Hints and Tips
OXFORD INSTRUMENTS EBSD PRODUCTSCMOS Detector RangeAZtecHKL Acquisition SoftwareAZtecCrystal Processing Software
Setting up an Electron Backscatter Diffraction (EBSD) system to collect suitable quality diffraction pattens for a particular application can be a challenge. For example, when is it necessary to work with the full detector resolution? How many frames should be averaged? Does the pattern need to look good quality to the human eye? What background correction settings should be used?
There are no single answers to these questions, but the settings that you choose should be governed by the application and the type and quality of data that you require.
The 2 EBSPs shown to the right were both collected from the same cracked duplex steel sample, one using high resolution and a very high electron dose, the other at relatively low resolution and a very small electron dose. Which is the most appropriate to use?
EBSPs from austenite grains in a duplex stainless steel. Left – full 1344x1024 resolution with a high electron dose (~10,000 nAms). Right – 158x128 resolution with a low electron dose (~3 nAms).
Both are appropriate for different analyses. The high resolution, high dose pattern has sufficient resolution and signal to noise to be used for high angular resolution analyses, for example to look at very small lattice rotations and even elastic strain measurements. The low resolution, low dose pattern is ideal for fast phase and orientation mapping with less emphasis on the angular precision of the data. The following maps were collected using these 2 detector settings from the cracked steel sample.
EBSD maps collected using different EBSP quality from a cracked steel sample. Left – high angular resolution map showing the distribution of plastic strain at a crack tip (shown using a kernel average misorientation map), collected with high resolution, high dose patterns. Right – phase map collected at high speed (3,355 patterns per second) using low resolution, low dose patterns.
In the tabs below there are recommendations about how to select the optimum EBSP resolution, the importance of the electron dose and what background correction settings should be used.
Most EBSD detectors can collect EBSD patterns in a range of different resolutions. These can be very low resolutions (as low as 40 x 30 pixels for heavily binned patterns from high-speed CCD-based detectors) or can be relatively high resolution (> 1 megapixel). Below we consider the benefits of working with high and low resolution EBSPs
The latest CMOS-based EBSD detectors, such as Oxford Instruments’ Symmetry S3 detector, can collect high resolution patterns at relatively high speeds; however, this does not necessarily make sense as an experimental strategy for a number of reasons. This includes the fact that higher resolution images (a) take longer to transfer (e.g. from the detector to the acquisition software) and (b) take longer to process (e.g. for background correction and Kikuchi band detection), but it may also provide little benefit if a high resolution EBSP is being processed using a much lower resolution Hough transform to detect the band positions. The list below indicates when higher resolution EBSPs should be collected:
High resolution (1344 x 1024 pixels) EBSP from a Zirconium Oxide sample.
Low resolution EBSPs are usually collected in order to speed up the whole data collection workflow. Smaller images can be transferred, saved and processed more quickly, yet typically contain sufficient information to be reliably indexed. However, the lower resolution of the patterns will reduce the potential angular precision of the data, as well as more advanced interrogations of the patterns to extract additional information such as elastic strain or unit-cell parameters.
Here is a list of applications for which low resolution EBSPs are sufficient:
Low resolution EBSP (156 x 128 pixels) from a ferritic steel sample
In many cases a compromise will be required, and a mid-resolution EBSP (e.g. 622 x 512 pixels) may provide the best balance between angular precision and speed. More information about the best way to set up for EBSD analyses can be found in many of the Oxford Instruments Nanoanalysis blogs here.
The electron dose is a more important variable than the beam current, and perhaps is also more important than the EBSP resolution. Essentially, while keeping the beam current constant, the electron dose for each pattern can be increased by (a) increasing the exposure time for each frame (but avoiding saturation) and (b) frame averaging.
The electron dose can be defined as:
Electron Dose = Beam Current x Exposure time (units: nAms)
Firstly, it is necessary to recognise that some materials (such as many lower atomic number phases, including a lot of common minerals) are weakly diffracting, and require a relatively high dose in order to generate an indexable diffraction pattern. A mineral such as plagioclase feldspar (CaAl2Si2O8) would usually require an electron dose of ~40 nAms to generate indexable patterns, whereas a Nickel sample (a strongly diffracting material) may only require a ~2 nAms dose.
However, after taking into account the diffraction behaviour of the material, increasing the electron dose will further improve the signal to noise ratio, which in turn will result in greater precision in the EBSD indexing. This is shown in the following example EBSPs from a steel sample. The first EBSP has been collected with a medium dose of 40 nAms. The pattern quality is high and there is enough detail for very good quality indexing. The second pattern was collected with a very high dose of 1000 nAms: here the pattern quality is exceptional (far beyond what is required for Hough-based indexing), but this EBSP could be used for HR-EBSD techniques to measure elastic strain, or for advanced patten matching analyses to determine variations in the unit cell ratios, for example.
EBSD patterns from a ferrite grain. Left – with an electron dose of 40 nAms. Right – with a dose of 1000 nAms.
As a final word, the Hough transform (used for regular indexing) is remarkably robust at detecting bands and indexing EBSPs when the signal to noise ratio is very low (i.e. for noisy patterns). Unless higher precision or additional information is required, there is no need to collect good quality patterns with high electron doses.
A more rigorous examination of electron dose is provided here.
The signal on the EBSD detector is dominated by backscattered electrons (BSEs) that carry no diffraction information, and the intensity varies significantly from a maximum usually close to the centre of the phosphor screen to much lower levels at the edges of the screen. Although the Kikuchi bands are visible in such an image, as shown below, for routine Hough-based indexing the results will be much improved if we (a) remove the signal from unwanted BSEs and (b) make the intensity levels much more uniform (“flat fielding”). This we can achieve by collecting a background image that contains no diffraction signal and then using a static background subtraction, or alternatively using a dynamic background correction process to enhance the final EBSD pattern.
EBSP background correction process. Left – raw EBSP. Centre – background image. Right – background corrected EBSP
The background image can be collected by scanning the beam rapidly over multiple different crystal orientations (e.g. across many different grains, usually at relatively low magnification) and integrating the signal on the EBSD detector. The advantage of using a static background correction process is that the background signal will also include any blemishes on the phosphor screen (such as the mark in the upper centre of the EBSPs, above) and will remove these from the final processed EBSP.
However, dynamic background correction processes do not require a stored background image, with each EBSP being used to generate its own background image. Although this won’t remove static artefacts, it will compensate better for changes in intensity due to the analysis of phases with different atomic numbers.
In general a combination of static and dynamic background correction will give good results for most materials.