Predictive device simulation requires, as a prerequisite, an accurate description of the impurity concentrations in silicon; this, in turn, asks for a reliable modelling of the primary fabrication processes, such as ion implantation and dopant diffusion, which determine the concentration-depth profiles. Due to the still partial understanding of the solid-state physics and chemistry which underlies process models, a careful calibration of a process-modelling tool is needed in order to achieve a reasonably good agreement between simulated results and experimental data. The validity of such a calibration procedure is often limited to a particular technology.In this work, by tacking into account a number of recent models, concentration-depth profiles from two different technologies have been successfully reproduced by using the TSUPREM-4 [ 11 process simulator with a unique set of fitting parameters. At first, boron and phosphorous channel profiles, as well as source/drain n+ profiles, have been simulated for a 0.35pm CMOS technology with a calibrated version of the simulator. Next, this version has been successfully checked by simulating profiles of the same dopant species for a 0.25pm CMOS technology.Process simulations have been carried out with the fully coupled dopant-defect diffusion model available in TSUPREM-4. Figures 1 and 2 show the simulated n-channel profiles in comparison with experimental Secondary Ion Mass Spectroscopy (SIMS) data. The asimplanted profiles have been modelled with the Monte Carlo approach and default diffusivities for boron have been assumed. The overall agreement is good in both cases, although minor deviations are present in the surface region. However, since results derived with conventional SIMS in the near-surface region are generally less accurate, it is worth noticing that simulations have been carried out by using the default model for the trapping of implanted boron at the silicon-silicon dioxide interface [2] without an attempt to include the data in the near-surface region into the fit procedure.Simulated and measured n+ junction profiles are shown in figures 3 and 4. To reduce the computational workload, arsenic and phosphorous as-implanted profiles have been modelled by a Pearson type IV distribution, neglecting the channelling contribution. As for the iongenerated damage distribution, the net excess of interstitials has been described by adopting an effective "+n" factor, as recently proposed in [3]. The junction depth is nicely predicted by the simulation. The calculated boron distribution in the n region shows some deviations with respect to the experimental profile, which, however, clearly do not affect the net doping concentration. As for the phosphorous profile, it turned out that is not possible to fit the entire curve by accounting for interstitials only, as in the intrinsic region; to match the SIMS profile, an additional vacancy mechanism has been introduced, as suggested in [4]. It is worth noticing that such modification does not prevent the possibility to matc...
