Assessing the challenges of EUV lithography

02/01/2007

Kurt Ronse, IMEC, Leuven, Belgium

It is expected that single-exposure 193nm immersion lithography will reach the ultimate physical limits at the 32nm half-pitch technology node. Extreme-UV lithography (EUVL) and very-high-NA 193nm immersion lithography are both high on the agenda of the semiconductor industry to extend scaling to the 32nm half-pitch node. The attractiveness of EUVL however, is without any doubt, its extendibility beyond 32nm half-pitch.

EUVL will cause many headaches before being introduced in chip manufacturing somewhere between 2009 and 2012. Its extremely small wavelength (13.5nm) implies absorption of light by any material, extremely high photon energy destroying chemical bonds, and the unwanted imaging of each small defect and particulate.

The first challenge of EUV lithography is the development of a suitable light source that emits at the 13.5nm wavelength. The essence of the problem is to make a light source that is sufficiently bright, but doesn’t damage the optics, two properties that have not yet been demonstrated together. Today, EUV-source development focuses on laser-produced plasma and discharge-produced plasma sources. High-energy ions and atoms (either Xe or Sn), created in the high-energy plasma in the light source, and debris collide with the mirrors and cause damage. A similar issue occurs on the other side of the optical path, where the high-energy photons tend to break chemical bonds in the resist, causing outgassing of certain resist compounds. These substances may be deposited on the optics, leading to another degradation mechanism.

The optics for EUVL are fundamentally different from optical lithography. Because light at the small EUV wavelength is absorbed by any material, reflective mirrors are used instead of light-transmitting lenses. Today, the peak reflectivity of each mirror is somewhere between 60 and 70%. This implies a loss of 30-40% of the light for each of the many mirrors used in a EUV optical system, giving a net throughput that is very low; a stronger light source has to solve this problem. The vacuum, required for EUV because air absorbs light at this short wavelength, should contain very low levels of moisture (water) or organics because of contamination risks for the masks and optics.

Further, imperfections in the optics cause light scattering, known as ‘flare,’ which becomes more critical at shorter wavelengths. Both lens manufacturers and the research community are tackling this problem. The former is improving lens smoothness, and the latter is compensating for the flare in the design layout.

As with the optics, EUV masks are reflective. Defect-free mask (blank) production and mask durability become critical issues at this small wavelength. Each little defect will be translated into a phase shift and eventually into a pattern inaccuracy.

Masks must be kept “clean” during handling and use, which is challenging considering that no pellicle can be used to protect the mask from contamination during exposure because it would absorb the light. Only during mask storage can a pellicle be used.

Finally, chemically amplified resists are known to reach some fundamental limits around 32nm. The ideal EUV resist should allow limited diffusion length of acids, a low line-edge roughness, and a high sensitivity. The problem with these three characteristics is that they counteract one another. Over the years, good progress has been made on resist development, with encouraging results on resist sensitivity and resolution. However, lowering line-edge roughness still has a long way to go. Some research groups, mainly at universities, are exploring alternative resist chemistries that should better combine the three critical factors mentioned above.

The interest by IC manufacturers in EUV lithography is clearly increasing this year, especially among pitch-driven memory manufacturers. It is felt that a technology to replace immersion ArF lithography will be needed in the not-too-distant future and that EUV is a serious candidate. A consolidation of efforts is needed in the coming years to make the required progress for EUV to be inserted in manufacturing in the 2010-2012 timeframe. It is clear that EUVL will be late for the initial insertion of the 32nm node (2009), but it may take over from the more expensive ArF double patterning solutions at the ramp-up to very large volume production.

For more information, contact Kurt Ronse, director advanced lithography program at IMEC, Kapeldreef 75, B-3001, Leuven, Belgium; ph 32/16 281336, e-mail [email protected], www.imec.be.