Electron affinities with GPU-accelerated density-fitting EOM-CCSD, approximate EOM-CCSD methods and EOM-CCSD with...

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Materials Science

Background:

• Calculating electron affinities (EAs) is crucial for understanding and designing organic photovoltaic materials.
• Traditional equation-of-motion coupled-cluster singles and doubles (EOMEA-CCSD) methods are computationally expensive.
• Developing efficient computational strategies is essential for accurate EA predictions in larger molecular systems.

Purpose of the Study:

• To develop and implement strategies for reducing the computational cost of EOMEA-CCSD calculations for electron affinities.
• To assess the accuracy of these optimized methods for medium-sized organic molecules relevant to organic photovoltaics.
• To investigate the impact of basis sets and approximation methods on EA calculations.

Main Methods:

• Developed an EOMEA-CCSD program incorporating density-fitting and GPU acceleration for reduced storage and computation time.
• Calculated EAs for 24 organic molecules using aug-cc-pVXZ (X = D, T, Q) basis sets.
• Employed approximate methods like corr-CIS(D∞) and EOMEA-CCSD with frozen natural orbitals (FNOs).

Main Results:

• Basis set incompleteness significantly affects EAs with the DZ set; TZ and QZ sets show good agreement.
• The corr-CIS(D∞) method achieved EAs with a mean absolute deviation (MAD) of ~0.18 eV using the QZ basis.
• FNOs derived from the electron-attached state wavefunction (eaNOs) reduced the MAD to 0.03 eV with ~30% virtual orbitals retained.

Conclusions:

• Optimized EOMEA-CCSD methods, including density-fitting, GPU acceleration, and FNOs, provide efficient and accurate EA calculations.
• The use of eaNOs in FNOs offers a significant reduction in computational cost with minimal loss of accuracy for EAs.
• Approximate methods like corr-CIS(D∞) and FNOs are viable alternatives for high-throughput screening of organic photovoltaic materials.