Removing constraints of 4D-STEM with a framework for event-driven acquisition and processing.

This paper focuses on the problem of how data representation influences the generalization error of kernel based learning machines like Support Vector Machines (SVM) for classification. Frame theory provides a well founded mathematical framework for representing data in many different ways. We analyze the effects of sparse and dense data representations on the generalization error of such learning machines measured by using leave-one-out error given a finite number of training data. We show that, in the case of sparse data representation, the generalization capacity of an SVM trained by using polynomial or Gaussian kernel functions is equal to the one of a linear SVM. This is equivalent to saying that the capacity of separating points of functions belonging to hypothesis spaces induced by polynomial or Gaussian kernel functions reduces to the capacity of a separating hyperplane in the input space. We show that sparse data representations reduce the generalization error as long as the representation is not too sparse, as in the case of very large dictionaries. Dense data representations, on the contrary, reduce the generalization error also in the case of very large dictionaries. We use two different schemes for representing data in overcomplete Haar and Gabor dictionaries, and measure SVM generalization error on bench mark data set. Moreover we study sparse and dense data representations with frame of data and we show how this leads to Principal Component Analysis.

Strain visualization using large-angle convergent-beam electron diffraction

Data representations and generalization error in kernel based learning machines

This paper focuses on the problem of how data representation influences the generalization error of kernel-based learning machines like support vector machines (SVMs). We analyse the effects of sparse and dense data representations on the generalization error of SVM. We show that using sparse representations the performances of classifiers belonging to hypothesis spaces induced by polynomial or Gaussian kernel functions reduce to the performances of linear classifiers. Sparse representations reduce the generalization error as long as the representation is not too sparse as with very large dictionaries. Dense data representations reduce the generalization error also using very large dictionaries. We use two schemes for representing data in data-independent overcomplete Haar and Gabor dictionaries, and measure the generalization error of SVMs on benchmark datasets. We study sparse and dense representations in the case of data-dependent overcomplete dictionaries and we show how this leads to principal component analysis.

The properties of nanoscaled materials can be changed by applying strain and as such there is an interest in the accurate measurement of deformation with nm‐scale resolution. This was until recently considered as a difficult problem. However, the last ten years has seen a great deal of development in techniques that can be used to measure deformation with the required resolution [1,2]. Today there are many different approaches which can be used to recover valuable information about the deformation. Each of these techniques has strengths and weaknesses and requires different set ups in the electron microscope [2]. In this presentation we will present dark field electron holography, nanobeam electron diffraction (NBED), precession diffraction (NPED) and the geometrical phase analysis (GPA) of TEM and high angle annular dark field (HAADF) scanning transmission electron microscopy (STEM) images. We will then discuss which technique is most suitable for different types of materials problems and benchmark their performance with respect to the accuracy of the measurements, precision and spatial resolution. Figure 1(a) shows a HAADF STEM image of a 100‐nm‐thick Si test specimen containing 10‐nm‐thick SiGe layers with different Ge concentrations. As the specimen has been grown epitaxially we expect no deformation to be measured in the ex direction. However, due to the expanded lattice parameter of the SiGe layers relative to the Si reference, tensile deformation is expected in the ez direction. Figure 1 shows deformation maps that have been acquired by (b) GPA of HAADF images (c) dark holography and (d) precession diffraction. These are compared to finite element simulations that are shown in Figure 1(e). These results reveal that all of the different techniques provide accurate measurements of the deformation. Figure 2(a) shows a HAADF STEM image of a SiGe test device structure with a gate length of 35 nm. Finite element simulations showing the expected deformation in the struture is shown in Figures 2(b) and (c). Deformation maps are shown in Figure 2(d) and (e) for precession diffraction, (f) and (g) for dark holography and (h) and (i) for GPA of HAADF STEM images [4]. Again, accurate measurements of the deformation are made, but the precision and spatial resolution depends on the experimental technique that has been used. As well as presenting state of the art results from a range of strained materials we will highlight improvements that are required for all of the different techniques in order to optimise their performance and provide the best possible measurements of deformation. Acknowledgements : These experiments have been performed on the platform nanocharacterisation (PFNC) at Minatec. The work has been funded by the ERC Starting Grant 306365 « Holoview ».

