The possible structural transformations during the sample preparations and the sample observations are important issues in electron microscopy. Several publications of High Resolution Electron Microscopy (HREM) have reported that structural transformations and evaporation of the thin parts of a specimen could happen in the microscope. Diffusion and preferential etchings could also occur during the sample preparation.Here we report a structural transformation of a germanium Σ=13 (510) [001] tilt grain boundary that occurred in a medium-voltage electron microscopy (JEOL 400KV).Among the different (001) tilt grain boundaries whose atomic structures were entirely determined by High Resolution Electron Microscopy (Σ = 5(310), Σ = 13 (320), Σ = 13 (510), Σ = 65 (1130), Σ = 25 (710) and Σ = 41 (910), the Σ = 13 (510) interface is the most interesting. It exhibits two kinds of structures. One of them, the M-structure, has tetracoordinated covalent bonds and is periodic (fig. 1). The other, the U-structure, is also tetracoordinated but is not strictly periodic (fig. 2). It is composed of a periodically repeated constant part that separates variable cores where some atoms can have several stable positions. The M-structure has a mirror glide symmetry. At Scherzer defocus, its HREM images have characteristic groups of three big white dots that are distributed on alternatively facing right and left arcs (fig. 1). The (001) projection of the U-structure has an apparent mirror symmetry, the portions of good coincidence zones (“perfect crystal structure”) regularly separate the variable cores regions (fig. 2).