# Heterointerface‐Functionalized Photoelectric Response of Metal‐Oxide Schottky Photodiode for Intelligent Fire Detection

## TL;DR

A new photodiode design improves fire detection by combining high UV sensitivity with machine learning for accurate material identification.

## Contribution

A heterointerface-functionalized Schottky photodiode enables high-performance UV detection and intelligent material recognition.

## Key findings

- In2O3 HF-SPDs achieve 27.75 A W−1 responsivity and 2.036 × 1013 Jones detectivity under 360-nm UV light.
- Integration with an MLP neural network enables 99.94% accuracy in identifying burning materials.
- The device exhibits bias-tunable photoelectrical characteristics and unique wavelength selectivity.

## Abstract

Ultraviolet photodetectors (UV‐PDs) based on wide‐bandgap oxide semiconductors (OSs) hold broad prospects in scientific, civil, and especially fire alarm applications. However, the hotly pursued flexible and hetero‐integration capabilities are hindered by the high processing temperature of mainstream Ga2O3 UV‐PDs, while the selective UV detection capability of low‐temperature OS (LT‐OS) is undermined by the pronounced contradictory effects of abundant native defects on electric and optoelectrical properties. Herein, a heterointerface‐functionalized Schottky photodiode (HF‐SPD) is proposed to realize high‐performance LT‐OS UV‐PD. Due to the sophistically modulated Schottky contact and defect distribution, ultralow dark current and remarkable photoresponse are simultaneously achieved in In2O3 HF‐SPDs. Under the typical 360‐nm UV illumination, the responsivity (R) and detectivity (D
*) respectively reach 27.75 A W−1 and 2.036 × 1013 Jones, enabling the early‐fire‐warning capability. Moreover, the HF‐SPD possesses unique wavelength selectivity and bias‐tunable photoelectrical characteristics. By feeding such multidimensional feature data into the multi‐layer perceptron (MLP) neural network, the pattern recognition of the spectrum delivers a remarkable accuracy of 99.94%, ensuring the precise identification of combustion materials. Such an intelligent fire warning system demonstrates the prospects of algorithm‐boosted LT‐OS HF‐SPDs in ubiquitous intelligent systems, such as wearable photodetection and in‐sensor computing (ISC) devices.

A heterointerface‐functionalized Schottky photodiode (HF‐SPD) is developed to explore the photoelectric capabilities of low‐temperature oxide semiconductors. Besides the remarkable UV detection capability, the bias‐tunable photo response endows HF‐SPD with the unique wavelength selectivity. Integrated with an MLP neural network, it attains 99.94% accuracy in burning material identification, demonstrating the potential of low‐temperature HF‐SPD for wearable photo‐sensing and in‐sensor computing applications.