Strain and the related stress in nanoelectronic devices play a critical role in semiconductor technologies. In general, strain arises from the lattice mismatch between dissimilar materials and thus strain is critical for the design and growth of epitaxial heterostructures in micro- and nano-electronics. In transistors, strained silicon channels provide a major boost to the performance of metal-oxide-semiconductor field-effect transistors (MOSFETs), e.g., the use of epitaxial strain of Si and Ge alloys in PMOS [1]. In the near term, the transistor technology is pushing towards the sub 10 nm technology node. At this length scale, resolving strain in the devices requires a resolution of sub nanometer with not only high resolution, also high accuracy and reliability. The use of novel architectures and alternative materials in the transistor technology also demands metrology capable of handling these structural and material changes. High resolution strain measurement can be performed using high energy electrons either in reciprocal space using convergent beam electron diffraction (CBED), scanning electron nanodiffraction (SEND), nanobeam electron diffraction (NBED) [2], or scanning precession electron diffraction (SPED) or in real space through direct imaging using high resolution electron microscopy (HREM), scanning transmission electron microscopy (STEM) [3] or dark-field electron holography (DFEH)[4]. The resolution and strain measurement sensitivities range from Å to several nm and 10-3 to 10-4respectively. Recent developments in aberration corrected STEM especially have improved the spatial resolution to sub- Å and the precision of measuring atomic column positions to picometers [3]. This talk will review high resolution and high sensitivity electron beam techniques for strain characterization in nanodevices and device materials. Examples include planar and a finfet transistors and III-V heterostructures. For finfet transistors, we demonstrate that NBED provides an effective approach for strain mapping (See figure below) and some of challenges will also be highlighted. In III-V heterostructures, we will describe atomic resolution Z-contrast imaging based methods for high resolution interfacial strain analysis and its applications for the investigation of interfacial intermixing and atomic layer-by-layer strain profiles in InAs/GaSb and InAs/InAsSb superlattices, targeted for middle- and long-wavelength IR detection.

It is very important to know the morphology and chemical properties of subscales of oxide layers on surface for controlling and understanding high temperature oxidation in electrical steel. In present work, the oxide layers were investigated by various methods of transmission electron microscopy (TEM) such as scanning transmission electron microscopy (STEM), nanobeam electron diffraction (NBD), energy dispersive X-ray spectrometry (EDS), and electron energy loss spectrometry (EELS). The high-angle annular dark field (HAADF) of STEM could be a useful analysis technique to study the morphology of the oxide layers. The main oxides formed in the subscales during the decarburization annealing were fayalite, iron oxides, and silica, which were identified by EDS, NBD and EELS. The crystalline fayalites were found both in the surface region within several tens nanometers and in the region within a micrometer surrounding silica, and the atomic configuration in the unit cell of the fayalite was presented. Amorphous silica was formed both in the upper region of the subscales with a spherical shape and in the interface between the spherical silica and the iron matrix with a lamellar shape. TEM could be useful technique to characterize morphologies, microstructures and elemental compositions of oxides, and to understand the oxidation mechanism for the manufacture of the high quality electrical steel.

