In Silico study of degradation in the host molecular materials of OLEDs

Abstract We perform an in-depth computational study of degradation processes occurring during OLED operation for three commonly used emissive-layer host molecules: TBSO (4-(m-tolylsulfonyl)-1,1-biphenyl), DPPCarba (9-(3,5-diphenylphenyl)carbazole), and DPPTriphen (2-(3,5-diphenylphenyl)triphenylene). The investigation is carried out from both static and dynamic viewpoints. In the static framework, we evaluate thermodynamic descriptors that allow us to analyze possible chemical degradation pathways in the ground state, charged states, first excited states, and after exciton--polaron annihilation (EPA) and exciton--exciton annihilation (EEA). In the dynamic framework, non-adiabatic molecular dynamics (NAMD) simulations are used to follow degradation starting from both the first excited state and a higher-lying excited state representative of an EEA process. Furthermore, machine-learning-accelerated NAMD (ML-NAMD) simulations are employed to enhance the sampling of degradation pathways in TBSO. Combining static descriptors with direct and ML-accelerated dynamics provides molecular-level insight into chemical degradation mechanisms in OLED host materials and highlights the importance of going beyond static stability criteria when assessing degradation. These findings can support the rational design of more stable host materials for next-generation OLEDs. Molecular dynamics trajectories data File Name Description dyn.out Molecular dynamics with geometries and velocities dyn.xyz Atomic coordinates en.dat Time step, potential energies (S0, S1, S2 ...), current electronic state potential energy, total energy geom Initial geometry nx.log Newton-X molecular dynamics log including gradients velc Initial velocity Molecular dynamics with machine learning trajectories data File Name Description en.dat Time step, potential energies (S0, S1, S2 ...) state.dat Current electronic state traj.xyz Atomic coordinates