A crucial step towards the optimization of the

A crucial step towards the optimization of the optoelectronic device structures is the investigation of the defect formation mechanisms. Hence, the detailed atomic study of the local structure at the interface between the semiconductor epilayers and the Si substrates is of particular interest. For this, transmission electron microscopy (TEM) is ideally suited as it provides the spatial resolution to study the structure of materials on the atomic level. Over the past 30 years, extended defects in semiconductor materials have been intensively studied by conventional and high-voltage high-resolution TEM [1,9–12]. However, it was experimentally challenging to resolve the exact atomic-scale structure due to the lack in spatial resolution and/or image delocalization in conventional phase contrast micrographs. Thus, “holographic” reconstruction techniques were indispensable to increase the spatial resolution [13–15]. Alternatively, image simulations based on structural models were performed to validate the image interpretation [11,12].
More recently, the implementation of spherical-aberration correctors in scanning transmission electron microscopes (STEM) [16–18] has lead to dramatic improvements in lateral resolution. These instruments are now capable of routinely producing images in the deep sub-Ångström range [19]. Thus, the dumbbell structure in crystalline Si along and samples can be clearly resolved, which is particularly useful for the characterization of defects [20]. Additionally, by using a high-angle annular dark-field (HAADF) detector it is possible to produce images which show contrast that approximately scales with the square of the atomic number, hence the name Z-contrast imaging. This allows to discriminate, for example, between the lighter Cd atomic columns from the heavier Te columns in the CdTe dpn semiconductor. Recent studies of low-dimensional semiconductor structures and devices by aberration-corrected STEM clearly demonstrate the power of this technique for investigating defects and interfaces [3,21–29]. It has proven to be decisive both in the detection of novel types of defects but also in the advancement of our understanding of seemingly basic crystal-structure defects.

Experimental details
Unless otherwise stated, all experimental images were taken using a double spherical aberration-corrected JEOL JEM-ARM200F microscope equipped with a cold field-emission electron source operating at 200kV. In STEM mode, a convergence semiangle of 25mrad was used in combination with an annular dark field (ADF) detector with inner and outer collection semiangles of 90 and 370mrad, respectively. A Gaussian low-pass filter for noise reduction was applied to all images.
Samples for the HAADF-STEM analysis were prepared by means of either a FEI Helios NanoLab 600i or FEI Helios NanoLab 450S focussed ion beam (FIB) operated at accelerating voltages of 30 and 5kV. Additionally, Si/Ge specimens [30] were prepared by mechanical polishing and dimple grinding, followed by ion-milling with Ar+ ions using a Fischione Model 1050 TEM-Mill operating at low voltages and grazing incidence to achieve electron transparency.

Configuration of planar defects
The most commonly observed planar (2D) defects in cubic semiconductors by using high-resolution TEM are twin boundaries (TBs) and stacking faults (SFs) [31]. They are caused by discontinuities in the …AaBbCcAaBbCc… stacking sequence of 111 -type close-packed layers in diamond or, its ordered variant, zincblende structure. Additionally, compound semiconductor layers grown on elemental semiconductor substrates, like Si or Ge, are also susceptible to form antiphase boundary (APB) defects. In the following sections, the characteristics of these defects are discussed.

Configuration of dislocations
Fig. 4 illustrates the presence of dislocations both inside a Ge crystal and at a Ge/Si interface, together with their corresponding ε strain field maps obtained by geometrical phase analysis (GPA) [53]. In particular, Fig. 4a and b show two parallel dislocations lying on perpendicular slip planes in close proximity within a Ge crystal and the strain field interaction between them: both dislocation cores exhibit a compression region (in blue) and a tensile region (in yellow) [30]. Similar butterfly-like shapes are observed at the Si/Ge interface (Fig. 4c and d). They are MDs (marked with white arrows) resulting from the 4.2% lattice mismatch between Ge and Si, and are identified as pairs of perfect glissile 60° and perfect sessile 90° MDs (a more detailed description of these type of dislocations is given in the following sections). The number of atomic planes between the MDs is not constant. It varies between 20 and 40 planes depending on the area of observation. Theoretically, for the total relaxation of the misfit strain one MD should be introduced every 24 planes [, n=number of planes, a=lattice constant, a=0.5658nm and a=0.5431nm]. Experimentally, the number of 111 planes between the MDs is slightly larger with an average of 26±0.5 planes between dislocations. Therefore, not enough MDs are incorporated to fully relax the strain plastically at the selected growth temperatures [54].