Service Communautaire d'Information sur la Recherche et le Développement - CORDIS

Quality specifications, contamination sources and VUV cleaning procedure for 157nm reticles

- The main result of the reticle printability work is that a proof-of-concept for VUV cleaning was reached. It is based on formerly published work of MIT Lincoln, ASML Wilton and TNO, and turned into a process usable at the wafer lithography environment. The installed methodology for VUV cleaning can be a base for further work that any partner may want to do in case of future application of 157nm exposure, because all the results were shared with the partners. The work and the understanding of airborne contamination is considered a good basis for similar studies in EUV, which will very likely be required. The main customers for this result are the semiconductor manufacturers using (or considering) an exposure wavelength of 157nm or below.

In our work a not yet optimised, but stable methodology has been shown to re-establish the transmission of 157nm unpelliclized reticles to within 0.25% of the clean state. Reticles exposed to ambient air accumulate a film of contaminants that are mostly of hydrocarbon origin. Upon 157nm wavelength exposure even a monolayer of organic contaminants or water can lead up to 1-2% transmission loss in the modified fused silica mask substrate, leading to an exposure variation, which would consume a too large part of the CD control budget. Contamination of mask substrates appears to be a competitive adsorption phenomenon whereby low molecular weight species (with low adsorption energy) are replaced over time by large molecular weight ones (with high adsorption energy).

The VUV cleaning procedure is based on an UVO RCS cleaner (obtained from ASML). The principle of this cleaner is based on reaction of the deposited hydrocarbons with oxygen radicals generated from by irradiation with 172nm light (xenon excimer lamp), both directly by dissociation of molecular oxygen and indirectly through the dissociation of ozone, which is unstable at this wavelength.

Transmission measurements were done on a Sopra PUV SE5, although this tool is rather designed as an ellipsometer, and hence it is not a double beam system. With an optimized measurement procedure the repeatability of a single point on the reticle could be improved to typically within ±0.25%.

The cleaning procedure was verified by lithographic imaging on the MS-VII. When using a reticle that had become contaminated over time by storage in cleanroom air, it could be shown that the transmission increased upon exposure to 157nm light of the laser and saturated around 1% higher after a dose equivalent to the exposure of eight 300mm wafers using a typical production type resist. In-situ cleaning during exposure in the MSVII was calculated to use a too large fraction of the CD budget across a minimum batch of 25 wafers. When pre-cleaning the reticle with the VUV process, the change in transmission caused by in-situ cleaning in the MSVII could be reduced to within 0.25%.

The mask fabrication, transportation, handling and inspection appear to contribute to contamination of 157nm masks, but no irreversible effects have been observed. The impact of storage environment and storage duration on contamination of masks was also investigated.

Limitations: The tools in use were not interfaced. Reticles that have received cleaning can get contaminated when they are unloaded from the PUV and loaded into a SMIF pod of the MSVII, before they are inserted into the MS-VII. This limitation is s estimated acceptable in view of the contamination rate found during the experiments, which were target towards a proof-of-concept. In a production environment corresponding tools would need to be integrated.

- As two side results mask specs and printability are described herebelow. They are generically applicable to any exposure wavelength. These results are, apart from wafer fabs, also very relevant for mask shops, as they clarify the mask quality needs and improve the communication between these two parties.

A list of evolutionary mask specs was composed, in which a certain ratio of the critical dimension is used to specify most of the quality parameters. Per CD and selected aggressiveness of the major quality parameters a so-called mask technology code is used.

On reticle quality printability, a formerly published study of mask quality printability (tackling printability of linewidth control, corner rounding), was extended, now focusing on printability of edge slope of the 180deg areas in AAPSM: It is a known point of attention that AAPSM topography induces an image intensity imbalance between the light propagating through the zero and pi-shifted space. We correlated printability results directly to the mask quality through the use of rigorous electro-magnetic field simulations including 3D mask topography. The amount of undercut or bias and the quartz etch depth of the pi-shifted space on the mask could be estimated. A through-pitch assessment gives an indication of the quality of the etch process during mask making.

More information on the UV2LITHO project can be found at:

Reported by

Kapeldreef 75
3001 Leuven
See on map