Flash-assist Rapid Thermal Processing (RTP) presents an opportunity to investigate annealing time and temperature regimes which were previously not accessible with conventional annealing techniques such as Rapid Thermal Annealing. This provides a unique opportunity to explore the early stages of the End of Range (EOR) damage evolution and also to examine how the damage evolves during the high temperature portion of the temperature profile. However, the nature of the Flash-assist RTP makes it is extremely difficult to reasonably compare it to alternative annealing techniques, largely because the annealing time at a given temperature is dictated by the FWHM of the radiation pulse. The FWHM for current flash tools vary between 0.85 and 1.38 milliseconds, which is three orders of magnitude smaller to that required for a RTA to achieve similar temperatures. Traditionally, the kinetics of the extended defects has been studied by time dependent studies utilizing isothermal anneals; in which specific defect structures could be isolated. The characteristics of Flash-assist RTP do not allow for such investigations in which the EOR defect evolution could be closely tracked with time. Since the annealing time at the target temperature for the Flash-assist RTP is essentially fixed to very small times on the order of milliseconds, isochronal anneals are a logical experimental approach to temperature dependent studies. This fact presents a challenge in the data analysis and comparison. Another feature of Flash-assist RTP which makes the analysis complex is the ramp time relative to the dwell time spent at the peak fRTP temperature. As the flash anneal temperature is increased the total ramp time can exceed the dwell time at the peak temperature, which may play a significantly larger role in dictating the final material properties. The inherent characteristics of Flash-assist RTP have consequently required the development of another approach to analyzing the attainable experimental data, such that a meaningful comparison could be made to past studies. The adopted analysis entails the selection of a reference anneal, from which the decay in the trapped interstitial density can be tracked with the flash anneal temperature, allowing for the kinetics of the interstitial decay to be extracted.

Junction formation in FinFET-based 3D-devices is a challenging problem as one targets a complete conformal doping of the source/drain regions in order to produce equal gate-profile overlaps (and thus transistor behavior) on all sides of the fins. Due to the lack of predictive modeling for several of the doping strategies explored (plasma immersion, cluster implants, vapor phase deposition, etc…) it becomes difficult to correctly predict the performance of the devices and hence, accurate 3D-doping profile determination is desired. Although several dopant/carrier profiling methods exist with excellent one- or two-dimensional resolution and properties, there is an urgent need to extend these towards a quantitative three-dimensional geometry. In this work, we use scanning spreading resistance microscopy (SSRM) with dedicated FinFET test structure to obtain three-dimensional information from successive two-dimensional scanning spreading resistance maps. We also assess the validity of our methodology by comparing various sections along the fins which represent the variability due to the processing and measurement procedure.

The bulk CMOS devices continue as the dominant player for at least another couple of technology nodes. This drives the increasingly contradicting requirements for the channel, source/drain extension, and heavily doped source/drain doping profiles. To analyze and optimize the transistors, it is becoming necessary to combine a number of effects that have been treated as decoupled so far. The temperature gradients, combined with stress engineering techniques such as embedded SiGe and Si:C source/drain and stress memorization technique, create non-uniform stress distributions determined by the layout patterns. The interaction of implant-induced damage with dopants, stress, and defect traps shapes up the dopant activation, retention of useful stress, and junction leakage. This work reviews recent trends in modeling these effects using continuum and kinetic Monte Carlo methods.

The activation behaviour of dopants in ultra-shallow junctions on strained silicon is investigated from a simulation vantage point. Process models available in commercial simulation tools are typically developed for junctions formed with high implantation energies (> 50 keV) and for long anneal times. Hence the question arises as to whether these models and parameter sets can accurately predict the active profile for highly doped, ultra-shallow junctions formed thin strained silicon layers using short rapid thermal anneals (RTA, <10 seconds) at temperatures below 800 °C. By incorporating the results from experimental data, we develop modified models allowing for improved predictions of antimony activation within both bulk and strained silicon.

