Hybrid‐Printed Single‐ChemFET Electronic Nose Enabled by Machine Learning

We present an electronic nose sensor based on a single chemically sensitive field-effect transistor (ChemFET). A novel hybrid printing process that combines the advantages of inkjet and aerosol jet printing is used to fabricate the ChemFET with a semiconducting sub-10 nm In$_2$O$_3$ thin-film channel. The sensor was first electrically characterized, showing an on/off ratio of 2.3 · 10$^5$ and a threshold voltage of 12.67 V. Three gases at 100 ppm concentration—isopropyl alcohol, benzene, and carbon monoxide–are used to test the sensing capability. We show that a single ChemFET generates gas-specific features across different drain-source (VDS) and gate-source voltages (V$_{GS}$). Two classification algorithms, linear discriminant analysis (LDA) and random forest (RF), are evaluated using leave-one-out cross-validation. Both algorithms can classify reference air and carbon monoxide without errors, but RF classifies isopropyl alcohol and benzene with higher precision and recall. The feature importance analysis reveals a correlation with V$_{GS}$ and not with V$_{DS}$. After feature importance-based feature selection, LDA and RF achieve F1-scores of 0.85 and 0.94, respectively. ... mehrOur findings indicate that the printed single ChemFET-based sensor is a promising candidate for gas-sensing applications.