A multiobjective optimization approach for extended interaction oscillators

Abstract

The development of high-frequency, wide-bandwidth, high-power extended interaction oscil-lators (EIOs) has always been the focus of researchers working on millimeter-wave and terahertz electronic devices. However, these design objectives are affected by many structural and operating parameters, and traditional manual optimization and local optimization are no longer suitable for solving these problems. In this paper, based on a one-dimensional, nonlinear, self-consistent program of EIOs, a multiobjective optimization method that employs the nondominated sorting genetic algorithm II (NSGA-II) is proposed to simultaneously optimize device output power, bandwidth, and structure length. By using this approach, the optimization process of a 95-GHz EIO is presented, and the corresponding Pareto solutions are obtained after 500 generations with a population size of 50. The results show that the beam-wave interaction and the coupling mechanism lead to synchronization of the structural parameters and the electrical parameters with each other, and the coexistence of multiple objectives guides the zonal distribution of the optimal solutions. That is, the oscillators with fewer gaps have shorter structure length and higher power, whereas those with more gaps are prone to start oscillation and have wider bandwidth. Several sets of optimization results obtained using the proposed method agree well with the results obtained in the CST-PIC solver, which proves that the proposed algorithm is effective for optimizing EIOs because it considers multiple design goals and can serve as a theoretical basis for engineering development.