Silicon layers containing B in excess of a few atomic percent create a supersaturation of Si self-interstitials in the underlying Si, resulting in enhanced diffusion of B in the substrate (boron-enhanced diffusion, BED). The temperature- and time-dependence of BED is investigated here. Evaporated-boron as well as ultra-low energy 0.5-keV B-implanted layers were annealed at temperatures from 1100 to 800°C for times ranging from 3 to 3000s. Isochronal 10s anneals reveal that the BED effect increases with increasing temperature up to 1050°C and then decreases. In contrast, a simulation of interstitial generation via the kick-out mechanism predicts a decreasing dependence on increasing temperature leading to the conclusion that the kick-out mechanism is not the dominant source of excess interstitials responsible for BED. The diffusivity enhancements from the combined effects of BED and TED (transient-enhanced diffusion), measured in 2×105cm−2, 0.5-keV B-implanted samples, show a similar temperature dependence as seen for evaporated B, except that the maximum enhancement occurs at 1000°C. The temperature-dependent behavior of BED supports the hypothesis that the source of excess interstitials is the formation of a silicon boride phase in the high B concentration silicon layer.

To meet the challenge of achieving ultra shallow p+/n source/drain extension junctions for 0.1 Oim node devices, ultra low energy boron implant and advanced annealing techniques have been explored. In this paper, we report the extended defect and boron diffusion behavior with various implant and annealing conditions. Boron implants were performed at energies from 0.25keV to lkeV and doses of 5 × 1014 cm−2 and 1 × 1015cm−2. Subsequent anneals were carried out in nitrogen ambient. The effect of energy, dose and oxide capping on extended defect formation and enhanced dopant diffusion was examined. It was observed that a thin screen oxide layer (35Å), grown prior to implantation, reduces the concentration of dopant in the Si by a significant amount as expected. This oxide also reduces the dislocation loops in the lattice and lowers diffusion enhancement of the dopant during annealing. The final junction depth can be optimized by using a low thermal budget spike anneal in a controlled oxygen ambient.

This paper describes the influence of the sub-nm screening and capping oxide layer of wet process on B diffusion for 50 nm shallow junction formation. The reason is proposed that the knock-on oxygen into Si makes B diffusion faster such as in the case of oxidation enhanced diffusion. To suppress the B enhanced diffusion, the screening oxide formed before implant should be thin to reduce the knock-on oxygen. The wet process before annealing strongly influences the B diffusion, in particular, thicker capping oxide with sub-oxide structure and knock-on oxygen has an important rule for the enhanced diffusion.

Over the last couple of years rapid thermal annealing (RTA) equipment suppliers have been aggressively developing lamp-based furnaces capable of achieving ramp-up rates on the order of hundreds of degrees per second. One of the driving forces for adopting such a strategy was the experimental demonstration of 30nm p-type junctions by employing a ramp-up rate of ≈400°C/s. It was subsequently proposed that the ultra-fast temperature ramp-up was suppressing transient enhanced diffusion (TED) of boron which results from the interaction of the implantation damage with the dopant. The capability to achieve very high temperature ramp-rates was thus embraced as an essential requirement of the next generation of RTA equipment.

In this paper, recent experimental data examining the effect of the ramp-up rate during spike-and soak-anneals on enhanced diffusion and shallow junction formation is reviewed. The advantage of increasing the ramp-up rate is found to be largest for the shallowest, 0.5-keV, B implants. At such ultra-low energies (ULE) the advantage arises from a reduction of the total thermal budget. Simulations reveal that a point of diminishing return is quickly reached when increasing the ramp-up rate since the ramp-down rate is in practice limited. At energies where TED dominates, a high ramp-up rate is only effective in minimizing diffusion if the implanted dose is sufficiently small so that the TED can be run out during the ramp-up portion of the anneal; for larger doses, a high ramp-up rate only serves to postpone the TED to the ramp-down duration of the anneal. However, even when TED is minimized at higher implant energies via high ramp-up rates, the advantage is unobservable due to the rather large as-implanted depth. It appears then that while spike anneals allow the activation of ULE-implanted dopants to be maximized while minimizing their diffusion the limitation imposed by the ramp-down rate compromises the advantage of very aggressive ramp-up rates.

