Processing of X-ray diffraction imaging data using remote sensing techniques

Abstract

Material properties are not only determined by the chemical composition but are modified during processing and use. The microstructure and thus the material properties are in most cases not homogeneous but exhibit spatial variations. To investigate these variations numerousμ-beam techniques have been developed during the last years [1], but even with powerful sources like those of synchrotron radiation facilities these techniques still require time consuming scans if larger areas shall be investigated. A novel technique, developed at HASYLAB, DESY [2] selects the probed region not from the primary side using a μ-beam but at the secondary side of the sample. Here a large sample area is illuminated and an array of parallel tube like collimators, a so called micro-channel plate (MCP), in front of a position sensitive detector (PSD) suppresses crossfire of radiation scattered from different regions of the specimen. Only radiation parallel to the channel axes reaches the detector yielding an entire image in a single shot. The channels also define the scattering angle 22 like a Soller collimator does in conventional diffraction experiments. If the arrangement of MCP and PSD is mounted on the 22-arm of a diffractometer scans can be performed by variation of the angles of the diffractometer or other parameters. These scans yield an entire image for each value of the scanning variable, while conventional experiments only measure one intensity for each value of a variable yielding one dimensional data sets. Using the imaging method a data set consists of images with more than 1 million pixels. A series of such images yields one spectrum per pixel. Even if the amount of data is reduced by integrating over several pixels or by restricting the analysis to a smaller region of interest (ROI) the number of spectra is still too large to be processed interactively one at a time. Furthermore, there exist strong correlations between spatial and spectral adjacent data making joint processing highly desirable. The aim of such an analysis is both the determination of the spatial distribution of different microconstituents in the sample and the extraction of their spectral properties. Images of the same scene as function of one parameter are also obtained in remote sensing where pictures are taken in different optical or infrared bands. In this field manifold techniques have been developed to extract information from such (hyperspectral) data [3] and commercial software packages are available. The similarity in data structure suggests the application of these methods also to data obtained by diffraction imaging. The experiments were performed at the 4-circle diffractometer at HASYLAB beamline G3 [4] carrying the detector arrangement consisting of a MCP in front of a Peltier-cooled CCD on its 22-arm. With this instrument manifold scans may be performed by variation of either the diffractometer angles or the conditions at the sample. Applications range from reciprocal space mapping [5] to kinetic studies [6]. The use of synchrotron radiation further allows variations of the energy of the incident radiation enabling spectroscopic investigations [7]. Here we will treat data obtained in ω− 22-scans in which the diffracted intensity is measured as function of the length of the scattering vector. The techniques used are, however, also applicable to data obtained by other scans. The new technique has been applied to a piece of railway rail which is of particular interest since the increase of axle loads in freight traffic as well as the increase in velocity of passenger trains within recent years led to an increase of wear of the rails, that appeared e.g. in an increase of frequency of structure modifications and defects in their surface layer. Among the structure modifications, the riffles and the white etching-layers (named according to their resistance towards metallographic etching) are of high practical as well as theoretical interest due to their high hardness that reaches up to 1200 HV [8, 9], their brittleness and due to the fact that their composition and crystallographic structure as well as their process of formation still are under vivid discussion [10–12]. In order to contribute towards clarifying