Effects of self-consistent extended Hubbard interactions and spin-orbit couplings on energy bands of semiconductors and topological insulators

Wooil Yang,Young-Woo Son
2024-06-18
Abstract:A first-principles computational method with self-consistent on-site and inter-site Hubbard functionals is able to treat local and non-local Coulomb interactions on an equal footing. To apply the method to understand solids with strong spin-orbit coupling (SOC), we have extended a psuedohybrid functional approach developed by Agapito-Curtarolo-Buongiorno Nardelli to implement self-consistent extended Hubbard energy functionals for noncollinear spin states. With this, energy bands of semiconductors with various SOC strengths such as Si, Ge, GaAs, GaSb, CdSe and PdO are obtained, agreeing with results from fully relativistic $GW$ approximation (FR-GWA) as well as experiments. We also compute energy gaps of HgTe, CuTlS$_2$, and CuTlSe$_2$ and assign them to be topological insulators correctly, unlike characteristic failures for judging topological properties from typical hybrid functionals. We demonstrate feasibility of our method to handle large systems by computing surface bands of topological insulators, Bi$_{\text{2}}$Se$_{\text{3}}$ and Bi$_2$Te$_3$ with varying thickness up to eight quintuple layers. Considering its low computational cost comparable to conventional {\it ab intio} methods and improved accuracy to FR-GWA, we expect that our method provides an opportunity to study large scale correlated systems with the strong SOC efficiently and reliably.
Materials Science
What problem does this paper attempt to address?
The main problem that this paper attempts to solve is the deficiencies of existing computational methods in dealing with semiconductors and topological insulators with strong spin - orbit coupling (SOC). Specifically: 1. **Limitations of the traditional DFT method**: Traditional methods based on density functional theory (DFT), such as LDA or GGA, have self - interaction errors (SIE) when dealing with materials containing heavy elements, which makes it difficult for them to accurately describe the electronic properties of these materials. 2. **Limitations of the DFT + U method**: Although the DFT + U method partially solves the SIE problem by introducing a local Coulomb repulsion term (U), this method is still insufficient for cases where orbital hybridization or non - local Hubbard interactions (V) need to be considered. 3. **The need for an extended Hubbard model**: In order to more accurately describe materials with strong spin - orbit coupling, a method is required that can handle both local and non - local Coulomb interactions simultaneously and can combine the spin - orbit coupling effect within a self - consistent framework. To this end, the authors propose a new computational method, that is, combining the extended Hubbard model (including U and V terms) and spin - orbit coupling (SOC) on the basis of DFT, and applying the Agapito - Curtarolo - Buongiorno Nardelli (ACBN0) pseudo - hybrid functional method to deal with non - collinear spin states. This method aims to provide an efficient and reliable ab - initio tool for studying large - scale correlated systems containing heavy atoms. ### Main objectives - **Improve the computational accuracy of band gaps and spin - orbit splitting**: By introducing extended Hubbard interactions and spin - orbit coupling, improve the computational accuracy of band gaps and spin - orbit splitting in semiconductors and topological insulators. - **Correctly judge topological properties**: Compared with traditional hybrid functionals, the new method can more accurately identify the topological properties of materials such as HgTe, CuTlS₂ and CuTlSe₂. - **Feasibility of dealing with large systems**: Verify the feasibility of this method in dealing with large systems when calculating the surface states of materials such as Bi₂Se₃ and Bi₂Te₃. Through these improvements, the authors expect their method to provide high - precision calculations consistent with the full - relativistic GW approximation (FR - GWA) and experimental results while maintaining a relatively low computational cost.