Unphysical solutions in coupled-cluster-based random phase approximation and how to avoid them

Area of Science:

• Quantum Chemistry
• Computational Chemistry
• Electronic Structure Theory

Background:

• Direct ring coupled-cluster doubles (drCCD)-based random phase approximation (RPA) is a powerful computational framework.
• A known issue involves unphysical solutions, limiting its applicability, especially for small-gap systems.

Purpose of the Study:

• To investigate the origin of multi-solution issues in drCCD.
• To develop methods for validating drCCD solutions and improving their stability.

Main Methods:

• Elucidation of the underlying cause of multi-solution problems in drCCD.
• Development of a practical validation criterion for drCCD solutions.
• Implementation of improved preconditioners using level shifting and regularized MP2 methods.

Main Results:

• The study successfully identified the origin of the multi-solution issue in drCCD.
• A robust validation criterion and effective preconditioners were developed.
• The approach demonstrated effectiveness on challenging systems like stretched bonds, conjugated systems, and metallic clusters.

Conclusions:

• The developed methods enhance the reliability and applicability of drCCD-based RPA.
• This work provides a foundation for stable, large-scale applications of advanced RPA methods.
• The approach is extendable to various advanced RPA formulations.