1: \begin{abstract}

2: In this work, we consider the time-harmonic Maxwell’s equations and their 

3:   numerical solution with a domain decomposition method. As an innovative feature, 

4:   we propose a feedforward neural network-enhanced approximation of the interface conditions 

5:   between the subdomains. The advantage is that the interface condition 

6: can be updated without recomputing the Maxwell system at each step.

7: The main part consists of a detailed description of the construction 

8: of the neural network for domain decomposition and the training process.

9: To substantiate this proof of concept, we

10:   investigate a few subdomains in some numerical experiments with low 

11:   frequencies. Therein the new approach is compared to a classical domain decomposition method.

12: Moreover, we highlight 

13:   current challenges of training and testing with different wave numbers 

14:   and we provide information on the behaviour of the 

15:   neural-network, such as convergence of the loss function, 

16:   and different activation functions.

17: \end{abstract}

18: