Zirconium belongs to the group of critical rare metals and is primarily used in industry. Its most important application, as the basis for specialized alloys, is in nuclear reactors, owing to its exceptionally very low thermal neutron absorption cross-section. Based on theoretical and experimental investigation, the potential for removing metallic (Al, Ti, Hf, V, Fe, Cr, Cu, Ni) and non-metallic (O, C) impurities from technogenic zirconium during electron beam melting (EBM) was assessed. The influence of temperature (ranging from 2350 K to 2750 K) and refining duration (10, 15, and 20 min) under vacuum conditions (1 × 10⁻³ Pa) was investigated concerning the degree of impurity removal, the microstructure, and the micro-hardness of the resulting ingots. It was established that under optimal EBM conditions for technogenic zirconium (T = 2750 K, τ = 20 min), the total refining efficiency reached approximately 87%, and the achieved Zr purity was 99.756%. Among the impurities present in the technogenic zirconium, the lowest removal efficiencies were recorded for Al (54.90%) and Cr (88.89%), with the lower refining efficiency for Al influencing the microstructure and micro-hardness of the ingots produced after EBM.

Electron beam lithography (EBL) using bilayer resists is influenced by electron scattering, proximity effects, resist thickness, pattern density, and exposure dose. Optimization based only on the top-view critical dimension does not capture linewidth variation through the developed resist thickness. This study presents an experimental and theoretical approach to model and optimize the depth-dependent sidewall geometry of bilayer PMMA exposed by a variable-shaped-beam EBL at 20 keV. Bilayers consisting of 950 PMMA A2 and 495 PMMA A6, with total thicknesses of 300 and 550 nm, on SiO₂/Si substrates were fabricated. We investigated isolated lines and line gratings with nominal line/space (L/S) dimensions of 100/500 and 100/700 nm over selected exposure dose ranges. Cross-sectional SEM measurements at five resist depths were used to develop regression models describing linewidth as a function of the exposure dose and resist depth. Two models for mean values and standard deviations were constructed for each of the considered experimental cases to account for external noise not included in the regression models, and the propagated uncertainties are quantified by Monte Carlo simulation. Four linewidth-difference criteria to identify doses producing near-vertical profile segments were introduced and applied. The results demonstrate that the selected optimum dose depends strongly on the bottom-layer thickness, pattern density, and the profile segment considered. The proposed approach provides a quantitative empirical method for predicting, analyzing, controlling, and optimizing EBL resist profiles, and can also be applied, with appropriate experimental calibration, to different resist systems, pattern geometries, exposure dose ranges, beam shot sizes, and acceleration voltages.