03 Jun 2005
European firms are making good progress in developing the optical equipment that will be needed to make the microchips of the future. Rob van den Berg finds out more.
Moore's law - the doubling of the number of transistors on a chip every two years - celebrates its 40th birthday this month. And it is still going strong. Even though the end of its "shelf life" has regularly been forecast, continual advances in the imaging resolution of optical lithography have kept the law on track by enabling the creation of ever-smaller features on semiconductor chips.
The key to improving lithography is the use of shorter-wavelength light, and lenses with a higher numerical aperture (NA). Since the mid-1980s, the wavelength used for lithography has steadily come down from the 436 nm g-line of mercury to the 193 nm emission of an ArF excimer laser, which is the current basis for chip manufacturing. While further incremental reductions in wavelength do not seem feasible, scientists are investigating immersing the lithography optics in a fluid such as demineralized water, to increase their NA from the current value of 0.93 to values beyond unity.
However, there's a problem: according to Moore's law and the Lithography Roadmap - a kind of industry consensus about where the technology is heading - feature sizes on semiconductor chips should reach 32 nm by the end of this decade. The challenge is that making features this small will require an entirely new approach to lithography.
In order to save Moore's law, researchers at Carl Zeiss Semiconductor Manufacturing Technologies (SMT), Germany, have been working with leading equipment manufacturer ASML Optics for the last 10 years to develop extreme ultraviolet (EUV) lithography. This ambitious approach performs lithography at a wavelength of just 13.5 nm - 14 times shorter than today's systems - and presents its own set of technical challenges. For a start, EUV radiation is absorbed by all materials, including air, so EUV lithography has to be performed in a vacuum. What's more, as there are no materials that can be used to make refractive optics (lenses) that operate at this wavelength, the light must be focused using only specially shaped reflective optics (mirrors).
These are just some of the challenges being addressed at Zeiss's brand-new production and research facility, which is located just outside the old town of Oberkochen.
Winfried Kaiser, manager of product strategy at Zeiss, explained in more detail: "In conventional refractive objective designs, lens elements are added in order to improve performance. A single lithographic 'lens' may therefore consist of up to 60 surfaces. In an EUV system the goal is to minimize the number of components, since even the best reflective surfaces at this wavelength reduce the number of photons by 30%."
One of the problems has been finding a suitable reflective coating for the mirrors, and Zeiss has settled on alternating layers of molybdenum (Mo) and silicon (Si). "These coatings were originally developed for astronomical telescopes, which were sent into the upper atmosphere with balloons to detect soft X-ray radiation," Kaiser told OLE. "Each Mo/Si layer only has a minor reflectivity, but, by adding up to 40 or 50 layers, an acceptable reflectivity of 70% is obtained."
Manufacturing the reflective optics presents another challenge. The mirrors have a special aspheric shape and are made by a computer-controlled figuring and polishing process. To complicate matters, the coating thickness must be adjusted across the mirror surface to correct for the change in the angle of incidence of the light.
Requirements for the optical quality of the mirror surfaces are also staggering. For a mirror 100 mm in diameter , the acceptable root-mean-square surface roughness is only 0.2 nm. In addition, the thickness variations of the individual layers of coating must be controlled to within 0.1% - a mere 7 pm.
In a painstaking process that may take weeks, polishing slurries and ultra-accurate ion beams are used to achieve the required shape and roughness. Kaiser explained: "Ten to fifteen years ago, we would rely on manual polishing - people with 'golden hands' - but with these kinds of requirements, an automated system is indispensable."
High-performance surface metrology equipment is required to check the figure and roughness of the mirror. To this end, Zeiss has developed special interferometers which can make measurements with a repeatability of 27 nm and a reproducibility of 46 pm - more than accurate enough to determine the characteristics of the mirrors.
To date, five of the six mirrors needed for a first-demonstration EUV lithography tool have been fabricated and have withstood reflectivity and lifetime tests. In order to suit use in a commercial system, the lithography optics need to be able to function for more than 30,000 h.
