Raman-calibrated image-based screening of ambient aging in CVD-grown HfS₂ thin films

HfS₂ is a layered semiconductor for electronic and optoelectronic applications, but humidity-assisted native oxidation limits its ambient stability. Raman spectroscopy probes HfS₂ degradation but is time-consuming and condition-sensitive. Here, we revisit a previously reported HfS₂ aging dataset to develop a Raman-calibrated image-based framework for rapid, non-destructive screening of CVD-grown HfS₂ films on c-plane sapphire. Photographs collected under a documented configuration yielded predefined color, histogram, texture, and spatial descriptors. A simple, interpretable regression using only CIE Lab b* with physical-range clipping achieved a within-library, query-excluded leave-one-day-out RMSE of 6.78 days for 21 queries from three optically evolving conditions: native HfS₂ at 35% and 70% RH and PMMA-capped HfS₂ at 70% RH. The b*-based model performed comparably to kNN (7.70 days) and significantly better than a no-information pool-mean baseline (9.84 days, p = 0.0030), indicating that more complex estimators provided no clear benefit in the present dataset. Under leave-one-pair-out evaluation of 20 image-Raman pairs, the combined image-derived estimate showed a trend-level relationship with measured normalized A_1g intensity (R² = 0.600), while ΔE gave the best individual result (R² = 0.701). The framework identified its limits: severe degradation of native HfS₂ at 70% RH was underestimated, whereas Al₂O₃-encapsulated HfS₂ showed insufficient optical evolution for reliable chronological estimation. Accordingly, this material- and configuration-specific method is a proof-of-concept screening framework, not a universal dating model or a substitute for Raman spectroscopy. Within this scope, controlled photography provides interpretable pre-screening of broad aging states and prioritizes samples for Raman or chemical analysis.