A near-linear scaling equation of motion coupled cluster method for ionized states

Abstract

In this work, a domain-based local pair natural orbital (DLPNO) version of the equation of motion coupled cluster theory with single and double excitations for ionization potentials (IP-EOM-CCSD) equations has been formulated and implemented. The method uses ground state localized occupied and pair natural virtual orbitals and applies the DLPNO machinery to arrive at a linear scaling implementation of the IP-EOM-CCSD method. The accuracy of the method is controllable using ground state truncation parameters. Using default thresholds, the method predicts ionization potential (IP) values with good accuracy (mean absolute error of 0.08 eV). We demonstrate that our code can be used to compute IP values for systems with more than 1000 atoms and 10 000 basis functions.

Single‐reference coupled cluster methods for computing excitation energies in large molecules: The efficiency and accuracy of approximations

“…Similar conclusions hold for the state‐specific PNO implementation as well as for the state averaged PNO EOM implementation of Valeev . Our own STEOM implementation depends on ground state PNOs, for which standardized parameter sets guaranteeing a certain accuracy are available, and it has been shown that the error in the ground state computation and in the subsequent IP/EA calculations can also be kept below similar limits . The question of accuracy is most complicated for multilayer embedding methods.…”

“…Using ground state PNOs in the IP/EA/STEOM cases can be pushed a step further. A DLPNO implementation of the IP–EOM is straightforward and requires only a few ground state intermediates in the ground state PNO basis to reach near linear scaling. The EA–EOM equations are more problematic because of the truncation of the leading term, as already noted by Krylov and coworkers .…”

“…Before turning to state‐of‐the‐art results with the STEOM–CCSD method, we briefly mention those obtained with the IP/EA–EOM method, since the efficiency of STEOM depends crucially on the performance of these. Figure a demonstrates the largest IP–EOM–CCSD calculation obtained with the near‐linear scaling DLPNO–IP method . Calculating two IP roots of the DNA decamer depicted here using more than 10,000 basis functions takes somewhat more than 9 days on a single core.…”

Solvation effect on the vertical ionization energy of adenine‐thymine base pair: From microhydration to bulk

“…In a recent study, we have reported a three-layer protocol [30] for treating environmental effects in IP, where the chemically relevant part of the system is treated with a multilayer IP-EOM-DLPNO-CCSD method, [30][31][32] the environment in the close vicinity of the active region is treated at the Hartree-Fock (HF) level, and the bulk environment is treated with molecular mechanics. The transition from the IP-EOM-DLPNO-CCSD to HF is smooth, and the partitioning avoids any explicit cutting of bonds, which makes the method equally applicable to the environment of both bonded and nonbonded nature.…”

Reduced‐scaling coupled cluster response theory: Challenges and opportunities

“…For properties related to ground‐state energies, for example, reaction enthalpies and other thermodynamic constants, for which the response or derivative of the wave function is not needed, methods such as PAOs, PNOs, and OSVs have proven to be invaluable for treating large systems with little loss of accuracy compared to canonical approaches. Even for properties such as excitation energies, much progress has been made, and a number of reduced‐scaling schemes have been introduced . However, the application of localization methods to higher‐order response properties is significantly more complex and thus has been much more limited.…”