Design of Fe2O3–ZnO heterostructured nanofibers integrated with machine learning for acetone and ethanol classification

Abstract

Reliable discrimination of volatile organic compounds (VOCs) with similar physicochemical properties, such as ethanol and acetone, remains a key challenge for semiconducting metal-oxide (SMO) sensors. Herein, α-Fe2O3–ZnO heterojunction nanofibers (HJ-NFs) were fabricated via double-jet electrospinning and calcination, with HJ density tuned by electrospinning duration (30-120 min). The optimized 60 min sample achieved the highest responses of 11.45 (ethanol) and 8.52 (acetone) at 450 °C, ~5× higher than pristine oxides, due to enhanced charge separation and increased active sites at well-defined interfaces. To address overlapping response signatures, PCA and LDA were employed, enabling clear discrimination with a ratio of 8.20. This work demonstrates that integrating HJ engineering with machine learning provides an effective strategy for high-performance, intelligent gas sensing.