This work shows a comprehensive atomistic model to describe amorphization and recrystallization, and its different effects on dopants in silicon. We begin by describing the physical basis of the model used, based on the transformation of ion-implanted dopants and generated point defects into amorphous pockets of different sizes. The growth and dissolution of amorphous pockets is simulated by the capture and recombination of point defects with different activation energies. In some cases, this growth leads to the formation of amorphous layers. These layers, composed of a set of amorphous elements, have an activation energy to be recrystallized. The recrystallization velocity is modeled not only depending on temperature, but also on dopant concentration. During the recrystallization, dopants move with the recrystallization front to simulate the dopant redistribution during solid phase epitaxial regrowth (SPER). At the edge of the amorphous-crystalline interface, the remaining damage forms end-of-range (EOR) defects.

Once the model is explained, we discuss the calibration methodology used to reproduce several amorphous/crystalline (A/C) experiments, including the dependencies of the A/C transition temperature on dose rate and ion mass, and the A/C depth on ion implant energy.

This calibrated model allows us to explore the redistribution of several dopants, including B, As, F, and In, during SPER. Experimental results for all these dopants are compared with relevant simulations.

Two recent papers by Pei et al. and Steen et al. have shown that the observed pile-up of arsenic at Si/SiO2 interfaces surprisingly does not seem to involve point defects as a major factor, causes local distortions that strain the Si in the pile-up region locally, and that the segregated arsenic atoms are deep donors. In this paper, we use ab-initio modeling to study possible configurations for high As concentrations that may fulfill these criteria. We find for a simple model structure that As nearest neighbors become stable in Si in the vicinity of the interface. We also have stud-ied dopant deactivation using bulk-Si models. Even without invoking point defects explicitly and starting from a purely substitutional arrangement, we find that the energetically most favorable configurations are most stable in the neutral charge state, indicating that high enough concentra-tions of arsenic atoms make them electrically inactive and hence result in dopant dose loss.

Low thermal budget annealing approaches, such as millisecond annealing or solid-phase epitaxy (SPE) of amorphized silicon, electrically activate implanted dopants while minimizing diffusion. However, it is also important to anneal damage to the crystal lattice in order to minimize junction leakage. Annealing experiments were performed on low-energy B implants into both crystalline silicon and into wafers pre-amorphized by Ge implantation. Some wafers also received As implants for halo-style doping, and in some cases the halo implants were pre-annealed at 1050°C before the B-doping. The B-implants were annealed by either SPE at 650°C, spike annealing at 1050°C, or by millisecond annealing with flash-assisted RTP™ (fRTP™) at temperatures between 1250°C and 1350°C. Residual damage was characterized by photoluminescence and non-contact junction leakage current measurements, which permit rapid assessment of damage removal efficacy. Damage from the heavy ions used for the halo and pre-amorphization implants dominates the defect annealing behaviour. The halo doping is the critical factor in determining junction leakage current. Millisecond annealing at high temperatures helps to minimize residual damage while limiting diffusion.

Nickel monosilicide (NiSi) is used for lowering the parasitic resistances in source/drain. However, NiSi has a disadvantage of lower thermal instability such as NiSi2 nucleation and agglomeration. We first reported Pt segregation was found at a NiSi/Si interface by Atom Probe (AP) analysis.[6] In this study, we found that the scheme of Ni silicide grain growth and the resultant NiSi crystal shape is strongly affected by the existence of Pt by utilizing AP analysis, TEM and FE-SEM. AP observations were carried on a Ni-Pt as sputterd sample and on a Ni-Pt as annealed sample. The depth profile for the sample after silicidation indicates Pt atoms are segregated at a NiSi surface and NiSi/Si interface. For the analysis of the Pt distribution in the sample after silicidation in more details, we thoroughly analyzed the 3D image as follows: a cylindrical part is extracted from the 3D image; it is divided into 5nm thick slices; and 2D images depicting density-distribution of Pt and As. From these 2D images, we found Pt atoms exist around NiSi grains. The Atom Probe result indicates that Pt atoms segregate at the NiSi surface, grain boundaries and NiSi/Si interface. Plane view TEM and FE-SEM observations were carried out on Ni(Pt) silicide and on Ni silicide without Pt to find the influence of the segregated Pt atoms on the microstructure of NiSi. The spindle shape of NiSi(w/o Pt) grains were observed on the former. On the other hand, we observed on the latter that the Pt addition affected the shape of the NiSi grains and changed NiSi grains to round polygonal shape and the average grain size became smaller. It can be said that the Pt addition suppresses a crystal growth along in a longitudinal direction of a grain. We speculate that fine round shape NiSi grains are formed as a result of the suppression of the anisotropic crystal growth by Pt segregation, and provide the improvement of the thermal stability of the NiSi film.

