Exploring new useful phosphors by combining experiments with machine learning

Area of Science:

• Materials Science
• Solid-State Physics
• Computational Chemistry

Background:

• Advances in solid-state lighting and displays necessitate the continuous development of new phosphors.
• Traditional methods for discovering new phosphors rely on time-consuming trial-and-error experiments.
• Computational approaches can significantly accelerate the identification of promising phosphor candidates.

Purpose of the Study:

• To explore a more practical and efficient approach for developing novel phosphors with targeted luminescent properties.
• To investigate the integration of computational science and machine learning in phosphor discovery.
• To identify new phosphor compositions and crystal structures with desirable optical characteristics.

Main Methods:

• Combining experimental investigations with machine learning algorithms.
• Focusing on key luminescent properties: emission wavelength, full width at half maximum (FWHM), and thermal quenching.
• Utilizing high-throughput experimentation for rapid screening of potential candidates.
• Exploring new chemical compositions and crystal structures for phosphor hosts.

Main Results:

• Machine learning models can predict phosphor properties, reducing experimental time.
• Integration of computational and experimental methods enables faster discovery of novel phosphors.
• Identification of potential new phosphor candidates with tailored emission characteristics.

Conclusions:

• Combining computational science and machine learning offers a significantly faster route to developing new phosphors.
• This integrated approach can lead to the discovery of unexpected and overlooked phosphor compositions.
• The methodology holds promise for advancing solid-state lighting and display technologies.