Partitioned EOMEA-MBPT(2): An Efficient N⁵ Scaling Method for Calculation of Electron Affinities

Abstract

We present an N(5) scaling modification to the standard EOMEA-CCSD method, based on the matrix partitioning technique and perturbative approximations. The method has lower computational scaling and smaller storage requirements than the standard EOMEA-CCSD method and, therefore, can be used to calculate electron affinities of large molecules and clusters. The performance and capabilities of the new method have been benchmarked with the standard EOMEA-CCSD method, for a test set of 20 small molecules, and the average absolute deviation is only 0.03 eV. The method is further used to investigate electron affinities of DNA and RNA nucleobases, and the results are in excellent agreement with the experimental values.

Electron attachment to DNA and RNA nucleobases: An EOMCC investigation

“…For the accurate prediction of EA, one need to use large basis set with diffuse functions and this makes calculation of EOMEA‐CCSD very expensive and unlike EOMIP‐CCSD method, EOMEA‐CCSD method involves 4‐particle intermediates, which has large storage requirement and slows down the calculations significantly because of I/O bottleneck. However, very recently, a lower scaling and 4‐particle intermediates free approximation to EOMEA‐CCSD is developed, which can be applied to large systems …”

Lower scaling approximation to EOM‐CCSD: A critical assessment of the ionization problem

“…The method is size‐extensive for every value of n and the lowest order of approximation to it leads to EOM‐CCSD(2) approximation with MBPT(1) ground state wave function and MBPT(2) ground state energy. Similar ideas were persuaded by Bartlett and coworkers in the context of excitation energy, and Dutta et al for electron affinity, and spin‐flip variants of EOM‐CC.…”

“…In P‐EOM‐CCSD(2) approach the doubles‐doubles block of the matrix is approximated as diagonal. It has been observed that partitioned version of EOM‐CCSD(2) method provides an improvement in results compared to the standard EOM‐CCSD(2) method for both EA and EE . Therefore, it would be interesting to extend the same idea to IP problem.…”

“…It is possible to further approximate the EOMIP‐CCSD(2) method, by using a diagonal approximation for the doubles‐doubles block. The idea is similar to the P‐EOM‐MBPT(2) method developed for EE and EA . Pal and coworkers have developed a more accurate version of EOMIP‐CCSD(2), which gives results comparable to the standard EOMIP‐CCSD for both IP and properties.…”

Perturbative order analysis of the similarity transformed Hamiltonian in Fock-space coupled cluster theory: difference energy and electric response properties

“…However, a second order approximation (n = 2) leads to EOM-CCSD(2) method and provides significant advantage in terms of reducing the scaling as well as storage requirements [59]. Therefore, the EOMCCSD(2) approximation, in its original form as well as with further modifications, has been extensively used to generate lower computational scaling methods [49][50][51][52] for the calculation of IP, EA and EE [55,57,60]. It has been shown to give excellent performance for potential energy surfaces, as well as other complex multireference situations, despite of its lower computational cost.…”