Abstract:A new method is presented that allows for efficient evaluation of spin-orbit coupling (SOC) in density-functional theory calculations. In the so-called second-variational scheme, where Kohn-Sham functions obtained in a scalar-relativistic calculation are employed as a new basis for the spin-orbit-coupled problem, we introduce a rich set of local orbitals as additional basis functions. Also relativistic local orbitals can be used. The method is implemented in the all-electron full-potential code \exciting. We show that, for materials with strong SOC effects, this approach can reduce the overall basis-set size and thus computational costs tremendously.
What problem does this paper attempt to address?
### Problems Addressed by the Paper
This paper proposes a new method aimed at efficiently handling spin-orbit coupling (SOC) in density functional theory (DFT) calculations. Specifically, this method uses Kohn-Sham functions obtained from scalar relativistic calculations as a new basis in the so-called second variational scheme, introducing a set of rich local orbitals as additional basis functions. Additionally, relativistic local orbitals can also be used.
### Background and Motivation
Spin-orbit coupling (SOC) is crucial for accurate electronic structure calculations of many materials. For example, SOC can explain the lifting of degeneracy of low-energy excitons in transition metal dichalcogenides (TMDCs), the opening of a small bandgap in graphene, and the significant reduction of the fundamental bandgap in halide perovskites. However, the impact of SOC is not limited to changing the characteristics of electronic bands; it also affects bond lengths, phonon energies, and can even convert deep defects into shallow ones.
In DFT calculations, different methods and codes handle SOC in various ways. The commonly used full-potential linearized augmented plane wave (LAPW) method is often used as a reference. For low-energy core orbitals, the standard method is to solve the radial four-component Dirac equation, assuming a spherically symmetric potential. For semicore and valence electrons, a two-step procedure is commonly used: first solving the Kohn-Sham problem under the scalar relativistic approximation (first variational, FV), and then using the FV wave functions as a basis to construct the solution of the full problem including SOC (second variational, SV).
While this two-step procedure is effective and efficient for many materials, some materials require more complex calculations. For example, Bi₂Te₃ requires considering unoccupied bands at least 8 Ry above the Fermi level to obtain reliable results. In halide perovskites, the full set of Kohn-Sham orbitals is needed for convergence. These examples indicate that SOC cannot be considered a small perturbation in some materials.
### Proposal of the New Method
This paper introduces a new method called the second variational local orbital method (SVLO), which leverages the strong relativistic effects around atomic nuclei. Compared to the standard SV method, the SVLO method improves computational efficiency by increasing the basis flexibility in these regions. Specifically, the SVLO method uses FV wave functions and a rich set of local orbitals to represent the solution of the full problem. All local orbitals are treated as explicit basis functions, including Dirac-type local orbitals obtained from solving the Dirac equation.
### Method Validation
Based on the implementation in the all-electron full-potential package exciting, the authors validated the effectiveness of the method in band structure and total energy calculations for Xe, MoS₂, PbI₂, γ-CsPbI₃, and Bi₂Te₃.
### Results
- **Xe**: The SVLO method keeps the total energy difference within 2 × 10⁻³ eV/atom when using a number of basis functions comparable to the number of local orbitals, whereas the SV method requires all available FV states to achieve a value an order of magnitude larger (7 × 10⁻² eV/atom).
- **MoS₂**: SOC only reduces the bandgap by 0.07 eV, but the splitting of the VBM (0.15 eV) is crucial, as neglecting this splitting would lead to an unphysical prediction of an indirect bandgap.
- **PbI₂**: SOC reduces the bandgap by 0.54 eV and is crucial for describing the optical spectrum at the K point of the Brillouin zone.
- **CsPbI₃**: SOC reduces the bandgap from 1.64 eV to 0.82 eV, mainly due to a 0.71 eV splitting of the CBm.
- **Bi₂Te₃**: SOC significantly alters the band structure, especially at the Γ point, where SOC leads to an inverted characteristic of the VBM and CB. Adding p₁/₂-type local orbitals further reduces the bandgap and makes it an indirect bandgap.
### Conclusion
The new method SVLO proposed in this paper can significantly reduce the overall number of basis functions when dealing with materials with strong SOC effects, thereby greatly reducing computational costs. The method has been validated in band structure and total energy calculations for various materials, demonstrating its advantages in practical applications.