We present a Lattice Kinetic Monte Carlo study of the atomic evolution leading to {311} defects formation upon annealing of a damaged Si-crystal. Self-interstitial (I) agglomeration is modeled by using local interaction and considering the energetic cost to under/over coordinate the Si atoms belonging to an I-complex. The static properties of the I aggregates as derived by molecular dynamics calculations, in the two extreme regimes of very small and very large clusters, have been mapped in the model. The typical evolution of an excess of Si ions is characterized by three distinct stages: (1) the formation of clusters consisting of few interstitials in a over-coordination state, (2) their redistribution in larger agglomerates containing a few of these small I clusters all preserving their original structure, (3) a transition leading to Is rearrangement along the ≪110> chains, which are the structural units of {311} defects. The duration of the preliminary stages critically depends on temperature and density of the added atoms.

The transient enhanced diffusion (TED) during activation annealing of ultra low energy implanted boron (0.5 keV & 1 keV, 1×1013/cm2 & 1×1014/cm2) in silicon is investigated in detail. Annealing in the temperature range from 450°C to 750°C is either performed directly after implantation or after the removal of a surface layer before annealing. The kinetics revealed two regimes of enhanced diffusion ruled by different decay constants and different activation energies. The dependence of these two processes on implantation energy and annealing temperature is described and explained from the microscopical point of view. The annealings performed after surface layer removal, revealed that the defects responsible for the faster diffusion are located deeper than the defects responsible for the slower process.

Formation of p-type shallow junctions for future generations of Si devices will require ion implantation of B at very low energies (< 1 keV). An alternative to implantation of monomer ions at very low energy is implantation of large molecular ions at a higher energy. In an ion beam of decaborane (B10H14) each of the B atoms carries only 9% of the ion kinetic energy. We have examined ionization properties of decaborane and built an experimental ion source and an implantation apparatus with magnetic mass analysis. Analyzed decaborane ion beams with energies from 2 to 10 keV and beam currents of several microamperes were obtained. Si samples were implanted with decaborane ions and the implanted dose measured by current integration was compared with B content obtained by nuclear reaction analysis. Experiments with electrostatic beam deflection show that the large ions survive the transport in the implanter environment and that neutralization is negligible. During implantation, the retained B dose is reduced in comparison with the nominal implanted dose due to sputtering. Dose loss is greater at 200 eV compared to 500 eV. The properties of decaborane ion beams and the prospects of using them for shallow implantation of B into Si are discussed.

The interaction between boron and excess silicon interstitials caused by ion implantation hinders the formation of ultra-shallow, low resistivity junctions. Previous studies have shown that fluorine reduces boron transient enhanced diffusion, however it is unclear whether this observed phenomenon is due to the fluorine interacting with the boron atoms or silicon self-interstitials. Amorphization of a n-type Czochralski wafer was achieved with a 70 keV Si+ implantation at a dose of 1×1015/cm2. The Si+ implant produced a 1500Å deep amorphous layer, which was then implanted with 1.12 keV 1×1015/cm2 B+. The samples were then implanted with a dose of 2×1015/cm2F+ at various energies ranging from 2 keV to 36 keV. Ellipsometry measurements showed no increase in the amorphous layer thickness from either the boron or fluorine implants. The experimental conditions allowed the chemical species effect to be studied independent of the implant damage caused by the fluorine implant. Post-implantation anneals were performed in a tube furnace at 750° C. Secondary ion mass spectrometry was used to monitor the dopant diffusion after annealing. Transmission electron microscopy (TEM) was used to study the end-of-range defect evolution. The addition of fluorine reduces the boron transient enhanced diffusion for all fluorine energies. It was observed that both the magnitude of the boron diffusivity and the concentration gradient of the boron profile vary as a function of fluorine energy.