Experimental and simulated P and As dopant diffusion profiles in Si:C epi films containing high C (>1 atomic %) are presented. A new set of physical effects were incorporated to accurately model P or As diffusion in the presence of high level of C. Evolution of substitutional C (Csub) profile in the Si:C epi film through dopant implant and activation anneal was characterized by high-resolution x-ray diffraction (HRXRD) technique. Three-layer analysis was utilized to obtain non-uniform Csub profile. Dependency of Csub retention on anneal thermal budget is studied. It is shown the initial Csub in the epi layer is lost during dopant implantation and conventional spike anneal sequence. Use of advanced millisecond (ms) laser anneal resulted in near 100% Csub retention in P-implanted Si:C epi film without compromising junction depth. Measured Csub (by HRXRD) and total C (by SIMS) profiles are compared with the ones predicted by the newly developed compact modeling in this study.

Boron enhanced H diffusion in amorphous Si (a-Si) layers formed by ion implantation is observed using secondary ion mass spectroscopy (SIMS). Constant concentrations of B were achieved using multiple energy B implantations into thick a-Si layers. The evolution of single H implanted profiles centered on the uniformly B-implanted regions was studied for partial anneals at temperatures in the range 380 – 640 °C. Boron enhanced diffusion is observed and the enhanced diffusion coefficient shows trends with temperature typically associated with a Fermi level shifting dependence. A modified form of the generalized Fermi level shifting model is considered in light of these results.

The role of the dopant activation on the segregation efficiency during the formation of platinum silicide (PtSi) is investigated in this paper. Using an implant before silicidation technique, we first demonstrate an important Schottky Barrier Height (SBH) modulation for As and B segregation. In the case of As, we highlight that an activation of the dopants before the silcidation does not impact the SBH modulation. On the contrary, an important impact of the dopant crystalline position is evidenced for Boron. Also, a comparison of conventional implant versus a PLAsma Doping (PLAD) highlights the suitability of the latter implantation tool for the SBH modulation. Those results are interpreted on the basis of SIMS depth profiling.

The fabrication of highly doped and ultra-shallow junctions in silicon is a very challenging problem for the materials scientist. The activation levels which are targeted are well beyond the solubility limit of current dopants in Si and, ideally, they should not diffuse during the activation annealing. In practice, the situation is even worse and when boron is implanted into silicon excess Si interstitial atoms are generated which enhance boron diffusion and favor the formation of Boron-Silicon Interstitials Clusters (BICs). An elegant approach to overcome these difficulties is to enrich the Si layers where boron will be implanted with vacancies before or during the activation annealing. Spectacular results have been recently brought to the community showing both a significant control over dopant diffusion and an increased activation of boron in such layers. In general, the enrichment of the Si layers with vacancies is obtained by Si+ implantation at high energy. We have recently developed an alternative approach in which the vacancies are injected from populations of empty voids undergoing Ostwald ripening during annealing. While different, the effects are also spectacular. The goal of this work is to establish a fair evaluation of these different approaches under technologically relevant conditions. The application domains of both techniques are discussed and future directions for their development/improvement are indicated.

Si:C layers are interesting candidates as stressor layers for NMOS transistors. Growth of such a Si:C layer has been realized by an expensive epitaxial growth process for devices to produce tensile strain in the channel, leading to enhanced mobility and device performance. Use of a monomer carbon ion implant in conjunction with a Ge pre-amorphizing implant (Ge-PAI) in Si is an alternative, lower cost approach to obtaining such SiC layers. This approach has not yielded desired device performance owing to low carbon substitutionality [C]sub, and also the presence of end-of-range (EOR) defects and large leakage currents due to the Ge-PAI implant. In this study we will show the formation of a Si:C layer using a ClusterCarbon approach that creates self-amorphization in Si thus avoiding an extra Ge-PAI implant step. We show that more than 2% substitutional carbon can be realized by using solid-phase-epitaxial-regrowth (SPER) and millisecond anneal. Si:C layers are characterized by using High Resolution X-ray Diffraction (HRXRD) and SIMS techniques.

This paper discusses the role of amorphisation and residual end-of-range defects in p-channel source/drain engineering. A comparison between preamorphisation and molecular implant approaches shows up some important common features of electrical activation, diffusion, and junction leakage, related to the formation and location of boron-interstitial and self-interstitial clusters. The success of these techniques depends on confining ‘end-of-range’ defects – whether TEM-visible defects or sub-microscopic clusters – within the narrow region between the boron implant peak and the source-drain/halo depletion region. This observation points to significant improvements that can still be made in implantation processing for ultrashallow junctions.

