In accordance with decrease of device size, ultra shallow junctions are required for realizing superior device performance. Enhanced diffusion caused by implantation is a crucial factor to realize ultra shallow junctions. Not only implant but also RTA conditions are key factors to suppress enhanced diffusion. In this paper, process conditions to minimize enhanced diffusion are discussed. Implant ion species, energy, dose and beam current parameters are investigated for implantation and temperature, time and ramping rate parameters are investigated for RTA. Important result is that optimization of not only implant but also RTA conditions should be carried out in order to fabricate ultra shallow junctions.

This paper describes ultra shallow junction formation using 0.5 keV B+/BF2+ implantation, which has the advantage of a reduced channeling tail and no transient enhanced diffusion. In the case of l × 1014 cm−2, 0.5 keV BF2 implantation a junction depth of 19 nm is achieved after RTA at 950°C.

Implantation of B cluster ions into Si using decaborane (B10H14) has been demonstrated. SIMS measurements show that the depth distribution of boron atoms implanted with a monomer ion is exactly matched by that of boron atoms implanted from decaborane ions, if the cluster ion has an order of magnitude larger acceleration energy. According to the Langmuir-Child equation, two orders of magnitude larger space-charge limited ion beam current is possible when decaborane ions are used. Implanted boron atoms from decaborane ions are electrically activated after annealing. Junction depth of the implanted layer with 3 keV decaborane ions is approximately 20nm after annealing at 900°C. Molecular dynamic caluculations show that implantation efficency of boron monomer ions and decaborane ions are the same. However, the number of displaced silicon atoms per implanted boron atom from a decaborane ion impact is 4 times larger than that by boron monomer impact so that a heavily damaged region is created near the impact zone by decaborane ion penetration.

We have investigated the origins of sheet resistance increase in ultra shallow junctions formed by low energy As or Sb implantation. The increase is mainly attributed to dopant loss during annealing due to pileup of dopant at Si02/Si interface. This problem is common to As and Sb and will become more significant as the implantation energies are decreased. We found that the pileup can be classified into two stages from the time dependence of Sb SIMS depth profile .In the early stage of annealing the pileup is very fast and is probably related to the transport of the dopants due to solid phase epitaxial growth of an amorphized layer formed by the implantation. In the later stage the pileup is much slower and is considered to be governed by dopant diffusion.

In this article it is shown that the reactions between dopants and point defects in silicon are slow enough to play a significant role for low temperature transient enhanced diffusion (TED). As a consequence, diffusion models based on the assumption of local equilibrium between dopants and dopant-defect pairs highly overestimate TED at temperatures below 800°C. Without this assumption, i.e. when full dynamic pairing of dopants is included in simulations, good agreement to experimental results is achieved.

The SIA roadmap predicts that junction depths of 500 angstroms are required for CMOS technology nodes of 0.18 μm or beyond by the year 2001. There are several ultra-shallow junction doping techniques currently under investigation. These include beamline ion implantation, plasma immersion ion implantation, and gas immersion laser doping. This study was based on beamline ion implantation of B, P, and As into single-crystal Si wafers at 0.25-2 keV to doses of (2- 10)×1014 at./cm2 with minimized beam energy contamination. Rapid thermal annealing was applied to the implanted wafers at 1000-1050 °C for 10-15 sec at ramp rates of 35- 50 °C/s in a N2 ambient. Transient enhanced diffusion was observed for all three implant species. For example, the depth of 0.25 keV B measured by SIMS increases from 250 to 520 A at a concentration level of l×1017 at./cm3 upon RTA. To minimize the TED, several schemes of defect engineering were applied prior to low energy implantation, including pre-amorphization and implantation of other species. A comparison of TED for different implantation conditions is given with the aim of process development for minimizing TED. The impact of energy contamination on ultra shallow junctions is also addressed.

Atomistic process modeling, a kinetic Monte Carlo simulation technique, has the interest of being both conceptually simple and extremely powerful. Instead of reaction equations it is based on the definition of the interactions between individual atoms and defects. Those interactions can be derived either directly from molecular dynamics or first principles calculations, or from experiments. The limit to its use is set by the size dimensions it can handle, but the level of performance achieved by even workstations and PC's, together with the design of efficient simulation schemes, has revealed it as a good candidate for building the next generation of process simulators, as an extension of existing continuum modeling codes into the deep submicron size regime. Over the last few years it has provided a unique insight into the atomistic mechanisms of defect formation and dopant diffusion during ion implantation and annealing in silicon. Object-oriented programming can be very helpful in cutting software development time, but care has to be taken not to degrade performance in the critical inner calculation loops. We discuss these techniques and results with the help of a fast object-oriented atomistic simulator recently developed.

In this paper we first review the main experimental results concerning boron diffusion in preamorphised silicon, focusing on the role played by the End Of Range defects. It is then shown that the application of the Ostwald ripening theory to the particular geometry of these defects permits to understand why and how they affect dopant diffusion. Contradictory experimental results can be reconciled if one considers that most of the diffusion enhancement occurs during the nucleation stage of the extended defects, and that the amorphous / crystalline interface is a perfect screen for the diffusion of the self-interstitials.

