Three diradical pyrazine isomers were characterized using highly correlated, multireference methods. The lowest lying singlet and triplet state geometries of 2,3didehydropyrazine (ortho), 2,5-didehydropyrazine (para), and 2,6-didehydropyrazine (meta) were determined. Two active reference spaces were utilized. The complete active space (CAS) (8,8) includes the σ and σ* orbitals on the dehydrocarbon atoms as well as the valence π and π* orbitals. The CAS (12,10) reference space includes two additional orbitals corresponding to the in-phase and out-of-phase nitrogen lone pair orbitals. Adiabatic and vertical gaps between the lowest lying singlet and triplet states, optimized geometries, canonicalized orbital energies, unpaired electron densities, and spin polarization effects were compared. We find that the singlet states of each diradical isomer contain two significantly weighted configurations, and the larger active space is necessary for the proper physical characterization of both the singlet and triplet states. The singlet−triplet splitting is very small for the 2,3-didehydropyrazine (ortho) and 2,6didehydropyrazine (meta) isomers (+1.8 and −1.4 kcal/mol, respectively) and significant for the 2,5-didehydropyrazine (para) isomer (+28.2 kcal/mol). Singlet geometries show through-space interactions between the dehydocarbon atoms in the 2,3didehydropyrazine (ortho) and 2,6-didehydropyrazine (meta) isomers. An analysis of the effectively unpaired electrons suggests that the 2,5-didehydropyrazine (para) isomer also displays through-bond coupling between the diradical electrons.

We introduce a free and open-source software package (PES-Learn) which largely automates the process of producing high-quality machine learning models of molecular potential energy surfaces (PESs). PES-Learn incorporates a generalized framework for producing grid points across a PES that is compatible with most electronic structure theory software. The newly generated or externally supplied PES data can then be used to train and optimize neural network or Gaussian process models in a completely automated fashion. Robust hyperparameter optimization schemes designed specifically for molecular PES applications are implemented to ensure that the best possible model for the data set is fit with high quality. The performance of PES-Learn toward fitting a few semiglobal PESs from the literature is evaluated. We also demonstrate the use of PES-Learn machine learning models in carrying out high-level vibrational configuration interaction computations on water and formaldehyde.

Hypohalous acids (HOX) are a class of molecules that play a key role in the atmospheric seasonal depletion of ozone and have the ability to form both hydrogen and halogen bonds.