Solutions Control Carbon Buildup on EUVL Optics
Aaron Hand, Managing Editor -- Semiconductor International, 5/1/2002
As is the case with so many other areas of semiconductor manufacturing, the lithography process is seeing increasing concern with contamination. With tightening specifications and skyrocketing tool costs abounding — particularly moving into the next-generation lithography (NGL) realm — contamination control is coming to the planning forefront.
Researchers at Sandia National Laboratories (Livermore, Calif.) recently presented results at the latest SPIE Microlithography conference, detailing solutions for controlling carbon contamination on extreme ultraviolet (EUV) optics. Sam Graham Jr. explained the results of studies of in situ carbon removal from molybdenum/silicon multilayer mirrors, and Michael E. Malinowski revealed a way to reduce the amount of carbon buildup on these structures in the first place.
The solution from Malinowski and his team is an apparently simple one. Typical EUV optics are mirrors consisting of a substrate, then about 40 alternating layers of molybdenum and silicon, all topped off with a silicon cap. The Sandia team found that simply making the silicon cap 3 nm thick rather than the typical 4.3 nm thickness that's been used in the EUVL development program has a considerable effect on carbon growth, and therefore the reflectivity of the mirrors.
The researchers experimented with caps ranging from 2 to 7 nm thick, their goal being to balance a high initial reflectivity with a low reflectivity loss (due to carbon buildup). Samples from each multilayer mirror (MLM) were exposed to a combination of EUV light and hydrocarbon vapors at the Advanced Light Source (ALS) synchrotron in Berkeley, Calif. The MLM with the 3 nm silicon cap not only has a better initial resistance to carbon buildup, but also the highest initial reflectivity.
Theory suggests — and the Sandia experiments support this theory — that carbon deposits on the EUV mirrors are caused by substrate photoelectrons cracking hydrocarbons. Manipulating the electric field intensity at the MLM surface (done by changing the thickness of the silicon cap) can minimize the production of the photoelectrons, thereby minimizing carbon growth. Although this does not completely stop carbon buildup, Malinowski noted, the technique allows a longer time between cleaning of the optics. The difference between a 3 and 4 nm cap, he said, corresponds to an extra year of tool uptime before cleaning.
Sandia researchers also investigated the cleaning process itself, removing carbon contamination from MLMs with remote rf-O2, rf-H2, and atomic hydrogen. Their experiments — which were done on Si wafers coated with 100 Å of sputtered carbon, as well as bare Si-capped and Ru-B4C-capped Mo/Si optics — found that atomic H cleaning had the highest carbon removal rate while also reducing the threat to MLM reflectivity. Removal rates (up to 20 Å/hr for sputtered carbon near room temperature) are sufficient for in situ cleaning, and extended exposures (up to 20 hr) show less degradation of the MLMs than rf-discharge cleaning.
The team experimented with rf plasma discharges at 100, 200 and 300 W. The rf-O2 cleaning technique proved much more effective than rf-H2 at removing carbon buildup (with a ~6× faster etch rate for a given discharge power), but rf-O2 also presented a greater risk to the optics by oxidation, Graham noted. This loss of reflectivity is linearly related to the growth of SiO2 on the surface of Si-capped MLMs. For example, in a three-hour experiment, rf-O2 removed 51 Å of carbon buildup, compared with 27 Å for rf-H2. However, reflectivity loss was 1.4% for rf-O2 vs. 1% for rf-H2. For the same experiment, atomic H had a removal rate of 60 Å, and showed no measurable reflectivity loss.
For additional information on lithography, go towww.semiconductor.net/lithography
