Multi-center decomposition of molecular densities: A numerical perspective.

Area of Science:

• Computational chemistry
• Quantum chemistry
• Theoretical chemistry

Background:

• Molecular density partitioning is crucial for understanding chemical properties.
• Iterative Stockholder Analysis (ISA) methods are widely used for this purpose.
• Existing ISA methods face challenges with numerical efficiency and accuracy, particularly for charged molecules.

Purpose of the Study:

• To systematically analyze various Iterative Stockholder Analysis (ISA) methods.
• To evaluate the numerical performance and accuracy of the Linear approximation of Iterative Stockholder Analysis (LISA) model.
• To compare LISA with other established ISA variants like Gaussian iterative stockholder analysis and Minimum Basis Iterative Stockholder analysis (MBIS).

Main Methods:

• Systematic derivation of iterative solvers for the unique LISA solution.
• Numerical evaluation of LISA performance on 48 diverse molecules (organic, inorganic, neutral, charged).
• Comparative analysis of LISA against Gaussian iterative stockholder analysis and MBIS.

Main Results:

• LISA-family methods demonstrate superior numerical efficiency and accuracy over comparative methods.
• LISA successfully avoids the anomalous negative atomic charges observed with MBIS for negatively charged molecules.
• Elevated entropy in LISA, caused by limited basis functions, can be mitigated by incorporating additional or supplementary basis functions.

Conclusions:

• LISA presents a promising and robust alternative for molecular density partitioning.
• The findings establish a foundation for future research into the efficiency and chemical accuracy of molecular partitioning schemes.
• LISA's ability to provide accurate atomic charges and its flexibility in basis function incorporation enhance its applicability.