Application of circuit analog optimization method in fast optimization of dynamically tunable terahertz metamaterial sensor

Abstract

The simulation design of terahertz metamaterial sensors with dynamically tunable parameters typically relies on manual parameter tuning for structural optimization. However, this method is often prone to subjective factors and suffer from issues such as frequent reconstruction of simulations and suboptimal optimization results. In this paper, we propose a circuit analog optimization method (CAOM), which constructs equivalent RLC parameters to achieve a highly fitted terahertz transmission spectrum frequency obtained from CST full-wave numerical simulation. To validate the effectiveness of the proposed model, we use a typical periodic structure unit, a double-nested split ring resonator (DSRR) terahertz metamaterial sensor, as the simulation object. DSRR is made of graphene, as a result, the opening size and Fermi level of two resonant rings are dynamically tunable. The results of the validation demonstrate that the adjustments of the sensor parameters can be effectively mapped by the changes of the equivalent RLC parameters. And the proposed equivalent circuit model has parameter substitutability in the simulation modeling of split ring resonator type sensors. The proposed equivalent circuit model exhibits parameter substitution in the simulation modeling of open resonant ring-type sensors. To achieve optimal sensing performance for the electromagnetically induced transparency (EIT)-like resonant peak (with a resonant frequency of f2) of the sensor under constrained conditions, we introduce the genetic algorithm (GA) into the equivalent circuit model to enable fast optimization of the opening sizes of the inner and outer resonant rings, as well as the Fermi level of the sensor. Moreover, the accuracy of the optimization results is verified by CST simulations. Finally, the optimization results show that the optimal FOM of the EIT-like resonant peak within the given parameter range is 0.712, which is greater than that of any randomly combined parameters. This numerical result demonstrates the effectiveness of the proposed CAOM.