Search

In this work, we compare the results of different Cliff–Lorimer (Cliff & Lorimer 1975) based methods in the case of a quantitative energy dispersive spectrometry investigation of light elements in ternary C–O–Si thin films. To determine the Cliff–Lorimer (C–L) k-factors, we fabricated, by focused ion beam, a standard consisting of a wedge lamella with a truncated tip, composed of two parallel SiO2 and 4H-SiC stripes. In 4H-SiC, it was not possible to obtain reliable k-factors from standard extrapolation methods owing to the strong CK-photon absorption. To overcome this problem, an extrapolation method exploiting the shape of the truncated tip of the lamella is proposed herein. The k-factors thus determined, were then used in an application of the C–L quantification procedure to a defect found at the SiO2/4H-SiC interface in the channel region of a metal-oxide field-effect-transistor device. As in this procedure, the sample thickness is required, a method to determine this quantity from the averaged and normalized scanning transmission electron microscopy intensity is also detailed. Monte Carlo simulations were used to investigate the discrepancy between experimental and theoretical k-factors and to bridge the gap between the k-factor and the Watanabe and Williams ζ-factor methods (Watanabe & Williams, 2006).

In this paper the electrical and structural characteristics of n-MOSFETs fabricated on 4H SiC with a process based on nitrogen (N) implantation in the channel region before the growth of the gate oxide are reported for low (5x1018 cm-3) and high (6x1019 cm-3) N concentration at the SiO2/SiC interface. The electron mobility and the free carrier concentration in the MOSFET channel were evaluated by Hall effect measurement. The MOSFETs with the higher N concentration had the best electrical characteristics in terms of threshold voltage and field effect mobility, in spite of a lowering of the electron mobility in the channel. The latter is a negative drawback of the fabrication process that probably can be ascribed to an incomplete recovery of the implantation damage or to a high density of interstitial N atoms present in the channel region. In fact, the MOSFETs with the superior electrical performances were fabricated with the higher N+ dose and the shorter thermal oxidation time. However, no evidence of extended defects, clusters or nano-particles in SiC at the interface with the gate oxide was found in every SiC MOSFETs devices observed by electron transmission microscopy

A previously developed Monte Carlo code has been extended to the X-ray microanalysis in a (scanning) transmission electron microscope of plan sections, consisting of bilayers and triple layers. To test the validity of this method for quantification purposes, a commercially available NiOx (x ∼ 1) thin film, deposited on a carbon layer, has been chosen. The composition and thickness of the NiO film and the thickness of the C support layer are obtained by fitting to the three X-ray intensity ratios I(NiK)/I(OK), I(NiK)/I(CK), and I(OK)/I(CK). Moreover, it has been investigated to what extent the resulting film composition is affected by the presence of a contaminating carbon film at the sample surface. To this end, the sample has been analyzed both in the (recommended) “grid downward” geometry and in the upside/down (“grid upward”) situation. It is found that a carbon contaminating film of few tens of nanometers must be assumed in both cases, in addition to the C support film. Consequently, assuming the proper C/NiOx/C stack in the simulations, the Monte Carlo method yields the correct oxygen concentration and thickness of the NiOx film.

A technology computer aided design homemade tool, IMPACT, is described. The mechanical models and the numerical implementation are presented. Details are given about the methodology to calibrate and validate the implementation. This tool is used for the analysis of the processing of a micromachined cantilever. The originality of the study lies in (i) the capability to simulate almost all the sources of stress taking into account the complex rheological behaviours of the materials, (ii) the experimental determination of the mechanical properties of various thin film materials, and (iii) the validation of the calculations by direct comparisons with the measured deformations of the cantilever.

