Developing Rare-Earth Substituted Inorganic Phosphors through Machine Learning




Journal Title

Journal ISSN

Volume Title



Replacing incandescent and fluorescent light bulbs with LED-based, solid-state white lighting devices is one of the most accessible ways to reduce energy consumption around the world. Solid-state white lights have advantages such as high efficiency, a long lifespan, a small physical size, and environmentally benign components. The development of rare-earth substituted inorganic phosphors, which are the central component of solid-state white lighting devices, accelerates the complete switch to solid-state lights. Historically, the development of phosphors relies on chemical intuition or trial-and-error synthesis. These approaches require a significant amount of starting reagents and are usually highly time-consuming. Recently, the employment of machine learning methods has provided an alternative avenue to accelerate the development of phosphors. It not only enables the fast identification of candidates but also provides insights into the composition-structure-property relationships within these materials. This contribution employs machine learning algorithms to predict optical properties of phosphors. First principle calculations and experimental methods are also used to verify the machine learning predictions and investigate the properties of the synthesized materials. First, the bandgap of inorganic compounds was predicted with a two-stage machine learning model. Then, a regression model was developed to predict the Debye temperature (D) of inorganic compounds as D is a proxy for structural rigidity, which is closely related to photoluminescent quantum yield (PLQY) of phosphors. Using this model, a sorting diagram was created to evaluate D and bandgap simultaneously. The blue-emitting phosphor, NaBaB9O15:Eu2+, was highlighted and the subsequent synthesis and characterization confirmed its high PLQY. Moreover, changing the synthetic route of NaBaB9O15:Eu2+ yielded a green-emitting phosphor. DFT calculations and synchrotron data confirmed the origin of the green emission stems from Eu2+ substituting on the aliovent Na+ site rather than the energetically favorable Ba2+ site. The application of machine learning can also be expanded to predict the T50 of Eu3+-doped phosphors, which is a value corresponds to thermal stability. Finally, combining compositional descriptors with local geometry information made the centroid shift be successfully predicted. This dissertation provides new methodologies for phosphor discovery. The fast prediction of optical properties with machine learning can undoubtedly accelerate the advancement of phosphors.



Machine learning, Phosphors, Luminescence


Portions of this document appear in: Zhuo, Ya, and Jakoah Brgoch. "Opportunities for next-generation luminescent materials through artificial intelligence." The Journal of Physical Chemistry Letters 12, no. 2 (2021): 764-772; and in: Zhuo, Ya, Aria Mansouri Tehrani, and Jakoah Brgoch. "Predicting the band gaps of inorganic solids by machine learning." The journal of physical chemistry letters 9, no. 7 (2018): 1668-1673; and in: Zhuo, Ya, Aria Mansouri Tehrani, Anton O. Oliynyk, Anna C. Duke, and Jakoah Brgoch. "Identifying an efficient, thermally robust inorganic phosphor host via machine learning." Nature communications 9, no. 1 (2018): 1-10; and in: Zhuo, Ya, Jiyou Zhong, and Jakoah Brgoch. "Controlling Eu2+ substitution towards a narrow-band green-emitting borate phosphor NaBaB9O15: Eu2+." (2019); and in: Zhuo, Ya, Shruti Hariyani, Edward Armijo, Zainab Abolade Lawson, and Jakoah Brgoch. "Evaluating thermal quenching temperature in Eu3+-substituted oxide phosphors via machine learning." ACS applied materials & interfaces 12, no. 5 (2019): 5244-5250.