Vertical Graphene Nanosheets on Flexible Substrates for ppb-Level NH₃ Sensing with Neural Network Acceleration

Vertical graphene nanosheets (VGN) have a promising future in gas sensing due to their high surface area and edge-rich structure, yet their synthesis typically requires high temperatures (>500 °C) or catalytic substrates, restricting flexible applications. Here, we introduce a catalyst-free, low-temperature chemical vapor deposition (CVD) under electron irradiation (threshold energy 60 eV) to directly grow VGN on flexible polyimide (PI) and silicon substrates, systematically excluding Joule heating effects. The resulting VGN-based sensor detects NH 3 at an ultralow limit of 36 ppb at room temperature, outperforming many flexible sensors (limit of detection 50–500 ppb) and rivaling rigid oxide-based systems. To address the slow response inherent to carbon-based sensors, we developed a physics-informed neural network (PINN) incorporating Langmuir adsorption kinetics, reducing detection time by 50% at 96% accuracy at 1 ppm of NH 3 . This synergy of scalable VGN fabrication, parts per billion-level sensitivity, and neural network-driven acceleration enables real-time, wearable NH 3 monitoring for health diagnostics (e.g., breath analysis) and environmental safety. Our approach establishes a versatile platform for direct VGN synthesis on flexible substrates, paving the way for wearable NH 3 sensors in health monitoring and environmental applications.