Test structures consisting of shallow trench isolation (STI) structures are fabricated using advanced silicon (Si) technology. Different process parameters and geometrical features are implemented to investigate the residual mechanical stress in the structures. A technology computer aided design homemade tool, IMPACT, is upgraded and optimized to yield strain fields in deep submicron complementary metal–oxide–semiconductor devices. Residual strain in the silicon substrate is measured with micro-Raman spectroscopy (m-RS) and/or convergent beam electron diffraction (CBED) for large (25 micron) and medium size (2 micron), while only CBED is used for deep submicron STI (0.22 micron). We propose a methodology combining CBED and technology computer aided design (TCAD) with m-RS to assess the accuracy of the CBED measurements and TCAD calculations on the widest structures. The method is extended to measure (by CBED) and calculate (by TCAD) the strain tensor in the smallest structures, out of the reach of the m-RS technique. The capability of determining, by both measurement and calculation, the strain field distribution in the active regions of deep submicron devices is demonstrated. In particular, it is found that for these structures an elastoplastic model for Si relaxation must be assumed.

The use of TiSi 2 layers as ohmic contacts in complementary MOS (CMOS) devices is expected to introduce large distortions in the underlying silicon which may degrade the device performances. In this work, the TEM technique known as convergent beam electron diffraction (CBED) has been employed to map the strain distributions across 〈1 1 0〉 cross-sectional specimens of sub-quarter-micron CMOS structures. Different experimental set-ups in terms of energy filtering, liquid nitrogen cooling and zone axis have been explored and compared to each other by using three differently equipped microscopes. The strain induced on silicon by TiSi 2 layers has then been investigated, quantified and compared with that due to the presence of an oxide isolation trench. Also, the effects of depositing a TiN capping layer on TiSi 2 have been analysed. Our results show that its presence results in a more tensile strain without altering the strain distribution trend.

In this work we present some recent results concerning the α-particles irradiation of Si1−yCy alloy epitaxially grown on silicon. The study of the damage process is interesting because of the extensive use of backscattering technique as a tool of characterisation of this kind of materials and because of the possibility of adding information about the transformations that this metastable material undergoes. We point out that the irradiation damage process causes a change in the material structure different from that due to the thermal treatments. The irradiation damage occurs at a rate much higher than in Si, however it involves only a silicon atom fraction that appears to be proportional to the substitutional carbon content.

The Large Angle Convergent Beam Electron Diffraction (LACBED) technique has been applied to determination of the tetragonal mismatch in coherent Si/Si1−xGex/Si heterostructures. Two-dimensional (2D) dynamical simulation of the LACBED patterns has been performed and compared with the corresponding experimental ones. A good agreement is found in the whole simulated area, particularly as regards the splitting of the Bragg contours, due to the strain field present in the TEM cross-sections.

A method for the fabrication of a wedge-shaped thin NiO lamella by focused ion beam is reported. The starting sample is an oxidized bulk single crystalline, ⟨100⟩ oriented, Ni commercial standard. The lamella is employed for the determination, by analytical electron microscopy at 200 kV of the experimental k(O-Ni) Cliff-Lorimer (G. Cliff & G.W. Lorimer, J Microsc103, 203–207, 1975) coefficient, according to the extrapolation method by Van Cappellen (E. Van Cappellen, Microsc Microstruct Microanal1, 1–22, 1990). The result thus obtained is compared to the theoretical k(O-Ni) values either implemented into the commercial software for X-ray microanalysis quantification of the scanning transmission electron microscopy/energy dispersive spectrometry equipment or calculated by the Monte Carlo method. Significant differences among the three values are found. This confirms that for a reliable quantification of binary alloys containing light elements, the choice of the Cliff-Lorimer coefficients is crucial and experimental values are recommended.

The convergent beam electron diffraction (CBED) technique has been applied to determine the lattice strain in Si1−xGex/Si heterostructures and in local isolation structures. Both plan and cross-sections have been investigated by transmission electron microscopy. In the heterostructures, the strain value obtained by CBED along the growth direction ɛCRED is affected by a relaxation induced by the thinning process in a direction normal to the cross-section plane, being generally smaller than the bulk tetragonal value ɛT This effect can be overcome using the large angle CBED technique on plan sections. In addition, in the heterostructures the Ge concentration has been determined by energy dispersive X-ray spectrometry, allowing the pseudomorphicity of the samples to be evaluated. All the values of strain and Ge concentration thus obtained are in good agreement with those deduced from Rutherford backscattering spectrometry. In isolation structures, the CBED technique has been applied to determine the d...

