ImageEval . A software for the processing, evaluation and acquisition of (S)TEM images

Abstract

Quantitative scanning or conventional TEM, as well as electron diffraction techniques usually require the processing of raw data, the comparison with simulated counterparts and a meaningful visualisation of results. Here we present ImageEval, a Matlab-based software which contains established techniques in a modularised manner with a graphical user interface. This enables the efficient, reproducible application of established techniques and assures both the access to and the transparent exchange of know-how among different groups after a pioneer methodical development. In the following we summarise central ImageEval features. The concept. ImageEval currently hosts 13 evaluation methods as separate modules. All modules can be used independently of each other but share a common data structure, so that results of one module can be further evaluated in another. Intermediate results, simulated reference or structural data are stored as meta data that is available for comparison or evaluation in all modules. The program reads Gatan (dm3) and FEI (emi/ser) file formats as well as common image formats. Large image series, e.g., 4-dimensional STEM data which often contain \sim106 images and more, can be handled by pointers to files so as to load images into workspace only when to be processed. ImageEval is available as a standalone executable Version for Windows/Unix or as the Matlab source code directly, the latter option enabling the adaptation of evaluations for individual tasks. STEM Z-contrast. This module allows for the quantitative composition and/or thickness mapping based on (High-resolution, HR-) STEM images either on the atomic lattice or on a regular user-defined grid. To this end, STEM intensities extracted from a Voronoi segmentation are compared with simulations to be loaded as reference data [Ultram. 109, 1171 (2009)]. For complex unit cells, the comparison can be distinguished with respect to an arbitrary number of sublattices for the different atomic sites, and intensities extracted from different images (e.g. bright & dark field acquired simultaneously) may be used simultaneously. High-resolution strain state analysis. HR(S)TEM images can be evaluated by detecting atomic column positions (intensity maxima or minima) or lattice fringes with subpixel accuracy, e.g., by parabolian or Gaussian fits. Various filtering options are available, namely a Wiener noise filter which preserves the contrast of the high-resolution image. The vector field of the displacement from a regular lattice, its projection along a given direction and the local strain can be calculated [Optik 102, 63 (1996)]. Strain analysis by nano-beam electron diffraction (SANBED). Series of parallel- or convergent-beam electron diffraction patterns (CBED) are evaluated as to the positions of the reflections using centre of gravity computation (spot pattern) as well as radial gradient maximisation, edge detection or cross correlation methods [Micr. Microanal. 18, 995 (2012)]. Typically these are 3D or 4D data sets, corresponding to the acquisition of diffraction patterns on a STEM scan line or area. Differential Phase Contrast (DPC). This module calculates the centre of gravity (average momentum [Nat. Comm. 5, 5653 (2014)]), the signals of a segmented quadrant detector, bright- and annular bright field signals from a 3D or 4D data set, i.e. ronchigrams recorded at each position of the STEM probe. Tools for angle calibration and calculating the charge density are also available. COM interface. ImageEval can communicate with the common object model (COM) of the Microscope and the acquisition software. This enables the efficient implementation of individual experimental procedures such as acquiring a STEM image series while automically changing imaging parameters subsequently (e.g. acceptance angles of the ADF detector). Basic Tools. Routinely needed tools such as cross-correlation of images to correct for relative shifts, (inverse) Fourier transform, line profiling, calculation of rotational average/sum, applying circular, annular or polygon masks, binning and image rotation are collected here. Further methods include the analysis of angle-resolved STEM data, composition evaluation by lattice fringe analysis (CELFA, [Ultram. 72, 121 (1998)]) exploiting chemically sensitive imaging, geometric phase analysis (GPA (Ultram. 74, 131 (1998)]), data reduction in spot diffraction patterns, parametric fits (``atom counting'' [Nature 470, 374 (2011)]), and processing of simulation results from the STEMsim software [Spr. Proc. Phys. 120, 169 (2007)] .