Integrated WTe ₂ @SnO ₂ Heterojunction Sensors and Deep Learning Architecture for Intelligent Multi-Gas Detection under Environmental Variations

Multigas detection in varying environmental conditions remains a critical challenge for conventional sensors, requiring innovative integration of advanced materials and intelligent algorithms. An integrated sensor-algorithm platform synergistically combines WTe₂@SnO₂ heterojunctions with a novel deep learning framework for intelligent multigas recognition. The heterojunction sensors, synthesized via liquid-phase exfoliation, are designed to provide complementary response characteristics that facilitate effective algorithmic discrimination. Enhanced sensing performance is achieved through increased surface area and interfacial charge transfer, creating distinct electronic signatures with remarkable metrics, response value of 37.3 to 8 ppm of NO₂, faster recovery time of 34 s, and sub-100 ppb detection limits. A custom-designed VAE-BiLSTM-SA deep learning architecture directly exploits these heterojunction characteristics by extracting temporal and amplitude features from sensor responses. The co-optimized sensor-algorithm system demonstrates exceptional performance in simultaneous classification and concentration prediction of NO₂, NH₃, and complex mixtures under varying humidity conditions, achieving 99.7% classification accuracy. Real-time detection capabilities surpass conventional sensor limitations through algorithmic enhancement, enabling immediate gas identification that exceeds physical response constraints.