Backscattering and ion channeling measurements have been carried out on 〈100〉 silicon wafers implanted with a dose of 3×1016 As+/cm2. These samples have been submitted to laser annealing and then to various thermal annealings in the 350–1000 °C temperature range. Three different crystallographic directions, 〈100〉, 〈110〉, and 〈111〉, have been explored. The deactivated As atom displacement with respect to the lattice sites has occurred already by the first stage of electrical deactivation at 350 °C. The average displacement is constant in the considered temperature range and has been found equal to 0.23±0.06 A making an angle of 23±8° with the [100] direction. By comparing this finding with the local atomic As arrangement, deduced from previously reported extended x‐ray absorption fine structure measurements performed on the same samples, new cluster models are presented. These models consist of As2V and As4V clusters during the very early stage of deactivation, and of several As4V clusters grouped together...

The thermal evolution of monoclinic SiAs precipitates at 1050° C in silicon samples implanted with 1 and 1.5×1017 As/cm2 was followed by transmission electron microscopy (TEM) and secondary neutral mass spectrometry (SNMS). These experiments show, for the first time, the coexistence of two different states of As in silicon, i.e., the electrically active and the inactive mobile dopant, in equilibrium with monoclinic SiAs precipitates. Moreover, they provide, for the saturation concentration of As in silicon, which includes both these states, a value of 3×1021 cm−3 at 1050° C.

The tetrahedral heterobimetallic cluster (C 5 H 5 )NiRu 3 H 3 (CO) 9 , 1 , when decomposed on KOH-modified γ-alumina, produces an effective heterogeneous catalyst for ammonia synthesis. Bimetallic particles have been evidenced on the support surface by analytical electron microscopy; their main size is 4.0 nm and their Ni/Ru atomic ratio is practically the same as that of the starting cluster. The treatment of unsupported 1 with alcoholic KOH gives a black, conductive, amorphous alloy powder. In this case, the ultrafine and bimetallic nature of the individual particles was also revealed by analytical TEM investigations.

Silicon wafers have been implanted with As+ ions at an energy of 100 keV and a dose of 1 × 1017 cm−2 and subsequently annealed at 1050°for 15 min. This results in a peak As concentration of 7 × 1021 cm−3, which is far beyond the solid solubility value of arsenic in silicon at this annealing temperature. Rod-like precipitates, dislocations, and small precipitate-like defects have been observed by transmission electron microscopy. From the analysis of several diffraction patterns taken on a number of rod-like particles at different tilt angles, it has been unambiguously found that they have the structure of the monoclinic SiAs compound previously reported in literature. The stoichiometry of the precipitates has been confirmed by x-ray microanalysis. To our knowledge, this is the first time that this SiAs phase is detected in As+-implanted silicon.

Molybdenum silicide is investigated because of its possible application in a self‐aligned process for ultra large scale integration technology. p+/n shallow junctions are fabricated by boron implantation through Mo films or MoSi2 layers and rapid thermal annealing. Both procedures enable us to obtain junctions as shallow as 130 nm with very good electrical characteristics (reverse current density 1 nA/cm2 at −1 V, ideality factor 1.05 and contact resistivity 1×10−6 Ω cm2). The influence of a surface preamorphization with different thicknesses on junction depth and electrical characteristics is reported.

A method to perform nanobeam diffraction (NBD) in a transmission electron microscope with high spatial resolution and low convergence angle is proposed. It is based on the use of a properly fabricated condenser aperture of 1 mu m in diameter, which allows an electron beam about 10 nm in size to be focused on the sample, with a convergence angle in the 0.1 mrad range. Examples of NBD patterns taken in an untilted < 110 > cross section of a silicon device are shown. Their quality is adequate for spot position determination and hence to obtain, in principle, quantitative strain information. (C) 2008 American Institute of Physics. [DOI: 10.1063/1.3003581]

Summary A method to extract accurate information on the displacement field distribution from split high-order Laue zones lines in a convergent-beam electron diffraction pattern of nanostructures has been developed. Starting from two-dimensional many beam dynamical simulation of HOLZ patterns, we assembled a recursive procedure to reconstruct the displacement field in the investigated regions of the sample, based on the best fit of a parametrized model. This recursive procedure minimizes the differences between simulated and experimental patterns, taken in strained regions, by comparing the corresponding rocking curves of a number of high-order Laue zone reflections. Due to its sensitivity to small displacement variations along the electron beam direction, this method is able to discriminate between different models, and can be also used to map a strain field component in the specimen. We tested this method in a series of experimental convergent-beam electron diffraction patterns, taken in a shallow trench isolation structure. The method presented here is of general validity and, in principle, it can be applied to any sample where not negligible strain gradients along the beam direction are present.

