Another Angle on Benchmarking Noncovalent Interactions

Area of Science:

• Computational Chemistry
• Quantum Mechanics
• Electronic Structure Theory

Background:

• Coupled cluster methods, particularly CCSD(T), are the benchmark for calculating noncovalent interactions.
• Fixed-node quantum Monte Carlo (FN-DMC) shows increasing divergence from CCSD(T) for larger systems, especially those with π-stacking.
• Higher-level post-CCSD(T) methods are computationally expensive, necessitating alternative approaches.

Purpose of the Study:

• To investigate the behavior of electron correlation methods for π-stacked systems.
• To propose a new method for estimating post-CCSD(T) contributions.
• To analyze the divergence of FN-DMC from CCSD(T) in relation to system size.

Main Methods:

• Analysis of the evolution of correlation energy with the number of subunits in π-stacked systems (acene and alkadiene dimers).
• Linear fitting of correlation energy versus subunit number to define a slope as a probe.
• Comparison with higher-level coupled cluster calculations (e.g., CCSDT(Q)) for benzene and naphthalene dimers.

Main Results:

• The correlation energy evolution for π-stacked sequences is found to be nearly linear.
• The slope of this linear trend serves as a reliable probe for electron correlation method behavior.
• CCSD(T) slightly overbinds in dimers, but less so than indicated by FN-DMC results.

Conclusions:

• The linear evolution of correlation energy offers a practical approach to assess electron correlation methods for π-stacked systems.
• This method provides a more accurate estimation of post-CCSD(T) contributions than FN-DMC for these systems.
• The findings refine our understanding of the accuracy of CCSD(T) and FN-DMC for noncovalent interactions in extended π-systems.