Conformational energies of microsolvated Na+ clusters with protic and aprotic solvents from GFNn-xTB methods

Performance of contemporary tight-binding semiempirical GFNn-xTB methods for the conformational energies of singly charged sodium clusters Na+(S)(n) (n = 4-8) with 3 protic and 8 aprotic solvents is examined against the reference RI-MP2/CBS method. The median Pearson correlation coefficients of rho = 0.84 (GFN2-xTB) and rho = 0.82 (GFN1-xTB) do not give the clear preference to any tested approach. GFN1-xTB method demonstrates more stable performance than its GFN2-xTB successor with the average mean absolute errors (MAEs)/mean signed errors (MSEs) of 1.2/0.2 and 2.3/1.6 kcal mol(-1), respectively. Conformational energies produced by the computationally efficient DFT functional PBE and double-zeta basis set complemented with -D3(BJ) dispersion correction are suitable for the preliminary sampling (median rho = 0.93), but should be used with a caution for the calculations of the average ensemble properties (MAE/MSE = 1.7/1.1 kcal mol(-1)). Higher-ranking PBE0-D3(BJ) and omega B97M-V with triple-zeta basis sets yield significantly lower MAEs/MSEs of 0.55/0.20 and 0.51/0.23 kcal mol(-1), respectively.

Automated summary

The performance of contemporary tight-binding semiempirical GFNn-xTB methods for the conformational energies of singly charged sodium clusters Na+(S)(n) (n = 4-8) with various solvents is compared against the reference method RI-MP2/CBS. The results show that the GFN1-xTB method is more stable and outperforms its successor, GFN2-xTB, in terms of average mean absolute errors and mean signed errors.