Extensive ab-initio calculations were performed to find formation energies of stable C complex configurations in silicon as function of stress. The results indicate that substitutional C is the lowest energy state, while the <100> split interstitial is the dominant mobile species. Investigation of small carbon/interstitial clustering suggests that these clusters are only significant under a substantial interstitial supersaturation. We studied the diffusion path for neutral C including the impact of stress. Through KLMC analysis of stress effect on diffusivity, we found that tensile biaxial strain enhances the effective C diffusivity, with a stronger stress dependence for C diffusivity in the out-of-plane direction.

The impact of low temperature plasma resist strip on doped silicon surface and microelectronic device performance is investigated using different chemical gas mixtures. In this investigation, different plasma treatments where applied on non-structured and structured silicon on insulator (SOI) wafers post ultra shallow surface implants .The main surface impacts dopant bleaching and oxide loss in conjunction with plasma enhanced re-oxidation are analyzed by time of flight secondary ion mass spectrometry (TOF-SIMS) and electrical measurements of microelectronic test devices. As the result, a long range plasma radiation induced dopant activation and deactivation is separated as the main effect from surface oxide loss and re-oxidation processes. This implies further optimization of plasma resist strip processes for device improvements.

Intrinsic point defects determine to a large extent the semiconductor crystal quality both mechanically and electrically not only during crystal growth or when tuning polished wafer properties by thermal treatments, but also and not the least during device processing. Point defects play e.g. a crucial role in dopant diffusion and activation, in gettering processes and in extended lattice defect formation.

Available experimental data and results of numerical calculation of the formation energy and diffusivity of the intrinsic point defects in Si and Ge are compared and discussed. Intrinsic point defect clustering is illustrated by defect formation during Czochralski crystal growth.

Maintaining or improving device performance while scaling semiconductor devices, necessitates the development of extremely shallow (< 20 nm) source/drain extensions with a very high dopant concentration and electrical activation level. Whereas solutions based on RTA with cocktail implants have been proposed in previous generations, sub-45 nm technologies will require even shallower junctions which motivates the research effort on milli-second anneal approaches as these hold the promise of minimal diffusion coupled with high activation levels [1]. Laser annealing is one of these concepts proposed to achieve the junction specifications and is typically described as a msec anneal process. Different from lamp based concepts which illuminate a full wafer simultaneously, the laser has an illuminated area which is much smaller than the wafer size thus necessitating a dedicated scanning pattern. In such a case one is potentially faced with areas subject to multiple overlaps and/or different temperatures and thus issues related to within wafer and within die uniformity need to be addressed.

In this work we use optimized metrology to probe such macro- and micro non-uniformity and determine the origin of the various components contributing to the observed non-uniformity patterns (laser stitching patterns, laser beam uniformity, optical path) and their impact on the local sheet resistance.

Plasma doping (PLAD) process utilizing PH3 plasma to fabricate n-type junction with supplied bias of −1 kV and doping time of 60 sec under the room temperature is presented. The RTA process is performed at 900 °C for 10 sec. A defect-free surface is corroborated by TEM and DXRD analyses, and examined SIMS profiles reveal that shallow n+ junctions are formed with surface doping concentration of 1021atoms/cm3. The junction depth increases in proportion to the O2 gas flow when the N2 flow is fixed during the RTA process, resulting in a decreased sheet resistance. Measured doping profiles and the sheet resistance confirm that the n+ junction depth less than 52 nm and minimum sheet resistance of 313 Ω/□ are feasible.