To continue scaling dimensions of transistors, higher dopant concentration levels are needed for ultra-shallow contacts. Therefore studies of dopant activation have been performed in preamorphized silicon wafers with various boron implant conditions to determine the maximum achievable dopant concentrations after Solid Phase Epitaxial Regrowth (SPER) alone. In the first experiment a silicon piece was preamorphized with a 30 keV, 1×1015 cm−2 and 90 keV, 1×1015 cm−2 Si+ implant followed by a 30 keV, 1×1015 cm−2 B+ implant. Solid phase epitaxial regrowth at 500 °C indicates that boron can be activated at low temperatures. Ultra Low Energy (ULE) implants were studied in the second experiment. Silicon wafers were implanted with 2.5 keV, 1×1015 cm−2 Si+ to amorphize and then B+ was implanted at 0.5 keV in the dose range of 1×1015 to 9×1015 cm−2. Samples were annealed in the temperature range of 500 to 650 °C. High concentrations of boron make it difficult to fully regrow amorphous layers and thus yield marginal electrical properties. Much of the boron remains inactive, particularly at the higher dose implants. In both experiments Variable Angle Spectroscopic Ellipsometry (VASE) is used to measure amorphous layer thickness and Hall effect measures active boron dose. For the first experiment, Secondary Ion Mass Spectrometry (SIMS) data compares chemical dose to active dose during the regrowth process. Sheet resistance data is obtained from a four point probe for the ULE implant experiment.

We propose a novel process module by using cryo-implantation and rapid thermal annealing (RTA). Boron or arsenic ions were implanted into a 8 inch (100) Si substrate which was cooled by using liquid nitrogen. The substrate temperature was controlled to be below at -160°C during ion implantation. It was found that an amorphous layer was formed by boron or arsenic implantation and the amorphous layer was completely recovered to a single crystal after annealing at 900°C for 30sec. No dislocation was observed in the implanted layer. It was also found that the thermal diffusion of boron was suppressed by cryo-implantation. PN junction depth was found to be about 10-20% shallower than that of room temperature implantation. These results suggest that transient enhanced diffusion of boron can be reduced by suppressing vacancy migration toward the surface during implantation. Cryo-implantation was found to be very effective in reducing defects and PN junction leakage was successfully reduced by one order of magnitude as compared with room temperature implantation.

Low temperature annealing combined with pre-damage (or preamorphization) implantation is a very promising method to overcome the activation barrier in ultra-shallow junction formation. We have made a 32 nm p+/n junction with sheet resistance of 290 /sq. using 20 keV 4×1014 Ω/cm2 Si followed by 2 keV 1×1015 at./cm2 B implantation and 10 minutes 550 °C annealing. This paper studies the boron activation mechanism during low temperature annealing. The result shows that placing B profile in the vacancyrich region has much better boron activation than placing B profile in interstitial-rich region or without pre-damage. It also shows that a significant portion of boron is in substitutional positions before annealing. The amount of substitutional boron is correlated to the amount of vacancies (damage) by the pre-damage Si implantation. The result supports our speculation that vacancy enhances boron activation.

In this work, we have studied the relative stability of perfect (PDLs) and faulted (FDLs) dislocation loops formed during annealing of preamorphised silicon. In particular, we have investigated the effect of the initial supersaturation of Si interstitial atoms (Si(int)s) created by the implantation process on their thermal evolution. Transmission Electron Microscopy analysis shows that in samples with a low Si interstitial supersaturation, FDLs are the dominant defects while PDLs appear as the most stable defects in highly supersaturated samples. We have calculated the formation energies of both types of dislocation loops and found that, for defects of the same size, FDLs are more energetically stable than PDLs, if their diameter is smaller than 80 nm and viceversa. The application of these calculations to the samples studied in this work indicates that a direct correspondence exists between the formation energy of the two defect families and the number of atoms bound to them. Moreover, we have shown that the relative stability of FDLs and PDLs depends on the initial supersaturation of Si(int)s created during the implantation process.

To investigate the effects of ramp rate on the transient enhanced diffusion of boron in silicon, laser thermal processing (LTP) in the nonmelt regime has been investigated. A nonmelt laser anneal has been performed on a 5 keV, 1e15 boron implant. The implant energy of 5keV was chosen to simplify analysis. A rapid thermal anneal (RTA) at 1000°C and furnace anneals at 750 °C were used to show the effect of post annealing on the LTPd samples. Results show the sheet resistance drops by up to a factor of two for samples receiving the nonmelt LTP and the RTA compared with the samples just receiving the RTA. An increase in the hall mobility was also observed for the samples receiving the LTP. The nonmelt LTP was also shown to strongly affect the extended defect density. During post anneals, a higher density of smaller defects evolved in the samples receiving the LTP.