Boron was implanted into n-type Si at energies from 500 eV to 1 keV and doses near 1 E14 cm-2and 1E51 cm−2. Electrical activation was achieved by rapid thermal annealing (RTA) in nominally pure N−2and 0.1% 02 with the fastest available heating rates of up to 150 °C/s, cooling rates up to 80 °C/s, and included “spike” anneals with minimum dwell time at peak temperature. Measurements of sheet resistance, Hall coefficient, and secondary ion mass spectroscopy profiling were used to determine dopant activation and diffusion. Surface oxidation was studied by film thickness ellipsometry. Analyses of electrical transport measurements are used to relate junction depths to sheet resistance and their dependence on annealing temperature and time. For spike annealing, junction leakage and adequate activation limits the minimum practical temperature while diffusion limits the maximum practical temperature for formation of shallow junctions.

Experiments have been caried out to form ultra-shallow (Xj <50nm) and abrupt (Xjs < 5nm/decade) P'junction for sub-50 nm CMOS devices using a combination of shallow implant, Ge preamorphization and high energy Si implant as an interstitial getter layer. Experimentally, it was observed that the Si getter layer, not only stopped the TED at the boron tail but also promoted enhanced diffusion close to the surface boron peak. These unique features have enabled the shallowest and sharpest box-like boron junction yet achieved by implant. With I kV BF2, Xj ∼ 23 nm, Xjs ∼ 48 A/decade, no Ge end of range damages and good dopant activation at the same time.The sheet resistance ρ − 1 kohm/sq is comparable to shallow BF2 + Ge and is better than the shallow BF 2 alone (ρ ∼ 2.38 kΩ/sq) or the shallow BF2 + Si implants (ρ ∼ 1.5 kohm/sq). Tests with device leakage test structures show that there is no additional junction leakage introduced by the Si getter layer.

We investigated the atomic transport properties and electrical activation of boron in crystalline epitaxial silicon after ultra-low energy ion implantation (0.25–1 keV) and rapid thermal annealing (750–1100 °C). A wide range of implant doses was investigated (3×1012-1×105/cm2). A fast Transient Enhanced Diffusion (TED) pulse is observed involving the tail of the implanted Boron, the profile displacement being dependent on the implant dose. The excess of interstitials able to promote enhanced diffusion of implanted boron occurs, provided the implant dose is high enough to generate a significant total number of point defects. The Boron diffusion following the fast initial TED pulse can be described by the equilibrium diffusion equations.

The electrical activation of ultra-shallow implants is hard to achieve, due to the high concentration of dopant and point defects confined in a very shallow layer that significantly contributes to the formation of clusters and complex defects. Provided a correct combination of annealing temperatures and times for these ultra-shallow implants is chosen, however, a sheet resistance 500 Δ/square with a junction depth below 0.1μm can be obtained, which has a noteworthy technological relevance for the future generations of semiconductor devices.

Formation of p-type shallow junctions for future generations of Si technology will require ion implantation of B at very low energies, i.e. below 1 keV, where the beam formation and transport at reasonably high currents are hindered by Coulomb repulsion of ions at high volume density. An alternative to implantation of monomer ions at a very low energy is implantation of large molecular ions at a higher energy. In the latter case, the implantation depth of the atoms corresponds to a fraction of the beam energy, partitioned between the atoms of the molecule. The decaborane molecule (B10H14) is of particular interest for implantation of p-type shallow junction in Si, because each of the B atoms carries only 9% of the molecule's kinetic energy. Experimental PMOS devices made using decaborane implantation have been demonstrated recently. It also has been shown that transient enhanced diffusion (TED) of B in Si implanted with B ions and with decaborane ions at the equivalent dose and energy are the same. The prospect for using decaborane in ion sources is examined, based on measurements of its ionization and dissociation properties. It is shown that decaborane molecules are effectively ionized by electron impact in the energy range near 100 eV.

As the drive towards the production of 100 nm CMOS devices pick up speed, the practical aspect of transistor shallow junction formation, including a large menu of process integration issues, must now be solved in a short order. The most direct path to 50 nm junction depths is through the sub-keV boron implantation and rapid thermal annealing.

