Analytical gradients of random-phase approximation plus corrections from renormalized single excitations
Muhammad N. Tahir, Honghui Shang, Xinguo Ren · 2025
Read Analytical gradients of random-phase approximation plus corrections from renormalized single excitations by Muhammad N. Tahir, Honghui Shang, Xinguo Ren (2025) 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-2505-07357.txt.
Introduction
Muhammad Tahir, Honghui Shang, and Xinguo Ren work out a missing piece of a popular method for computing the energies of molecules and materials. The method, the random-phase approximation, tends to bind molecules a little too weakly, and a correction called renormalized single excitations fixes much of that. Formulas for how the plain method's energy changes as atoms move were already known, which allows shape searches and simulations, but the correction's version was missing.
The team derives and codes that missing gradient within an atomic-orbital framework. Tests show the correction slightly strengthens the plain method's overbinding for small linked molecules, while fixing its behaviour for pairs held together only by weak attraction. On water clusters the gains are modest but real, cutting the error on a standard test set.
Catalog record
| CREATOR | Muhammad N. Tahir, Honghui Shang, Xinguo Ren |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 79 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-07357
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Analytical gradients of random-phase approximation plus corrections from renormalized single excitations Muhammad N. Tahir, Honghui Shang, Xinguo Ren Abstract The random-phase approximation (RPA) formulated within the adiabatic connection fluctuation-dissipation framework is a powerful approach to compute the ground-state energies and properties of molecules and materials. Its overall underbinding behavior can be effectively mitigated by a simple correction term, called renormalized single excitation (rSE) correction. Analytical gradient calculations of the RPA energy have become increasingly available, enabling structural relaxations and even molecular dynamics at the RPA level. However, such calculations at the RPA+rSE level have not been reported, due to the lack of the rSE analytical gradient. Here, we present the first formulation and implementation of the analytical gradients of the rSE energy with respect to the nuclear coordinates within an atomic-orbital basis set framework, which allows us to assess the performance of RPA+rSE in determining the molecular geometries and energetics. It is found that the slight overestimation behavior of RPA for small covalently bonded molecules is strengthened by rSE, while such behavior for molecules bonded with purely dispersion interactions is corrected. We further applied the approach to the water clusters, and found that the energy difference between the low-energy isomers of water hexamers is almost unchanged when going from RPA to RPA+rSE geometries. For the bigger WATER27 test set, using the RPA+rSE geometries instead of the RPA ones leads to a slight reduction of the mean absolute error of RPA+rSE from 0.91 kcal/mol to 0.70 kcal/mol, at the complete basis set. keywords American Chemical Society, L a T e X † † affiliation: Institute of Physics, Chinese Academy of Sciences, Beijing 100190, China † † email: [email protected] † † affiliation: State Key Laboratory of Precision and Intelligent Chemistry, University of Science and Technology of China, Hefei 230026, China † † email: [email protected] † † email: [email protected] † † affiliation: Institute of Physics, Chinese Academy of Sciences, Beijing 100190, China † † affiliation: The NOMAD Laboratory at the Fritz Haber Institute of the Max Planck Society, Faradayweg 4-6, D-14195 Berlin, Germany † † abbreviations: IR,NMR,UV † † suppinfo: T 1 Introduction Today, density functional theory (DFT) is the most popular method of choice for describing molecules and materials from first principles. However, the accuracy and reliability of DFT calculations depend crucially on the actual approximations to the underlying exchange-correlation (XC) functional. The XC approximations sitting on the first four rungs of the “Jacob’s ladder” Perdew and Schmidt 2001 , including local/semi-local Kohn and Sham 1965 ; Perdew et al. 1996 and hybrid functionals Becke 1993 ; Heyd et al. 2003 , do not capture nonlocal electron correlations which are indispensable for describing, e.g., van der Waals (vdW) interactions. Therefore, practical