Genetic algorithm-based interpretation of anomalous susceptibility in correlated double perovskites

We present a genetic algorithm (GA)-based approach to the analysis of temperature-dependent AC susceptibility in double perovskite Ba 2 FeMoO 6 (BFMO). Conventional descriptions, including nonlinear least-squares (NLS) regression of phenomenological two-tanh forms, critical power-law fits, and two-sublattice mean-field models, fail to capture the full experimental response, particularly the subtle shoulder near 332 K that accompanies the main ferromagnetic transition. In contrast, the GA fit successfully reproduces both the sharp drop at T c and the anomaly, providing a quantitatively robust and physically consistent representation. Although the GA yields a slightly larger plain RMSE than NLS, it achieves the lowest weighted error in the anomaly region and avoids systematic residuals. The mean-field attempt produces poor agreement, confirming that the anomaly cannot be explained within a simple exchange framework. Our results demonstrate the utility of evolutionary optimization for magnetic susceptibility analysis in complex oxides, where non-analytic features and rugged parameter spaces frustrate deterministic fitting methods. The methodology is readily extendable to other correlated electron systems, multifunctional materials, and anomalous response functions, establishing GA-based data analysis as a valuable complement to conventional theoretical models. • Genetic algorithm applied to AC susceptibility of Ba 2 FeMoO 6 . • Reveals both ferrimagnetic transition and 332 K compensation-like anomaly. • Achieves stable fits in complex, non-mean-field magnetic behavior. • Outperforms conventional nonlinear and power-law models. • Provides a general framework for analyzing correlated oxides.