Optimally incorporating intra-cascade spatial correlations into mean-field cluster dynamics for radiation damage modeling

Understanding radiation damage and its relation to materials degradation is critical for the construction of safe and long-operating nuclear reactors. The mean-field cluster dynamics (CD) model predicts microstructure evolution over longer spatio-temporal scales and damage accumulation up to high doses, which is typically used to relate damage production to material properties; however, its inability to treat intra-cascade spatial correlations (SC) limits the accuracy of radiation damage modeling. Beyond the current solutions of using an ambiguous annealing time for cascade homogenization, we develop a novel characteristic-time annealing methodology that optimally embeds intra-cascade spatial correlations into the source term, establishing the CD-SC model within the continuum framework. This approach enables unified modeling of defect evolution under both monoenergetic ion bombardment and continuous neutron spectrum irradiation. CD-SC achieves kinetic Monte Carlo-level accuracy with high efficiency, reproducing experimental size distributions of dislocation loops/voids and dose trends under self-ion/neutron irradiations. The characteristic times for iron and tungsten are in the orders of μs and ps, respectively, breaking the traditionally accepted values (∼ns) under room-temperature conditions. We further reveal that the characteristic time is proportional to the relative positions of primary defects and inversely proportional to the competition between defect reaction and sink absorption, which are related to damage energy as well as temperature and material properties (sink character and migration energy), respectively. Finally, CD-SC is applied to predict neutron radiation damage under fission and fusion conditions. The present approach builds an accurate defect production term, enabling reliable evaluation of in-service materials and facilitating the development of next-generation materials.