Kinetic Monte Carlo simulations of boron activation in implanted Si under laser thermal annealing

International audienceWe investigate the correlation between dopant activation and damage evolution in boron-implanted silicon under excimer laser irradiation. The dopant activation efficiency in the solid phase was measured under a wide range of irradiation conditions and simulated using coupled phase-field and kinetic Monte Carlo models. With the inclusion of dopant atoms, the presented code extends the capabilities of a previous version, allowing its definitive validation by means of detailed comparisons with experimental data. The stochastic method predicts the post-implant kinetics of the defect-dopant system in the far-from-equilibrium conditions caused by laser irradiation. The simulations explain the dopant activation dynamics and demonstrate that the competitive dopant-defect kinetics during the first laser annealing treatment dominates the activation phenomenon, stabilizing the system against additional laser irradiation steps. T he interactions between defects and implanted impurity atoms in semiconductor substrates play a central role in the dopant activation mechanism during an-nealing processes. In past decades, the study of uniform temperature treatments revealed the relationship between defect kinetics, impurity diffusion, and activation in Si, leading to the discovery of defect-driven transient phenomena. 1) Modified thermal processes are currently needed for ultra-shallow junctions in micro-and nanodevices. In particular, submicrosecond laser thermal annealing (LTA), which supplies very localized thermal budgets in a submicrosecond time interval, could satisfy new requirements for junction engineering. Indeed, the localized and transient annealing can lead to activation without diffusion in the solid phase irrespective of whether melting occurs. 2-4) In the molten region, dopant activation is a consequence of impurity trapping, 5) whereas in the solid phase, it is expected to be the result of ultrafast kinetics in the highly damaged matrix. 6) The kinetic Monte Carlo (KMC) method should be the key formalism for accurate computational study of this thermally activated kinetics 7,8) in suitably large systems. However, the conventional KMC algorithms developed so far reliably predict the dynamics of dopant-defect system only for the uniform temperature case; consequently, they cannot be applied to the simulation of LTA processes, which induce highly non-uniform thermal fields (the temperature typically drops by a hundred degrees from the hottest regions to the coldest ones) as well as phase changes when melting takes place. In this paper, we use an original phase-field KMC (PF-KMC) methodology to investigate the kinetics of B implants in Si in the extremely far-from-equilibrium conditions caused by a sequence of implantation and laser irradiation. The particles involved in the system evolution are boron atoms, implantation damage, and boron-interstitial clusters (BICs). A key feature of our method is coupling of the KMC code with a phase-field model, 2,4) which also properly simulates the thermal field evolution in irradiated samples when the liquid-solid phase transition occurs. The PF-KMC simulations correctly predict the experimentally measured local dopant activation levels, demonstrating the central role played by the highly damaged crystalline matrix and the importance of the competitive dopant-defect kinetics in the dopant activation dynamics. The PF-KMC code applied in this research extends the capabilities of a previous version that was restricted to simulation of only the implantation damage, i.e., interstitial-(I) and vacancy (V)-type defects. 9) The advantage of the PF-KMC code over conventional KMC schemes is the proper statistical simulation of a system of interacting particles in non-uniform thermal and phase fields T(t, r), ¯(tr), respectively , whereas the conventional KMC works only for a uniform temperature T(t) and homogeneous phase. We note that for the LTA processes considered here (inducing a temperature gradient on the order of ³1 K/nm), any approximation of T(t, r) using an ad hoc uniform field T(t) + is not feasible because it dramatically affects the position-dependent rates of the different phenomena considered in the simulation. Hence, it is mandatory to couple the stochastic method with a phase-field model that properly simulates the rapidly varying non-uniform thermal and phase (if melting occurs) fields. 9) The current version includes dopant atoms that can reside in a substitutional position B s or form a complex with interstitial defects, i.e., BICs. 10) Without loss of generality, we assumed that I and V point defects as well as the BI complex (one dopant atom bound to one I point defect) are the only mobile species, and that defect clusters X m (X = I or V) as well as the dopant-defect cluster B n I m (a cluster formed by n dopant atoms and m interstitials) can only absorb or emit mobile defects. 11) Hence, we introduce the following reactions in our PF-KMC code: I þ B s $ BI; ð1Þ I þ B n I m $ B n I mþ1 ; ð2Þ V þ B n I m $ B n I mÀ1 ; ð3Þ BI þ B n I m $ B nþ1 I mþ1 : ð4Þ Concerning the B n complexes, they can eventually form by Eqs. (2) or (4) and annihilate only through interaction with the mobile BI or I particles