Solid acid fuel cells (SAFC) working at intermediate temperature (250°C) have many advantages compared to the lower temperature proton exchange membrane fuel cell (PEMFC), such as increased catalyst activity and are more resistant to CO poisoning.1, 2 The state-of-the-art cathode is comprised of porous protonic conductor coated by Pt.3, 4 There are two major roles of Pt in the cathode: (1) the oxygen reduction reaction (ORR) catalyst and (2) the electron conductor. To maintain a highly electron conductive network, a high fraction of Pt has to be loaded in ‘conventional’ SAFC electrodes. However, to generate wider application, the cost of the cathode must be reduced significantly. The function of Pt as electron conductor can be replaced by the catalyst support to achieve this reduction in loading. The catalyst support must have high electron conductivity, corrosion-resistance and high surface area. The carbon nanostructures are the common catalyst supports used in the fuel cells. In our previous studies, we demonstrated that multi-wall carbon nanotubes are much more stable during fuel cell testing compared to the single wall carbon structures.5 Indeed, pure single layer or few-layer carbon supports are well known to be particularly vulnerable to corrosion in the present of water. However, previous research showed that substitutional boron can lower the electron density of the reactive carbon, leading to a reduction in the rate of O2 chemisorption.6 Meanwhile, very small amounts of boron (~1%) have been demonstrated to enhance the conductivity of the single carbon structure.7In our recent studies, we synthesize the pure single wall carbon, carbon structures in hydrogen gas and with boron loading. Surprisingly, the boron loading induces the growth of few-layer graphenes (FLGs), which exhibit much better corrosion resistance than the single wall carbon and carbon synthesized in hydrogen gas which consists of multi-wall and single-wall carbon structure. In this research, we focus on the structural investigation of those FLGs. Characterization of the FLGs are performed by high resolution transmission electron microscopy (HRTEM) imaging, monochromated electron energy-loss spectroscopy (EELS), nano-beam electron diffraction (NBED), aberration corrected scanning transmission electron microscopy (STEM) imaging, X-ray photoelectron spectroscopy (XPS), nuclear magnetic resonance (NMR), Raman spectroscopy and X-ray diffraction (XRD). Representative STEM images in a low and high magnification are shown in Figure 1 (a)-(b). A significantly structure deviation compared to pure graphene/graphite is evident due to 1-3% boron loading. Those results indicate the few-layer graphene does not stack in a normal graphitic way in the c direction. The EELS, XPS and NMR show complicated electronic states of the boron in the carbon, implying boron does not have to be in the substitutional site. For example, Figure 1(c) show a representative EELS spectrum taken from a FLG. The featureless boron K-edge looks like to be taken from amorphous boron instead of substitutional boron. This result is also supported by the NBED and XRD, showing a non-flat features of the FLGs. Finally, we used the density functional theory (DFT) to explain the excellent oxygen resistance of FLG with this unique structure. Acknowledgments This work is supported by ARPA-E via cooperative agreement DE-AR0000499. The synthesis science is supported by BES-MSED. We thank JIAM and CNMS for microscopy use. Reference 1. K. Ishiyama, F. Kosaka, I. Shimada, Y. Oshima and J. Otomo, Journal of Power Sources, 2013, 225, 141-149. 2. A. B. Papandrew, R. W. Atkinson Iii, R. R. Unocic and T. A. Zawodzinski, Journal of Materials Chemistry A, 2015, 3, 3984-3987. 3. A. B. Papandrew, C. R. I. Chisholm, R. A. Elgammal, M. M. Özer and S. K. Zecevic, Chemistry of Materials, 2011, 23, 1659-1667. 4. A. B. Papandrew, C. R. I. Chisholm, S. K. Zecevic, G. M. Veith and T. A. Zawodzinski, Journal of The Electrochemical Society, 2013, 160, F175-F182. 5. A. B. Papandrew, R. A. Elgammal, M. Tian, W. D. Tennyson, C. M. Rouleau, A. A. Puretzky, G. M. Veith, D. B. Geohegan and T. A. Zawodzinski, Journal of Power Sources, 2017, 337, 145-151. 6. B. Yuan, W. Xing, Y. Hu, X. Mu, J. Wang, Q. Tai, G. Li, L. Liu, K. M. Liew and Y. Hu, Carbon, 2016, 101, 152-158. 7. M. Harada, T. Inagaki, S. Bandow and S. Iijima, Carbon, 2008, 46, 766-772. Figure 1

Fast pixelated detectors incorporating direct electron detection (DED) technology are increasingly being regarded as universal detectors for scanning transmission electron microscopy (STEM), capable of imaging under multiple modes of operation. However, several issues remain around the post-acquisition processing and visualization of the often very large multidimensional STEM datasets produced by them. We discuss these issues and present open source software libraries to enable efficient processing and visualization of such datasets. Throughout, we provide examples of the analysis methodologies presented, utilizing data from a 256 × 256 pixel Medipix3 hybrid DED detector, with a particular focus on the STEM characterization of the structural properties of materials. These include the techniques of virtual detector imaging; higher-order Laue zone analysis; nanobeam electron diffraction; and scanning precession electron diffraction. In the latter, we demonstrate a nanoscale lattice parameter mapping with a fractional precision ≤6 × 10−4 (0.06%).

