Fotoresist metrologie voor hoge numerieke apertuur (NA) extreme ultraviolet lithografie (EUVL)

The transition to High Numerical Aperture Extreme Ultraviolet Lithography (High NA EUVL) necessitates the use of ultra thin photoresists (<30 nm) to achieve the patterning requirements. While enabling pattern fidelity at small pitches, thin resists pose significant challenges for electron beam metrology, particularly in critical dimension scanning electron microscopy (CDSEM). Reduced imaging contrast, lower signal-to-noise ratio (SNR), and beam-induced effects, such as charging and shrinkage, degrade the accuracy, precision and throughput of critical dimensions (CD), roughness, and other measurands for the wide array of lithographic features of interest. This thesis investigates the origin of contrast degradation in thin resist stacks, quantifies materials and thickness dependent effects, and develops methodologies to enhance metrology performance. Monte Carlo simulations are combined with experimental measurements to provide comprehensive understanding. For thin resists in the range of 15 nm, both the resist and underlayer properties strongly influence the imaging contrast. Experimental yield curves are measured, and show a higher metal oxide resist (MOR) yield compared to chemically amplified resist (CAR), thus confirming the superior visibility of MOR when imaged using CDSEM. Different strategies are explored to overcome those SNR limitations present in low visibility resists. First, switching to low voltage SEM (LVSEM) can enhance the contrast in some stacks, but it requires careful optimization to balance the charging and shrinkage effects. Second, machine learning denoising techniques, either supervised or unsupervised, enable accurate roughness extraction from images captured with low number of frames, matching the images captured with a higher number of frames (hence 16 times improved throughput). Third, using high probe current can also increase the emitted signal and improve unbiasing fidelity, as shown by using power spectral density (PSD) analysis matching methodology between two CDSEMs with different columns design. Those strategies are compared to the existing standard benchmark criteria, stating that the SNR should be above 2 to ensure the reliable roughness extraction, and the measurement precision should be below 1% of the measured CD value itself. These criteria are applied to the above strategies to benchmark them to the existing approaches. Considering other features, besides line and spaces (LS), such as contact hole (CH) and tip-to-tip (T2T), metrology becomes more complex for such features. For contact holes, optimal scanning strategies, beam energies, and complementary metrology confirm the interplay between charging, shrinkage, and feature profile effects, paving the way to the use of new measurement approaches such as the random measurement acquisition (RMA). Monte Carlo simulations support these findings and shows material-dependent charge polarity reversal induced by beam energy. T2T features exhibit intrinsically higher variability, with measurement and process conditions, such as pixel size, resist thickness, and post exposure bake (PEB) temperature, all significantly influencing LCDU. Analysis of variance across lithographic conditions identifies key contributors to T2T variability, including stochastic effects, mask contributions, and metrology limitations. The findings of this thesis provide both fundamental understanding of thin resist electron e-beam interactions and practical strategies for enhancing CDSEM metrology in the High NA EUVL era. The proposed approaches combining material-based process optimization/investigation, hardware tuning, simulations, and artificial intelligence (AI)-enabled image enhancement allow accurate, efficient, and low-dose measurements, ultimately supporting the reliable characterization of scaled resist features for the next generation semiconductor manufacturing.