Machine Learning Assistant Discovery of Stable Lead-free Perovskites for Extreme Low-dose X-ray detection

Hybrid lead halide perovskites with benchmark compositions of APbX3 has emerged recently aspromising building blocks for X-ray detector with high detectivity, and easier fabrication.

In particular,its extremely low detection dose rate limit enables the fabrication of low dose thin film detectors forsafe and versatile X-ray imaging in medical diagnoses. However, the toxicity of lead and the limitedstability restrict their commercialization.

The toxicity issue can be solved by Pb substitution, while theinstability can be solved by forming 2D structures embedding hydrophobic cations.

Combining theabove two strategies, however, would generate millions of candidate 2D lead-free perovskites to beselected, which is manually impossible. Therefore, we will utilize the intelligent machine learning (ML)process to implement the massive materials screening.

The ML screening task will be integrated withdensity functional theory simulation, automated robotic sample synthesis, and off-site devicefabrication to predict, optimize, and finally materialize the optimal 2D lead-free perovskites with thebest X-ray detector performance.

This new methodology can be used to accelerate the design of biocompatiblematerials, from in-body sensors to artificial skeletons.