The material aspects of the process integration centers on: (1) CMOS devices for shallow, highly-activated and abrupt junctions (involving the choice of ion species [B, BF, B10H1 4, BSi2, etc.], substrate materials [ Cz, Epi, SOI], anneal conditions [ramp rate, soak time, ambient gas], etc.) and (2) Defect-dopant interactions during annealing (including surface reactions of high concentration species [B, F], diffusion and carrier trapping by background and co-implanted species [C, 0, F, etc.].

Process data for atomic and electrical activity profiles as well as defect and interface structures will be presented to illustrate progress towards understanding these complex process interactions. A particular focus will be the effects of anneal ambient and rapid temperature rise times approaching the “pike” anneal ideal.

The global scaling down of device dimensions requires process technologies which are able to create ultra-shallow junctions. Besides using ultra-low implant energies for shallow junction creation we present an alternative approach for the creation of MDD (Medium Doped Drain) junctions by using moderately low implant energies. Our approach employs the dopant/point-defect interaction mechanism to retard dopant diffusion as well as dopant de-activation in the tail of the diffusion profiles to achieve suitable shallow junctions. The silicon preimplant allows fabrication of 90nm arsenic, 150nm phosphorus, and 140nm boron metallurgical junctions for a 40keV arsenic, 20keV phosphorus, and 8keV boron implant.

We have compared the electrical characteristics and the depth profile of an ultrashallow junctions formed by boron implantation at 0.5 keV and BF2 implantation at 2.2 keV. The modeling of the boron profile was performed using the Monte Carlo method for an as-implanted profile and the computationally efficient method for transient-enhanced diffusion. A junction depth of BF2 is shallower than that of boron after annealing. HF dipping prior to rapid thermal annealing causes a significant loss of dopant and high sheet resistance. Considering the 0.1 νn metal-oxide-semiconductor field-effect-transistor (MOSFET) application, the optimizations of implantation and annealing conditions are necessary to satisfy the requirement ofjunction depth and sheet resistance.

At the current pace of semiconductor technology development, transistor dimensions in advanced IC products will approach the range of a few tens of nanometers within the next decade. This presents a major challenge for our understanding of defects and diffusion in these tiny devices during processing. In response, an almost explosive growth in research on process physics has taken place at universities, national institutes and industry research labs worldwide. The central issue is the phenomenon of nonequilibrium diffusion driven by processing steps such as oxide growth, high concentration gradients of impurities, and annealing of damage caused by ion implantation. Nonequilibrium diffusion arises from perturbations to the natural thermal equilibrium concentrations of point defects - interstitial atoms and vacancies - in the silicon crystal. This paper gives a snapshot of our current understanding of the atomic-scale interactions between point defects and impurity atoms, extended defects and interfaces, as revealed by recent experimental and theoretical studies. The paper emphasizes the important role played by defect cluster ripening during transient enhanced diffusion and dopant activation.

In this work we investigate boron diffusion as a function of the Fermi-level position in crystalline silicon using ab-initio calculations and the nudged elastic band method to optimize diffusion paths. Based on our results, a new mechanism for B diffusion mediated by Si self-interstitials is proposed. We find a two-step diffusion process for all Fermi-level positions, which suggests a kick-out with a directly following kick-in process without extensive B diffusion on interstitial sites in-between. Our activation energy of 3.47 – 3.75 eV and diffusion-length exponent of -0.55 to -0.18 eV are in excellent agreement with experiment.

Self-diffusion in silicon has been studied using epitaxially grown isotopically enriched structures under nonequilibrium concentrations of intrinsic point defects created by thermal oxidation and nitridation. Comparison of identical anneals for self, antimony, and phosphorus diffusion in silicon enables us to determine bounds on the fractional contributions of microscopic mechanisms for Si self-diffusion in the temperature range 800–1100°C. We obtain direct experimental evidence for a dual vacancy-interstitial mechanism of self-diffusion in silicon and show that the fractional contribution of each mechanism has a weak dependence on temperature.

