COO) of commercial systems (see Table 2) is very high at 2.5-3.0? that of 248-nm lithography used at 180 nm. But COO improves with time; the natural development progression is to improve throughput and laser power, plus resist process maturity, thus driving yield up and COO down. Today`s COO compares, and perhaps confuses, 193-nm process development tools with 248-nm production tools. It is expected that 193-nm COO will come down to ~1.5? 248 nm in the next few years, when there is a drive toward volume production. Comparison must properly consider that 248-nm COO will be driven up as it is enhanced with OPC and PSM.
 |
Figure 1. Resist images, dual polysilicon exposure on oxide patterned wafers, used in DSP IC fabrication demonstrate the capabilities of 0.53 NA, 248-nm DUV with design level application of PSM from Numerical Technologies. (Source: Lucent Bell Labs)
 |
Figure 2. An AFM image of an etched 74-nm, 100-nm thick polysilicon gate from actual device fabrication with 193-nm DUV lithography at Lucent Technologies, demonstrating manufacturability of Lucent-Arch Chemicals SL 193-nm resist platform. Inset shows device lot linewidth control with 193-nm exposure.
An infrastructure for volume manufacturing of CaF2, the required lens material, has to develop. This is certainly part of bringing COO down. Progress is being made on both verification of material quality and supply. For example, Schott Glass has reported lens blanks up to 233-mm dia. with <2 ppm homogeneity, roughly twice that required for 193-nm lenses. According to Richard George, "ASML has an adequate supply of CaF2, which we also use in illuminator optics of our 248-nm system (model 700B), for all 1999 shipments of 193-nm systems. We are working with our suppliers to ramp up for 2000." John Bruning, president ofTropel, adds, "The optical material lifetime for 193 nm is now well understood and not the show stopper it was once."
Historically, resist development tends to lag tool development, but 193-nm resist solutions are coming on strong. The biggest change with 193-nm lithography is the need to abandon
phenolic polymer based novolak and hydroxystyrene resists with their aqueous base, good adherence, and excellent etch resistance. A worldwide effort has occurred over the last five years, creating new classes of resists. Bob Allen at IBM`s Almaden Research Center notes, "Much of today`s emphasis is on the cyclic olefin chemical platform, which affords an excellent way to transition the industry towards an entirely new chemical platform." Cyclic olefin polymer materials offer the potential for superior etch properties, surpassing the most robust 248-nm resists. In terms of etch resistance for polysilicon and oxide, cyclic olefin polymer materials approach if not surpass that of novolak-based mid-UV resists.
Om Nalamasu, technical manager for optical lithography and imaging materials research at Lucent`s Bell Labs, notes that for 193-nm lithography, the combination of etch stability and reflectivity control is particularly important (Fig. 3) because of the relatively thin resist thicknesses (0.6-0.4 ?m) required for 150- and 130-nm design rules. Nalamasu says, "Because of the inherent `paradigm shift` in technology with 193-nm DUV resist development, many of the issues have been addressed up front and this is speeding up progress."
Murrae Bowden, R&D director for Olin Microelectronic Materials, says, "These concepts are being actively exploited by resist vendors with first-generation products expected early in 1999, along with the first tools."
It seems that major manufacturers, particularly those with good infrastructures, have taken full grasp of 193-nm DUV devel opment. Wiesner notes, "During 1999, with 193-nm systems shipping into fabs for process development, the remaining
questions will be addressed. Given the fact that the node here is roughly 2001-2002, fabs have at least a year to work on issues. If they make a buy decision in 2000-2001, they will still have time to have systems for the 130-nm node."
 |
Figure 3. Researchers at Lucent Bell Labs have used an innovative multilayer ARC CVD dielectric to control substrate reflectivity during 193-nm lithography with PSM. SEMs compare 80-nm lines, in 0.2-?m steps of focus, with reflectivity control (top row) and without (bottom row).
157-nm?
With so much emphasis and remaining potential for 248- and 193-nm DUV, it seems that the relatively narrow production process window for 157-nm DUV is simply too narrow to make economic sense, as the argument often goes. In addition, 157 nm has some seemingly overriding "show stoppers," for example, the high coefficient of thermal expansion of CaF2 reticles makes them difficult to build and use.
Just when the industry again projects the ultimate limit to optical lithography, however, it becomes evident that the fundamental research from which lithography rises is not static; nor is the calculation of the economic advantages of the various possibilities for the future. Ultimately, if the narrow window of 157 nm makes economic sense from the standpoint of DOF and process latitude, then there are those in the industry who will do it. Indeed, the list of 157-nm work is growing; consider, for example, that:
Recent work from Corning Glass shows some promise with fluorine-doped fused silica; it has enough transmission to make it practical at least for consideration as an alternative to CaF2.
Intel, a major supporter of NGL EUV technology, sees 157 nm as possibly more cost-effective than any NGL methods. It is at least funding a nine-month study with MIT Lincoln Labs to identify potential 157-nm show stoppers.
Late in 1998, SVGL announced its intention to develop a 157-nm lithography system to fill in the gaps between 193 nm and EUV, which the company is also developing for 70 nm and beyond.
Ware says, "157 nm is a prime example of just how erratic and unreliable the SIA Roadmap has become. The death knell for optics has been sounded so many times that it is almost impossible to gauge when the end will actually come."
Acknowledgment
The author thanks all who contributed dissertations or just passing thoughts during the development and review of this article. Space does not allow for individual recognition.
References
1. J.S. Petersen, et al., "Assessment of a Hypothetical Roadmap that Extends Optical Lithography Through the 70-nm Technology Node," Proceedings of SPIE, 3546-32, Sept. 1998
2. Petersen Advanced Lithography is an autonomous division of FINLE Technologies. The division`s mission is "to help companies make the transition from the k1 of today to the small k1 of tomorrow." For more information, contact: [email protected]
PIETER "PETE" BURGGRAAF has 25 years of experience in the semiconductor industry, including work at Motorola, Siemens, and ASM. He can be reached at 875 S. Yucca Drive, Wickenburg, AZ 85390; ph/fax 520/684-1265, e-mail [email protected]