Electron Backscatter Diffraction (EBSD) is not a very accurate technique. The accumulated errors associated with cutting and mounting samples, affixing them to the appropriate sample holders for each scanning electron microscope (SEM) and then the process of tilting the sample to a high angle and collecting and indexing the diffraction patterns all contribute to a relatively poor level of accuracy. In most cases, the accuracy will not be better than 2° relative to an external reference frame (such as the rolling and transverse directions in a rolled metal sheet), and in many cases will be significantly poorer.
However, EBSD can be a very precise technique. The concepts of precision and accuracy are often mixed up and are best illustrated using the following image.
Schematic illustration to highlight the difference between accuracy and precision
In terms of EBSD, the precision means that it is possible to measure the differences in orientation between points in a dataset very accurately. Most published studies suggest that EBSD can measure orientations with a precision of ~0.5° using conventional Hough-based indexing techniques, although the latest generation of commercial EBSD systems may be able to collect data at high speeds with significantly better precision than this (e.g. <0.2°). There also exist a number of approaches to EBSD analyses that can provide significantly improved precision and these are explained in the tabs below.
High precision orientation data can provide very useful information about the nature of microstructures that have arisen due to localised strain, such as dislocation arrays or low angle boundaries. This will in turn enable a better understanding of deformation processes or, in the case of semiconductors and thin films, an effective characterisation of the quality of the component.
The simplest way to improve the precision of EBSD data is to generate more precise results during the initial indexing stage. This can be achieved by using a combination of higher quality (and higher resolution) EBSD patterns along with the use of a higher resolution Hough transform, but the precision will always be limited by the peak detection routines in the Hough image. Improved band detection routines, with localised higher resolution peak fitting in the Hough space, are commercially available and can improve the angular precision significantly below the 0.1° level. Details about the optimised band detection routines within the Oxford Instruments AZtecHKL EBSD software can be found here.
However, further improvements to the angular precision can be achieved by directly refining the fit of the simulated Kikuchi band positions to the Kikuchi bands themselves: i.e. performing the refinement in the true EBSP image space rather than in the Hough space. This is the approach of the Refined Accuracy band detection and indexing approach within the Oxford Instruments AZtecHKL software. Following initial band detection and preliminary indexing, the software will refine the positions of the Kikuchi band edges, taking into account their curved geometry, by adjusting the band orientation to match with the highest gradients in the EBSD pattern itself (i.e. the sharp edges of the individual Kikuchi bands).
The Refined Accuracy process is explained in more detail here.
Refined Accuracy gives significantly improved results in real-time, with angular precision down to the 0.01° level. This allows conventional EBSD to be used to examine very small orientation changes, such as those associated with single threading dislocations in GaN thin films, as shown in the below example. The downside of this approach is that it requires good quality EBSPs with sharp Kikuchi bands, so will perform well on samples that have undergone relatively low levels of deformation. For highly strained materials with more blurry EBSPs, the benefits of Refined Accuracy will be small.
Electron channelling contrast image showing individual threading dislocations.
Dislocation density map (using the Weighted Burgers Vector technique) of the same area, collected using Refined Accuracy. The orientation change across the dislocations is significantly less than 0.1°.
On the EBSD Indexing Techniques page the relatively new “Dictionary Indexing” approach was introduced. This involves the cross correlation of experimental EBSD patterns with simulations, and has proven extremely effective for indexing very poor quality patterns, when conventional Hough-based indexing struggles.
A similar workflow can also be used to improve the precision of EBSD results but in this case it can work in combination with standard Hough-based indexing as well. The concept follows these basic steps:
The simulated patterns can be produced to improve the orientation precision, but they can also target other properties that may produce changes in the EBSD patterns. These can include specific pseudosymmetry-related orientations, crystal polarity, slight variations in the unit-cell dimensions (e.g. related to carbon content in martensite in steels) or crystal chirality. The example below shows the use of pattern matching to resolve severe pseudosymmetry indexing problems in a metal halide perovskite sample, revealing the presence of twin domains.
Using conventional Hough-based indexing, with significant misindexing due to the pseudosymmetry of the perovskite structure.
Refined results using pattern matching, revealing the presence of twin domains. Modified from Liu et al., ACS Nano 2021, 15, 4, 7139–7148.
This approach has sometimes been referred to as “forward modelling”, and may become a common extension of the conventional EBSD workflow. This is particularly likely for cases where a full dynamical simulation of the EBSD is not required; if a kinematical simulation (which is much quicker to generate) can be used during the matching process, then there is significant potential for these forward modelling approaches to be applicable during live acquisition of EBSD datasets.
High resolution EBSD (generally referred to as HR-EBSD) is an increasingly popular technique that is used to deliver high angular resolution results (a more apt name would be “High Angular Resolution EBSD” to avoid confusion with high spatial resolution EBSD).
The technique was developed by Wilkinson et al., 2006 (refer to the Useful References Page for the full information) with the primary goal of measuring elastic strain using EBSD. This is achieved by measuring very small changes in the interplanar angles, enabling calculation of the deviatoric components of the elastic strain tensor. A high level of precision down to ~0.005° is achieved using cross-correlation functions as follows:
This process can determine the deviatoric lattice strain to a precision of 1 x 10-4, but the improvement in the precision of lattice rotation measurements also enables a better measurement of dislocation densities.
An example analysis is shown below: here a deformed and heat-treated Al-Mg alloy has been analysed using HR-EBSD to investigate the dislocation cell structures.
Disorientation colouring map, showing low angle boundaries (>0.4°) in red. Sample courtesy of Ali Gholinia (University of Manchester).
Remapped elastic strain maps from HR-EBSD. Results calculated using CrossCourt4, courtesy of Graham Meaden (BLG Vantage).
Total geometrically necessary dislocation density map from HR-EBSD.
There are several challenges associated with HR-EBSD. Firstly, the technique benefits from high resolution, high quality EBSD patterns. This generally necessitates using a very high electron dose (e.g. >1000 nAms), and storing the patterns at the highest resolution possible (ideally > 1 megapixel), so the speed of analysis is rarely quicker than a few analyses per second. Secondly, an undistorted reference pattern is required for each grain – this may be impossible to achieve in more highly strained samples. Finally, the processing of the patterns can take many hours and so HR-EBSD studies are usually targeted at small areas of interest rather than for larger scale investigations.
The HR-EBSD field is continually developing, with significant focus on the use of high-quality dynamical simulations as reference patterns, although determining the calibration (geometry) parameters with sufficient precision is still an outstanding challenge.