Noriaki Endo

Scanning moiré fringe imaging for quantitative strain mapping in semiconductor devices

The development of a method for the precise measurement of strain fields in semiconductor devices has become a critical requirement because the electrical performances of the devices are greatly influenced by the strain induced in their structures. We applied scanning moire fringe imaging to demonstrate the quantitative strain mapping of a Si/Si1−xGex interfacial layer. The strain field was measured at a nano-meter scale spatial resolution, with a detection precision of 0.1%. The maximum value of the strain was measured to be 1.1% ± 0.1%, which is consistent with the direct measurement by high-resolution scanning transmission electron microscopy image.

Ferric iron content in (Mg,Fe)SiO3 perovskite and post-perovskite at deep lower mantle conditions

Abstract We have determined the Fe3+/ΣFe ratio of Al-free (Mg,Fe)SiO3 perovskite, post-perovskite, and (Mg,Fe)O ferropericlase synthesized at 99 to 187 GPa and 1830 to 3500 K based on the electron energyloss near-edge structure (ELNES) spectroscopy. The results demonstrate that post-perovskite includes minor amounts of ferric iron with Fe3+/ΣFe ratios of 0.11 to 0.21. These values are substantially lower than those of Al-rich post-perovskite (Fe3+/ΣFe = 0.59 to 0.69) reported in a previous study, suggesting that the Fe3+-Al3+ coupled substitution is important in post-perovskite, as in the case of perovskite. The Al-bearing post-perovskite in a pyrolitic mantle composition likely contains a considerable amount of ferric iron, which affects various physical properties in the lowermost mantle.

Accuracy of Strain in Strain Maps Improved by Averaging Multiple Maps

Several methods have been proposed and tried for strain analysis of a crystal on nano-scale. Diffractometry such as nano-beam diffraction [1] and convergent-beam electron diffraction [2][3], and the phase imaging methods such as dark-filed electron holography [4] and high-resolution electron microscopy [5] have been applied to this issue. In scanning transmission electron microscopy (STEM), when the raster spacing is close to the one of a crystal lattice or the multiple of it and the probe size is smaller than the crystal lattice spacing, we sometimes observe the moiré fringes due to the under-sampling effect. We applied the fringe to detect strain in practical semiconductor devices and succeeded to map and quantify the strain [6]. However, the required accuracy of the strain measurement in the semiconductor industry is as high as or better than 0.2 %. In this paper, we report how we improve the accuracy of the strain measurement by using multiple strain maps.

Way to Reduce Electron Dose in Pseudo Atomic Column Elemental Maps by 2D STEM Moire Method

Atomic column elemental mapping, by EELS and/or EDS, is useful for material research, because we can clarify the elemental specie and atomic site simultaneously in these maps [1]. Though, the method requires high electron dose density on a sample, since the elemental signal is weaker than the image signal due to small ionization cross section for core electrons of atoms. Many efforts to acquire more elemental signal has been made for improvement of X-ray detection system in recent years, that is, the system with multiple silicon drift detectors (SDD) that have large sensor area, has been realized. And total solid angle of X-ray detection is > 1.75 sr in a detection system with two SDD detectors [2]. However, this improvement is insufficient for very electron-beam-sensitive materials. We need to find other way to have an atomic column elemental map with low dose density on samples. We recently succeeded in displaying the pseudo atomic column elemental map utilizing a 2D STEM moiré by EDS [3] and EELS [4]. This method allows the atomic column mapping with the dose density < 1 % of one for the conventional method, since a pixel interval to draw the pseudo lattice fringe is sparser than that for a real lattice. We already reported the results on typical ceramics (Si3N4) [5] and very fragile mineral (Aquamarine: Be3Al2Si6O18: Fe 2+ ) [6].

Concentration at Detection Limit of Dopant for Semiconductor Samples Using Dual SDD Analysis System

An energy dispersive spectrometer (EDS) analysis system using a Si drift detector (SDD) can enhance sensitivity of compositional analysis by using a large solid angle detector [1,2]. The use of multiple EDS detector system dramatically enlarge the effective area of the detector, resulting in the short measurement period. Up to now, transmission electron microscopy (TEM)-EDS has not been greatly used in the analysis of elements for semiconductor doping. This is because the detection limit by Si (Li) EDS is of the order of 1500-2000 ppm, which is insufficient for semiconductor industry. There has consequently been a strong desire to see “limit of detection (LOD)”, that is, how small an amount of elements we can be detected, using a multiple SDD analysis system. In this study, we explore the concentration at LOD for As dopant in Si device using FETEM with a dual SDD analysis system.