In this work the Si interstitial contribution of F+ implants in crystalline Si is quantified by the analysis of extended defects and B diffusion in samples implanted with 25 keV F+ and/or 40 keV Si+. We estimate that approximately 0.4-0.5 Si interstitials are generated per implanted F+ ion, which is in good agreement with the value resulting from the net separation of Frenkel pairs obtained from MARLOWE simulations. The damage created by F+ implants in crystalline Si may explain the presence of extended defects in F-enriched samples and the evolution of B profiles during annealing. For short anneals, B diffusion is reduced when F+ is co-implanted with Si+ compared to the sample only implanted with Si+, due to the formation of more stable defects that set a lower Si interstitial supersaturation. For longer anneals, when defects have dissolved and TED is complete, B diffusion is higher because the additional damage created by the F+ implant has contributed to enhance B diffusion.

We report continuous-wave (CW) infrared-semiconductor laser annealing of silicon implanted with boron atoms with assistance of diamond-like carbon (DLC) films as optical absorption layer in order to form shallow junctions. BF2 ions were implanted at 10 keV at doses of 1.5×1015 cm−2 thorough an 8-nm thick SiO2 layer. The effective boron dose implanted into silicon was 7.5c1015 cm−2 a half of the implantation dose. The initial boron distribution had a peak concentration of 6×1020 cm−3 at the silicon surface and a concentration of 1×1019 cm−3 at a depth of 27 nm. The samples were coated with 200-nm-thick DLC films. The samples were annealed by irradiation with a 940 nm continuous wave laser at 80 kW/cm2 with a beam diameter of 180 μm for 2.6 ms. Heat flow analysis estimated that the sample surface was heated to 1350°C for 1.5 ms by laser annealing. Laser annealing markedly reduced the sheet resistance to 531 Ω/sq. Boron atoms were almost completely activated with a carrier density close to the effective boron dose of 7.5×1014 cm−2. The in-depth profile of boron concentration hardly changed within 3 nm for laser annealing for 2.6 ms. The intermediate SiO2 layer effectively blocked carbon incorporation to a level below 1017 cm−3. These results show that the present laser annealing method is suitable to form shallow junction of a high dopant activation ratio. We will report low energy implantation of boron-cluster ions followed by the present infrared-semiconductor laser annealing in order to form a shallow junction with a depth less than 15 nm.

We have carried out an ab initio simulation study of boron in amorphous silicon. In order to understand the possible structural environments of B atoms, we have studied substitutional-like (replacing one Si atom in the amorphous cell by a B atom) and interstitial-like (adding a B atom into an interstitial space) initial configurations. We have evaluated the Fermi-level dependent formation energy of the neutral and charged (±1) configurations and the chemical potential for the neutral ones. For the interstitial-like boron atom, we have find an averaged formation energy of 1.5 eV. For the substitutional case, we have found a dependence of the chemical potential on the distance to Si neighbors, which does not appear for the interstitial ones. From MD simulations, we could observe a diffusion event for an interstitial-like boron atom with a migration barrier of 0.6 eV.

FinFET is one of the leading candidates to replace the classical planar MOSFET for future CMOS technologies due to the double-gate configuration of the device leading to an intrinsically superior short channel effect (SCE) control. A major challenge for FinFETs is the increase in parasitic source-drain resistance (Rsd) as the fin width is scaled. As fins must be narrow in order to control SCEs, Rsd reduction is critical. This work will deal with the challenges faced in the use of ion implantation for the low-ohmic source-drain contacts. Firstly a new technique to characterize fin sidewall doping concentration will be introduced. We will have a closer look at the Rsd dependency upon fin width for different fin implant conditions and investigate how the implant conditions affect FinFET device performance. It will be shown that the cause of the device degradation upon fin width scaling is related to the fundamental issues of silicon crystal integrity in thin-body Si after amorphizing implant and recrystallization during source-drain activation.

We present the results of extensive ab initio simulations for phosphorus clustering and diffusion in silicon and the application of these results in a state-of-the-art process simulator. The specific defects and the parameters that are investigated are selected according to the needs of diffusion and activation models, taking into account the availability of experimental data, the capabilities of current ab initio methods and the requirements for advanced technology development. The calculated formation energies, binding energies and migration barriers are used to determine a good starting point for the calibration of a new charged cluster model implemented in the process simulator. The defect species V, I, P, PV, PI, PI2, P2, P2V, P2I, P3, P3V, P3I and P4V are considered in all relevant charge states. The ab initio results are discussed as well as the transfer of this information into the process simulation model and the impact on model quality.