TSMC Confronts Mask Defects from EUV Hydrogen Plasmas

With the highest volume of EUV exposures in the world, TSMC definitely gets to see the most of all the known defect mechanisms associated with EUV lithography. While their findings are mostly not published at SPIE conferences, key hints can be found in their patent applications. A particularly troublesome aspect of EUV lithography is the use of hydrogen as a cleansing agent for the mirrors in the EUV systems [1-3]. The hydrogen is ionized by the EUV radiation during the exposure process, giving rise to the infamous EUV-induced hydrogen plasma [2,3]. A number of TSMC patent applications report that the hydrogen infiltrates the EUV mask layers, forming blisters and causing them to peel [4-6]. In particular, a key detail is mentioned in [6]:

The hydrogen molecules become radicals by interaction with the EUV radiation, which then penetrate into the photomask. The penetrated hydrogen (or hydrogen radicals) diffuses into the absorber layer and accumulates in the photomask, in particular, at the interface between the reflection mirror layer and the capping layer. The accumulated hydrogen may cause a blister defect, which is a peeling-off of the capping layer from the reflection mirror layer. For example, when the number of wafers exposed by the EUV radiation tool using the photomask exceeds 10,000, the blister defects tend to occur.

Assuming a tool throughput ~3000 wafers a day, this means that blister defects will become a problem every 3 days or so, around twice a week.

As one application proposes, some areas of the mask may be excluded from patterning but be reserved for hydrogen release through targeted openings [4]. Besides limiting productivity, this cannot be applied to products that have to use the entire exposure field.

Alternatively [6], the mask is taken out and annealed above 100 deg C to get the hydrogen out before 10000 exposures are reached. Note that re-seating the mask is nontrivial due to the EUV pattern position dependence on the mask’s Z-position [7]. A vertical offset of some tens of nm could lead to a wafer pattern shift of 1 nm.

Without following this procedure, the mask, which must remain uninspected in the EUV tool while running, risks being scrapped after 10000 exposures, along with wafers that were exposed by the blistered or peeled mask. It goes to show that practicing EUV lithography requires the utmost attention to detail and diligence.

References

[1] S.M. Zyryanov, A.S. Kovalev, D.V. Lopaev, E.M. Malykhin, A.T. Rakhimov, T.V. Rakhimova, K.N. Koshelev, V.M. Krivtsun, Plasma Physics Reports, 37, 881 (2011).

[2] Y-H. Huang, C. J. Lin, Y-C. King, Discover Nano 18:22 (2023).

[3] M. van de Kerkhof, A. M. Yakunin, V. Kvon, A. Nikipelov, D. Astakhov, P. Krainov, V. Banine, Rad. Effects and Def. in Solids, 177, 486 (2022). https://doi.org/10.1080/10420150.2022.2048657

[4] C-H. Liao, P-M. Shih, US20210335599.

[5] C-H. Liao, P-M. Shih, US12025922. This was called out previously for having the erroneous teaching that there was air in the EUV system, upon which the hydrogen floats. A portion of this teaching was left in [6].

[6] C-H. Chang, M-W. Chen, A-J. Ma, C-Y. Chen, US20220382168.

[7] T. Schmoeller, T. Klimpel, Proc. SPIE 6921, 69211B (2008).