Machine Learning Enables Predictive Modeling of 2-D MaterialsDecember 9, 2016
Machine learning (ML) techniques have contributed towards development of the first atomic level model to accurately predict the thermal properties of stanene, a 2-D atomic sheet of tin. ML techniques train computers to recognize patterns in data and make new predictions. This study applied ML methods for materials modeling for the first time, and was more accurate at predicting material properties than past models for stanene.
Researchers at the Center for Nanoscale Materials and the Advanced Photon Source introduced a bond order potential (BOP) for stanene (single-layer tin) based on an ab-initio derived training data set. The potential is optimized to accurately describe the energetics, as well as thermal and mechanical properties of a free-standing sheet, and used to study diverse nanostructures of stanene, including tubes and ribbons. As a representative case study, using the potential, we perform large-scale molecular dynamics simulations to study stanene’s structure and temperature-dependent thermal conductivity. The room temperature structure of stanene was found to be highly rippled, far in excess of other 2-D materials (e.g., graphene), owing to its low in-plane stiffness (stanene: ~ 25 N/m; graphene: ~480 N/m). The extent of stanene’s rippling also shows stronger temperature dependence compared to that in graphene. Furthermore, stanene-based nanostructures were found to have significantly lower thermal conductivity compared to graphene based structures owing to their softness (i.e. low phonon group velocities) and high anharmonic response. The newly developed BOP will facilitate the exploration of stanene based low dimensional heterostructures for thermoelectric and thermal management applications.
Mathew J. Cherukara, Badri Narayanan, Alper Kinaci, Kiran Sasikumar, Stephen K. Gray,§ Maria K.Y. Chan, and Subramanian K. R. S. Sankaranarayanan*, “Ab Initio-Based Bond Order Potential to Investigate Low Thermal Conductivity of Stanene Nanostructures” J. Phys. Chem. Lett. 2016, 7, 3752.