Issue



Technology News


05/01/2004







Software uses lens aberrations to enhance images

Lithographers have fought long and hard to diminish the unwanted effects of lens aberrations on optical imaging in advanced semiconductor processes, but Nikon Corp. is using software to reverse this situation. The company is selectively using aberrations for enhanced printing of critical features on an application-by-application basis.

Nikon's aberration optimization software could allow wafer fab operators to tune and "detune" projection lenses on scanners for increased product yields and higher-speed devices.

"If you are doing dense equal lines and spaces, one sort of image might be optimal," explains Gene Fuller, principal engineer at Nikon Precision Inc. "But for isolated features, it might be useful to apply a little bit of stigmatism in the lens in order to get the absolute best imaging of a single critical isolated line. There are many options like that."


TAO software applies linewidth abnormality (LWA) measurements to a lens (top) and then predicts the results of aberration tuning (middle). Measured values of adjustments show the simulation's prediction almost identically matched the results at 150nm LWA (bottom). (Source: Nikon Corp.)
Click here to enlarge image

The new Technology for Aberration Optimization (TAO) software is being deployed this year in production fabs to enable IC makers to extend the use of field-tunable projection lenses in Nikon's 248nm KrF and 193nm ArF step-and-scan tools. The TAO software, running on a PC, quickly simulates and predicts the results of lens adjustments for aberration tuning and desired critical features from patterns on photomasks.

The genesis of TAO came from Nikon's own internal needs to simulate and precisely adjust lens-cell setting in scanners to the lowest possible aberration levels. Nikon's KrF and ArF scanners have incorporated field-tunable projection lenses for a number of years, but the rollout of TAO software extends the ability to automate and customize aberration tuning for specific illumination conditions, reticle patterns, and applications in production fabs.

"What's new in TAO is an integrated approach, including the software, the aberration analysis, understanding of pattern requirements, and tying all of this together with the mechanics of the tools to make the adjustments that are required," Fuller says. "It allows the end users to customize the image so that they get the best depth of focus [DOF] and best exposure latitude."

Nikon says TAO will enable optimization of scanner imaging capabilities on an application-by-application basis by incorporating wave-front measurements as input and then allowing tool users to specify which patterns are the most critical in a mask layer. The software simulates imaging performance before optimization using Zernike calculations. Nikon says the TAO software substantially reduces the time it takes to tune optical lens groups in a projection system, with the final output for the best "lens prescription" being sent to scanners for automatic positioning.

In addition to potentially enhancing the imaging of critical device features — such as transistor gates in logic or repetitive lines in memory arrays — the aberration optimization software could allow chipmakers to extend the use of scanners to next-generation processes and avoid the retooling of fabs, according to Nikon. The concept is somewhat akin to resolution enhancement techniques (RET) used to support subwavelength lithography.

"I don't believe anyone will initially adjust the lens for every mask layer, but this is a capability that could be used more and more in the future, just like OPC [optical proximity correction] started out only on the most critical levels and expanded," Fuller speculates. "Aberration control will most likely follow the same trend."

In aberration tuning, various groups of lens elements are adjusted with nanometer movements. "Because the lens designs are well computed, we know precisely what will happen to various aberrations, distortions, stigmatisms, or Coma," Fuller explains. "TAO closes the loop between knowing exactly how the lens will perform and the kind of pattern desired on the wafer."

The overall benefits of leveraging aberrations in scanners will be tested by production fabs this year for the first time. No estimates are available yet for yield improvements or device speed enhancements, but Fuller believes the potential is significant. "If we can tighten up the CD control by one nanometer — as a generic guess — we could probably improve the speed sort of microprocessors by several percentage points, which would mean fewer parts in the bargain bucket," he says, referring to higher average selling prices for faster processors.

The TAO functionality requires a license from Nikon and company support for initial simulations. The capability is being deployed on Nikon's NSR-S205C (248nm) and NSR-S305B (193nm) scanners as well as newer-generation tools and SF200 models. —J.R.L.


Immersion litho hot topic at SPIE

Dominating many of the presentations and discussions at this year's Microlithography Symposium and Exhibition was immersion lithography. Thus far, no "show-stopper" has emerged that would prevent further development and eventual production use of immersion technology. While immersion development has come a long way in the past year, several issues remain to be resolved.

Metrology challenges

While the hopes of the microlithography community are clearly focused on liquid immersion techniques, it was clear at SPIE that dry lithography would remain the workhorse through the next expansion, and that immersion would not solve the problems in ancillary fields like metrology.

