Abstract:

In the modern age of data science, there is more catalogued and query-able materials data available than ever before. Here, the speaker will illustrate two examples of how data-driven materials science enables the research team to both discover new materials, as well as design rational synthesis routes to these predicted compounds. First, they use high-throughput computational materials discovery techniques to survey uncharted chemical spaces for novel synthesizable materials, constructing large stability maps to guide exploratory synthesis of new ternary nitride materials. By combining machine-learning algorithms with new electronic descriptors for solid-state bonding, they can rationalize the complex interplay between chemistry, composition, and electronic structure in governing large-scale stability trends across broad materials spaces.1Next, to design synthesis routes to computationally-predicted materials, they demonstrate how Natural Language Processing algorithms can extract materials synthesis recipes from the scientific literature.2 By analyzing trends and anomalies in these synthesis recipes, they can learn new fundamental principles on how to navigate the thermodynamic and kinetic energy landscape to optimize synthesis pathways to complex functional oxides.3

1   W. Sun, "A map of the inorganic ternary metal nitrides." Nature Materials (2019).

2   O Kononova et al., "Text-mined dataset of inorganic materials synthesis recipes." Scientific Data(2019).

3   A. Miura et al., "Observing and Modeling the Sequential Pairwise Reactions that Drive Solid‐State Ceramic Synthesis." Advanced Materials (2021).

Biography: https://mse.engin.umich.edu/people/whsun

Registration link:

https://hkuems1.hku.hk/hkuems/ec_hdetail.aspx?guest=Y&ueid=84785