Harnessing Transfer Deep Learning Framework for the Investigation of Transition Metal Perovskite Oxides with Advanced...

Area of Science:

• Materials Science
• Chemical Engineering
• Artificial Intelligence

Background:

• Transition metal perovskite oxides (TMPOs) are crucial for gas sensing in environmental and healthcare applications.
• Traditional TMPO research faces challenges in data analysis and material optimization.
• Developing advanced TMPO gas sensors requires innovative predictive and design methodologies.

Purpose of the Study:

• To develop a predictive deep learning framework for designing high-performance TMPO gas sensors.
• To identify optimal precursor combinations and understand ligand effects on sensing performance.
• To synthesize and experimentally validate a novel TMPO-based gas sensor for specific VOC detection.

Main Methods:

• A deep learning framework combining natural language processing (Word2Vec) and crystal graph convolutional neural networks (CGCNN) was employed.
• A comprehensive dataset of 1.2 million literature abstracts and 110,000 crystal structures was utilized.
• The Zn(II)/Co(III)/1-methyl-1H-imidazole-2-carboxylic acid complex was synthesized and characterized.

Main Results:

• The framework predicted optimal zinc-cobalt bimetallic ion precursors for perovskite oxide sensors.
• Ligand concentration was found to regulate p-n transformation, optimizing sensing performance.
• The synthesized sensor demonstrated high sensitivity and selectivity for 3-hydroxy-2-butanone (3H-2B) at 25 ppb.

Conclusions:

• The developed deep learning framework efficiently predicts and aids in the precise design of TMPO gas sensors.
• Fundamental understanding of sensing mechanisms was enhanced through novel modeling approaches (HSLIM-n and PSLRM-p).
• Experimental validation confirmed the framework's effectiveness and the sensor's superior performance in complex environments.