Near Shadowless EDS Tomography for Sliced Sample Realized by X-ray Collection with One Large Sized SDD Detector

Three-dimensional (3D) elemental mapping by energy dispersive X-ray spectroscopy (EDS) is getting popular for characterizations of samples having 3D structures to be solved such as semiconducting devices or blended polymers, since the method enables us to see the atomic species and 3D distribution of sample simultaneously. For the reconstruction of the 3D elemental maps, the EDS tomography is developed combining electron tomography and the two dimensional (2D) elemental mapping by EDS [1]. In the previous X-ray detection system composed of two EDS detectors, the detectors locates symmetrically with respect to the tilt axis of the sample holder. Therefore, some portion of the generated X-rays are blocked by the sample holder or supporting mesh in specific tilting angle range. The shadowing on the configuration may make artifacts in resulting 3D elemental maps. In order to avoid this problem, it is necessary to correct the measured intensity of EDS maps after the acquisition, according to the expected or pre-measured detection efficiency.

Accelerating Voltage and Probe Current Dependence of Electron Beam Drilling Rates for Silicon Crystal

Energy dispersive X-ray spectroscopy (EDS) and/or electron energy-loss spectroscopy (EELS) in scanning transmission electron microscopy (STEM) is very popular for analysis of semiconductor devices, since we require chemical information of them other than their shapes and dimensions. However, it requires more electron dose onto a sample than that of imaging due to small signal intensity. It is unavoidable that the sample lamella suffers the electron beam damage from electrons of an STEM probe in EDS and/or EELS analysis. The damage could be categorized into two types, one is structure deformation of sample and the other is beam drilling caused from etching and/or migration of sample atoms [1]. The latter is crucial for elemental analysis, since it significantly affect to results of compositional and/or quantitative chemical analyses. Several results on the beam drilling effect have been reported so far [2, 3]. In this paper, we report beam drilling of Si crystal which depends on accelerating voltage and electron probe current using EDS, because it could be a crucial problem and should be avoided for long time or high magnification analysis.

Effective Method for Decreasing Detection Limit of Dopant Concentration in Semiconductor Using Dual SDD Analysis System

An energy dispersive spectrometer (EDS) analysis system using a Si drift detector (SDD) can perform a highly sensitive elemental analysis due to its flexibilities of detector shape and size and high speed processing. In addition, the use of multiple SDD sensors can dramatically enlarge the effective area of the detector, which should shorten the measurement period while increasing detection sensitivity. Up to now, transmission electron microscopy (TEM)-EDS has not been greatly used in the analysis of elements in semiconductor doping. This is because the limit of detection (LOD) by Si(Li)-EDS is of the order of 1500-2000 ppm, which is a level of performance insufficient for dopant analysis in semiconductor industry. There has consequently been a strong desire to determine just how small an amount of elements can be detected (detection limit) using an analysis system with two SDDs. Theoretical estimation of LOD is determined with competition between the net counts for dopant element and the fluctuation of background tail of Si peak at the dopant peak energy [1]. When we measure the net counts of dopant element is three times larger than the fluctuation (sigma) of the background, it confirms the presence of the dopant with 99.7% certainty. Thus, we define a dopant concentration (Cdopant) at LOD as

Three-Dimensional Magnetic Domain Analysis by Lorentz Tomography

Recently, TEM tomography is used well in many fields (such as biology, polymer, and catalyst). In this method, a series of step by step tilted TEM images are acquired and a three-dimansional image is reconstructed by a Computerized Tomography (CT). The reconstructed image is analyzed in three-dimensions [1]. We have developed Lorentz tomography. In this new method, a three dimensional image is reconstructed from a series of acquired step by step tilted Lorentz microscope images by the CT method. Magnetic domain structures are analyzed in three-dimensions. By this new method, a structure of a magnetic domain wall and the relationship between a fine structure and a magnetic domain structure can be revealed three-dimensionally. Moreover, we expect that this new method is useful for the analysis of a magnetic characteristic and the development of a new magnetic material. We show an experimental method and an example of the Lorentz tomography. A series of step by step tilted Lorentz microscope images were acquired with a JEM-2100F. In the case of observing a magnetic domain, the structure was photographed without exciting an objective lens, because a magnetic field of an objective lens tends to destroy an original magnetic domain of a specimen. A specimen was prepared by Bulk Pick-up method with a JEM-9320FIB [2]. A conventional Lorentz microscope image of a SmCo permanent magnet is shown in Fig.1(a) and (b). These images were taken by a Fresnel mode. Bright and dark lines correspond to the position of magnetic domain walls. Three-dimensional reconstructed image of SmCo magnetic domain walls is shown in Fig. 2. Green lines are reconstructed from Fresnel mode images acquired in an under-focus, and White lines are reconstructed from Fresnel mode images acquired in an over-focus. As shown in Fig. 3, magnetic domain walls are observed in the shape of a plane. As a result, the three-dimensional knowledge of a magnetic domain structure can be displayed.