The structure of the Lennard-Jones solid, obtained by molecular-dynamics simulation of crystallization in the supercooled liquid, may be fee, although the hep structure is energetically more favorable. This could derive from the cubic symmetry of the fee lattice, allowing lattice defects that are not possible in the hep arrangement, but are essential to crystal growth in the simulated liquid. Two crossing stacking faults in a small fee crystallite can produce nonvanishing, growth-promoting, but stacking-faultresisting, surface steps. PACS numbers: 61.20.Ja, 36.40. +d, 61.50.Cj It is well known that the Lennard-Jones (LJ) potential favors the hep structure over the fee structure for the solid [1]. The difference in cohesive energy (^0.01%) appears to be too small, however, to provide a basis for an explanation of an observed preference for one of the two structures in molecular-dynamics (MD) simulations of crystallization in the supercooled LJ liquid. In fact, until very recently, such a preference has never been found [2][3][4][5][6][7], suggesting the inadequacy of the LJ potential to model the interatomic interactions in a simulation of either fee or hep crystal growth.It is the purpose of this Letter to investigate the role that lattice defects may play in the simulated crystallization process, and, in particular, to demonstrate that growth-stimulating defects are much more probable to occur in fee crystallites than in hep crystallites. Moreover, it will be shown that such defects exclusively stimulate fee growth, without further assumptions regarding the interatomic potential other than that it is isotropic and short ranged.That growth characteristics can be decisive in the choice of crystal structure of a substance, rather than a difference in cohesive energy, can be illustrated by a comparison with the method of static lattice energy calculations, aimed at structure prediction. Here, the evaluation and subsequent minimization of lattice sums involves inclusion of all interactions between a representative central atom (the reference atom) with all its close and more distant neighbors within a limiting sphere. However, the reference atom has been incorporated in the crystal lattice under completely different conditions, notably in the absence of at least half of its ultimate neighbors. Moreover, according to accepted theories of crystal growth [8], the motions of surface-migrating atoms are governed by short-range forces and are rather insensitive to the detailed shape of the potential. Trapping sites that are (almost) equally favorable (but, possibly, not equivalent, as on close-packed faces of fee or hep crystals) will have equal a priori occupation probability. The possibility of complete layers of atoms shifting to "better" positions in response to the arrival of new neighbors can be ruled out.Both the fee and the hep structures consist of plane hexagonal arrays of atoms that are stacked in an orderly way, with atoms in one layer over three-coordinated sites in the preceding layer. Consequently, each l...