Random phase approximation-based local natural orbital coupled cluster theory.

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Materials Science

Background:

• Fragment embedding and local correlation methods are crucial for accurate molecular simulations.
• The choice of low-level theory significantly impacts the performance of these methods.
• Second-order Møller-Plesset perturbation theory (MP2) is widely used but has limitations for certain systems.

Purpose of the Study:

• To evaluate Random Phase Approximation (RPA) and Second-Order Screened Exchange (SOSEX) as alternatives to MP2 in local natural orbital-based coupled-cluster (LNO-CC) methods.
• To assess the performance of RPA and SOSEX in fragment embedding for various chemical systems.

Main Methods:

• Implementation and application of RPA and SOSEX within the LNO-CC framework.
• Benchmark calculations on noncovalent complexes, crystals, reaction barriers, and bulk metals.
• Comparison with MP2-based LNO-CC calculations at coupled-cluster singles doubles triples [CCSD(T)] and coupled-cluster singles doubles (CCSD) levels.

Main Results:

• RPA- and SOSEX-based LNO-CC methods closely match MP2 performance at the LNO-CCSD(T) level.
• At the LNO-CCSD level, RPA and SOSEX demonstrate significantly faster convergence towards the canonical CCSD limit compared to MP2.
• These improvements are particularly notable for metallic systems near the thermodynamic limit.

Conclusions:

• RPA and SOSEX are effective and promising alternatives to MP2 for low-level theory in fragment embedding and local correlation methods.
• The choice of low-level theory is critical for the accuracy and efficiency of local correlation methods.
• RPA emerges as a compelling alternative to MP2, offering enhanced performance, especially for challenging systems.