Random phase approximation-based local natural orbital coupled cluster theory

Ruiheng Song, Xiliang Gong, Aamy Bakry, Hong-Zhou Ye · 2026

Read Random phase approximation-based local natural orbital coupled cluster theory by Ruiheng Song, Xiliang Gong, Aamy Bakry, Hong-Zhou Ye (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-00131.txt.

Introduction

Ruiheng Song, Xiliang Gong, Aamy Bakry, and Hong-Zhou Ye work on faster quantum chemistry for large molecules. Local correlation methods split a system into pieces, but they depend heavily on a cheaper low-level theory that defines each piece and handles far-off effects. The standard low-level choice, called MP2, is known to fail for some systems. Their 2026 paper tests a different choice inside a local natural orbital coupled-cluster method.

The alternative, the random phase approximation, matches the standard version's accuracy for systems with healthy energy gaps. For metallic systems it converges much faster toward the full coupled-cluster answer, above all as the system grows large. The authors argue the low-level choice deserves more care, and name their alternative as a strong candidate.

Catalog record

CREATORRuiheng Song, Xiliang Gong, Aamy Bakry, Hong-Zhou Ye
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE45 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-00131

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Random phase approximation-based local natural orbital coupled cluster theory Ruiheng Song, Xiliang Gong, Aamy Bakry, Hong-Zhou Ye Abstract Practical applications of fragment embedding and closely related local correlation methods critically depend on a judicious choice of a low-level theory to define the local embedding subspace and to capture long-range electrostatic and correlation effects outside the embedding region. Second-order Møller-Plesset perturbation theory (MP2) is by far the most widely used correlated low-level theory; however, its applicability becomes questionable in systems where MP2 is known to fail either quantitatively or qualitatively. In this work, we present the random phase approximation (RPA) as a promising alternative low-level theory to MP2 within the local natural orbital-based coupled-cluster (LNO-CC) framework. We demonstrate that RPA-based LNO-CC closely matches the performance of its MP2-based counterpart for systems with sizable energy gaps, while delivering significantly faster convergence toward the canonical coupled-cluster limit for metallic systems, particularly as the thermodynamic limit is approached. These results highlight the critical role of the low-level theory in fragment embedding and local correlation methods and identify RPA as a compelling alternative to the commonly used MP2. I Introduction Quantum embedding and fragment-based local correlation methods mitigate the steep computational cost of high-accuracy electronic structure theories such as coupled cluster Bartlett and Musiał 2007 (CC) by applying these methods only to chemically meaningful local fragments Sun and Chan 2016 ; Herbert 2019 ; Lee et al. 2019 ; Wasserman and Pavanello 2020 . Long-range electrostatic and correlation effects outside the fragments are instead treated with a low-level theory, rendering the overall approach computationally tractable. Representative examples include the use of one-body effective potentials in wavefunction-in-density functional theory embedding Manby et al. 2012 ; Daday et al. 2014 ; Jacob and Neugebauer 2014 ; Libisch, Huang, and Carter 2014 ; Lee et al. 2019 , mean-field or correlated bath orbitals in density matrix embedding theory Knizia and Chan 2013 ; Wouters et al. 2016 ; Cui, Zhu, and Chan 2020 (DMET) and its variants Ye et al. 2019 ; Cho et al. 2025 ; Nusspickel and Booth 2022 ; Shee et al. 2024 ; Hermes, Pandharkar, and Gagliardi 2020 ; King et al. 2025 , and correlated natural orbitals in the local natural orbital Rolik and Kállay 2011 ; Rolik et al. 2013 ; Nagy and Kállay 2017 ; Nagy and Kállay 2019 ; Nagy 2024 ; Ye and Berkelbach 2024a (LNO) extension of cluster-in-molecule Li, Ma, and Jiang 2002 ; Li and Li 2004 ; Li et al. 2006 ; Li et al. 2009 ; Wang et al. 2019 ; Wang et al. 2022 (CIM) methods. In all such approaches, the choice of the low-level theory plays a pivotal role, as it controls the rate of convergence with fragment size and thus largely determines the practical efficiency of the method. Second-order Møller-Plesset perturbation theory Møller and Plesset 1934 (MP2) is by far the most widely used correlated low-level theory in