Microstructure Evolution of Nanosecond Laser Annealed Si/Si1-XGex/Si Structures

Abstract

Si1-xGex has a liquidus temperature lower than the melting point of Si [1], and the liquidus temperature decrease with the Ge concentration. Here, we investigate the possibility of selectively melting a buried Si1-xGex layer in Si during annealing. However, as Si and Ge are miscible over the entire composition range [2], Ge can diffuse into the Si matrix before such melting occurs. To avoid this effect, we used a nanosecond Laser Annealing (nLA) which allows raising the temperature at a fast heating rate (1010 K/s).Pseudomorphic Si1-xGex layers [3] of 20 nm for x = 0.3 or 0.4, and of 15 nm for x= 0.5 (corresponding to Delta Temperature = Tmelt Si – Tliq Si1-xGex of 60°C, 95°C, and 139°C respectively) were grown on 300 mm (001)-Si wafers. The Si1-xGex layers were covered with 17 nm of Si. Selected areas of the wafers were irradiated with nLA (308 nm wavelength and pulse duration of 160 ns) at different energy densities (ED). After laser annealing, the samples were characterized by X-Ray Diffraction (XRD), SP2 Haze, Atomic Force Microscopy (AFM), and Scanning Electron Microscopy (SEM).In Fig. 1a, we show the in-situ time-resolved reflectometry (TRR) recorded during the laser annealing. For ED lower than the ones indicated by the dotted lines, the surface reflectivity remains unchanged during annealing. For higher ED, the reflectivity increases sharply for a transient period ranging from 100 ns to 380 ns. This is typically associated with a melting process [4].In Fig. 1b, we display the (004) XRD scans for the three studied structures. For low ED, the Si1-xGex diffraction pattern may present a gradual reduction in intensity. For an ED higher than a threshold ED (different for each substrate), we report on a sharp transition of the Si1-xGex peak to a characteristic position of lower Ge concentration. This new characteristic peak increases in intensity with ED. The peak gradually shifts towards lower Ge concentrations for the higher tested EDs.In Fig. 1c, we show the measurements of the SP2 haze and AFM Root Mean Square (RMS) roughness. For ED lower than a threshold ED (zone 1), the roughness is low, characteristic of an unmodified surface. For intermediate ED (zone 2), the roughness rises abruptly. It then gradually falls with increasing ED to enter a third zone showing constant values.In Fig. 1d, we display SEM micrographs using secondary electrons (SE) and backscattered electrons (BSE). The SE provides topographic information and BSE provides additional chemical information from a higher depth range in the substrate. In the stacks with Si0.7Ge0.3 and Si0.6Ge0.4 layers, four-fold structures with symmetry along <110>-directions are observed in the images obtained using SE and BSE. For the stack with Si0.5Ge0.5, these formations are observed only with the BSE.The presented results show that using AFM, Haze, and SEM, surface modifications are detected at a lower ED compared to intensity rise obtained by TRR, providing then a higher sensitivity to identify the melt onset. By XRD, the first modifications are detected at a higher ED compared to TRR, wherein a larger modified volume is likely required to provide enough diffraction intensity. The characteristic Si1-xGex peak in the diffraction diagrams (Fig. 1b), abruptly appearing at a threshold ED for each stack, is associated with the formation of the four-fold structures in the SEM micrographs (Fig. 1d). Therefore, it is a signature of its composition/structure. The SEMs have also shown that for the Si1-xGex layers with up to 40 at% Ge, the Si1-xGex and top Si layers are both molten at the detected melt onset. In contrast, for the stack with Si0.5Ge0.5, the four-fold structure is only detected with BSE, evidencing a buried transformation. Therefore, for our experimental setup, a minimum Ge content of 50 at% or a Delta Temperature of 139°C was necessary to obtain a buried melted layer.