Advancing Non-Atom-Centered\nBasis Methods for More\nAccurate Interaction Energies: Benchmarks and Large-Scale Applications

Recent advances in local electron correlation approaches\nhave enabled\nthe relatively routine access to CCSD(T) [that is, coupled cluster\n(CC) with single, double, and perturbative triple excitations] computations\nfor molecules of a hundred or more atoms. Here, approaching their\ncomplete basis set (CBS) limit becomes more challenging due to extensive\nbasis set superposition errors, often necessitating the use of large\natomic orbital (AO) basis sets with diffuse functions. Here, we study\na potential remedy in the form of non-atom-centered or floating orbitals\n(FOs). FOs are still rarely employed even for small molecules due\nto the practical complication of defining their position, number,\nexponents, etc. The most frequently used FO method thus simply places\na single FO center with a large number of FOs toward the middle of\nnoncovalent dimers; however, a single FO center for larger complexes\ncan soon become insufficient. A recent alternative uses a grid of\nFO centers around the monomers with a single s function per center,\nwhich is currently applicable only for H, C, N, and O atoms. Here,\nwe build on the above advantages and mitigate some drawbacks of previous\nFO approaches by using a layer of FO centers and 4–9 FOs/center\nfor each monomer. Thus, a double layer of FOs is placed between the\ninteracting subsystems. When extending the double-ζ AO basis\nwith this double layer of FOs, the quality of conventional augmented\ndouble-ζ or conventional triple-ζ AO bases can be reached\nor surpassed with less orbitals, leading to few tenths of a kcal/mol\nbasis set errors for medium-sized dimers. This good performance extends\nto larger molecules (shown here up to 72 atoms), as efficient local\nnatural orbital (LNO) CCSD(T) computations with only double-ζ\nAO and 4 FOs/center FO bases match our LNO–CCSD(T)/CBS reference\nwithin ca. 0.1 kcal/mol. These developments introduce FO methods to\nthe accurate modeling of large molecular complexes without limitations\nto atom types by further accelerating efficient correlation calculations,\nlike LNO–CCSD(T).