quantum embedding and local correlation methods. MP2-based natural orbitals and long-range correlation corrections have a long history in local CC theories Neese, Hansen, and Liakos 2009 ; Neese, Wennmohs, and Hansen 2009 ; Hansen, Liakos, and Neese 2011 ; Huntington et al. 2012 ; Liakos et al. 2015 ; Werner and Schütz 2011 . More recently, MP2-derived natural orbitals have also been adopted as correlated bath states in DMET Nusspickel and Booth 2022 ; Nusspickel, Ibrahim, and Booth 2023 and dynamical mean-field theory Li and Zhu 2024 (DMFT), where they alleviate the disentanglement problems associated with mean-field baths Ye and Van Voorhis 2019 and enable an effective description of dynamical correlation in molecules and insulating solids. Despite these successes, the use of MP2 as a low-level theory becomes questionable in systems where MP2 is quantitatively inaccurate or even formally invalid. Two prominent examples, which are the focus of the present work, are noncovalently bound molecules and molecular solids—where MP2 significantly overestimates intermolecular interaction energies Distasio JR. and Head-Gordon 2007 ; Cybulski and Lytle 2007 ; Beran 2016 ; Liang, Ye, and Berkelbach 2023 —and bulk metals, for which the MP2 correlation energy diverges in the thermodynamic limit Shepherd and Grüneis 2013 ; Masios et al. 2023 ; Neufeld and Berkelbach 2023 (TDL). The random phase approximation Bohm and Pines 1953 ; Gell-Mann and Brueckner 1957 ; Furche 2001 (RPA) provides a promising alternative to MP2 in both of these contexts Ren et al. 2012 ; Chen et al. 2017 ; Zhang and Xu 2021 . By resumming the direct particle-hole ring diagrams to infinite order, RPA incorporates dynamical screening of the bare Coulomb interaction, leading to improved descriptions of noncovalent interactions Harl and Kresse 2008 ; Lu et al. 2009 ; Del Ben, Hutter, and VandeVondele 2013 ; Klimeš 2016 ; Yao and Kanai 2021 ; Hellgren and Baguet 2021 ; Pham, Modrzejewski, and Klimeš 2023 ; Liang et al. 2025 and metallic systems Ren, Rinke, and Scheffler 2009 ; Harl and Kresse 2009 ; Schimka et al. 2010 ; Olsen et al. 2011 ; Rohlfing and Bredow 2008 ; Ma et al. 2011 ; Kim et al. 2012 ; Schmidt and Thygesen 2018 ; Sheldon et al. 2024 . Within embedding frameworks, RPA has been most extensively explored in the form of constrained RPA (cRPA) Aryasetiawan et al. 2004 ; Şaşıoğlu, Friedrich, and Blügel 2011 ; Chang et al. 2024 ; Scott and Booth 2024 for Hamiltonian downfolding, where the RPA polarizability is used to construct effective screened interactions for a chosen active space. More recently, Schäfer and co-workers have demonstrated that RPA natural orbitals, obtained from first-order Green’s functions Ramberger et al. 2019 , can be used to accelerate embedding-based CC calculations Schäfer et al. 2021a ; Schäfer et al. 2021b . Callahan and co-workers have demonstrated the use of RPA-based composite energy corrections to accelerate the convergence of CC calculations for uniform electron gas Callahan, Lange, and Berkelbach 2021 . In this work, we present an alternative route for incorporating RPA into fragment embedding and local correlation theories. Our approach is based on the direct ring coupled-cluster doubles (drCCD) formulation of RPA Scuseria, Henderson, and Sorensen 2008 , which provides direct access to RPA wavefunction amplitudes and can therefore be integrated seamlessly into existing MP2-based embedding and local correlation frameworks. We report an efficient implementation of RPA-based natural orbitals based on the factorized drCCD equation by Heßelmann Heßelmann 2012 and Kállay Kállay 2015 with asymptotic cost scaling more favorable than that of conventional MP2-based approaches. The effectiveness of RPA as a low-level theory is demonstrated within the LNO-CC framework, originally developed by Kállay and co-workers for molecules Rolik and Kállay 2011 ; Rolik et al. 2013 ; Nagy and Kállay 2017 ; Nagy and Kállay 2019 and recently extended to both insulating and metallic solids by one of the authors Ye and Berkelbach 2023 ; Ye and Berkelbach 2024b ; Ye and Berkelbach 2024a . Our numerical results for noncovalently bound molecules, molecular solids, and bulk metals indicate that RPA-derived LNOs are of comparable quality to their MP2-based counterparts, while composite energy corrections based on RPA and the closely related second-order screened exchange Grüneis et al. 2009 (SOSEX) approximation lead to systematically faster convergence with respect to LNO truncation than MP2-based

Read the complete text