Toward Spectroscopic Accuracy for the Structures of Large Molecules at DFT Cost: Refinement and Extension of the Nano-LEGO Approach

Abstract

The SE100 database collecting accurate equilibrium geometries of medium size molecules obtained by the semiexperimental (SE) approach has been extended to species containing Br and I atoms. This has allowed the determination of accurate linear regressions between DFT and SE values for all the main bonds and angles involving H, B, C, N, O, F, P, S, Cl, Br, and I atoms. An improved Nano-LEGO tool has been developed, which is based on suitable hybrid and double hybrid functionals and combines in a fully coherent way the templating molecule and linear regression approaches. A number of case studies show that the new Nano LEGO tool provides geometrical parameters on par with state-of-the-art composite wave function methods, but can be routinely applied to medium-to large-size molecules. The accuracy reached for structural parameters is mirrored on rotational constants that can be predicted with an average error within 0.2%.

Hunting for Complex Organic Molecules in the Interstellar Medium: The Role of Accurate Low-Cost Theoretical Geometries and Rotational Constants

“…Improved results can be obtained correcting the density functional theory (DFT) geometrical parameters by bond-specific parameters derived from a large database of accurate molecular structures , [the so-called linear regression (LR) approach], or with reference to suitable fragments whose accurate geometries are already known [the so-called template molecule (TM) approach] . Recently, the LR and TM approaches have been combined with remarkable success in the nano-LEGO model , (also called LEGO-brick , ). However, the use of a large number of parameters remains quite unsatisfactory and, above all, suitable fragments are not always available. , Based on these premises, it is the purpose of this paper to show that much better results can be obtained by the recent Pisa composite scheme (PCS), ,− which improves the accuracy of current approaches for medium-sized molecules by nearly an order of magnitude, without any significant increase of computational cost, especially when core–valence (CV) correlation is accounted for by a simple one-parameter approach (PCS/Bonds).…”

“…As a consequence, state-of-the-art methods are needed, which are, unfortunately, applicable to systems containing at most a dozen atoms. − For larger systems, the current standard is to use B3LYP or MP2 computations with medium-size basis sets. − However, the intrinsic errors of these approaches (0.5–1%) are too large to allow unbiased analyses without additional experimental pieces of information (e.g., quadrupole coupling constants in the presence of nuclei, in which they do not vanish) . Improved results can be obtained correcting the density functional theory (DFT) geometrical parameters by bond-specific parameters derived from a large database of accurate molecular structures , [the so-called linear regression (LR) approach], or with reference to suitable fragments whose accurate geometries are already known [the so-called template molecule (TM) approach] . Recently, the LR and TM approaches have been combined with remarkable success in the nano-LEGO model , (also called LEGO-brick , ).…”

Rotational spectra and semi-experimental structures of furonitrile and its water cluster

“…These structures will contribute to the extension of the SE127 database, which now collects more than a hundred semi-experimental equilibrium structures of small-/medium-size molecules. 23,56…”

“…If the number of available isotopologues of a molecular system is not sufficient to allow the determination of all structural parameters, a subset of structural parameters can be kept fixed at the computed values, which can be further improved by using the linear-regression or template-molecule models. 23,56…”

Accurate Thermochemical and Kinetic Parameters at Affordable Cost by Means of the Pisa Composite Scheme (PCS)

“…In the case of kinetic parameters, this recipe can be employed to trace the intrinsic reaction coordinate along which more accurate energy computations can be performed to locate an improved transition state and the corresponding energy barrier by the so-called IRCmax approach . Also in this case, molecules containing up to about 50 atoms can be routinely studied. , For larger systems the results delivered by hybrid functionals in conjunction with double-zeta basis sets can be further improved by linear regression (LR), templating molecule (TM), , or, possibly, machine learning approaches.…”