A few recent applications of scanning transmission electron microscopy (STEM) methods to problems of interest for nanoelectronics are reported. They include nanometer-scaled dopant profiles by Z-contrast and strain mapping by convergent beam diffraction.

Summary We applied the parametric bootstrap to the X-ray microanalysis of Si-Ge binary alloys, in order to assess the dependence of the Ge concentrations and the local film thickness, obtained by using previously described Monte Carlo methods, on the precision of the measured intensities. We show how it is possible by this method to determine the statistical errors associated with the quantitative analysis performed in sample regions of different composition and thickness, but by conducting only one measurement. We recommend the use of the bootstrap for a broad range of applications for quantitative microanalysis to estimate the precision of the final results and to compare the performances of different methods to each other. Finally, we exploited a test based on bootstrap confidence intervals to ascertain if, for given X-ray intensities, different values of the estimated composition in two points of the sample are indicative of an actual lack of homogeneity.

The effects of 2 MeV Si+ implantation on silicon-on-insulator layers uniformly doped with B at concentrations 1.0 and 1.8 x 10(20) cm(-3), and the kinetics of damage recovery were investigated by carrier density, mobility measurements, and transmission electron microscopy (TEM) observations. High energy implantation reduces the hole density by about 98%; the mobility is also reduced at an extent which increases with B concentration. Isochronal and isothermal annealings show that recovery of the hole density takes place in three stages: the first stage (alpha) is accompanied by a mobility decrease and is followed by the second stage (beta) where mobility increases attaining values close to the ones of the reference undamaged samples. Mobility keeps nearly constant in the third recovery stage (gamma), which takes place above 800 degrees C. As a characterizing feature the mobility values for each B concentration only depend on the hole density, irrespective of the thermal history of the samples. Experiments and TEM observations allowed us to distinguish defect recovery from SiB3 precipitation, which can take place at temperatures higher than 700 degrees C. Recovery stages are discussed, and it is concluded that dissolution of B rich clusters in stage (alpha) modifies the concentration, or the charge state, of the defects responsible of the second (beta) stage. These defects are identified as boron interstitial clusters in consideration of their mobility behavior and of the activation energy E-beta for their recovery process, which results to be 3 +/- 0.2 eV.

Using a combination of the CBED and the LACBED techniques in the transmission electron microscopy (TEM), we have investigated the strain field in the silicon active region of a shallow trench isolation structure, underlying a TiSi 2 layer. Starting from the analysis of the deformation in a sample, thinned for TEM analysis, we have reconstructed the displacement field, simulating the split HOLZ lines visible in the experimental CBED patterns. From the comparison between the experimental LACBED patterns, taken in a suitable sample orientation to evidence the stressors distribution in the polycrystalline silicide layer, and the corresponding dynamically simulated ones, we have reproduced the strain field in the unthinned, bulk sample.

Silicon-germanium thin films have been analyzed by EDS microanalysis in a field emission gun scanning transmission electron microscope (FEG-STEM) equipped with a high angular dark-field detector (STEM/HAADF). Several spectra have been acquired in the same homogeneous area of the cross-sectioned sample by drift-corrected linescan acquisitions. The Ge concentrations and the local film thickness have been obtained by using a previously described Monte Carlo based “two tilt angles” method. Although the concentrations are in excellent agreement with the known values, the resulting confidence intervals are not as good as expected from the precision in beam positioning and tilt angle position and readout offered by our state-of-the-art microscope. The Gaussian shape of the SiKα and GeKα X-ray intensities allows one to use the parametric bootstrap method of statistics, whereby it becomes possible to perform the same quantitative analysis in sample regions of different compositions and thicknesses, but by doing only one measurement at the two angles.

