Theoretical studies of three-dimensional potential energy surfaces using neural networks and rotational spectra of the Ar-N-2 complex

A new three-dimensional potential energy surface (PES) of the Ar-N-2 van der Waals complex is constructed using the neural network method based on ab initio data points at the CCSD(T) level. The aug-cc-pVQZ basis set is employed for all atoms with midbond functions. The vibrationally averaged PES V-00 is characterised by a global T-shaped minimum which occurs at R= 3.715 angstrom, = 90.0 degrees with a well depth of 98.779 cm(-1). Based on our three-dimensional PES, bound-state calculations are performed for three isotopomers of Ar-N-14(2), Ar-N-15(2), and Ar-(NN)-N-14-N-15, and several intermolecular vibrational states are assigned by analysing the wavefunctions. Moreover, the averaged structural parameters are calculated and the pure rotational transition frequencies with J = 0--6 are predicted. The spectroscopic constants are determined by fitting the rotational energy levels. The theoretical results are in good agreement with experimental data and this work gives more accurate results than those determined previously for the Ar-N-2 complex.