Undoped, 30 nm thick pseudomorphic Si0.8Ge0.2 layers were grown at 550°C on (100) Si substrates in a Reduced Pressure CVD reactor. They were subsequently laser annealed in a SCREEN LT-3100 platform operating at 308 nm with a 160 ns pulse duration. The laser energy density was increased from 1.40 up to 2.40 J/cm² in order to investigate various regimes, from SiGe layer sub-melt to a point at which the whole SiGe layer and part of the Si substrate reach the melt. Time Resolved Reflectometry (TRR) and SP2 haze measurements combined with Atomic Force Microscopy (AFM) were used to detect melt threshold and quantify surface roughness evolutions, respectively. High Resolution X-Ray Diffraction and Time-Of-Flight-Secondary Ions Mass Spectrometry characterizations were performed to assess layer composition, strain and crystalline quality as a function of the laser energy density.
Melt threshold was detected at 1.55 J/cm² with SP2 Haze measurements (Fig.2) and at 1.58 J/cm² by TRR (Fig. 1). When increasing the laser energy density, three different regimes were identified. In the sub-melt regime, i.e. at energy densities lower than 1.55 J/cm², the annealed samples showed neither surface nor volume differences compared to the as-grown sample. For higher energy densities, part of the Si0.8Ge0.2 layer melted. At first, the melt remained very shallow: TOF-SIMS at 1.59 J/cm² (Fig.3) and Reciprocal Space Map (RSM, not shown here) were identical to those for the reference sample. However, a strong increase in haze and roughness indicated that the surface was modified. For a deeper melt (annealing at 1.8 J/cm²), the germanium was redistributed over the melted depth and segregated towards the surface, creating a top layer with a higher germanium concentration (cf. Fig.3). The solidification and segregation were accompanied by a partial relaxation of the layer up to 25% (the SiGe layer peak in Fig. 4.b is broadened due to mosaicity and not at the same Qx coordinate than that of the Si substrate) while the surface stayed rough. At higher energy density values, for instance 2.0 J/cm² (still in the partial melt regime), germanium segregation still occured and resulted in an even higher germanium concentration near the surface, up to 40%. However, the quality of the layer strongly improved compared to that of the 1.8 J/cm² sample. Roughness and haze were indeed much lower and close to their initial level. The RSM in Fig.4.c indicates that the layer is pseudomorphic, with a rather good crystalline quality, which will be further assessed by TEM observations. This sub-regime with strained SiGe layers exhibiting a high concentration of germanium near the surface could be favorable to the formation of low-resistance contacts in in-situ doped layers [2]. For higher energy densities, the whole SiGe layer and part of the Si substrate were melted: we were then in the full melt regime. For an annealing at 2.4 J/cm², the deeper melt and the intermixing resulted in the formation of a pseudomorphic 85 nm-thick Si0.93 Ge0.07 layer, with an almost uniform germanium profile. The film seems to have a very good crystalline quality as seen in Fig.4.d while RSM, Roughness and haze values were similar to those of the as-grown sample.
It is therefore possible, by selecting the appropriate energy densities, to form pseudomorphic SiGe layers with a high germanium content near the surface, which would be favorable for electrical contact formation. Among all the annealing regimes, it seems to be the most promising one for further investigations with doped layers.
References
[1] Ni C.-N. et al., Proceedings of symposium on VLSI Technology (2016) 7480531.
[2] Rosseel E. et al., ECS Transactions, 75 (8) 347-359 (2016)