High order Laue zone lines present in the central disk of convergent beam electron diffraction patterns have been used to determine the bulk mismatch in thinned, cross‐sectioned heterostructures, where a relaxation occurs along the thinning direction. The position of these lines is sensitive to lattice parameters along different crystallographic directions, so that information on the residual strain along the growth and the thinning directions can be extracted from a single diffraction pattern. This information has been properly combined using the isotropic elasticity theory to give the bulk mismatch. The results are in good agreement with independently obtained bulk measurements.

In this paper we discuss the physical mechanism and the critical factors for the dislocation generation in device processing. The mechanisms of the stress development are identified. Elastic stress is estimated from the convergent beam electron diffraction (CBED) measurements using transmission electron microscopy (TEM), and the defect generation is monitored by electrical measurements of a specific monitor structure. The implantation conditions and the implantation damage recovery is shown to be another key factor in the defect generation. Thermal oxidations of the STI structure are shown to be a major responsible of the stress increase. In addition to the process the pattern geometry strongly affects the mechanical stress. Minimum geometry structures with a complex layout are prone to defect generation.

Silicon on insulator (SOI) substrates, uniformly n doped with As or P at different concentrations in the range between 1.9 and 8 × 10 20 cm −3 , have been irradiated with increasing doses of high energy (2 MeV) Si + ions. The effects of irradiation and the subsequent recovery have been followed by electrical conductivity and carrier mobility measurements and transmission electron microscopy observations. Isothermal annealings have been performed at temperatures in the range 600–800 °C. It was found that recovery takes place in two distinct stages which have been evidenced by using rapid thermal annealing and furnace heating. Our results show a very similar behavior for the two donors and indicate that electron trapping is the responsible of the reduction of the carrier density upon irradiation both in As and P doped samples.

Convergent beam electron diffraction is a technique available in any modern transmission electron microscope. It allows the determination of the strain tensor in crystals at the nanometer scale. The basic principles of the method and examples of application to microelectronic devices are given.

A method of obtaining quantitative two-dimensional (2D) maps of strain by the convergent beam electron diffraction technique in a transmission electron microscope is described. It is based on the automatic acquisition of a series of diffraction patterns generated from digital rastering the electron spot in a matrix of points within a selected area of the sample. These patterns are stored in a database and the corresponding strain tensor at each point is calculated, thus yielding a 2D strain map. An example of application of this method to cross-sectioned cells fabricated for the 0.15 μm technology of flash memories is reported.

In this paper, three techniques are discussed that provide information on process-induced local mechanical stress in silicon: the convergent beam electron diffraction technique of transmission electron microscopy, X-ray micro-diffraction and micro-Raman spectroscopy. We discuss the principles of these techniques, their spatial resolution, the ease-of-use, the information that can be obtained, the required sample preparation, the measurement time, and the complementarities of these techniques. We demonstrate this for stress induced by shallow trench isolation and correlate the results to finite element analysis results.

The convergent beam electron diffraction (CBED) technique of transmission electron microscopy (TEM) has been applied to 〈1 1 0〉 cross-sections of shallow trench isolation structures in silicon, in order to evaluate the strain field distribution in 0.22 μm wide active areas. Different spot sizes (1 and 10 nm) and sample temperatures (room temperature with energy filtering, liquid nitrogen cooling without filtering) have been employed. It has been found that the regions of the active area closer than about 100 nm to the padoxide/substrate interface can be analysed only by using a 1 nm spot size. Moreover, the use of an energy filter to reduce the inelastic scattering improves the contrast of the diffraction lines in the CBED pattern, thus allowing the analysis to be performed at room temperature.

The convergent beam electron diffraction technique (CBED) of the transmission electron microscopy (TEM) has been employed to determine the strain distribution along a cutline parallel to the padoxide/Si interface in a 0.80 micron wide recessed-LOCOS structure. The values of the components of the strain tensor so obtained have been compared with those computed by two simulator codes. It has been found that both the LOCOS morphology and the strain distribution deduced from TEM images and TEM/CBED patterns, respectively, were in agreement with the simulation results, if some oxidation-related parameters were modified.

