New Non-Local XC Functionals and High-Throughput Studies of 2D Materials

This thesis introduces a new type of exchange-correlation functional called the Weighted Local Density Approximation (WLDA), describes the development of a variety of tools for high-throughput studies, and finally concludes with a high-throughput study of bilayer materials. It is increasingly important to develop new approximation methods to handle highly correlated materials as well as other systems that are beyond the reach of most commonly used electronic structure methods.

WLDA attempts to improve on correlation effects by matching both the energy and the correlation hole of the homogeneous electron gas. This yields a fully nonlocal functional. Different variants are proposed and several perform comparably to the PBE functional on atomic energies, atomization energies, and lattice constants.

Suggestions for future development and improvement of WLDA are also discussed. When Density Functional Theory (DFT) methods are not sufficient for the task at hand, methods such as many-body perturbation theory must be used. Typically such methods are very computationally expensive which often prohibits the use f these methods in high-throughput projects.

In this thesis, one of the most commonly used many-body perturbation methods, namely G0W0 , in a high-throughput context by analysing the results of more than 60.000 individual self-energy calculations. Problematic materials that require careful handling in a potential high-throughput study are identified.

Errors in the quasiparticle energies stemming from the linear approximation to the quasiparticle equation are investigated and several error-reduction schemes are proposed. The validity of the extrapolation of the plane-wave cutoff to an infinite cutoff and the scissor-operator approximation are also discussed.

As science advances, increasingly complex calculations need to be performed in order to extract the interesting information from the material under investigation. Much of this burden is alleviated by the various electronic structure codes that are available. These typically implement DFT, Quantum Monte Carlo, or Dynamical Mean Field Theory, to name a few.

However, many interesting properties have to be calculated by combining or processing the results of an electronic structure calculation, often in nontrivial ways. This thesis dicsusses the Atomic Simulation Recipes (ASR) Python package and contributions made to ASR. ASR standardizes common electronic structure computational tasks by implementing recipes to perform such tasks.

Finally, this thesis describes a high-throughput screening project for bilayer materials. Monolayers are extracted from the Computational 2D Materials Database and bilayer structures are generated from a heuristic stacking algorithm. A workflow is implemented that calculates bilayer binding energies and various electronic structure properties of the most interesting candidate bilayers.

The candidates are then analysed for emergent and for switchable properties, i.e. properties that change when going from monolayer to bilayer, and properties that change between different stacking orders, respectively.