• Physique/Physique/Agrégats Moléculaires et Atomiques
  
• Physique/Physique/Chimie-Physique
  
• Physique/Physique/Physique Numérique
  
• Chimie/ou physique
  
• Physique/Physique Quantique
  
• Physique/Astrophysique/Astrophysique galactique
  
• Physique
  
• Physique/Astrophysique
  

We investigate the possibility to use an artificial neural network in order to generate a large number of accurate state-to-state rate constants, from the available rates obtained at different accuracy levels, including small numbers of accurate rates and large numbers of low-accuracy rates.

We report a large set of state-to-state rate constants for the H + HD reactive collision, using Quasi-Classical Trajectory (QCT) simulations on the accurate H$_3$ global potential energy surface of Mielke {\it et al.} [J. Chem. Phys. \textbf{116}, 4142 (2002)]. High relative collision energies (up to $\approx 56,000$~K) and high rovibrational levels of HD (up to $\approx 50,000$~K), relevant to various non thermal equilibrium astrophysical media, are considered. We have validated the accuracy of our QCT calculations with a new efficient adaptation of the MCTDH method to compute the reaction probability of a specific reactive channel.