Volker Heine Young Investigator Award

Predicting material properties with the help of machine learning

A central goal of computational physics and chemistry is to predict material properties using first-principles methods based on quantum mechanics. However, the high computational costs of these methods typically prevent rigorous predictions of macroscopic quantities at finite temperatures, such as chemical potential, heat capacity and thermal conductivity.

In this talk, I will first discuss how to enable such predictions by combining statistical mechanics with data-driven machine learning interatomic potentials. I will give example applications on the systems of water and high-pressure hydrogen. Besides thermodynamic properties, I will also talk about how to compute the heat conductivity of liquids just from equilibrium molecular dynamics trajectories.

During the second part of the talk, I will rationalize why machine learning potentials work at all, and in particular, the locality argument.

Strong light-matter coupling in molecular and extended systems from first principles

Recent research at the interface of material science, chemistry, and quantum optics has offered new possibilities to study light-matter interactions. Combining concepts from these fields presents an opportunity to create a predictive theoretical and computational approach that describes the correlated dynamics of electrons, nuclei, and photons on the same quantized footing.

In this talk, we discuss quantum-electrodynamical density-functional theory (QEDFT) to describe these systems, new exchange-correlation potentials and its approximations. We discuss linear-response theory for QEDFT to access excited states and the incorporation of losses. By considering electrons, nuclei and photons explicitly, we find polaritonically induced vibrational mode mixing and show how chemical reactivity is altered. Beyond molecular systems, we discuss how strong light-matter coupling makes nonlinear-phonon processes more efficient and discuss first principle methods to characterize single-photon emitters. Our work highlights the novel abilities to alter material properties and chemical reaction pathways using strong light-matter coupling.

Invariance principles and topology in the ab initio charge transport of ionic fluids

According to the Green-Kubo theory of linear response [1], the electrical conductivity can be extracted from the time-correlation function of the charge flux in equilibrium molecular dynamics simulations.

I combine a recently-formulated gauge-invariance principle of transport coefficients [2] with arguments from charge-transport quantization [3] to prove that the same ab initio electrical conductivity of ionic fluids is exactly obtained if the time-dependent Born charge tensors, entering the definition of the ab initio charge flux according to the modern theory of polarization (MTP), are replaced by the time-independent integer oxidation numbers of the atoms [4]. Besides providing an alternative method to compute the ab initio electrical conductivity with no need for MTP calculations, this result offers a topology-based definition of oxidation numbers in disordered materials, as I show through numerical experiments on molten potassium chloride. I finally discuss non-stoichiometric systems, where quantized charge transport may occur without a net ionic displacement [5].

The exact factorization approach: From strongly correlated electrons to polaritonic chemistry

The exact factorization approach has recently generated a growing interest for the description of coupled electron-nuclear dynamics and the development of mixed quantum-classical methods. However the idea of separating the wave function into subsystems can be generalized to a plethora of other systems, from N-body electronic systems to the inclusion of photonic degrees of freedom. Here, we first present an exact-factorization-based quantum electronic embedding method to calculate static properties of a many-electron system, and demonstrate its success in capturing strong electron correlation. The method is exact in principle but the practical power lies in utilizing input from a low-level calculation on the entire system in a high-level method computed on a small fragment, as in other embedding methods. In a second part, we present what insights can be obtained in polaritonic chemistry from exact factorization and discuss how it could be used to derive mixed quantum-classical schemes for coupled light-matter problems.

Full Cell Quantum Embedding for Correlated Materials

The quantitative prediction of quantum properties in correlated materials requires simulations without empirical truncations and parameters. We present a method to achieve this goal through a new ab initio formulation of quantum embedding. Instead of using small impurities defined in a low-energy subspace, which require complicated downfolded interactions which are often approximated, we describe the full cell dynamical mean-field theory (DMFT), where the impurities comprise all orbitals of atoms in a unit cell or supercell of the crystal. Our formulation results in large impurity problems, which we treat here with efficient quantum chemistry impurity solvers including the coupled-cluster theory and DMRG, combined with a GW treatment of long-range Coulomb interactions. We apply our approach to study correlated transition metal oxides such as nickel oxide and cuprate superconductors, and find the predicted spectral properties are in good agreement with the experiment.