Doping of germanium by ion implantation and flash lamp annealing

In the past the lack of stable native germanium oxide for surface passivation and gate dielectrics as well as the inability to epitaxially grow sufficiently thick defect-free germanium layers on silicon hindered the integration of germanium into the mainstream Si-based technology. Recent developments, such as high-k dielectrics and germanium-on-insulator substrates have made germanium a promising candidate for future high mobility devices. Therefore electrical doping of germanium by ion implantation and subsequent annealing has drawn a renewed interest. Investigations on the formation of ultra shallow junctions by ion beam processing have shown that p+-doping using B yields junctions that meet the requirements for the 22 nm technology node, whereas the formation of n+-junctions by P or As is complicated by the high diffusivity and the low solubility of the dopants. Recently, the concentration-dependent diffusion of n-dopants like P, As and Sb has been explored, and it has been found that doubly negatively charged vacancies are the mobile species responsible for the migration of the dopant atoms. The application of conventional rapid thermal annealing (RTA) with durations of some seconds and temperatures above about 500 °C leads to the activation of the n-dopants but their fast concentration-dependent diffusion can generally not be prevented. On the other hand it has been shown that both the diffusion and the activation of the dopants does not depend significantly on the implantation damage, i.e. using the defect engineering schemes known from Si technology seems not to be promising. Therefore, in order to control junction depth and dopant activation ultra-short annealing by flash lamps or lasers are currently under investigation.
The present work deals with the application of millisecond flash lamp annealing (FLA) to samples containing an implanted surface layer of about 100 nm thickness. P or As ions were implanted at an energy of 30 or 90 keV, respectively, and a fluence of 3x1015 cm-2. The investigations are focused on solid phase epitaxial recrystallization, dopant redistribution and dopant activation. The dependence of these effects on the heat transfer to the sample during FLA as well as on pre-amorphization and pre-annealing treatment is discussed. The results are compared to typical data achievable by RTA. Different characterization methods were employed. Channeling Rutherford backscattering spectrometry and cross-sectional transmission electron microscopy (XTEM) were used to monitor the recrystallization of the amorphous layers formed during implantation. The depth distributions of P and As were measured by secondary ion mass spectrometry. In order to determine the sheet resistance variable probe spacing and micro four point probe measurements were utilized. Selected samples were studied by XTEM to search for precipitates and end-of-range defects. While in RTA the concentration dependent dopant diffusion hinders the formation of ultra-shallow n+ layers, FLA does not cause any diffusion. The maximum activation obtained by FLA is about 6x1019 and 2x1019 cm-3 for P and As, respectively. This is about 3-4 times higher than under typical RTA conditions. However, the activation and the sheet resistance achieved by FLA do not yet fulfill the ITRS requirements for the 22 nm technology node. Possible mechanisms responsible for dopant deactivation are discussed.