IMX Seminar Series - Dynamical materials: From atomic scale modeling via machine learning to experiments

The dynamical behavior of materials at the atomic scale, i.e., the motion of individual atoms, is crucial not only for their thermodynamic stability but directly impacts their electronic, optical, and transport properties. Detailed insight into these dynamics is therefore fundamental for understanding and designing materials. By the very nature of the problem the length and time scales involved are extremely short. As a result, atomic scale modeling plays an important role in guiding and interpreting experimental studies, and discovering novel phenomena and mechanisms.
Traditional approaches of condensed matter physics are built on perturbation theory and typically assume a regular (crystalline) reference lattice. As the complexity of materials increases, for example, through dimensional engineering as in the case of layered materials or by integrating organic and inorganic components such as in hybrid perovskites, these approaches reach their applicability limit, both due to the explosion of the degrees of freedom and a failure of the underlying assumptions. Here, the combination of atomic scale simulations, correlation function based analysis, and machine learned potentials is emerging as a tool set that can lead to a paradigm shift in how we approach these questions.
In this presentation, I will show some recent work from my group that showcases the approach and illustrates its potential. In the first part I will focus on extremely anisotropic thermal conductors based on large-area van-der-Waals thin films with random interlayer rotations. They can produce among the highest room-temperature thermal anisotropy ratios, which can be used for very efficient thermal management in electronic devices at the nanometer scale. Using a combination of molecular dynamics simulations, neuroevolution potentials, and correlation function analysis, we are able to quantitatively explain experimental data and reveal a one-dimensional glass-like thermal transport that is concurrent with a two-dimensional crystalline transport mechanism. We show that this behavior is transferable between chemistries and identify a simple descriptor that allows one to predict the dependence of the through-plane conductivity on the rotation angle.
The second part of the presentation will be concerned with the atomic dynamics in perovskites, a very large class of materials with wide-ranging applications in, for example, actuators, sensors, energy harvesting, and optical devices. I will show recent work concerned with the systematic construction of transferable and accurate models for these materials. The latter enable one to quantitatively analyze the dynamics associated with the phase transitions that are pivotal for the unique properties of perovskites. In particular, one can show that the so-called soft modes associated with these transitions exhibit overdamped behavior already hundreds of Kelvin above the actual transition temperature. This gives rise to a pronounced feature in the vibrational density of states in the zero-frequency limit, which is confirmed by quasi-elastic neutron scattering experiments. These results have implications for our understanding of the local structure in these materials, which is important for the electronic and optical properties.
Bio: Paul Erhart graduated from Technische Universität Darmstadt (Germany) in 2006. Starting in 2007 he was first a post-doctoral researcher and later a staff member at Lawrence Livermore National Laboratory in California (USA). He joined the faculty at the Department of Physics at Chalmers in 2011.