[3E12-16] Surface Analysis

2:15 PM - 2:45 PM JST(5:15 AM - 5:45 AM UTC)

[3E14【依頼講演】] Angle-resolved depth-profile analysis for laboratory Ga Kα hard X-ray photoelectron spectroscopy: accelerated reconstruction with integrated deep-learning and statistical denoising

*Satoshi Toyoda¹ (1. VACUUM PRODUCTS CO. LTD. (Japan))

Laboratory angle-resolved hard X-ray photoelectron spectroscopy with a Ga Kα source enables non-destructive depth profiling of buried interfaces in multilayer films, but extracting the depth profile from angle-dependent intensities is an ill-posed inverse problem, and high-throughput or short-exposure data demand both speed and noise robustness. We report a workflow combining accelerated computation with deep-learning and statistical processing. Rebuilt from a MATLAB interface to a Python pipeline with memory-free streaming and GPU-accelerated fitting, it processes over 10 TB on a workstation, and a neural-network surrogate for the depth inversion cuts reconstruction from 111 h to 1.3 s. For noisy short-exposure data, bin-pool resampling and self-supervised denoising prove complementary: the denoiser dominates under severe photon starvation while resampling matches it, training-free, at higher signal-to-noise ratio. The workflow transfers to synchrotron microbeam AR-HAXPES.

Password required to view