Third-order algebraic diagrammatic construction theory for electron attachment and ionization energies: Conventional and Green's function implementation

TL;DR

This paper develops and benchmarks second- and third-order algebraic diagrammatic construction (ADC) methods for accurately computing electron attachment and ionization energies, introducing efficient algorithms including a Green's function approach for large systems.

Contribution

The work introduces new implementation strategies for EA-/IP-ADC(2,3), including a Green's function algorithm, and demonstrates their accuracy and efficiency on diverse molecular systems.

Findings

EA-/IP-ADC(2,3) are accurate for small molecules and nucleobases.

Green's function algorithm efficiently computes density of states.

Methods successfully estimate band gaps in large hydrogen chains.

Abstract

We present implementation of second- and third-order algebraic diagrammatic construction theory for efficient and accurate computations of molecular electron affinities (EA), ionization potentials (IP), and densities of states (EA-/IP-ADC(n), n = 2, 3). Our work utilizes the non-Dyson formulation of ADC for the single-particle propagator and reports working equations and benchmark results for the EA-ADC(2) and EA-ADC(3) approximations. We describe two algorithms for solving EA-/IP-ADC equations: (i) conventional algorithm that uses iterative diagonalization techniques to compute low-energy EA, IP, and density of states, and (ii) Green's function algorithm (GF-ADC) that solves a system of linear equations to compute density of states directly for a specified spectral region. To assess accuracy of EA-ADC(2) and EA-ADC(3), we benchmark their performance for a set of atoms, small molecules, and five DNA/RNA nucleobases. As our next step, we demonstrate efficiency of our GF-ADC implementation by computing core-level K-, L-, and M-shell ionization energies of a zinc atom without introducing core-valence separation approximation. Finally, we use EA- and IP-ADC methods to compute band gaps of equally-spaced hydrogen chains Hn with n up to 150, providing their estimates near thermodynamic limit. Our results demonstrate that EA-/IP-ADC(n) (n = 2, 3) methods are efficient and accurate alternatives to widely used electronic structure methods for simulations of electron attachment and ionization properties.