The lifetime of the optics is reduced by surface oxidation, which is irreversible, and the deposition of carbon, which can, in principle, be removed. Preliminary tests under continuous synchrotron exposure at the synchrotron BESSY in Hamburg are so far promising, but it's still early days.
Another big technical challenge for EUV lithography is to develop an appropriate light source that can generate sufficient output power at 13.5 nm. The source of EUV radiation is a hot plasma, which is generated either by a laser or by a gas discharge.
Having an EUV source with sufficient output power is critical because it directly relates to production throughput. Cost-effective chip production calls for the exposure of about 120 wafers per hour. "We need to get the EUV throughput up to a level that customers are accustomed to getting from their tools now [based on 193 nm excimer lasers]," said Noreen Harned, vice-president of ASML .
Several manufacturers are working hard to deliver the more than 100 W of EUV required at the entrance of the illumination system. This may sound like a lot, but, after several reflections in the projection optics, there are only a few millijoules per square centimetre left at the wafer.
One leading manufacturer is XTREME technologies: a joint venture between Lambda Physik in Göttingen and Jenoptik in Jena that was founded with the purpose of to develop plasma sources for EUV lithography. In September of last year, XTREME generated a record 200 W of power from a xenon gas discharge plasma.
General manager Uwe Stamm explained how the EUV light is generated: "With a couple of electrodes we generate a peak current of 60,000 A in a low-pressure, pre-ionized xenon gas. The hot, dense plasma that results has a temperature of 200,000 °C, and emits light at 13.5 nm. Only 1-3% of the input energy gets converted in this way, however, so taking away the excess heat and preventing the electrodes from melting is a major problem." As a result, XTREME has put a lot of effort into developing novel cooling solutions using porous metals and increasing the plasma-wall distance to enlarge the cooling surface.
According to Stamm, even higher output powers are, in principle, not a problem: "With tin vapour we have even reached 400 W. Overall, there has been an increase in power by almost three orders of magnitude over the last three years."
XTREME is not the only firm considering the use of a tin-vapour discharge to generate high output-powers. Philips Extreme UV - a joint venture between Philips Lighting and the Fraunhofer Institute - also believes that tin has much to commend it.
"The broadband characteristics of xenon sources result in too much energy absorption outside the mirrors' narrow reflective band," said Joseph Pankert, general manager of Philips Extreme UV. "Tin sources radiate in a much narrower bandwidth and require much less input power to provide the necessary exposure on the wafer, as compared to xenon." Pankert's firm has recently demonstrated a tin-plasma discharge source, which generates 40 W of useable EUV power, and is well on course to developing industrial-scale EUV sources.
However, tin vapour does have a drawback: it can condense on the collector optics and contaminate them, resulting in a loss of reflectivity. This could cause problems for production tools which have to operate for at least one year, during which time the source is on for some 3000 h or 100 billion pulses. To get around the problem, Philips is experimenting with the use of a buffer gas between the source and the optics that sweeps the tin vapour onto a foil trap.
Once both the EUV source and optics are ready, there is the question of using them to construct a complete lithography system. In order to prove that this will be possible, smaller prototype lithography machines have recently been built.
Exitech, UK, has produced so-called EUV micro-exposure tools (with XTREME's gas-discharge plasma sources and Zeiss optics) and installed them at Intel and International SEMATECH. Such micro-exposure tools deliver the required chip-resolution of 32 nm, but at a much smaller field-of-view than in a commercial production tool. Zeiss and ASML are currently finishing the assembly of a larger alpha-demo tool, which contains a Philips source. By the first quarter of 2006, this will be delivered for testing to the Interuniversity Microelectronics Center (IMEC) in Belgium - the largest independent microelectronics research and development centre in the world. Another one will be installed at around the same time in the US.
It is hoped that the demonstration of these prototypes will boost the confidence of the semiconductor industry to make the switch from 193 nm and immersion optics to EUV technology. "The only way that immersion [optics] will impact on EUV is if there is the invention of a new fluid that is cost-effective and continues to extend immersion," said ASML's Harned. "Otherwise, our leading-edge customers feel that 2010 is a reasonable timeframe to have EUV tools in volume production."