Salatan (he/him) grew up in Bangkok, Thailand. He received his Ph.D. in Chemical Engineering from Vidyasirimedhi Institute of Science and Technology, Thailand. His Ph.D. thesis focused on the theoretical and experimental investigations of energy storage materials, especially lithium-ion and lithium-sulfur batteries. He is currently interested in accelerating materials discovery for solid state electrolytes. Aside from the laboratory, Salatan also enjoys swimming, playing board and video games, cycling, and photography.
Finding novel materials for solid-state electrolytes with high stability and ionic conductivity is one of the main challenges for next-generation batteries and is particularly hard for first-principles-based modelling due to a high computational cost of dynamic properties such as ion diffusion.
Machine learning (ML) models have shown great promise in predicting the outcome of first-principles calculations for a given atomic structure. However, currently available state-of-the-art ML methods work well with large datasets of high quality(10k-100k samples) but fail to generalize if trained on smaller data. This is particularly the case for Li conductors, which are key to enabling the solid-state battery era. Moreover, training on the available chemical databases can only predict structures that will be similar to already known materials.
We are aiming to facilitate an ML model that can effectively screen for targeted property from entire chemical possibility across the periodic table, including those materials that have never been discovered, synthesized, and characterized. This can be accomplished by using the ML model trained with physically motivated generic descriptors. This approach will allow us to screen all chemical possibilities and pave the way to predict the entirely new kind of structures. We also aim to synthesize the best-predicted materials.