Accurate parameterization of the kinetic energy functional

Area of Science:

• Computational Chemistry
• Materials Science
• Quantum Mechanics

Background:

• Orbital-free density functional theory (OF-DFT) is computationally efficient but lacks accurate kinetic energy density functionals.
• This deficiency hinders its application to diverse chemical and material systems.

Purpose of the Study:

• To propose a novel, data-aided learning approach for accurately modeling kinetic energy density.
• To enable the development of more reliable OF-DFT methods.

Main Methods:

• Utilized a data-aided learning paradigm with a dictionary of local and nonlocal functional forms.
• Employed linear regression to determine coefficients for these functionals.
• Introduced two new nonlocal functionals and used basis function expansion to model their kernels.

Main Results:

• Successfully modeled kinetic energy densities and total kinetic energies for various molecular and periodic systems (H2, LiH, LiF, H8 chain).
• Demonstrated the ability to reconstruct kernels using data from only a few structures.
• Achieved accurate predictions for kinetic energy density across different system types.

Conclusions:

• The proposed data-driven method provides a reliable prescription for kinetic energy density modeling.
• This work paves the way for advancing orbital-free density functional theory applications.
• The approach is generalizable and efficient, requiring minimal structural data.