Antimony implantation is a promising technique for fabricating ultra-shallow n+/p junctions for extensions of sub-100-nm n-MOSFETs. By increasing the Sb+ implantation dose to 6×1014 cm−2, sheet resistance (Rs) of an implanted layer was reduced to 260 /sq. for rapid thermal annealing (RTA) at 800°C. The obtained junction depth of 19 nm is suitable for sub-100-nm MOSFETs. However, the reduction in the sheet resistance showed a tendency to saturate. No pileup at the Si-SiO2 interface, which was the major origin of dopant loss in lower dose cases was, observed in Sb depth profiles in this case. However, in the case of 900°C RTA, Sb depth profiles indicated that Sb was nearly immobile in the region where Sb concentration exceeded 1×1020 cm−3. These results imply that the major limiting factor of Rs reduction under the highdose condition is Sb precipitation, which is different from the lower dose cases.

It has been shown recently that Au labeling can be used to profile vacancy-type defects located near half the projected range (½Rp) in MeV-implanted Si. In this work we have quantified the technique by determining the ratio of vacancies annihilated to decrease in the number of Au atoms trapped (calibration factor ‘k’) for the Au labeling technique. The 3 step experiment involved: 1) a high-energy Si-self implant (HEI) followed by an anneal to form stable vacancy clusters, 2) a controlled removal of vacancies via a medium energy Si self implant and interstitial-cluster dissolution anneal, and finally 3) Au labeling to count the change in vacancy concentration in the near surface region (0.1-1.6μm). It is seen that the Au concentration decreases linearly with increasing interstitial injection and the slope of this decrease determined the number of vacancies per trapped Au atom. The value of k was determined to be 1.2±0.2 vacancies per trapped Au atom.

Understanding the coupling between native point defects and dopants at high concentrations in silicon will be key to ultra shallow junction formation in silicon technology. Other effects, such as transient enhanced diffusion (TED) will become less important. In this paper, we first describe how thermodynamic properties of the two native point defects in silicon, namely vacancies and self-interstitials, have been obtained by studying self-diffusion in isotopically enriched structures. We then discuss what this tells us about dopant diffusion. In particular, we show that the diffusion of high concentration shallow dopant profiles is determined by the competition between the flux of mobile dopants and those of the native point defects. These fluxes are proportional to the interstitial or vacancy components of dopant and self-diffusion, respectively. This is why understanding the microscopic mechanisms of silicon self-diffusion is important in predicting and modeling the diffusion of ultra shallow dopant profiles. As an example, we show experimental data and simulation fits of how these coupling effects play a role in the annealing of shallow BF2 ion implantation profiles. We conclude that relatively low temperature furnace cycles following high temperature rapid thermal anneals (RTA) have a significant effect on the minimum junction depth that can be achieved.

The purpose of this study is to illustrate the effects of the structure of the silicon surface on low-energy implants commonly used in current silicon technology. T o this aim Molecular Dynamics simulations with classical forces are used to study implantation of argon and boron with a kinetic energy in the range from 5eV to 30 eV into a silicon stepped surface. In addition to these calculations, the binding energy of boron atoms embedded in small clusters with a step shape hav e been studied with a quantum mechanical method of a semiempirical type. The purpose of these calculations is to compare the stable sites generated by theimplan ts with the ones attained during thermal treatments under near-equilibrium conditions.

Monte Carlo simulation codes such as TRIM or MARLOWE show a net displacement of interstitials with respect to vacancies from the Frenkel pairs produced in implant cascades. As a result, defect profiles after recombination have an excess of vacancies (Vex) in the shallow region of these implants, with excess interstitials near the projected range Rp. Typically, accurate estimates of the Vex distribution are computer intensive as the complete history of each damage cascade must be recorded. In this work we introduce a method for fast estimation of the Vex profile based on the derivative of the recoil distribution obtained from binary collision simulations using the Kinchen-Pease approximation for damage production. Also, Au labeling has been used to quantitatively analyze and compare the experimental depth profiles of Vex produced by implants of B, Si, and Ge into Si for a range of doses. Specific conditions like matching of the projected range of the implanted ions and matching of the number of Frenkel pairs created have been compared. This detailed study and comparison with simulation provides the first systematic, quantitative measurements of the Vex concentrations over such a range of species and doses. In analogy to the “+ number” for excess interstitials from ion implantation, we determine the range of “minus (-) numbers”, that is the number of excess vacancies per implanted ion for the different species. For Si it is constant in the dose range studied at -0.05, while for B it is -0.001 to -0.004 and for Ge, -0.1 to -0.2. The general shapes of the Vex profiles are similar to those obtained from simulation. Evidence is presented that one significant difference, namely an experimentally observed vacancy depletion near the surface, may be due to the diffusion and annihilation of vacancies at the surface. In addition, an increased nucleation probability of vacancy defects close to Rp may account for the growth of excess vacancies from close to Rp towards the surface with increasing implant dose.