The interactions between end of range dislocation loops and {311} defects as a function of their proximity was studied. The dislocation loops were introduced at 2600 Å by a dual 1 × 1015 cm−2, 30 keV and a 1 × 1015 cm−2 , 120 keV Si+ implantation into Silicon followed by a anneal at 850 °C for 30 minutes. The depth of the loop layer from the surface was varied from 2600 Å to 1800 Å and 1000 Å by polishing off the Si surface using a chemical-mechanical polishing (CMP) technique. A post-CMP 1 × 1014 cm−2, 40 keV Si+ implantation was used to create point defects at the projected range of 580 Å. The wafers were annealed at 700, 800 and 900 °C and plan-view transmission electron microscopy (TEM) study was performed. It was found that the number of interstitials in {311} defects decreased as the projected range damage was brought closer to the loop layer, while the number of rectangular elongated defects (REDs) increased. Experimental investigation showed that REDs are formed at the end-of-range. It is concluded that the interstitials introduced at the projected range are trapped at the end-of-range dislocations. The REDs are formed due to the interactions between the interstitials and the pre-existing loops.

We have studied by SIMS the diffusion of boron in Ge-preamorphised silicon over a range of anneal temperatures and times, focusing on the influence of the depth of the boron profile relative to the crystalline-amorphous (c/a) interface. It is shown that, for all durations, transient enhanced diffusion (TED) occurs on both sides of the c/a interface. For short annealing times, the amplitude of TED varies by about two orders of magnitude between the surface and the end-of-range (EOR) defect band, formed just below the original c/a interface. We propose a model in which TED arises from the coupling between a Si-interstitial supersaturated « box », the EOR defect region, whose supersaturation decreases with time as the EOR defects grow, and a surface whose recombination efficiency is close to that of a perfect sink. The model successfully describes the different behavior of deep and shallow boron profiles, without requiring the existence of a diffusion barrier.

The evolution of defects in Czochralski and epitaxial p- and n-type silicon wafers following irradiation with He. Si or Ge ions at 80 K has been investigated by in situ deep level transient spectroscopy (DLTS). Defect annealing and formation reactions have been observed over the temperature range 80–350 K. In p-type silicon, new species-dependent levels are observed immediately after implantation, but these levels anneal out at or below room temperature. The wellknown divacancy and interstitial defects, usually reported after room temperature implantation, are revealed in the DLTS spectra only upon annealing at 160–200 K. In n-type silicon, vacancy-oxygen pairs are observed immediately after implantation. However, vacancy-related defects continue to form over a broad temperature range in samples implanted with Si or Ge. These observations are consistent with a model whereby vacancies and interstitials are released from defect clusters at temperatures >200 K to form divacancies and other defect pairs which are stable at room temperature.

Positron annihilation spectroscopy has been used to profile the distribution of defects following implantation of 120keV Ge+ into (100) Si in the dose range l x 1010 - lx104 cm−2 . The openvolume defect profiles can be adequately fitted assuming a simple rectangular block distribution extending to 350nm. Using anodic oxidation and etching, a procedure is described which allows details of the defect tails beyond the range of the implanted ion, usually inaccessible to positron -2 annihilation measurements, to be determined. For a time averaged dose-rate (Jr) of 0.02μA cm−2 and incident angle of 7°, open-volume defects are found to exist at concentrations exceeding 1016cm−3 at depths upto 600nm whereas the peak of the depth distribution of the implanted ions (Rp) is 76nm, measured using SIMS. When the time-average dose-rate is increased by a factor of 10, defects persist at concentrations in excess of 1017cm−3 beyond lμm and the Rp increases to 101nm. The open-volume defect profiles are compared to those deduced from Rutherford backscattering-channeling using the fitting routine DICADA.

Ultra-shallow p+ -n junctions have been obtained from Ge+ -preamorphized and crystalline <100> silicon substrates. B+ and BF2+ dopants have been used. Boron was implanted at low energy 3 keV / 1015 cm−2 while an equivalent energy of 15 keV / 1015 cm−2 was chosen for BF2+. Rapid Thermal Annealing (RTA) for 15 s at 950 °C was then used for dopant electrical activation and implantation damage removal. Electrically active defects in these samples were characterized using Deep Level Transient Spectroscopy (DLTS) and isothermal transient capacitance (ΔC(t, T)). Two electron traps were detected in the upper half of the band gap at Ec – 0.20 eV and Ec – 0.45 eV, respectively. These traps are shown to be induced by the Ge+ preamorphization stage. Dopant implantation along with RTA result in the formation of a depth distributed energy continuum for B+ and BF2+ implants. Each continuum has been ascribed to annealing residual defects. Low energy B+ implantation is seen to induce twice as many defects as BF2+, implantation. From isothermal transient capacitance (ΔC(t, T)), reliable damage concentration profiles have been obtained, revealing that preamorphization induces not only defects in the regrown silicon layer but also a relatively high concentration of electrically active defects up to 3.5 μm into the bulk.

