CTAB modified SnO PEDOT PSS heterojunction humidity sensor with enhanced sensitivity stability and machine learning evaluation.

This study presents the development of a high-performance resistive humidity sensor based on a cetyltrimethylammonium bromide (CTAB)-assisted tin oxide (SnO₂) nanostructured thin film integrated with a Poly(3,4-ethylenedioxythiophene): Poly(styrenesulfonate) (PEDOT: PSS)/SnO₂ heterojunction. The sensor design incorporates CTAB at varying weight percentages (0%, 6%, 11%, 16%, and 20%) during the hydrothermal synthesis of SnO₂ to regulate crystal growth, morphology, and surface area. The sample with 20 wt% CTAB (SnO-5) exhibited a flower-like stacked nanostructure, confirmed via field emission scanning electron microscopy (FESEM), which significantly enhanced water molecule adsorption and charge transport pathways. X-ray diffraction (XRD) analysis confirmed the tetragonal rutile phase of SnO₂ with decreasing crystallite size from 12.2 nm (nm) to 4.8 nm as CTAB concentration increased. The incorporation of PEDOT: PSS, a p-type conducting polymer, onto the SnO₂ layer via spin coating formed a p-n heterojunction, which improved charge separation and reduced recombination, thereby enhancing electrical conductivity and sensor performance. Electrochemical impedance spectroscopy (EIS) and current-voltage (J-V) measurements demonstrated that SnO-5 exhibited a low internal resistance (1.1 kilo ohms (kΩ)), a minimal cut-in voltage (0.071 Volts (V)), and a high current response (2.645 micro Amps.(µA)), indicating efficient carrier transport. The optimized SnO-5 sensor achieved a high sensitivity of 85.7%, a rapid response time of 14 s (s), and a quick recovery time of 7 s, with low hysteresis (1.60%) across a broad humidity range (5-97% Relative Humidity (RH)), outperforming several existing humidity sensing platforms. The synergistic effects of CTAB-induced nanostructuring and heterojunction engineering played a pivotal role in improving moisture interaction, charge mobility, and structural stability. Furthermore, to validate real-time application feasibility, machine learning (ML) algorithms were implemented to model and predict sensor behavior. Among the tested models, Random Forest (RF) Regression achieved the highest predictive accuracy (R² = 0.99), confirming the sensor's robustness and reproducibility in dynamic environments. The proposed sensor's outstanding performance, in combination with ML-enhanced evaluation, positions it as a promising candidate for next-generation humidity monitoring systems in industrial, environmental, and biomedical applications, including respiratory diagnostics and non-invasive health monitoring.