Deciphering Extreme Ultraviolet Lithography

By: Katherine Derbyshire, Semiconductor Online.

The era of optical lithography is coming to an end. While researchers recently printed features as small as 100 nm with 248 nm light, and 70 nm features with 193 nm light (See "Optical Lithography Lives On") the optical proximity correction and phase shift mask techniques required for sub-wavelength features reduce the process latitude. At some point, optical exposure tools will be unable to print critical features at an acceptable cost or with acceptable control.

The four leading candidates to succeed optical lithography are extreme ultraviolet (EUV), proximity x-ray, scattering with angular limitation projection electron-beam lithography (SCALPEL), and ion beam projection lithography. Of these, EUV gained momentum earlier this month when the newly formed US Advanced Lithography LLC (San Jose, CA) entered an agreement with the EUV LLC (Santa Clara, CA, consisting of Motorola, Intel, and AMD) to develop and commercialize EUV exposure tools (See "USAL, EUV LLC Team for EUV Lithography Development").

EUV lithography uses 13 nm radiation, close to the 10 nm used by x-ray proximity lithography. In both cases, the short wavelength greatly increases the resolution budget. For sub-100 nm features, both technologies should have larger process windows with more exposure latitude than optical wavelengths. Unfortunately, all known optical materials absorb strongly at these wavelengths, so conventional refractive optics cannot be used. The key difference between the two technologies lies in their solutions to the problem.

Proximity x-ray avoids reduction optics entirely, using 1× masks in close proximity to the wafer. Resolution depends on careful control of the gap. The success of proximity x-ray will depend largely on the availability of affordable 1× masks.

EUV avoids these challenges by using 4× reduction masks, but in doing so confronts the need for reduction optics using reflective elements. Perfect mirrors are not available (See "Optical Coatings," below); the optics themselves will absorb some radiation. To maximize throughput, an EUV system must both minimize the number of elements and maximize the reflectance of each.

Optical design
The EUV LLC is funding research at Lawrence Livermore (Livermore, CA), Sandia (Albuquerque, NM), and Lawrence Berkeley (Berkeley, CA) National Laboratories to demonstrate the basic technologies required in an EUV exposure tool. As part of this project, the laboratories are building an alpha-class tool, the Engineering Test Stand (ETS), scheduled for completion in late 1999 or early 2000. Don Sweeney and coworkers from Lawrence Livermore National Laboratory described one potential optical design for the ETS at the SPIE Microlithography meeting in February. As Sweeney explained, the basic specifications for the ETS camera are:

• 100 nm critical dimension (70 nm for isolated features) based on a numerical aperture (NA) of 0.1, K₁=0.77, and coherence factor s=0.7.
• Ring-field imaging with a cord-length of 26 mm at the wafer.
• Fully stationary imaging with better than 1% intensity uniformity.
• Depth of focus of ±0.5 mm.
• Reduction factor of 4:1 with residual magnification control of ±20 ppm and magnification resolution of 0.1 ppm.
• Telecentric imaging at the wafer.
  
• Total dynamic distortion of less than 5 nm over the full field.

The group's proposed design meets the specification for stationary imaging, but minimizes distortion by scanning the mask and wafer in synchrony. As with an optical step and scan system, this approach allows use of a smaller, well-corrected exposure field, but requires precise control of scan velocities. The design uses a condenser with six parallel channels and four imaging mirrors.

Basic layout of the ETS optics. Components are roughly to scale, but the shapes are more complex than shown. C4, a toroid, images the condenser pupil onto the projection optics entrance pupil. M3 is a sphere, while the other three projection mirrors are aspheric.

Each of the six condenser channels can be divided into three functional groups: collector mirror (C1), roof mirror pairs for field rotation (C2 and C3), and a field mirror (C4) common to all six channels. The six channels together collect 30% of the radiation emitted by the source. With such low source efficiency, losses in the optics can substantially reduce the radiation impinging on the wafer.

In order to achieve the design specifications, the mirrors must be fabricated with an RMS figure error of less than 0.25 nm and better than 0.2 nm RMS roughness, simultaneously. These requirements are more severe than those faced by EUV telescopes.

Optical coatings
Between the source and the wafer, the EUV radiation reflects from nine different surfaces. It strikes seven of these (the mask, M1-M4, C1, and C3) with near-normal incidence, which maximizes absorption. The system throughput is therefore a strong function of the mirror reflectance. As Claude Montcalm and coworkers at Lawrence Livermore National Laboratory explained, in work presented at the SPIE Microlithography Meeting, the mirrors use alternating coatings of either molybdenum and silicon (Mo/Si) or molybdenum and beryllium (Mo/Be). The interfaces between layers reflect EUV radiation, while the layers themselves absorb it. The thickness of the layers is determined by the tradeoff between maximizing the constructive interference of the beams reflected at each interface and minimizing the overall absorption. Achieving acceptable throughput and cost of ownership will require 70% or higher reflectance.

The ETS design, Sweeney wrote, specifies identical multilayer coatings of Mo(2.92 nm)/Si(4.0 nm) × 40 for M1, M2, and M4, while M3 requires Mo(2.92 nm)/Si(4.12 nm) × 40. Montcalm's group fabricated films with 40 Mo(2.8 nm)/Si(4.1 nm) bilayers, as well as the equivalent Mo(2.3 nm)/Be(3.4 nm) stack. Montcalm reported 67.5% reflectance for the Mo/Si multilayers at 13.4 nm and 70.2% reflectance for the Mo/Be multilayers at 11.4 nm.

High reflectance requires careful control of substrate quality, layer thicknesses, multilayer materials, interface quality, and surface termination. Control of the surface figure also demands accurate knowledge of the multilayer stress. In addition, the reflectance peaks of all the mirrors in a system must be aligned to within 0.05 nm, meaning that the bilayer thickness of each element must be controlled to within an accuracy of 0.025 nm. While Montcalm and coworkers were able to achieve this level of control on flat 100 mm substrates, optics for an EUV lithography system will be up to 270 mm in diameter. The researchers plan to extend the sputter deposition technique to larger substrates and more complex shapes next.
Back to "Perfect mirrors..."

Masks, sources, and other issues
Masks for EUV lithography will require similar multilayer reflective coatings, with an additional metal layer containing the pattern. They will also require unprecedented care during fabrication, as no techniques for repair of damaged bilayers are known or expected.

Design of the radiation source is likewise an open question. The most likely approach uses a laser to create a plasma at a metal target. The plasma emits EUV radiation. According to Andrew Hawryluk and co-authors from Ultratech Stepper (San Jose, CA), writing in the August 1997 issue of Solid State Technology, important source design issues include:

• efficient conversion from laser to EUV radiation;
• need for a high-average power laser; and
• need for a debris source.

The first condenser lens (C1) looks directly at the source, so debris ablated from the target could degrade the optic. Art Zafiropoulo, president and CEO of Ultratech Stepper and acting president of USAL, calls control of particulates in the optics one of the most serious obstacles to EUV lithography.

While this article reviews many difficulties and open questions facing EUV lithography, it's important to keep them in perspective. Post-optical lithography is not needed in production until the 100 nm generation, expected in 2006, according to the 1997 SIA Roadmap. A lot of research is needed, but the time and resources to do it seem to be available.