This work comprise two distinguish arguments, both important in RF engineering. The first one exploits the Method of Moments in Time Domain (MoM-TD). Such method can be implemented, as for frequency domain, either resorting to a thin wire approximation or to a surface patch model. For heterogeneous structures both implementations might be needed, and the problem of the attachment between a perfectly conducting thin wire and a perfectly conducting surface is hence relevant. In this work attachment modes are introduced in MoM-TD and a uniform formulation of the method itself is presented for the three models (WGM, SPM and ATT). The second part of this work exploits the optimization techniques in antenna design. In this contribution an improvement on existing multi-objective algorithm is performed and more precisely on the Nondominated Sorting Genetic Algorithm (NSGA-II). The modified algorithm is applied to the design of Horn antennas for discrete lens applications to achieve a Pareto-optimal solution. Horn design is aimed not only towards higher possible efficiency but also towards large band and compactness.