Field-localization-driven gas discrimination in porous silicon defect cavities

Porous-silicon photonic crystal cavities are widely used as refractive-index-based optical sensing platforms, but their response is often interpreted through a single observable, typically the resonance wavelength shift. This single-feature readout limits the ability of a single cavity to distinguish between analytes that induce similar refractive-index perturbations. In this work, we present a computational multi-feature optical characterization framework for gas discrimination in porous-silicon defect cavities. The objective is not to claim a chemically selective or experimentally validated trace-gas sensor, but to evaluate how multiple spectral observables can be jointly used to improve analyte-response separability in a theoretical design study. The proposed framework extends conventional sensing approaches by simultaneously exploiting multiple optical observables, including resonance wavelength shift, transmission variation, and linewidth modulation. Three defect architectures—single symmetric, split-offset, and asymmetric—are systematically investigated to tailor electromagnetic field localization within the sensing region and thereby control the light-matter interaction. A hybrid sensing model combining refractive-index overlap with weak spectral gas modulation is introduced to preserve physically meaningful responses while incorporating analyte-dependent features. Discrimination performance is evaluated using multi-dimensional distance metrics in raw, normalized, and uncertainty-normalized feature spaces. Nominal results show that the single symmetric defect provides the largest raw multi-feature separation, while the asymmetric defect exhibits comparable performance with slightly higher mean sensitivity. An alignment-robustness analysis confirms that the discrimination behavior remains stable under variations in analyte spectral positioning. More importantly, fabrication-tolerance analysis based on Monte Carlo simulations shows that the symmetric and asymmetric defect architectures become statistically comparable under realistic variations in porosity and layer thickness, while both clearly outperform the split-offset design. These findings establish a computational design and characterization framework linking defect architecture, field localization, multi-observable spectral response, and fabrication tolerance. The results support the use of porous-silicon defect cavities as model platforms for high-concentration gas-response discrimination, while also identifying the need for future experimental validation and more realistic gas-surface interaction models before trace-sensing claims can be made.