OICC Press International Journal of Energy and Environmental Engineering 2251-6832 14 4 (December 2023) 2022 09 26 High efficient solar cells through multi-layer thickness optimization using particle swarm optimization and simulated annealing 10.1007/s40095-022-00541-1 EN Hamed Kargaran Faculty of Engineering, Shahid Beheshti University, Tehran, IR Elahe Bayat Faculty of Engineering, Shahid Beheshti University, Tehran, IR Aliakbar Hassanzadeh Faculty of Engineering, Shahid Beheshti University, Tehran, IR Ghasem Alahyarizadeh Faculty of Engineering, Shahid Beheshti University, Tehran, IR Journal Article 2022 09 26 Abstract In a comparative study, a solar cell structure’s layer thicknesses were optimized using particle swarm optimization (PSO) and simulated annealing (SA). The ideal short-circuit current density was considered as the cost function for both algorithms to minimize the number of function evaluations (NFEs) needed to obtain the optimal thicknesses of the structure layers (ZnO and MoOx layers), separately and simultaneously, being single- and multi-layer optimization. The results were compared to those of genetic algorithm (GA) and brute-force methods that have been reported in the literature. In the single-layer optimization, the results obtained by PSO indicated that the maximum required NFEs for optimizing ZnO was 33.1 ± 23.16 compared to 81 for the brute-force method and 78.16 ± 1.65 for the GA. The PSO method needed 19.6 ± 11.7 NFEs for optimizing the MoOx layer, while, as was reported in Ref. Vincent (E 13:1726 2020), GA optimized this layer by 13.05 ± 3.24 in the best manner by the Roulette wheel method. Both single-based and population-based approaches were implemented for the SA. The results obtained by SA indicated that the required NFEs were estimated higher than that of the GA due to the small search space. The required NFEs for two-layer optimization using PSO amount to 111.27 ± 60.1, which is considerably lower than the 1758.77 ± 39.75 of GA. The NFEs are significantly lower than the similar value obtained using the brute-force approach and GA, even with the highest SD value. In the case of two-layer optimization, SA estimated 65.63 ± 21.1 and 575.76 ± 301.64 NFEs using single-based and population-based methods, respectively.