Stochastic Electromagnetic Correlation Green's Functions for Wireless Communications --- Part II: Computational Model and Applications

This paper provides comprehensive numerical validation of the source-based stochastic electromagnetic (EM) correlation framework established in Part I. The alternative reciprocity theorem is verified against method of moments (MoM) benchmarks, demonstrating that the reciprocity-based formulation accurately captures the response of coupled antenna arrays and revealing non-negligible correlation pathways from self-current corrections and cross-antenna induced currents. Two stochastic source-generation methodologies are introduced: the random spherical surface sprinkling scheme, which reproduces far-field (FF) correlation results consistent with the Clarke-Jakes power angle spectrum (PAS), and the random spherical volume sprinkling scheme, which models bounded near-field (NF) scattering clusters. The simulations confirm that NF correlation differs fundamentally from FF predictions and depends strongly on source distance, frequency, and cluster geometry. CGF-derived correlation matrices are incorporated into a Kronecker channel model to evaluate MIMO performance metrics, including average effective degree of freedom (EDoF) and ergodic capacity. The results show that NF source statistics, array size, element polarization, mixed NF/FF distributions, and mutual coupling reshape the channel eigenvalue structure and affect spatial multiplexing. This work establishes a physics-compliant foundation for NF-aware MIMO performance prediction and optimization.