The denitrogenation of cyclic diazines toward carbon-based bicycles has been extensively studied by thermolysis and photolysis. However, the mechanism of borodiazene denitrogenation is currently unknown. We used multiconfigurational complete active space self-consistent field with the second order perturbation theory correction, SA(5)-CASPT2(12,11)/cc-pVDZ//SA(5)-CASSCF(12,11)/cc-pVDZ, quantum chemical calculations to understand the photochemical denitrogenation of borodiazene. We performed 1 ps non-adiabatic molecular dynamics (NAMD) on 926 Wigner-sampled geometries of 1-methyl-1,3,4-borodiazene and identified that 10% of all simulations lead to denitrogenation. Of those landing in S0 after 1ps, we located a diyl intermediate (2), borirane (4), and diazaborete (5). We used Python Rapid Artificial Intelligence Ab Initio Molecular Dynamics (PyRAI2MD) to overcome the 1ps computational limit on NAMD and fully enumerate the reaction pathways of 1-methyl-1,3,4-borodiazene and run machine-learned nonadiabatic molecular dynamics (ML-NAMD). We performed ML-NAMD on 5,000 Wigner-sampled geometries initial conditions with curvature-driven time derivative coupling (kTDC) computed on-the-fly. We observed most ML-NAMD trajectories lead to diyl (2). Both diazaborete (5) and borirane (4) were present in the ML-NAMD trajectories.
Christian Salguero