The three-dimensional composition profiles of individual SiGe/Si(001) islands grown on planar and pit-patterned substrates are determined by atomic force microscopy (AFM)-based nanotomography. The observed differences in lateral and vertical composition gradients are correlated with the island morphology. This approach allowed us to employ AFM to simultaneously gather information on the composition and strain of SiGe islands. Our quantitative analysis demonstrates that for islands with a fixed aspect ratio, a modified geometry of the substrate provides an enhancement of the relaxation, finally leading to a reduced intermixing.
KeywordsSiGe Island Alloying Wet etching Tomography AFM Lateral ordering
The lattice mismatch between Si and Ge drives the formation of SiGe quantum dots (QD) during strained layer heteroepitaxy [1, 2]. For large-scale integration technologies , the position of such islands needs to be accurately controlled on the substrate surface . A viable process relies on the fabrication of lithographically defined pits, which act as a sink for the deposited adatoms, allowing the exact positioning and addressability of individual QDs. In addition, a precise control of the chemical composition of the SiGe islands is required, since the three-dimensional (3D) composition profile ultimately determines their electronic behavior and optical properties. However, little work has been done on this topic, and the different intermixing mechanisms sustaining the growth and evolution of Ge islands in presence of a surface with an extrinsic morphology are still debated [5–7]. It has been shown that SiGe islands grown on patterned areas have larger volumes than those on the surrounding planar surfaces [8, 9]. These observations are corroborated by a recent comparison of X-ray measurements and finite element calculations, which suggests a different compositional state with a larger intermixing and relaxation on the patterned substrates . However, the compositional differences at the single dot level were not yet considered.
In this letter we address the issue of the impact of substrate patterning on shape, composition, and strain relaxation at the single dot level by using atomic force microscopy (AFM)-based nanotomography (NT-AFM). Following Ref. , we have recently extended the capabilities of NT-AFM to quantitatively determine the full 3D composition profiles of strained SiGe islands . In this study, we compare lateral and vertical composition gradients of individual SiGe islands grown on pit-pattern and planar Si(001) substrates. Above all, by combining structural data with the average island compositions as obtained by NT-AFM, we are able to determine island strain only by means of an AFM analysis. The experimental ability to map the chemical composition at the nanoscale helps indeed to shed new light on the driving forces governing alloying. Our findings provide direct experimental evidence that a nanostructured surface plays a major role in determining strain relaxation and therefore in defining the compositional profiles of the islands.
Results and Discussion
A possible, additional explanation for the aforementioned discrepancy is based on the different Ge profiles of the islands of the two ensembles as shown in Fig. 3a. Since the scaling law used in the present discussion is derived for island with a uniform composition, our simple approach holds more likely for islands grown on the pit-patterned surface, because of their more homogeneous Ge distribution (see Fig. 3a). Therefore, the δ value provided here has to be considered as an upper limit for the relaxation enhancement, since a realistic non-homogeneous Ge distribution for the coherent islands on the planar surface could lead to a more effective elastic energy reduction .
Nevertheless, some islands in Fig. 1d still do not follow the volume rescaling. According to their volumes, those islands are most probably plastically relaxed. In this case, the system lowers its total free energy by introducing dislocations. Therefore, the misfit inV(f) has to be replaced by the residual misfitf = ε − d, whered is the plastic strain and the sign has been assigned according to the actual compressive stress.
Remarkably, the overall average Ge content is 0.361 ± 0.005 for islands grown on the pit-pattern and 0.36 ± 0.01 for islands on the flat surface area. As expected, despite an equal meanxGe, the standard deviation is a factor of two larger for the latter case, reflecting the larger composition fluctuations. Finally, the average Ge content of the individual islands increases monotonically with the island height (see Fig. 3b). This behavior can be rationalized as a result of the island evolution towards steeper and more relaxed morphologies. We can therefore compare islands with the same Ge content, e.g., about 0.36. These islands have the same height, i.e., about 55 nm (Fig. 3b), but different aspect ratio (Fig. 4b). As a consequence, the island base and the volume are larger for islands grown on a pit-patterned surface. This can again be explained by the different residual strain of the two island ensembles.
In conclusion, an AFM-based nanotomography approach was used to gather in-depth information about the alloying and relaxation mechanism on both flat and pit-patterned substrates. The 3D compositional profiles reveal that islands forming on pit-patterned areas have a more uniform Ge distribution and are slightly Ge-richer than their counterparts forming on flat areas. These periodic Ge-rich island arrays are therefore appealing candidates for efficient local stress engineering in next generation Si field effect transistors  for ultra large-scale integration technologies.
The authors acknowledge financial support by the EU D-DOTFET project (012150) and DFG (FOR 730).