Overlapped embedded fragment stochastic density functional theory for covalently-bonded materials

Area of Science:

• Computational chemistry
• Quantum mechanics
• Materials science

Background:

• Stochastic density functional theory (DFT) offers linear-scaling computational advantages.
• Current methods struggle with delocalized electrons in covalently bonded systems.
• Statistical noise is a key challenge in stochastic DFT.

Purpose of the Study:

• To improve the accuracy of stochastic DFT for covalently bonded systems.
• To address the limitations of existing fragmentation schemes.
• To reduce statistical noise in electronic structure calculations.

Main Methods:

• Developed a novel fragmentation scheme for stochastic DFT.
• Introduced overlapping fragments to interpolate between system parts.
• Analyzed the correlation between statistical error, density, and density matrix.
• Tested the approach on bulk silicon with large supercells.

Main Results:

• The new fragmentation scheme significantly reduces statistical noise.
• Overlapped fragments are effective even for systems with delocalized density matrices.
• Performance was validated on bulk silicon up to 16,384 electrons.
• Demonstrated improved accuracy for covalently bonded materials.

Conclusions:

• The proposed overlapping fragmentation scheme enhances stochastic DFT.
• This method overcomes previous limitations for delocalized electron systems.
• The approach provides a more robust and accurate linear-scaling DFT method.