Dimer strings, anisotropic growth, and persistent layer-by-layer epitaxy

We propose that molecular-beam epitaxy of Si on Si(001) can proceed by nucleation and anisotropic growth of dimer ``strings.'' Such anisotropic growth, which changes orientation from layer to layer, imposes topological constraints of growth and can explain recently observed, puzzlingly persistent oscillations in surface diffraction intensities. We speculate that a complete picture for such epitaxy can be based on strings, coupled with an understanding of nucleation rules at steps and on terraces.