Title: Modelling Thermal Transport and Machine Learning for Nanomaterials
Speaker: Dr. Ankita Katre, LITEN, CEA-Grenoble, France
Abstract: Nanomaterials consist of different kinds of intrinsic and extrinsic defects. These defects can have large effects on thermal conductivity of materials, which is immensely crucial for cutting edge technologies like thermoelectrics, photovoltaics, LEDs etc. However, it has been a challenge to theoretically predict these effects extensively with a profound understanding. Using our ab initio approach based on Green’s function formalism implemented in almaBTE software [1,2], we find excellent agreement of thermal conductivities for various technologically relevant defective materials with experiments [1,3,4]. Furthermore, unique behaviours of defects are also revealed in our studies, which are not captured by commonly used simple modeling approaches to date. Our study on thermoelectric half-Heusler compound ZrNiSn unveils that Ni/vacancy antisites (not the previously claimed Sn/Zr antisites) are the dominant defects affecting its thermal transport . Another study on cubic SiC shows an exceptionally strong effect of BC substitution on its thermal conductivity as compared to other defects as NC and Al Si. We find that such striking behaviour of BC defect arises from a unique pattern of ‘resonant phonon scattering’ caused by the broken structural symmetry around the boron impurity atom. In my talk, I will elaborate on these interesting findings for ZrNiSn and SiC and their implications on materials designing and characterization, along with an introduction to the novel ab-initio approach for defects. Furthermore, I will demonstrate my recent work on machine learning model for prediction of novel layered compounds and the future prospect of this approach.
 Katre et al., Physical Review Letters, 119, 075902 (2017).
 Carrete, Vermeersch, Katre et al., Computational Physics Communications, 220C, 351 (2017).
 Katre et al., Journal of Materials Chemistry A, 4, 15940 (2016).
 Katre et al., Physical Review B, 93, 155203 (2016).