Automatic Molecule Fragmentation for Density Matrix Embedding Theory.

Area of Science:

• Quantum Chemistry
• Computational Chemistry
• Materials Science

Background:

• High-accuracy quantum chemical calculations for large molecules are computationally expensive.
• Quantum embedding methods like Density Matrix Embedding Theory (DMET) and Bootstrap Embedding (BE) reduce computational cost.
• DMET's accuracy and cost depend on manual molecular fragmentation, limiting its practical application.

Purpose of the Study:

• To develop a Graph-based Automatic molecule Fragmentation (GAF) technique for easier application of DMET.
• To evaluate the accuracy and computational efficiency of GAF-DMET compared to atom-based BE (ABE).
• To demonstrate GAF-DMET's suitability for chemical binding energy and reaction calculations.

Main Methods:

• Representing molecular structures as graphs with interatomic interaction edge weights.
• Solving graph partitioning problems to determine optimal molecular fragmentation.
• Developing metrics for interatomic interactions across different basis sets and automatic fragment number adjustment.
• Comparing GAF-DMET and ABE performance on 14 small molecules, including binding energy and S_N2 reaction calculations.

Main Results:

• GAF successfully identifies accurate molecular fragmentation patterns.
• GAF-DMET achieves comparable or higher accuracy than ABE with reduced wall-clock times.
• GAF-DMET demonstrates superior accuracy in binding energy calculations and S_N2 reaction simulations compared to ABE.

Conclusions:

• GAF provides an automated and effective fragmentation strategy for DMET.
• GAF-DMET offers a computationally efficient and accurate alternative to existing embedding methods.
• The proposed GAF technique enhances the applicability of DMET for complex chemical systems and reactions.