Temperature-transferable tight-binding model using a hybrid-orbital basis.

Area of Science:

• Materials Science
• Computational Physics
• Condensed Matter Physics

Background:

• Finite-temperature calculations are crucial for understanding material properties but are computationally intensive.
• Existing methods like tight-binding and machine learning struggle with temperature transferability and data efficiency.
• Accurate modeling of temperature-dependent electronic properties remains a significant challenge.

Purpose of the Study:

• To develop a novel tight-binding model for efficient and accurate computation of temperature-dependent semiconductor properties.
• To enhance the temperature transferability of electronic structure models in a data-efficient manner.
• To provide a computationally feasible alternative to expensive first-principles calculations for materials at elevated temperatures.

Main Methods:

• Developed a physics-informed tight-binding model using hybrid-orbital basis functions.
• Numerically integrated atomic orbitals to determine distance-dependent matrix elements.
• Optimized model parameters using density functional theory and tested with molecular dynamics simulations.

Main Results:

• The proposed tight-binding model requires a minimal set of parameters, easily optimized.
• The model demonstrates good temperature transferability when applied to molecular dynamics trajectories without explicit temperature fitting.
• Accurate prediction of electronic properties at elevated temperatures for gallium arsenide was achieved, with thermal expansion effects on onsite terms being crucial.

Conclusions:

• The developed tight-binding model offers an efficient and accurate approach for calculating temperature-dependent properties of semiconductors.
• Physics-informed design, particularly incorporating thermal expansion, is key for achieving high accuracy at elevated temperatures.
• This method presents a promising avenue for accelerating materials discovery and design by reducing computational costs.