Solving response expressions in the ADC/ISR framework

Area of Science:

• Computational chemistry
• Quantum chemistry
• Theoretical chemistry

Background:

• Calculating molecular response properties is crucial for understanding chemical reactions and material properties.
• Existing methods, such as algebraic-diagrammatic construction (ADC), can be computationally intensive, particularly for higher-order models.
• Efficient algorithms are needed to make these calculations feasible for larger systems.

Purpose of the Study:

• To develop and implement a memory-efficient computational approach for molecular response properties.
• To investigate a novel ansatz for the second-order ADC (ADC(2)) model that avoids storing large double excitation amplitudes.
• To compare the performance and convergence of different numerical algorithms within this new framework.

Main Methods:

• Implementation of the algebraic-diagrammatic construction (ADC)/intermediate state representation approach.
• Development of a memory-efficient ansatz for the ADC(2) model, specifically targeting the reduction of memory footprint by avoiding storage of double excitation amplitudes.
• Comparative analysis of various numerical algorithms for solving the response equations associated with the ADC(2) model.

Main Results:

• The proposed memory-efficient ansatz for ADC(2) significantly reduces computational resource requirements.
• The new implementation demonstrates improved convergence behavior for the investigated numerical algorithms compared to standard approaches.
• The developed routines are integrated into an accessible, open-source Python library, facilitating broader use in the scientific community.

Conclusions:

• The developed computational approach offers a more efficient and practical method for calculating molecular response properties using ADC.
• The improvements in memory efficiency and convergence pave the way for applying these methods to more complex chemical systems.
• The open-source implementation democratizes access to advanced computational chemistry tools.