The scanner prints a grating of lines by letting diffraction orders from the mask pass through the projection optics. An order only gets through if it falls inside the pupil, a circle of radius NA. No 1st order, no pattern. This is a real Abbe (partially coherent) imaging calculation of a 1:1 line/space mask.
Rayleigh: CD = k₁·λ/NA. The hard limit for a grating is k₁ = 0.25 (half-pitch), reached when the 0th and 1st orders sit at opposite edges of the pupil. Mask 3D effects, polarization and resist chemistry are left out.
Light arrives as photons. An EUV photon carries 92 eV, so a given dose delivers ~14× fewer photons than at 193 nm. Each square nanometre of resist absorbs only a handful, and their random count (Poisson shot noise) roughens the edges of every line.
This simulates one exposure on a 1 nm grid using the aerial image from the imaging tab, random photon absorption, electron/acid blur and a development threshold set for dose-to-size.
Absorption fractions and blur values are representative (≈), not a specific commercial resist.
No material reflects much EUV at near-normal incidence. A stack of alternating molybdenum and silicon layers does: each interface reflects a tiny bit, and when the period is about half the wavelength all reflections add up in phase (a Bragg mirror). This is a Parratt recursion using the optical constants of Mo and Si at 13.5 nm.
n(Mo) = 0.9238 + 0.0064i, n(Si) = 0.9990 + 0.0018i at 13.5 nm, held constant across the plotted band (≈).