Computational Material Design powered by Machine Learning
Computational science has been of paramount importance in providing physical understanding as well as technological advancement for material and device development. Recently, the field has been propelled by the fusion of computational and machine learning (ML) techniques. Several approaches were proposed ranging from directly modeling atomistic structure-property relations, using ML as a surrogate model, to incorporating directly ML models within simulations. Although different spatial scales in materials (quantum, atomistic, continuum) face different obstacles, many challenges are common to all scales such as the need for sufficiently broad datasets that are simultaneously as small as possible. Knowledge transfer between computational science subfields can therefore boost advances in both method development and application.
ARTEMIS: Artificial Intelligence powered Multifunctional Material Design
PIs: A. Casini, A. Gagliardi, A.S. Bandarenka, R.A. Fischer, S. Günnemann, P.-S. Koutsourelakis, O. Lieleg, P. Mela, B. Rieger, B. Wolfrum; TUM Innovation Network