Membrane remodeling is a crucial process for various cellular pathways like signaling, viral assembly and clathrin-mediated endocytosis. This process is governed by a network of protein-protein interactions that perform specific tasks with high spatial and temporal accuracy. An important type of these proteins is the BAR domain family, proteins of which, self-assemble on the membrane surface and induce curvature. The localization of proteins to the 2D membrane surface from 3D solution reduces their search space, and our lab has shown how proteins can exploit this dimensionality reduction to trigger self-assembly. Further, the dimerization of BAR proteins, which is crucial in mediating protein-protein and protein-lipid interactions, is poorly understood from a quantitative perspective. We address this gap by applying Molecular Dynamics simulations to investigate the thermodynamics of binding in solution and on membrane surface. We employ metadynamics to evaluate the free energy of protein dimerization and obtain the binding affinity (KD). In addition, we aim to deconvolute the energetic and entropic contributions to the binding process and characterize the intermediate pathways of dimerization. Our results would quantify the role of dimensionality reduction on the binding affinity and also contribute to the theoretical framework of our reaction-diffusion software, NERDSS, that enables self-assembly simulations of a broad range of biological systems.