Agenda: Identifying materials with ultra-low thermal conductivities κ is the pivotal challenge in the development of more efficient thermoelectric devices. One strategy to achieve this goal is to find materials with a high level of anharmonicity, and therefore reduced phononlifetime and κ. To help discover such materials, we calculate the anharmonicity of materials ranging from simple binary compounds to complex perovskites using the high-throughput framework FHIvibes [1]. The framework automatically generates an accurate harmonic model for a material’s vibrational properties, from which we determine its anharmonicity by statistically comparing the harmonic and ab initio forces of thermally displaced structures. Our screening not only demonstrates that anharmonicity is more prevalent in material space than previously thought, but also shows that the developed metric strongly correlates with various thermal properties. Using classes of simple binaries as an example, we show that the anharmonicityof a material can be related to its atomic, bulk, and harmonic properties via the sure independence screening and sparsifying operator (SISSO) approach [2], thus facilitating an even more efficient screening.
[1] https://vibes.fhi-berlin.mpg.de
[2] R. Ouyang, et al. Phys. Rev. Mat. 2, 083802 (2018)
Bio: Thomas Purcell began his scientific career at New York University where he received his B.S. in Chemistry and worked as an undergraduate researcher for Prof. Mark Tuckerman calculating protein-ligand binding energies using an enhanced sampling molecular dynamics technique, driven-Adiabatic Free Energy Dynamics (d-AFED). He completed his Ph.D. at Northwestern University under Prof. Tamar Seideman. There he focused on developing classical and semiclassical models to describe the coupling between quantum emitters and plasmonic nanoparticles. Tom then went on to become an Alexander von Humboldt postdoctoral fellow at the Fritz Haber Institute. His current research focuses on building high-throughput computational workflows and machine learning models for understanding thermal transport properties of materials, and developing the new Python-based package FHI-vibes.