Towards the interpretation of a shift of the central beam in nano-beam electron diffraction as a change in mean inner potential

Evaluation of two-dimensional strain distribution by STEM/NBD

To facilitate investigations into the impacts of acquisition and reconstruction parameters on quantitative imaging, radiomics and CAD using CT imaging, we previously released an open-source implementation of a conventional weighted filtered backprojection reconstruction called FreeCT_wFBP. Our purpose was to extend that work by providing an open-source implementation ofa model-based iterative reconstruction method using coordinate descent optimization, called FreeCT_ICD. Model-based iterative reconstruction offers the potential for substantial radiation dose reduction, but can impose substantial computational processing and storage requirements. FreeCT_ICD is an open-source implementation of a model-based iterative reconstruction method that provides a reasonable tradeoff between these requirements. This was accomplished by adapting a previously proposed method that allows the system matrix to be stored with a reasonable memory requirement. The method amounts to describing the attenuation coefficient using rotating slices that follow the helical geometry. In the initially proposed version, the rotating slices are themselves described using blobs. We have replaced this description by a unique model that relies on trilinear interpolation together with the principles of Joseph's method. This model offers an improvement in memory requirement while still allowing highly accurate reconstruction for conventional CT geometries. The system matrix is stored column-wise and combined with an iterative coordinate descent (ICD) optimization. The result is FreeCT_ICD, which is a reconstruction program developed on the Linux platform using C++ libraries and released under the open-source GNU GPL v2.0 license. The software is capable of reconstructing raw projection data of helical CT scans. In this work, the software has been described and evaluated by reconstructing datasets exported from a clinical scanner which consisted of an ACR accreditation phantom dataset and a clinical pediatric thoracic scan. For the ACR phantom, image quality was comparable to clinical reconstructions as well as reconstructions using open-source FreeCT_wFBP software. The pediatric thoracic scan also yielded acceptable results. In addition, we did not observe any deleterious impact in image quality associated with the utilization of rotating slices. These evaluations also demonstrated reasonable tradeoffs in storage requirements and computational demands. FreeCT_ICD is an open-source implementation of a model-based iterative reconstruction method that extends the capabilities of previously released open-source reconstruction software and provides the ability to perform vendor-independent reconstructions of clinically acquired raw projection data. This implementation represents a reasonable tradeoff between storage and computational requirements and has demonstrated acceptable image quality in both simulated and clinical image datasets.

We report on the development of a nanometer scale strain mapping technique by means of scanning nano-beam electron diffraction. Only recently possible due to fast acquisition with a direct electron detector, this technique allows for strain mapping with a high precision of 0.1% at a lateral resolution of 1 nm for a large field of view reaching up to 1 μm. We demonstrate its application to a technologically relevant strain-engineered GaAs/GaAsP hetero-structure and show that the method can even be applied to highly defected regions with substantial changes in local crystal orientation. Strain maps derived from atomically resolved scanning transmission electron microscopy images were used to validate the accuracy, precision and resolution of this versatile technique.

Due to increasing demand for faster computations, deeply pipelined processor architectures are being employed to meet desired system performance. Functional validation of such pipelined processors is one of the most complex and expensive tasks in the current systems-on-chip design methodology. While language-based validation techniques have proposed several promising ideas, many challenges remain in applying them to realistic pipelined processors. This paper describes two practical challenges in this methodology: test generation and equivalence checking. The time and resources required for test generation using the existing approaches can be extremely large for today's pipelined processors. Similarly, traditional equivalence checkers are not useful in the context of language-driven model generation and functional validation. This paper outlines our plan to address these challenges using satisfiability checking

Pyrrhotites, characterized by the chemical formula Fe1–δS (0 < δ ≤ 1/8), represent an extended group of minerals that are derived from the NiAs-type FeS aristotype. They contain layered arrangements of ordered Fe vacancies, which are at the origin of the various magnetic signals registered from certain natural rocks and can act as efficient electrocatalysts in oxygen evolution reactions in ultrathin form. Despite extensive studies over the past century, the local structural details of pyrrhotite superstructures formed by different arrangements of Fe vacancies remain unclear, in particular at the atomic scale. Here, atomic-resolution high-angle annular dark-field imaging and nanobeam electron diffraction in the scanning transmission electron microscope are used to study natural pyrrhotite samples that contain commensurate 4C and incommensurate 4.91 ± 0.02C constituents. Local measurements of both the intensities and the picometer-scale shifts of individual Fe atomic columns are shown to be consistent with a model for the structure of 4C pyrrhotite, which was derived using X-ray diffraction by Tokonami et al. (1972). In 4.91 ± 0.02C pyrrhotite, 5C-like unequally sized nano-regions are found to join at anti-phase-like boundaries, leading to the incommensurability observed in the present pyrrhotite sample. This conclusion is supported by computer simulations. The local magnetic properties of each phase are inferred from the measurements. A discussion of perspectives for the quantitative counting of Fe vacancies at the atomic scale is presented.