DFT calculations with conventional functionals have to be complemented with empirical or semi-empirical vdW corrections Grimme 2006 ; Becke and Johnson 2007 ; Tkatchenko and Scheffler 2009 to describe systems where non-covalent interactions are important. However, for such schemes, a clean treatment of the intermediate, density-overlapping regime is always a challenge. A seamless, automatic inclusion of non-local electron correlations can be attained by going to the fifth rung of the Jacob’s ladder, where a prominent functional is the random phase approximation (RPA) Bohm and Pines 1953 ; Gell-Mann and Brueckner 1957 formulated via the framework of adiabatic connection fluctuation dissipation theorem (ACFDT) Langreth and Perdew 1977 ; Gunnarsson and Lundqvist 1976 . It has been shown that RPA provides a balanced description of different bonding interactions in complex chemical environments Ren et al. 2012 ; Eshuis et al. 2012 . In particular, RPA excels at discerning the ordering of the adsorption energies at different surface sites Ren et al. 2009 ; Schimka et al. 2010 and the energy differences between different isomers/polymorphs of the molecular and crystalline systems Lebègue et al. 2010 ; Zhang et al. 2019 ; Sengupta et al. 2018 ; Zhang et al. 2018 ; Cazorla and Gould 2019 ; Yang and Ren 2022 ; Chedid et al. 2021 ; Tahir et al. 2022 . Due to its attractive performance, it is natural to go beyond single-point energy calculations by enabling structural relaxation and molecular dynamics simulations based on the RPA potential energy surface. To this end, much effort has been devoted to developing computational formalisms and computer codes to evaluate the analytical gradient of RPA energies with respect to atomic displacements Rekkedal et al. 2013 ; Mussard et al. 2014 ; Burow et al. 2014 ; Ramberger et al. 2017 ; Beuerle and Ochsenfeld 2018 ; Chedid et al. 2021 ; Tahir et al. 2022 ; Tahir et al. 2024 . Efficient implementations of the RPA forces not only allow direct determination of the RPA structural and dynamical properties, but also facilitate the development of machine-learning force fields at the RPA level Liu et al. 2022 ; Verdi et al. 2023 . However, the standard RPA scheme still suffers from some well-known shortcomings. Most notably, RPA shows a systematic underestimation of the binding energies of both molecules and solids. To address this issue, various schemes to correct the RPA have been developed, from various perspectives such as ACFDT Klimeš et al. 2015 ; Erhard et al. 2016 ; Görling 2019 , many-body perturbation theory (MBPT) Ren et al. 2011 ; Paier et al. 2012 ; Bates and Furche 2013 , time-dependent DFT Olsen and Thygesen 2012 , and coupled cluster theory Grüneis et al. 2009 ; Paier et al. 2010 . Among these, a very simple and effective correction to RPA is the renormalized single excitation (rSE) correction scheme. The concrete expression of rSE can be derived from MBPT Ren et al. 2011 ; Ren et al. 2013 or the framework of ACFDT Klimeš et al. 2015 , and it accounts for the variation of the density matrix along the adiabatic correction path. In terms of practical performance, RPA+rSE cures the underbinding problem of RPA to a large extent for both molecules and solids Ren et al. 2012 ; Paier et al. 2012 ; Ren et al. 2013 . It particularly shows excellent performance for describing the hydrogen bonding Ren et al. 2013 , on par with the coupled cluster method. It should be noted that rSE achieves this with a marginal extra cost. As such, RPA+rSE can be taken as a natural starting point to develop more sophisticated computational schemes Pham et al. 2024 . In this connection, it would be highly desirable to calculate the analytical gradients at the level of RPA+rSE. To the best of our knowledge, such an implementation is not yet available. Since the rSE term is computationally rather cheap, one would expect that evaluating RPA+rSE gradients does not increase the computational overhead much. Efficient implementation of RPA+rSE gradients will allow us to examine whether the improved binding energies will lead to better structural properties. Previously, we have shown that the binding energies of water clusters have been significantly improved by adding the rSE correction to RPA Tahir et al. 2022 ; Tahir et al. 2024 . However, in these studies, the structures of the water clusters are still determined at the RPA level. The question is: What will happen if the structures are also consistently determined by RPA+rSE? To address these