This study presents the use of atomistic process simulations to optimize p+/n ultra-shallow junctions fabrication process. We first bring to the fore that a high injection of interstitials close to the boron profile decreases the sensibility of boron diffusion to thermal budget. Preamorphization of the substrate is thus necessary to decrease boron diffusion by thermal budget reduction, the latter being obtained by the use of the thermal conduction tool (Levitor) instead of the classical lamp-type rapid thermal annealing. At the same time we show that the use of Levitor does not enhance boron activation, the substrate being preamorphize or not. So Levitor anneal can improve sheet resistance/junction depth trade-off only with preamorphization implant. Experiments are performed that confirm the predictions of our simulations. Further discussions explain activation path of boron during temperature cycle, as a function of amorphous depth, and for both lamp-type and Levitor anneal.

We conducted molecular dynamics (MD) simulations of solid phase epitaxial growth of As-doped Si using a Tersoff potential characterized via comparison to DFT calculations, including energies of AsnV clusters. The Si:As systems were initialized by amorphizing the surface region of crystalline silicon via Si ion implantation and/or selective melting. The remaining crystalline region provides dual function of controlling temperature in system without perturbing regrowth and providing seed for recrystallization. After recrystallization, isolated As atoms occupy substitutional sites, with the average number of nearest neighbors for As changing from about 3.3 in amorphous Si to 4 after crystallization. We observe V incorporation associated with high As concentrations. A small fraction of isolated As atoms have associated vacancies, while vacancies are incorporated in the majority of cases in which there are sites with two As neighbors. These observations are consistent with our previous model developed to explain kinetics of As shallow junction formation which assumed V incorporation at sites with 2 or more As nearest neighbors to account for experimental data.

The behavior of point defects within silicon can be changed significantly by controlling the chemical state at the surface. In ultrashallow junction applications for integrated circuits, such effects can be exploited to reduce transient enhanced diffusion, increase dopant activation, and reduce end-of-range damage.

The use of millisecond annealing to meet ultra-shallow junction requirements for sub-45nm CMOS technologies is imperative. In this study, the effect of flash anneal parameters is presented. Reduced dopant diffusion and lower sheet resistance Rs is achieved for intermediate temperature Tint = 700°C (vs. 800°C). Significantly lower Rs is achieved with peak temperature Tpeak = 1300°C (vs. 1250°C). Multiple shots provide for lower Rs, albeit at the expense of increased dopant diffusion. Based on a simple quantitative model, an optimal flash anneal can achieve 82% dopant activation efficiency.

Laser annealed junctions and advanced ultra shallow junctions are studied in both atomistic modeling and experiments. SIMS and sheet resistance measurement for spike-RTA + Laser annealing show that additional laser annealing after spike-RTA (“+Laser”) improve the dopant activation level without increasing in junction depth. “+Laser” effect become effective in the combination of low spike-RTA temperature and high laser temperature. This effect is significant for As doped layer. Spike-RTA based junction has a limitation in viewpoint of Rs-Xj trade-off. Laser-only annealing is promising candidate to overcome this limitation. Boron diffusion with laser-only annealing is investigated. As atomistic kinetic Monte Carlo modeling show that BnIm complexes and End-of-Range (EOR) defects are formed during sub-millisecond annealing time range. Impact of F co-implant on Boron diffusion and EOR defect evolution during sub-millisecond annealing are also investigated.

The kinetics of intrinsic and dopant-enhanced solid phase epitaxy (SPE) are studied in thick amorphous germanium (a-Ge) layers formed by ion implantation on <100> Ge substrates. The SPE rates for H-free Ge were measured with a time-resolved reflectivity (TRR) system in the temperature range 300 – 540 °C and found to have an activation energy of (2.15 ± 0.04) eV. Dopant enhanced SPE was measured in a-Ge layers containing a uniform concentration profile of implanted As spanning the concentration regime 1 – 10 × 1019 cm3. The generalized Fermi level shifting model shows excellent fits to the data.

