Machine learning (ML) has emerged as a powerful tool to navigate polymer structure–property relationships. Despite recent progress, data sparsity is a major obstacle hindering the practical application of ML in polymer science. In this work, we explore functional monomer design by developing the first comprehensive database of monomer-level chemical and physical properties for approximately 12M synthetically accessible polymers. We generated diverse monomer-level properties by integrating quantum chemistry calculations with active learning to efficiently probe a vast chemical space of synthetically feasible polymers. Monomer-level property descriptors are benchmarked against both higher level computational predictions and experimental data to the extent possible, demonstrating their relevance to polymer design. Our results show that many monomer-level properties are weakly correlated, implying a strong freedom for functional design such that multiple physical properties can be simultaneously optimized by monomer selection. Moreover, the synthetically accessible nature of this chemical space allows targeted monomers to be considered by common polymerization mechanisms to facilitate their synthetic realization. Overall, this work opens new avenues for creating synthetically accessible polymers and provides new insights for designing next generation polymeric materials.
