Developing a machine learning model capable of predicting the melting points of inorganic crystals allows for rapid melting point predictions, eliminating the need for experimental synthesis and complex theoretical calculations. We improve the accuracy of these predictions on a data-scarce melting point database by applying transfer learning to a crystal graph neural network. By pretraining the model with a large atomization energy database and then transferring this learning to the melting temperature database, we achieve a satisfactory root mean squared error of 218 K. Additionally, we find that the accuracy improvement for each crystal category is significantly influenced by the physical relationships between the pretraining database and the target property. This work was published at Computational Materials Science 2023, 234, 112783

This study focuses on understanding the degradation mechanisms of Pt₃Co nanoparticles, which are crucial for fuel cell catalysts. We develop a machine-learned potential (MLP) for simulating the degradation processes, comparing its accuracy to density functional theory calculations. Utilizing off-lattice kinetic Monte Carlo simulations with the MLP, we replicate experimental trends and offer insights into atomic orderings. The findings suggest design principles for high-activity, durable Pt₃Co nanoparticles and validate the method's applicability to other alloy systems like Pt₃Ni and Pt₃Co_0.5Ni_0.5. This research provides a guideline for developing MLPs for alloy electrochemical catalysts and contributes to the design of more effective fuel-cell catalysts. This work was published at ACS Catalysis 2023, 13, 16073

Ternary metal oxides have played a crucial role in advancing technologies, but the low throughput of experimental synthesis hinders the discovery of new oxides. In this study, we utilize SPINNER, which combines neural network potential and evolutionary algorithm, to scan 181 ternary metal oxide systems that have not yet been reported and discover 45 systems containing novel oxides that are more stable than related binary compounds or elemental phases. By directly searching up to 60,000 structures, SPINNER overcomes the limitations of material discovery with data mining methods that only search for prototype structures. This work paves the way for more accurate discovery of new multielement compounds that can be synthesized prior to experimentation at modest computational cost. This study was published in Journal of the American Chemical Society, 2023, 145, 35, 19378–19386

The implementation of GaN-based micro-LEDs is a promising technology for next-generation displays. However, the e-h recombination on the sidewalls of device deteriorates the efficiency due to surface defects leading to Shockley-Read-Hall recombination. We demonstrate the carrier dynamics of Ga and N vacancies found on nonpolar GaN m-planes, which can greatly affect the efficiency of the device because they are produced in one-dimensional GaN structures. Using hybrid functional, we elucidate that N vacancies are crucial for device efficiency by calculating the formation energy and carrier capture coefficients. This study provides a potential solution for the creation of successful GaN-based micro-LEDs and was published in Physical Review Applied 2023 19, 014018.