Doping of thin body Si becomes very essential topic due to increasing interest of forming source/drain regions in fully depleted planar silicon-on-isolator (SOI) devices or vertical Fin field-effect-transistors (FinFETs). To diminish the role of the short-channel-control-effect (SCE) the Si layers thicknesses target the 10 nm range. In this paper many aspects of thin Si body doping are discussed: dopant retention, implantation-related amorphization, thin body recrystallization, sheet resistance (Rs) and carrier mobility in crystalline or amorphized material, impact of the annealing ambient on Rs for various SOI thicknesses. The complexity of 3D geometry for vertical Fin and the vicinity of the extended surface have an impact on doping strategies that are significantly different than for planar bulk devices.

Atomistic process models are beginning to play an important role as direct simulation approaches for front-end processes and materials, and also as a pathway to improve continuum modeling. Detailed insight into the underlying physics using ab-initio methods and classical molecular dynamics simulations will be needed for understanding the kinetics of reduced thermal budget processes and the role of impurities. However, the limited sizes and time scales accessible for detailed atomistic techniques usually lead to the difficult task of relating the information obtained from simulations to experimental data. The solution consists of the use of a hierarchical simulation scheme: more fundamental techniques are employed to extract parameters and models that are then feed into less detailed simulators which allow direct comparison with experiments. This scheme will be illustrated with the atomistic modeling of the ion-beam induced amorphization and recrystallization of silicon. The model is based on the bond defect or IV pair, which is used as the building block of the amorphous phase. It is shown that the recombination of this defect depends on the surrounding bond defects, which accounts for the cooperative nature of the amorphization and recrystallization processes. The implementation of this model in a kinetic Monte Carlo code allows extracting data directly comparable with experiments.

Ab-initio calculations of dislocation and {311} defect structures in silicon were performed in order to investigate the formation energies as functions of geometry, including the effects of applied strain, with a simple model. Predictions were made concerning the size at which a {311} defect becomes less favorable than a dislocation loop, and it was shown that this is affected by applied strain.

In the trend of scaling down metal-oxide-semiconductor field effect transistors (MOSFETs), reduction of contact resistance at the silicide/silicon (Si) interface will be essential for higher performance. Nickel silicide (NiSi) is considered as a substi-tute for a present electrode material in MOSFETs, cobalt silicide (CoSi2), because silicidation temperature can be reduced as compared with the case of the conventional CoSi2. Hence, we have focused on the NiSi/Si Schottky interface. An ordinary method to increase the dopant concentration at the interface is ion implantation before silicidation process. The dopant atoms are consequently condensed around the interface by snowplow effect, leading to the effective lowering of the Schottky bar-rier height (SBH) because of the band bending enhancement of the Si layer. However, this band bending technique does not reduce the SBH in further scaled MOSFETs. In this context, we studied another possibility of SBH modulation technique, based on the first-principles calculations. Throughout our calculations, we found that a large atomic-scale dipole between impurity and silicide atoms is generated across the interface. Impurity atoms are expected to be condensed because of a large energy gain at the interfaces, leading to the dramatic reduction of the SBH. Based on these results, we proposed a novel di-pole comforting Schottky (DCS) junction. We have also found that the thickness of the Si layer interfacing with the NiSi layer can be 1nm or less. In the present work, we applied this idea to the actual process through experimental techniques. The calculated results suggest that B implantation after silicidation leads to larger B concentration at the interface than that before silicidation, and thereby larger SBH modulation due to interface dipoles can be produced. Then, the NiSi/Si Schottky diodes were formed by ion implantation after silicidation process for dopants (As, B). We evaluated the interface dipoles contribution to the measured SBH reduction. As a result, the dopant atoms were found to be condensed beyond solubility limits on the interface Si side and we confirmed the generated interface dipoles actually reduces the SBT. Furthermore, we explored the other possibility of another type of impurity atoms applicable to the DCS junction. Among some other impurity atoms (Al, In, Mg), the calculated SBH modulation due to dipoles generated around these impurity atoms were found to be further enhanced in some cases. Based on these understandings, we propose a principle for choosing dopants towards ulti-mate lowering of the contact resistance in ultimately scaled MOSFETs.

In this paper, we discuss the accuracy of ab-initio calculations for self-interstitial and boron dif-fusion in silicon in light of recent experimental data by de Salvador et al. and Bracht et al. Map-ping the experimental data onto the activation energy vs. Fermi level representation commonly used to display ab-initio results, we show that the experimental results are consistent with each other. While the theoretical LDA value for the boron activation energy as a function of the Fermi level agrees well with experiment, we find for the self-interstitial in line with other calculations an underestimation of the experimental values, despite using total-energy corrections.

