Coupling between thermochemical contributions of subvalence correlation and of higher-order post-CCSD(T) correlation effects — a step toward ‘W5 theory’

Aditya Barman, Gregory H. Jones, Kaila E. Weflen, Margarita Shepelenko, Jan M. L. Martin · 2026

Read Coupling between thermochemical contributions of subvalence correlation and of higher-order post-CCSD(T) correlation effects — a step toward ‘W5 theory’ by Aditya Barman, Gregory H. Jones, Kaila E. Weflen, Margarita Shepelenko, Jan M. L. Martin (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-2601-02294.txt.

Introduction

Aditya Barman and four co authors, including Jan M. L. Martin, refine how chemists compute the total atomization energy of a molecule, the sum of all its bond energies. Very accurate methods build the answer from many correction layers. The question here is whether two layers, correlation from inner shells and higher order correlation beyond the standard CCSD(T) method, simply add up or interact.

They find the two layers do interact, most strongly when several second row atoms sit next to each other, and in systems with strong static correlation. Recomputing molecular shapes with the inner shell layer included raises energies across the board. From the results they propose a first W5 theory protocol and give revised reference energies for several key molecules, matching the active thermochemical tables well.

Catalog record

CREATORAditya Barman, Gregory H. Jones, Kaila E. Weflen, Margarita Shepelenko, Jan M. L. Martin
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE73 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-02294

