Impact of Nanosecond Laser Annealing on the Electrical Properties of Highly Boron-Doped Ultrathin Strained Si0.7Ge0.3 Layers

Abstract

The CMOS scaling beyond 10 nm technology node requires high active dopant concentrations in source/drain modules to minimize contact resistance. Pulsed laser annealing has been targeted by chip manufacturers as a future option to enhance the activation level inside highly doped SiGe:B and Si:P regions, mostly used in PMOS and NMOS transistor fabrication, respectively. Indeed, this annealing process allows reaching high temperatures (above melt threshold), locally (~100nm below the surface) and with extremely fast temperature ramp rates (>109 °C/s), so that high doping levels have been demonstrated both in pure Si and Ge [1]. In a recent study, structural investigations allowed identifying the best conditions to obtain fully strained and defect-free undoped SiGe layers by liquid phase epitaxial regrowth (LPER) [2]. In this work, we report an analysis on the activation of boron dopants in similar layers. In-situ boron-doped 30 nm thick pseudomorphic Si0.7Ge0.3 layers were grown on p-type bulk Si (100) by CVD. Three different boron concentrations were incorporated inside these layers: 7.8x1019 (A), 1.4x1020 (B) and 2.3x1020 cm-3 (C). By combining the boron chemical profiles with the corresponding Hall effect measurements (Hall scattering factor: 0.35), it was possible to estimate the activation rate inside the as-grown strained SiGe layers and the impact of the possible inactive dopants on the transport properties. From the lowest to the highest boron chemical concentration, we found activation rates of ~100%, ~80% and ~60%, with no significant carrier mobility degradation, even in the sample with the highest fraction of inactive dopants.The SiGe layers were subsequently laser annealed in a SCREEN-LT3100 platform operating at 308 nm (XeCl laser) with a pulse duration around 160 ns. The laser energy densities (ED) ranged from 1.20 to 2.40 J/cm2 in order to investigate all the various annealing regimes. Results obtained using several characterization techniques were combined to determine the laser annealing regimes, quantify surface roughness and assess the layers’ composition, strain and crystalline quality. These results were compared to electrical measurements performed to analyse the evolution of the electrical parameters as a function of the laser anneal conditions, particularly the activation rate.For the lowly-doped and fully activated layer (A), the sheet resistance increases rapidly at the melt threshold (1.5 J/cm2, cf. Fig. 1, red curve), concomitantly with the appearance of a partial relaxation inside the layer (Fig. 2) and the formation of extended defects. The defects may induce a local dopant deactivation, while strain relaxation can result in a modification of the transport properties of the material (modified Hall scattering factor). Both phenomena can be therefore responsible for the observed increase of the sheet resistance. In contrast, for laser EDs allowing complete melt of the layer (i.e. beyond 2.0 J/cm2), the sheet resistance decreases with increasing ED and full activation is achieved (Fig. 3, red curve) together with strain recovery (Fig. 2) and no observable defects.For the highly-doped and partially activated layer (C), partial relaxation also occurs at the melt threshold (Fig. 2). However, thanks to the strain compensation effect of the small boron atoms, the relaxation level is lower compared to the lowly-doped sample and more quickly recovered when increasing the ED (no defects observed at 1.95 J/cm2, Fig. 4). In addition, the sheet resistance is found to continuously decrease as a function of the ED (Fig. 1, blue curve) independently of the strain state of the structure. This suggests that, in addition to the previously described phenomena, the initially inactive dopants are progressively incorporated into substitutional positions by LPER. Indeed, when full melt and strain recovery is achieved, a 100% dopant activation is also observed (Fig. 3, blue curve). Characterizations made on sample B suggest a similar behaviour to that of sample C.Finally, further results will be reported from additional experiments aiming at (i) better understanding the impact of strain relaxation on dopant activation and (ii) optimizing the laser annealing process to avoid relaxation. These experiments include the use of a shorter laser pulse for the annealing of the strained SiGe layers, as well as the comparison with results obtained from fully-relaxed boron-doped SiGe annealed under similar laser conditions.Acknowledgements:This work was supported by the European Union’s Horizon 2020 research and innovation program under grant agreement No. 871813 MUNDFAB.