2026 International Conference on Sensing, Measurement & Data Analytics in the era of Artificial Intelligence (ICSMD 2026)

Quartz wet etching exhibits strong crystallographic anisotropy because bond-breaking characteristics vary substantially with crystal orientation, while existing structural descriptors mainly characterize local atomic coordination. This paper proposes a feature construction method based on full-space broken bond density curve. On this basis, five basic curve features (density_avg, density_max, cuts_max, mean_cuts, min_cuts_density) and one secondary structural feature (amplitude_ratio) were extracted hierarchically, and unified modeling was performed with directional harmonic fusion. In the leave-one-out crossvalidation of 198 experimental directions, the balance accuracy of the unified model for fast/slow etching classification is 0.823, the accuracy is 0.859, and the F1 is 0.900, which is significantly better than that of only direction harmonics (0.710).The proposed structural descriptors improve fast/slow etching classification and provide a structural characterization framework for orientation-dependent quartz wet etching. Continuous etching-rate regression is presented as a supplementary analysis with acknowledged limitations.
 

quartz wet etching, anisotropy, bond breaking density curve, statistical modelling,fast/slow etching classification