Materials-Algorithm Co-Optimization for Specific and Quantitative Gas Detection

Abstract

Rapid, reliable, and quantitative formaldehyde detection has become increasingly important in the processing industry and environmental protection. As an intelligent electronic instrument, the realization of electronic noses (e-noses) for quantitative gas detection relies on enhanced specificity. Here, we propose a materials-algorithm co-optimization (MACO) method that enables quantitative detection of formaldehyde in e-nose. This approach employs thermokinetic feature engineering to optimize data quality and algorithm selection, thereby reducing dependence on data scale and computing power resources. Specific thermokinetic activation patterns for formaldehyde can be generated through a single materials processing strategy. Through a combination of thermokinetic feature-driven machine learning, we demonstrated an e-nose—comprising only five Co₃O₄-based gas sensors—capable of discriminating formaldehyde from ethanol. The mathematical model reveals that the physicochemical mechanism of odor coding logic in our e-nose is dictated by the mass action law. A quantitative detection of formaldehyde in 0.1–20 ppm with a precision of 5% full-scale (F.S.) has been demonstrated. We also showcase the adaptability of e-nose for binary mixture analysis. The detection model of the MACO-driven e-nose is simple and interpretable, showing broad prospects to achieve quantitative gas detection rapidly and at a low cost.