The diffusion of antimony in Si and SiGe under the influence of grown-in point defects is studied in this work. The layers were grown by very low temperature MBE, and the diffused profiles measured using SIMS analysis. Fast diffusion of Sb is observed in both Si and SiGe. and is thought to be caused by grown-in point defects, specifically vacancies, arising from the very low temperature MIBE growth. The vacancies may be removed by heat treatments. and it was found that the processing the samples received before the heat treatments caused the most substantial diffusion. Comparison of the diffusivities for inert diffusion at 850°C following the heat treatments suggests 30 minutes at 800°C is sufficient to remove the majority of the grown-in defects. The diffusion of Sb in Si is thus demonstrated to be a very sensitive monitor of grown-in vacancies, and dopants which diffuse at least partly by vacancy mediated processes would be expected to be affected by this effect.

We have studied the formation of surface blisters in <100> n-type silicon following coimplantation with boron and hydrogen. In this work we study the effect of having the boron depth larger than the hydrogen depth. The silicon substrates had four different dopant levels, ranging from 1014 cm−3 to 1019 cm−3. The Si substrates were first implanted with B ions at 240 keV (Rp = 705nm) to a dose of 1015cm−2. Some of the B implanted samples were left in their as-implanted state, others were given a rapid thermal anneal (RTA) at 900°C. Boron implanted and virgin n-type Si samples were then implanted at cryogenic temperatures with 40 keV H+ (Rp= 475nm) to a dose of 5 x 1016 cm−2. Following H+ implantation, all the samples were vacuum annealed at 390°C for 10 minutes and examined for hydrogen blister formation. No blistering was observed in samples implanted with H only. In all cases, the blister depth was consistently found to be strongly correlated with H damage profile rather than the H or B concentration profiles. These results will be discussed in terms of defect-dopant interactions and the influence of doping on H kinetics and H-bubble nucleation.

We have carried out electrical characterization of defects in heavily damaged silicon, where damage is created by MeV heavy ions at doses near but below amorphization threshold and samples are subjected to different annealing conditions. Defect characterization is carried out using combination of deep level transient spectroscopy (DLTS) and isothermal time analyzed spectroscopy (TATS). The defect spectrum is observed to be sensitive to I) furnace annealing between 400–600°C, ii) low temperature oven annealing at 160°C, and iii) forward injection current. The observed changes in spectra are indicators of extreme sensitivity of dominant defects to relaxation of the disordered medium, and recombination enhanced reactions. Surprisingly the defect spectra is usually dominated by a single peak with unbroadened lineshape characteristic of discrete energy level in the bandgap, though electrical signatures keep varying for different processing conditions. In the light of these measurements, we discuss the nature of stability and metastability of the defects believed to be due to intrinsic defect clusters.

An improved model of {311} defect evolution is developed based on the rate equations approach. A new expression for the reaction rate constant is presented that is based on the assumption that interstitials may react with the {311} defects along their whole surface. The energetics of {311} defects is treated by calculating the strain energy within the framework of the theory of dislocations in isotropic continua. Using the core energy of the atoms in the defects as the only fit parameter, we explain a wide range of experimental data. Furthermore, we apply the model to investigate closure assumptions used in moments models and propose a new two-moments model that uses the rate equations solver as a pre-processor.

We have studied the damage annealing process using kinetic lattice Monte Carlo (KLMC) and molecular dynamics (MD) with initial damage distribution from Monte Carlo ion implant simulations. MD calculations find a long range interstitial vacancy interaction, as also seen in previous tight-binding molecular dynamics (TBMD) simulations.1 The influence of the long range interaction as well as the initial spatial correlations present in the implant damage are then analyzed with KLMC in the form of corrections to the +1 model. We find that both long range interactions and the initial spatial correlations are significant at low doses, while the effects disappear at high doses.

We present new experimental results on the transient enhanced diffusion (TED) of boron after ion implantation. The investigation is focussed on effects that influence TED of shallow profiles in the absence of {311}-defects. Under these conditions, TED is mainly determined by the formation of boron-interstitial complexes (BIC). In addition, effects from the proximity of the surface become more and more important. Insight into the behavior of the dopant atoms is obtained by the comparison with simulations.