Machine-Learning-Assisted Discovery of Cr³⁺-Based Near-Infrared Phosphors.

Area of Science:

• Materials Science
• Solid-State Chemistry
• Computational Materials Science

Background:

• Cr³⁺-substituted inorganic phosphors emit near-infrared (NIR) light (650-950 nm).
• Emission properties depend on the crystal field splitting (quantified by Dq/B ratio) of Cr³⁺ ions.
• Designing phosphors with specific Dq/B ratios is challenging due to complex material interactions.

Purpose of the Study:

• To develop a machine-learned model for predicting the Dq/B ratio in Cr³⁺-substituted phosphors.
• To accelerate the discovery of novel NIR phosphors for applications like biomedical imaging and sensing.

Main Methods:

• A regression model was trained on 193 experimentally determined Dq/B values and material features.
• The model predicted Dq/B values for over 6060 inorganic structures.
• Eight predicted phosphor hosts were synthesized and characterized to validate the model.

Main Results:

• The machine-learned model accurately predicted Dq/B values.
• Synthesized phosphors exhibited measured Dq/B values closely matching model predictions.
• The study identified promising new hosts for Cr³⁺-based NIR phosphors.

Conclusions:

• Machine learning provides a powerful tool for predicting and designing Cr³⁺-substituted NIR phosphors.
• This approach significantly accelerates the discovery of materials with tailored optical properties.
• The developed framework enables efficient exploration of the materials space for specific applications.