SEM image denoising with unsupervised machine learning for better defect inspection and metrology

Abstract

Moore's Law states that transistor density will double every two years, which is sustained until today due to continuous multidirectional innovations (such as extreme ultraviolet lithography, novel patterning techniques etc.), leading the semiconductor industry towards 3 nm node (N3) and beyond. For any patterning scheme, the most important metric to evaluate the quality of printed patterns is edge placement error, with overlay being its largest contribution. Overlay errors can lead to fatal failures of IC devices such as short circuits or broken connections in terms of pattern-to-pattern electrical contacts. Therefore, it is essential to develop effective overlay analysis and control techniques to ensure good functionality of fabricated semiconductor devices. In this work we have used an imec N-14 BEOL process flow using litho-etch-litho-etch (LELE) patterning technique to print metal layers with minimum pitch of 48nm with 193i lithography. Fork-fork structures are decomposed into two mask layers (M1A and M1B) and then the LELE flow is carried out to make the final patterns. Since a single M1 layer is decomposed into two masks, control of overlay between the two masks is critical. The goal of this work is of two-fold as, (1) to quantify the impact of overlay on capacitance and (2) to see if we can predict the final capacitance measurements with selected machine learning models at an early stage. To do so, scatterometry spectra are collected on these electrical test structures at (a) post litho, (b) post TiN hardmask etch, and (c) post Cu plating and CMP. Critical Dimension (CD) and overlay measurements for line/space (L/S) pattern are done with SEM post litho, post etch and post Cu CMP. Various machine learning models are applied to do the capacitance prediction with multiple metrology inputs at different steps of wafer processing. Finally, we demonstrate that by using appropriate machine learning models we are able to do better prediction of electrical results.

Near-field infrared nanoscopic study of EUV- and e-beam-exposed hydrogen silsesquioxane photoresist

“…Therefore, even though the AFM provides clear images, it does not accurately measure the linewidths. Meanwhile, SEM often offers weak contrast images for thin films, so it is difficult to precisely measure roughness using SEM [ 29 – 33 ]. For example, in our results, the LER values calculated from the two clear contrast images (HP 100 and 500 nm) are almost the same as the AFM values.…”

“…However, SEM images are not suitable for the pre-development inspection of latent photoresist images because they only provide morphological information. In addition, as the film thickness becomes thinner, the image contrast of SEM is often insufficient to estimate the patterning performance of photoresists [ 29 – 33 ]. On the other hand, although FTIR spectroscopy, Raman spectroscopy, and X-ray photoelectron spectroscopy (XPS) can provide information on chemical compositions and structures, they cannot be used for pattern inspection due to their poor spatial resolution.…”

Extraction of roughness measurements from thin resists with low signal-to-noise-ratio (SNR) SEM images by applying deep learning denoiser

“…We have trained our previous U-Net architecture-based unsupervised denoising machine learning algorithm 11 with the SEM images. The network has been trained on Lambda TensorBook with NVIDIA RTX 2080 MAX-Q GPU.…”

Mask defect detection with hybrid deep learning network

“…To even further suppress the contribution of image noise to the error of the defect shape detection and generation, deep learningbased tools have proven to be quite successful. [17][18][19] Therefore, the question of whether deep learning-based methods can successfully be applied to mask repair applications is of high relevance.…”

“…[14][15][16] A general property of SEM images, the noise, has been addressed in further works, with goal of reducing the noise to obtain higher accuracy in the image analysis. [17][18][19] Other application fields of machine learning methods operating on wafer and layout data are hot spot detection, layout classification, and pattern similarity detection. [20][21][22] The methods are also used for tasks that are not targeted at the processing and use of image data.…”