We investigated the performance of 65nm pFETs whereby the source and drain extensions (SDE) were implanted with Carborane, (C2B10H12) a novel form of molecular species. The high atomic mass of this molecule (146 a.m.u.) and the number of boron atoms transported per ion enables the productivity at low energy required for manufacturing of ultra shallow junctions for advanced scaling. In this investigation, Carborane was implanted at 13 keV to produce a Boron profile near equivalent to that produced by the reference BF2 implant. Results of electrical measurements did not exhibit any compromise in the I-V characteristics in terms of Id-Vg and Id-Vd and Ion-Ioff. External resistance and Vt roll-off shifted slightly with respect to the reference devices. This is attributed to a deeper junction with Carborane due to slight offset in the profile matching. It will be shown that with fully matched profiles, a perfect match of the device characteristics can be achieved.

Germanium has regained attention in the semiconductor industry for MOSFET application because of the higher mobility of carriers – two times higher mobility for electrons and four times for holes – as compared to silicon. In the opposite of the Silicon, the major issue with Germanium is to limit the n-dopant diffusion. Usual n-dopants (Phosphorus and Arsenic for example) are not electrically activated at an acceptable level without a large diffusion of the doping profile and a substantial dose loss. In this work, we have studied the influence of low energy and dose implant (15KeV to 40KeV @ 8E13 to 1E15at.cm−2) and low temperature anneal (515°C to 600°C) on diffusion, exodiffusion and activation of the phosphorus dopant into Germanium. The annealing steps were made in RTP furnace, the chemical profile and electrically active profiles were extracted by using Secondary-Ion-Mass Spectroscopy (SIMS) and sheet resistance measurement (Rs). To investigate the implantation-induced defects in depth, cross-sectional micrographs were made by using Transmission Electron Microscopy (TEM). Experimental results show that we achieved an efficient activation level by tuning both dose implant and anneal temperature, limiting the exodiffusion with pratically no diffusion of the dopant. We also show that very abrupt profile can be achieved with appropriate implant and thermal annealing conditions. To limit the leakage current in devices, we suppose we have to limit the defects generated during the implantation. Specially for dopant activation temperature anneal below 550°C, we have shown and observed by cross-sectional micrograph that the defects are totally removed by addition of a pre step of annealing at 400°C.

The diffusion of B atoms in crystalline and amorphous Si has been experimentally investigated and modeled, evidencing the indirect mechanism of these mass transport phenomena. The migration of B occurs after interaction with self-interstitials in crystalline Si (c-Si) or with dangling bonds in amorphous Si (a-Si). In the first case, an accurate experimental design and a proper modeling allowed to determine the microscopic diffusion parameters as the B-defect interaction rate, the reaction paths leading to the diffusing species and its migration length. Moreover, by changing the Fermi level position, B atoms are shown to interact preferentially with neutral or doubly positively charged self-interstitials. As far as the amorphous case is concerned, B diffusion is revealed to have a marked transient character and to depend on the B concentration itself. In particular, boron atoms can move after the interaction with dangling bonds whose density is transiently increased after ion implantation or permanently enhanced by the presence of boron atoms themselves. Unexpectedly, B diffusivity in a-Si is seen to be orders of magnitude above than in c-Si and to depend on the thermal history, i.e. the relaxation status of the amorphous phase. These data are presented and their implications discussed.

The ion implantation steps used in fabricating field effect transistors in ultrathin (6 to 30 nm) silicon-on-insulator (UTSOI) substrates present many challenges. Deep source/drain (S/D) implants in UTSOI layers are a particular concern, since it can be difficult to implant the desired dose without amorphizing the entire SOI thickness. In a first study, we investigated the effect of implant temperature (20 to 300 °C) on the sheet resistance (Rs) of 28 nm thick SOI layers implanted with As+ at an energy of 50 keV and a dose of 3 × 1015 /cm2, and found Rs values after activation sharply lower for samples implanted at the highest temperature. In a second study, on 8 nm thick SOI layers implanted with As+ at an energy of 0.75 keV and doses in the range 0.5 to 2 × 1015/cm2, the benefits of the elevated implantation temperature were less clear. Explanations for these effects, supported by microscopy, medium energy ion scattering (MEIS), and optical reflectance data, will be discussed.