A novel approach to the quantitative analysis of thinned samples, which exploits the finite and variable width of the incident beam of an analytical electron microscope (AEM), is reported. For a binary alloy AB, the method requires two measurements of the I(AKα)/I(BKα) X-ray intensity ratios, obtained with two different beam diameters. The digital image of the beam is also recorded by a slow-scan CCD camera; its pixel intensities are converted into probability densities by our Monte Carlo code, which has been modified to simulate the electron trajectories crossing the vertical boundaries of the sample. The result of the simulation consists of two thickness t vs concentration C matrices for the two different spot sizes; the unique t-C combination, corresponding to the analyzed region, is obtained through the convergence routine described in our previous papers.

The effect on spatial resolution of the spherical aberration of the objective prefield of a Philips CM30 (S)TEM has been investigated. For this purpose, the lattice mismatch of a 20 nm thick Si-Ge film in a cross-sectioned Si/Si0.9Ge0.1/Si heterostructure has been determined by convergent beam electron diffraction (CBED), performed at 100 kV with a spot size of 20 nm. It has been found that, if the disk of minimum confusion is focused onto the specimen plane, the measurement is not affected by a change in the size of the condenser-2 aperture. Therefore, the mismatch information is contained within the Gaussian part of the electron probe. In strain profile determinations by CBED, this allows one to use higher beam intensities (larger C2-apertures) with improved signal/noise ratios and reduced sample drift due to shorter acquisition times.

The Monte Carlo simulation has been applied to the determination of composition and thickness of thin Si1-x Ge x alloys in cross-sectioned, pseudomorphic Si1-x Ge x /Si heterostructures. The experimental intensity of the SiKα, GeLα and GeKα X-rays have been measured in an analytical electron microscope and ratioed. Two methods have been applied: the ‘2-tilt angle method’, which is based on the convergence of the I(SiKα)/I(GeLα) ratios at two different tilt angles, and the ‘2-ratio method’, which uses the I(SiKα)/I(GeKα) and the I(SiKα)/I(GeLα) ratios at a fixed angle. It is found that both methods give solutions, but not in the whole Ge concentration range investigated (from 5 up to 30 at.%). The best way is to apply the ‘2-tilt angle method’, but using I(SiKα)/I(GeKα) ratios.

The LACBED technique has been applied to the determination of the tetragonal distortion in Si 1-x Ge x /Si heterostructures, which are of great interest in the device technology. The strain determination has been performed on plan sections in an analytical electron microscope. The agreement between this strain value and the tetragonal distortion is influenced mainly by the local sample flatness and the acceleration voltage.

An overview is given of transmission electron microscopy techniques to address strain with nm scale spatial resolution. In particular the possibilities and limitations of (large angle) convergent beam electron diffraction ((LA)CBED) and electron diffraction contrast imaging (EDCI) techniques are discussed in detail. It will be shown by a few case studies that unique and quantitative information on local strain distributions can be obtained by the combined use of both (LA)CBED and EDCI in correlation with three dimensional finite element simulations of the strain distributions in the thinned specimen.

Using isotropic elasticity theory it is possible to determine the bulk mismatch in thinned, cross-sectioned heterostructures, where a relaxation occurs along the thinning direction. This is accomplished by measuring, in the central disk of a single Convergent Beam Electron Diffraction (CBED) pattern, the position of the High Order Laue Zone lines, which are sensitive to lattice parameters along different crystallographic directions. The results obtained in both uniform and graded Sil-xGex/Si heterostructures are in good agreement with bulk values deduced from independent techniques.

It is now widely accepted that one of the major issues for the future deep sub-micron integrated circuit (IC) technologies regarding yield, device performance and stability and device and product reliability, is the mechanical stress built up in the layers and substrate. It is therefore important to give a quantitative account of these stresses and this can only be achieved by disposing of reliable, performant (as to spatial resolution) and quantitative techniques for the local stress determination in the substrate, of adequate and dedicated process simulation tools and of sensitive methods to analyse the stress effects on device performance.The only experimental technique presently available which allows the strain field distribution in sub-micron CMOS devices is the convergent beam electron diffraction technique of the transmission electron microscopy. Its sensitivity is of the order of 10-4 and its spatial resolution is of 1 nm if a TEM/FEG is employed.