ETD

• excited states
• machine learning
• molecular dynamics
• quantum chemistry

Modeling chemical reactions in embedded systems poses a significant challenge within computational chemistry. Hybrid quantum mechanics / molecular mechanics (QM/MM) methods offer a powerful framework to tackle this task; however, their computational cost often limits their practical use, especially when combined with molecular dynamics. This Thesis overcomes this bottleneck by investigating and developing strategies to accelerate hybrid QM/MM dynamics along two complementary directions. First, retaining the quantum approach, approximate methods and extrapolation techniques are employed to reduce the computational cost of the calculations. Second, machine learning models are leveraged to capture the quantum behavior at a fraction of the cost. The resulting methods substantially broaden the applicability of hybrid simulations, enabling the study of complex chemical processes with greater computational efficiency and for longer timescales.