Non-parametrized functionals with empirical dispersion corrections: A happy match?

Assessment of Density Functionals for Computing Thermodynamic Properties of Lanthanide Complexes

“… GGA: PBE‐D3, OPBE, BP86, B97‐D3; Meta‐GGA: M06‐L, TPSS; Hybrid: PBE0‐D3, B3LYP‐D3, B3PW91, TPSSH, B97; Global‐hybrid meta‐GGA: M06, M06‐2X; Long range corrected hybrid: CAM‐B3LYP, ωB97, ωB97X, ωB97X‐D3, LC‐BLYP; Double hybrid: B2PLYP with dispersion corrections (NL or D3BJ), PWPB95 …”

Computation of covalent and noncovalent structural parameters at low computational cost: Efficiency of the DH‐SVPD method

“…F I G U R E 4 Signed bond length deviations (Å) computed over a selection of density functionals and basis sets vs the (top) H-bond CCSD(T) equilibrium reference distance of the formamide dimer, [39] (middle) mixed electrostatics/dispersion CCSD(T)/CBS reference distance of T-shaped ethyne aggregate, [42] and (down) dispersion-dominated stacking CCSD(T)/CBS reference distance of the methane-methane aggregate [42] In the r(OÁ Á ÁH cis ) H-bond length of the formamide dimer example (Figure 4), is related to the semilocal approximation used by construction in the DH, the PBE [53] exchange density functional capturing better dispersion forces than other semilocal approximations. [31,54] The other weakly interacting aggregates ruled by mixed electrostatics/dispersion or dispersion-dominated stacking interactions, in the case of the T-shaped ethyne or methane dimers, respectively, confirm the above-mentioned trend (Figure 4). For all the basis sets investigated, the errors are, however, about one order of magnitude larger with respect to those measured with H-bond interactions.…”

“…It remains, however, questionable when it enrolls nonempirical DHs as it often leads to overbinding situations when the size of the weakly interacting systems is not large enough. [31] To overcome these limitations while keeping homogeneous performances independent of the system size, we recently proposed an alternative to empirical dispersion corrections based on the use of a tailored atomic basis set. [32] Like second-order Møller-Plesset (MP2) perturbation theory, DHs exhibit a larger binding behavior in weakly interacting systems when using a smaller basis set (ie, double-vs quadruple-ζ or larger, respectively).…”

The diene isomerization energies dataset: A difficult test for double-hybrid density functionals?

“…Among them, we will select here some of the most modern expressions developed so far: PBE0-DH, 21 PBE0-2, 22 PBE-QIDH and TPSS-QIDH, 23 together with the XYG3, 24,25 XYGJ-OS, 26 and xDH-PBE0 27 models. In doing so, we aim to systematically investigate for these functionals: (i) the role played by intra-molecular interactions (by comparing the results of dispersion-corrected PBE0-DH-D3(BJ) 28 with those of uncorrected PBE0-DH); (ii) the role played by the nature of the exchange-correlation density functionals used (by studying PBE-QIDH and TPSS-QIDH models, the latter built with the TPSS exchange-correlation functional 29 instead of PBE); (iii) the role played by the underlying orbitals (density) used to evaluate the double hybrid energy (by analyzing the performance of the XYG3, XYGJ-OS, and xHD-PBE0 models, which use the converged orbitals of a B3LYP or PBE0 calculation for evaluating all the terms, including the EXX and perturbative energies 30 ); (iv) the role played by the specific weight given to the various energy terms in the final expression, which might be related to the self-interaction error of common functionals; 31 and (v) the role played by different spin-scaling approaches applied to the perturbative term (e.g., by comparing the XYG3 and the XYGJ-OS models, the latter of which neglects same-spin correlation). In such a way, we hope to systematically disentangle the effect(s) driving the final accuracy of the results, thus providing some robust guidelines for the future improvement of these two families of theoretical expressions.…”