Data-driven materials science often titled the “4th paradigm”, has opened an avenue towards material discovery through statistically driven machine learning approaches. Due to the continued increase in computing power and improvements of theoretical methods, the accuracy of predicted material properties has reached reliability comparable to experiments while significantly surpassing them in terms of speed and cost. This gave rise to a rapid increase in available open-source material databases, facilitating material discovery at an unprecedented pace.
In this talk, Peter will describe how a data-driven approach can facilitate discovering new materials with tuned properties and how AI can be a powerful tool to augment (however not replace) our physics/chemistry-based intuition. To illustrate these concepts, he will describe how his research enables the discovery of new ultra-low work function materials for thermionic energy conversion. He will demonstrate that creating data by high-throughput quantum chemistry calculations forms the basis on which a statistically driven machine learning model can be established. This model can then predict the property of interest with an accuracy comparable to the first principles calculations while being significantly less computationally expensive. Further, Peter will provide his perspective of where he sees the machine learning paradigm in materials science and sustainable energy in the near future.
To register for the seminar, click here.