Machine learning-based hydrogen recycling model for predicting rovibrational distributions of released molecular hydrogen on tungsten materials via molecular dynamics simulations

Understanding the hydrogen recycling process is crucial for comprehending the behavior of detached plasma in nuclear fusion devices. To achieve this, a molecular dynamics (MD) model is being developed to predict the distribution of translational energies and rovibrational states of hydrogen atoms and molecules released from the plasma-facing materials. Neutral transport simulations, utilizing distributions obtained from the MD model as boundary conditions, are also a powerful tool for analyzing the impact of recycled hydrogens on edge plasma. However, the MD model requires significant computational resources to obtain distributions under varying material and irradiation conditions such as material temperature and incident energy. Therefore, developing effective models that seamlessly integrate neutral transport simulation with hydrogen recycling models is crucial. Machine learning techniques are employed to develop predictive models capable of forecasting distributions of energies and rovibrational states of released hydrogen atoms and molecules. Furthermore, a model considering the incident energy distribution (shifted-Maxwellian) is developed by integrating the monochromatic distribution with the shifted-Maxwellian distribution. • Hydrogen Recycling in Nuclear Fusion: Understanding hydrogen recycling is crucial for comprehending detached plasma behavior in fusion devices. • Development of Molecular Dynamics (MD) Model: A MD model calculates the distribution of transrational energies and rovibrational states of hydrogen released from plasma-facing materials. The distribution are used for the boundary condition of neutral transport simulation for edge plasma. • Computational Challenges: MD model requires significant computational resources, urging the need for seamless integration with neutral transport simulations. • Machine Learning Approach: Machine learning techniques predict distributions of released hydrogen’s energies and states, addressing computational challenges. • Shifted-Maxwellian Distribution: Integration of machine learning model with shifted-Maxwellian distribution incorporates incident energy distribution. • Impact of Electron Temperature: High electron temperature favors hydrogen atom emission, while low temperature favors hydrogen molecule emission.