The results of several recent experiments aimed at assessing the room temperature migration properties of interstitials (D) and vacancies (V) in ion implanted crystalline Si are reviewed. We show that combining the results of ex-situ techniques (deep level transient spectroscopy and spreading resistance profilometry) and in-situ leakage current measurements new and interesting information can be achieved. It has been found that at room temperature I and V, generated by an ion beam, undergo fast long range migration (with diffusivities higher than 10−1 cm2/sec) which is interrupted by trapping at impurities (C, O) or dopant atoms and by recombination at surface. Analysis of two-dimensional migration of point defects injected through a photolithographically defined mask shows that a strong I recombination (characterized by a coefficient of 30 μm−1) occurs at the sample surface. Moreover, we have found that the strain field induced by an oxide or a nitride mask significantly affects defect migration and produces a strong anisotropy of the defect diffusivity tensor. Finally, using in-situ leakage current measuremens, performed both during and just after ion irradiation, the time scale of point defect evolution at room temperature has been determined and defect diffusivities evaluated. The implications of these results on our current understanding of defect and diffusion phenomena in Si are discussed.

Accurate modeling of extended defect kinetics is of primary importance for predictive modeling of transient enhanced diffusion (TED). Our previously developed model accurately accounts for extended defects and can be used predictively for TED. Using some experimental knowledge about the distribution of the extended defect population we can simplify our model. We demonstrate that reducing the number of solution variables by one doesn't affect the predictive capabilities of the model for extended defect kinetics and TED. However, some caution has to be used when applying the same principles to modeling of dopant deactivation.

Some of the key parameters used in process simulators are formation energy, diffusivity and the concentration of Si self-interstitials. Unfortunately, experimental verification of these parameters is lacking. This work is therefore designed to determine the equilibrium concentration of self-interstitials at various temperatures and thus compute the formation energy for selfinterstitials in Si. Samples with thin (10 nm) buried boron layers were grown by MBE and samples with a buried type II dislocation loop layer were produced by Ge+ implantation into undoped MBE. These samples were subjected to a 40 KeV 1E14 /cm2 Si+ implant followed by anneals at temperatures between 685°C and 815°C for varying times. The loop layer was investigated to monitor the total flux of the interstitials as a function of time while the broadening of the B layer profile was analyzed by SIMS to determine the interstitial supersaturation. A combination of these two values provides an estimate of the equilibrium concentration of the Si interstitials. The results at various temperatures are then used to extract the enthalpy of formation of the Si interstitial.

Interdiffusion of Si/Si0.85Ge0.15 heterojunctions subjected to annealing in inert and oxidizing ambients was investigated as a function of temperature (900 to 1200 °C) and time, allowing comparison between intrinsic diffusion and diffusion under interstitial injection. The Ge diffusivity was extracted using the process simulation program FLOOPS. A time-independent diffusivity was observed for all temperatures. The calculated Ge diffusivity in oxidizing ambient was comparable to that in inert ambient indicating that the interstitial concentration plays a minimal role in interdiffusion. A fractional interstitial component, f1, equal to 0.10 is estimated for annealing temperatures in the range 900 to 1100 °C, while f1 increases to approximately 0.17 at 1200 °C. This may indicate a change in diffusion mechanism at a temperature greater than 1100 °C

Crystalline quality of locally oxidized silicon wafers has been studied. Wafers from different supply sources were found to be differently susceptible to stress-induced dislocation generation, although they had been produced to the same specification. On the basis of the analysis of a depth distribution of the dislocations, critical resolved shear stress of dislocation movement in the bulk areas of the wafers was determined. It varied from about 1.65 to 5.12 MPa and correlated positively to the surface defect density. The results show that uncontrollable variations of bulk silicon properties may significantly influence the stress-induced defect nucleation on the surface of wafers during processing.

Channeling profile analysis is simulated using the dynamic binary collision code Crystal- TRIM. A good agreement between theoretical and experimental data is found for silicon targets which were predamaged by Si+ ions of different energies and analyzed by 140 keV B+ ions. For each example the depth profile of the defects relevant for the dechanneling of the analyzing ions is given. An estimate of the annealing of such defects is obtained by comparison of results for as-implanted and annealed samples.

A new physically-based model for reverse short channel effects has been developed. This kinetic model considers three species: neutral interstitial, immobile dopant and mobile interstitial-dopant pairs. To consider ion-implantation damage and stress effects on the Si/SiO2 interface, a non-uniform sink strength at the Si/SiO2 interface for interstitials has been assumed. ALAMODE, a PDE solver, was used to solve the model. Lateral boron distributions of NMOS devices in the channel near the Si/SiO2 interface have been simulated. Significant boron pile-up was found at the gate edges which is in quantitative agreement with the doping profiles extracted from experimental C-V data. The mechanism of the B diffusion in the channel region is discussed.