Edge Roughness Differences Among EUV Resists

Metal oxide resists offer better absorption but lack a key smoothing mechanism

EUV resists are a key component in the implementation and optimization of EUV lithography. By converting a limited number of absorbed EUV photons into a variable number of released migrating electrons, the resist becomes the final determinant of resolution. There are two kinds of resist which are seriously considered: chemically amplified resists (CARs) and metal oxide resists (MORs) [1].

Due to EUV resist feature dimensions approaching the molecular size (~ 2 nm), the effects of limited photon absorption are enhanced [2]. Moreover, the variable electron blur and electron yield per photon aggravate the noise already present from the Poisson statistics of photon absorption. CARs possess a smoothing mechanism in the form of acid blur which can act after the electrons have migrated, while MORs lack this mechanism. Consequently, as shown in Figure 1, MORs show more high spatial-frequency roughness at the edges (despite 3x higher absorption in 40 nm thick resist), while for CARs this is smoothed out, leaving only low spatial-frequency roughness.

Figure 1. Left: Thresholded acid density image for 20 nm half-pitch in 40 nm thick chemically amplified resist (absorption 5/um; sigma=5 nm Gaussian acid blur). Right: Thresholded electron density image for 20 nm half-pitch in 40 nm metal oxide resist (absorption 20/um). Variable electron blur taken from [2]; variable electron yield = 5-9 electrons/photon. Incident dose is 60 mJ/cm2 averaged over the 40 nm pitch.

The visual effect of smoothing by a hypothetical Gaussian blur function applied to a metal oxide resist is shown in Figure 2.

Figure 2. Thresholded electron density image for 20 nm half-pitch in 40 nm metal oxide resist (absorption 20/um). Variable electron blur taken from [2]; variable electron yield = 5-9 electrons/photon. Incident dose is 60 mJ/cm2 averaged over the 40 nm pitch. Left: before smoothing. Right: after smoothing by a Gaussian blur function (sigma=5 nm).

The remaining low spatial-frequency edge roughness after smoothing still has an amplitude on the order of the molecular size, which in this case is 10% of the feature size (20 nm). For smaller pitches and feature sizes, the variation on the order of the molecular size will become relatively more severe, due to less absorption in thinner resist and larger sampling of wider electron yield range (from adjacent exposed regions) by the smoothing blur function.

EUV resist sensitivity to low-energy electrons is the underlying cause of stochastic edge roughness. Low-energy electrons operate at the basic energy scale of chemistry. A non-chemical, atomic mechanism at ~90 eV energy scale insensitive to photoelectrons or secondary electrons would be needed for a more robust EUV resist.

References

[1] P. Leray et al., “NA0.33 EUV extension for HVM: Testing single patterning limits,” Proc. SPIE 13424, 1342403 (2025); https://doi.org/10.1117/12.3052244.

[2] F. Chen, Facing the Quantum Nature of EUV Lithography; The Quantum Nature of EUV Lithography.