The formation of prismatic dislocation loops is an important factor leading to radiation damage of metals. However, the formation mechanism and the size of the smallest stable loop has remained unclear. In this Rapid Communication, we use electronic structure calculations with millions of atoms to address this problem in aluminum. Our results show that there is a cascade of larger and larger vacancy clusters with smaller and smaller energy. Further, the results show that a seven vacancy cluster on the ͑111͒ plane can collapse into a stable prismatic loop. This supports the long-standing hypothesis that vacancy clustering leads to a prismatic loop, and that these loops can be stable at extremely small sizes. Finally our results show that it is important to conduct calculations using realistic concentrations ͑computational cell size͒ to obtain physically meaningful results. The embrittlement of metals subjected to radiation is a long-standing problem in various applications including nuclear reactors. As the irradiation dose increases above a certain threshold, a significant population of prismatic dislocation loops ͑dislocation loops whose Burgers vector has a component normal to their plane͒ has been experimentally observed to arise in metals. [1][2][3][4][5] It is widely believed these prismatic loops form through the clustering of vacancies that are generated randomly by irradiation.6 Specifically, the vacancies diffuse and eventually cluster on specific planes. Once there is a large enough planar cluster, the atoms on the two faces collapse onto each other leaving behind a prismatic dislocation loop.However, the formation mechanism and the size of the smallest stable loop remain unclear: there is no direct experimental observation of the process, and the theoretical investigations are inconclusive. Recent molecular dynamics simulations 7 support the hypothesized mechanism for iron, but these calculations used the Finnis-Sinclair empirical atomistic potentials whose validity is uncertain in situations involving changing atomic bonds. 8 In contrast, calculations for aluminum using quantum mechanical density-functional theory 9,10 show that divacancies-a complex of two vacancies-are either energetically unfavorable if they are aligned along the ͗110͘ direction or barely favorable with negligible binding energy if aligned along ͗100͘. If two vacancies can barely bind, it seems doubtful that they can be stable and grow to form clusters that can turn into prismatic loops. While these density-functional theory ͑DFT͒ methods are far more accurate, the computational effort is extremely large and consequently these calculations were limited to less than 100 atoms. This corresponds to an unphysically high concentration of vacancies. Furthermore, the results are in variance with experiments 11,12 that are indicative of a high binding energy of divacancies and a significant concentration of divacancies, especially at elevated temperatures.We study vacancy clustering and prismatic loops by performing electronic structure c...