Harry Levinson of AMD reviewed the methods used in the four past waves of semiconductor innovation in his keynote address to the metrology, inspection, and process control conference and outlined options for the fifth wave, which would take the industry from 100nm to 40nm.

According to Levinson, metrologists have suffered from an 11-year acceleration of the Roadmap, worse than any other field and taking them into the regime where no solutions were known. In particular, there is no known way to achieve the Roadmap's current 0.6nm precision demand for line edge roughness (LER), etc.

Defect metrology also is at an impasse, with new tools needed capable of surveying wafers for 50nm defects at a rate of 100s of cm2/hr. New capabilities would also be needed for next generation lithography (NGL) systems, when and if they appeared.

David Joy of the University of Tennessee, Knoxville, TN, described a method to overcome the inevitable distortions due to charging in SEM metrology of masks by filling the inspection chamber with gas at 8torr or so.

Images made using back-scattered electrons also lacked the edge-enhancing artifacts of back-scattered electrons, making it easier to determine the actual edge location. However, the low speed of the current low vacuum backscattering detectors slowed data collection.

Another advantage of low vacuum SEMs for mask inspection is that local injection of a suitable gas (WF6 or XeF2) could effect repairs when excited by the electron beam, according to Joy.

Problems with resists

While immersion lithography at 193nm was not expected to pose serious challenges for those well-developed resists, formulations for other wavelengths still seemed to be a problem.

In particular, Shigeo Irie of Selete reported that fluorinated resists tended to outgas when illuminated with 157nm radiation, depositing up to 6nm of material on the last element of a projection lens after only 4 wafers (30J/cm2 total dose). That reduced the transmission by 5% and inevitably added to the flare.

Experiments with the Canon FPA5800-FS1 full-field 157nm scanner reported by Akihiko Otoguru, also of Selete, showed 6nm line edge roughness for 60nm lines when exposed with an alternating PSM.

Such small resist lines are difficult to develop under the best of circumstances, in part because capillary forces tend to pull them over during the drying step after aqueous development.

While surfactants added to the rinse water to reduce surface tension promise some improvement, elimination of resist pattern collapse probably requires complete elimination of the liquid surface through some supercritical fluid transition.

Click here to enlarge image

The first Frits Zernike Award for Microlithography was presented at SPIE to Burn Lin, second from left, from Taiwan Semiconductor Manufacturing. Presenting the award to Lin were, left to right: Anthony Yen, Cymer Inc.; Chris Mack, KLA-Tencor; and Marylyn Hoy Bennett, Texas Instruments. The award, sponsored by ASML, Canon, and Cymer and accompanied by a cash award, is given out for distinguished, innovative research in the field of microlithography. Lin has long been an avid proponent of immersion lithography.

A new scheme employing heavy fluids such as SF6 and C2HF5 instead of CO2 was presented by Hideo Namatsu of NTT Basic Research labs (Japan).

Polarization anomalies

Liquid immersion promises to reduce the effective wavelength of 193nm lithography to 134nm, but there are several subtle effects that must be considered if optical lithography is really to achieve sub-50nm resolution.

Prominent among them are polarization anomalies, and Donis Flagello of ASML, Mesa, AZ, described the technological choices that may have to be made in order to print real circuit features (where lines have ends, etc.) in that regime.

For example, while y polarization was best for x-oriented 45nm lines, the minimizing line-end retraction required azimuthal illumination! Because of such trade-offs, Flagello advocated a double exposure strategy for difficult cases and showed how a typical SRAM "brick" pattern could be printed in that way with k1<0.15, roughly 30nm!

The contact layer will soon become the most difficult chip pattern and several presentations described innovative ways of printing them optically. Robert Socha of ASML Masktools (Santa Clara, CA) elaborated upon the interference mapping technique optimizing placement of nonprinting assist features.

Other proposed contact hole methods included assist-ring-only structures, double exposure with line space masks (but with the resist hard-baked in between to yield a waffle-like image), and further development of the vortex via process, which employs negative resist.

Sony presented low-energy electron projection lithography (LEEPL) as a nonoptical option for contacts.

Mask complexity is clearly not going to be reduced any time soon. In fact, there were several proposals to add another tone to the mask to allow gray-scale imaging and unresolved opaque "zebra" patterns in CPL to optimize exposure. — M.D.L.