Dynamics Study of the CaC (X3∑-) + C(3Pg) → Ca+C2 (∑v) Reaction: Based on a Full-Dimensional Neural Network Potential Energy Surface of CaC2.

Abstract

The CaC₂ molecule, as an interstellar species that has already been detected, has attracted significant attention. To date, studies on the potential energy surface (PES) and the reaction dynamics of CaC₂ are largely lacking. In this work, ab initio energy values were obtained for 3877 configurations using the icMRCI+Q method, and these energies were subsequently fitted using a neural network approach. During parameter optimization, the trust region framework (TRF) method, which has superior performance compared to the previously used Levenberg-Marquardt (LM) method, was used. The root-mean-squared error (RMSE) for both the training and testing sets meets the requirement for chemical accuracy (error less than 1 kcal/mol). Using the neural network PES, we identified one stable structure and two metastable structures for the ground state (X̃¹A') of CaC₂. The stable structure is T-shaped, while the two metastable structures are linear. The potential well depths of the stable structure and the two metastable structures are -10.98, -9.75, and -4.70 eV, respectively. Based on the obtained full-dimensional neural network PES, we investigated the CaC(X³Σ^-) + C(³P_g) → Ca + C₂ (Σv) reaction dynamics under different initial conditions. Under the condition that all other parameters remain unchanged, the reaction cross section and rate constant were found to be largest when the initial condition was v = 0 and j = 0. These findings indicate that the reaction rate is fastest when the CaC molecule is in its ground state.