Browse the full library · About the Tower of Babel

Full text

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

Coupling between thermochemical contributions of subvalence correlation and of higher-order post-CCSD(T) correlation effects — a step toward ‘W5 theory’ Aditya Barman, Gregory H. Jones, Kaila E. Weflen, Margarita Shepelenko, Jan M. L. Martin Abstract We consider the thermochemical impact of post-CCSD(T) contributions to the total atomization energy (TAE, the sum of all bond energies) of first- and second-row molecules, and specifically their coupling with the subvalence correlation contribution. In particular, we find large contributions from (Q) when there are several neighboring second-row atoms. Otherwise, both higher-order triples T 3 T_{3} –(T) and connected quadruples (Q) are important in systems with strong static correlation. Reoptimization of the reference geometry for core-valence correlation increases the calculated TAE across the board, most pronouncedly so for second-row compounds with neighboring second-row atoms. We present a first proposal for a "W5 theory" protocol and compare computed TAEs for the W4-08 benchmark with prior reference values. For some key second-row species, the new values represent nontrivial revisions. Our predicted TAE 0 values (TAE at 0 K) agree well with the ATcT (active thermochemical tables) values, including for the very recent expansion of the ATcT network to boron, silicon, and sulfur compounds. I Introduction Accurate thermochemistry is seeing a modest renaissance in recent years. This is thanks to the tandem efforts of, on the one hand, the ATcT (Active Thermochemical Tables 68 ; 71 ; 70 ; 67 ) team at Argonne National Laboratory, and, on the other hand, the development of high-accuracy computational protocol families such as HEAT by an international consortium centered around the late lamented John F. Stanton, 80 ; 6 ; 22 ; 85 Weizmann-4 (W4) theory developed by our group, 37 ; 41 ; 75 and the FPD (Feller-Peterson-Dixon) strategy. 11 ; 14 For a recent review, see Karton 43 . These latter techniques (and to a lesser extent, reduced-cost variants such as W4-F12 79 ; 45 , W3X-L, 8 and Wn-P34 9 ) now offer kJ/mol accuracy on a semi-routine basis. Of late, the ATcT project has been moving into the second row of the Periodic Table, and hence the ATcT team expressed a desire for ab initio TAEs (total atomization energies) of second-row species accurate to 1 kJ/mol (0.24 kcal/mol; 83.59 cm -1 ) or better. The W4-17 benchmark 39 contains a fair amount of second-row species — but owing to computational cost and resources limitations, any post-CCSD(T) corrections were limited to the valence electrons, aside from W4.4 data 41 for a handful of second-row diatomics (such as \text{Cl}{\vphantom{\text{X}}}_{\smash[t]{\text{2}}} and \text{S}{\vphantom{\text{X}}}_{\smash[t]{\text{2}}} ). HEAT 80 ; 6 ; 22 ; 85 was developed with first-row systems in mind and makes no effort to separate valence from inner-shell correlation (a.k.a., subvalence correlation), while W4 and its predecessors 51 ; 61 ; 52 ; 5 were aiming at second-row systems from the start and hence, of necessity, separated valence and inner-shell correlation from the ground up. Still, otherwise HEAT and Weizmann- n have pretty much converged to each other. 22 In Ref. 78 , subvalence correlation was considered near the complete basis set limit at the CCSD(T) level using three different basis set families. Among other things, the authors of that study found that first-row and second-row molecules behave qualitatively differently. While in first-row molecules, core-valence correlation predominates, and is almost exclusively attractive, in second-row molecules with multiple adjacent second-row atoms, a repulsive core-core contribution may partly compensate for an attractive core-valence contribution. (Triple substitutions were found to be attractive throughout.) Neither W4 nor HEAT, in their unmodified form, treat post-CCSD(T) contributions to the inner-shell correlation, although W4.3 and W4.4 do so. At the time (2004-2008) the two ‘competing’ approaches (and for that matter FPD) were developed, inclusion of such contributions would have been computationally intractable for second-row species beyond diatomics. Fortunately, two decades of hardware evolution have removed this obstacle. Therefore, in the present paper, we seek to address the importance of post-CCSD(T) subvalence correlation in detail. We will also address two subsidiary questions that arise when comparing HEAT and W4-type approaches. (a) Both use CCSD(T) optimized reference geometries in correlation consistent 12 ; 13 quadruple-zeta basis sets. However, while W4 freezes subvalence electrons in the CCSD(T)/cc-pV(Q+d)Z geometry optimization (again, a choice made in the interest of being able to treat second-row species with limited computational resources), the original HEAT protocol did include subvalence correlation. 1 1 1 For reasons of computational expediency, the original HEAT papers 80 ; 6 ; 22 prescribed a cc-pVQZ basis set for the all-electron optimization; evidently, as first pointed out by Taylor, 81 this is less than ideal considering the cc-pVQZ basis set is of just minimal basis quality in the subvalence orbitals. Hence we used core-valence optimized basis sets. Now while for first-row elements, the difference between core-optimized cc-pCVQZ 91 and core-valence weighted cc-pwCVQZ 64 is essentially one of semantics, the latter are definitely favored for second-row elements. The HEAT team raised the question to what extent this would affect computed TAEs, all else being equal. (b) The HEAT team has traditionally preferred UHF (unrestricted Hartree-Fock) references, while previous Weizmann- n editions eschewed it in favor of ROHF (restricted open-shell Hartree-Fock), sidestepping various spin contamination artifacts. If any subvalence orbitals are ‘frozen’ (constrained to be doubly occupied), ROCCSD(T) 87 contains an ambiguity discussed in the Appendix to Ref. 37 : whether to semicanonicalize the ROHF orbitals after integral transformation (and any dropped cores, the MOLPRO 89 choice) or prior to it (the path followed in CFOUR 60 and most other coupled cluster codes). We will depart from past practice and use UHF references exclusively in the present work for energy calculations, but the question remains whether ROCCSD(T) or UCCSD(T) is preferable for the reference geometries of open-shell species. II Computational methods Most calculations in this paper were carried out using a development version of the CFOUR program system 60 running on the CHEMFARM high-performance computing facility of the Faculty of Chemistry at Weizmann. Selected additional calculation were carried out using release versions of MOLPRO 2024 89 and MRCC 2024 33 . The molecules considered are the 200-species W4-17 thermochemical benchmark 39 . These span a range of inorganic and organic molecules, first-row and second-row (including ‘pseudohypervalent’ species in which the 3d acts as an ‘honorary valence orbital’ ( 55 ; 58 , and references therein) ), and range from essentially purely dynamical correlation (such as H 2 O and SiF 4 ) to strong static correlation (such as O 3 , S 4 , C 2 , and BN). In the present work, we focus mostly on the W4-08 40 subset of W4-11 34 (and, in turn, W4-17). Basis sets considered are the Dunning-Peterson ‘correlation consistent core-valence n -tuple zeta’, cc-pCV n Z ( n =T,Q,5,6), and their core-valence weighted variant, cc-pwCV n Z ( n =T,Q,5), both described in Ref. 64 , in combination with the standard cc-pVnZ basis sets 12 ; 90 on hydrogen. The original W4 theory employed aug-cc-pV(n+d)Z basis sets 31 on second-row atoms, as otherwise, major SCF-level errors are made (reaching 50 kcal/mol in \text{HClO}{\vphantom{\text{X}}}_{\smash[t]{\text{4}}} !) 55 for species including a second-row atom in a high oxidation state. Effectively (see Ref. 55 and references therein) an additional ‘tight’ (high-exponent) d

Read the complete text