AlF-AlF sticking time and prospects for ultracold dimers

We report on the sticking time of the AlF dimer in the ultracold regime. First, we construct a general full-dimensional potential energy surface for the AlF dimer by training a machine learning model on \textit{ab initio} points using an active learning scheme, which requires no long-range information and only $\lesssim 0.01\%$ of the configurations to be calculated \textit{ab initio}. We use molecular dynamics simulation to select the \textit{ab initio} training points to choose suitable configurations for different collision energies and map the relevant part of the potential energy landscape \cite{liu2023molecular}. We then use our general full-dimensional potential energy surface to compute the density of states using a semi-classical counting method. Next, using the Rice-Ramsperger-Kassel-Marcus (RRKM) theory, we determine an RRKM sticking time of 216.0 $\pm$ 4.8 ~ns, which is shorter than that of other previously reported non-reactive dimers. Then, based on the available potential energy surfaces and RRKM sticking time data for a few dimers, we observe that the RRKM sticking time correlates with the ratio of the dimer dissociation energy to the dissociation energy of its parent molecule. Thus, we propose this ratio as a parameter to estimate RRKM sticking times. Furthermore, using \textit{ab initio} methods, we optimize and calculate the energies of the equilibrium geometries of AlCl-AlCl, CaH-CaH, and MgF-MgF, thereby determining the ratio of the dimer dissociation energy to the parent molecule dissociation energy. Based on this ratio, we propose a hierarchy for ordering dimers' sticking lifetimes in the ultracold regime.