Creating Benchmarks for Lithium Clusters and Using Them for Testing and Validation.

Area of Science:

• Computational chemistry
• Quantum mechanics
• Materials science

Background:

• Accurate computational benchmarks are crucial for validating electronic structure theories.
• Small metal clusters, like lithium, offer a tractable system for developing and testing these theories due to their limited electron count.
• Existing benchmarks for cluster energetics are often limited in scope or convergence.

Purpose of the Study:

• To establish fully converged energetic benchmarks for small lithium clusters (Li4 and Li5).
• To rigorously test a diverse range of quantum chemical methods against these benchmarks.
• To provide reliable data for assessing and improving computational methods for metal clusters.

Main Methods:

• Optimization of stationary structures using Kohn-Sham density functional theory (KS-DFT).
• High-level single-point energy calculations using beyond coupled cluster (CC) methods.
• Benchmarking against various methods including KS-DFT, Møller-Plesset (MP) theory, CC theory, composite methods (e.g., G4, WM, W2X, W3X, W3X-L), MC-PDFT, CASPT2, and NEVPT2.

Main Results:

• Identified rhomboid and trigonal bipyramid (TBP) as the most stable structures for Li4 and Li5, respectively.
• Calculated mean unsigned deviations for various methods using the W3X-L method as the reference.
• Determined binding energies and M diagnostics for all evaluated cluster structures.

Conclusions:

• The established benchmarks provide a critical assessment of the accuracy of current electronic structure theories for metal clusters.
• The findings are valuable for the development of new density functionals and machine learning models in computational chemistry.
• This work highlights the importance of high-level, converged calculations for reliable theoretical predictions.