Boron, a P-type dopant, experiences Transient Enhanced Diffusion (TED) via interstitials. The Boron TED and {311} dissolution rates are explored as a function of implant energy dependence. Silicon implants of 1014/cm2 at 20, 40, 80, and 160 keV were used to damage the surface of a wafer with an epitaxially grown boron marker layer. Samples were annealed at 750°C for 15 to 135 minutes and 800°C for 10 to 30 minutes to observe the diffusion exhibited by the marker layer and to correlate this with the dissolution of {311} type defects. The diffusion enhancement depends strongly on implant energy but the {311} dissolution rate is weakly dependent.

Density functional theory was employed to explore the diffusion mechanism of boron in amorphous silicon oxide. The oxide was modeled with clusters of various sizes, and both neutral boron atoms and cations were considered. Three stable structures were found where B (or B+) was inserted into oxide: one in which B (or B+) is divalent and two in which B (or B+) is trivalent. Boron diffusion through silicon oxide proceeds as a sequence of B hops from one inserted position to another. For neutral boron the rate limiting step is B hop from one of trivalent structures to a divalent one with activation energies (Ea) in the range of 2.0-3.1 eV, depending on the model cluster. In the case of a cation the rate-limiting step is the B+ hop over the O atom in a divalent structure Si-B+-O-Si with calculated Ea of 2.4-2.8 eV. Experimental activation energies for B diffusion in silicon oxide are in the 2.3 - 4.2 eV range. Our results suggest that both neutral and cation boron can participate in B diffusion in oxide.

An expression for concentration dependence of diffusion coefficient of ionized impurity in semiconductors which is derived which is valid for the whole range of carrier degeneracy under the condition that the impurity concentration greatly exceeds the intrinsic carrier density. Due to reduction of impurity field screening by mobile carriers at degenerate conditions the diffusion coefficient is shown to be a monotonously increasing function of impurity concentration. It is proved that taking band gap narrowing into account results in a reduction of the diffusion coefficient in comparison with the case of unperturbed band structure. In addition, a decreasing concentration dependence of the diffusion coefficient can be realized at relatively low impurity concentrations.

CVD grown boron marker layers were used to monitor the release of silicon interstitials from an arsenic doped surface region that was subsequently implanted with silicon. These structures were annealed for various times at 750°C under a nitrogen gas flow. A comparison of boron spike enhancement and defect dissolution is made. It is shown that enhancement values from the Si+ implant were reduced at short times for samples containing arsenic compared to samples implanted with Si+ alone or As+ alone. The TEM results showed that defect densities were dramatically reduced for the samples containing As. These results imply that the previously reported reduction in {311} formation observed in As doped wells is most likely not a Fermi level effect and is consistent with the formation of As interstitial clusters (AsIC's). The data shows that AsIC's form and control extended defect formation and slow the enhanced diffusion.

The possibility to suppress undesirable diffusion of the base dopant boron in siliconbased bipolar transistor structures by the incorporation of a high concentration of carbon has lead to renewed interest in the behavior of carbon in crystalline silicon. The present paper will review essential features of carbon in silicon including solubility, diffusion mechanisms and precipitation behavior. Based on this information the possibilities to use carbon to influence diffusion of dopants in silicon by the introduction of non-equilibrium concentrations of intrinsic point defects will be discussed as well as the reason for the relatively high resilience against carbon precipitation. Interactions between carbon and oxygen will be mentioned, especially in the context of an as yet unexplained fast out-diffusion of carbon close to the surface.

The early use of ion bombardment of semiconductors for forming doped regions was viewed as a room-temperature process by solid-state scientists. Many interesting, but relatively useless devices were made by implanting species such as Na and Cs ions to form pn junctions from radiation damage or interstitial impurities. The revolutionary idea that one could implant Group III and V dopants into semiconductors and then heat the implanted substrate to above 800C didn't appear until 10 years after Shockley's 1954 patent. At that time, implantation damage became relatively unimportant as processes evolved with high temperature, long time diffusions. With the advent of rapid thermal processing, the attention shifted back to implantation-induced defects to explain transient-enhanced-diffusion effects. Today's challenges in forming ultra-shallow junctions by ion-implantation are in controlling and minimizing the damage structures that dominate junction activation and diffusion. Low-energy implants have been effective in this regard.