Tutorial B – Data Collection: Practical Considerations
1. Collection of High Quality EBSD Patterns
High quality, high resolution EBSD patterns are the most important ingredient for successful strain measurement; these EBSD patterns should always be collected at the maximum camera resolution for best results.
Pattern quality can generally be improved by :
- Using a camera with a detector containing at least 1000×1000 pixels and preferable with 12-bit greyscale resolution.
- Increasing the exposure time of the camera to the point where the brightest point in the image is almost saturated; in a 12-bit camera this would correspond to an intensity level of 4096. Do not overexpose. Overexposure may lead to problems associated with specimen drift and contamination.
- Having a bright and scratch free phosphor.
- Increasing the electron beam current, this may reduce spatial resolution because of the increased electron probe size.
- Selecting an appropriate electron beam voltage. Higher voltage improves pattern contrast but increases the sample depth of beam penetration.
- Tilting the Sample to 75º will give stronger EBSD pattern contrast. However, this will increase the electron probe footprint on the specimen surface and, therefore, add uncertainty to the pattern’s origin.
2. Avoid Scratches on the Phosphor
Ideally, the phosphor screen should be in perfect condition. Any scratches or other blemishes on the phosphor represent areas of static, i.e. unmoving, content when comparing images. Static image content may cause a shift between strained and reference images with images appearing smaller than they are or even resulting in a zero-shift measurement.
The usual EBSD procedure of removing the phosphor’s background image from all recorded images of the EBSD pattern has several advantages in this context as it not only boosts the contrast of the patterns but may also remove phosphor defects. However, this procedure is not always successful and obtaining good background images may be difficult, especially if the sample is a single crystal. In certain circumstances, e.g. in the case of measurements near a sample’s edge where electrons emitted from both surfaces at the edge contribute to the EBSD pattern, the resulting change in pattern contrast could lead to erroneous strain measurements.
3. Avoid Pattern Over Saturation
If there are areas with saturated pixels in the patterns, especially in the background pattern used for background removal, these areas will represent static features in the pattern which can cause an incorrect measurement of zero-shift as also described with phosphor blemishes above.
4. Good Vacuum and Clean Samples
Significant sample contamination, by the generation of a thin film of carbon deposit on the surface, drastically reduces both the pattern quality and the ability to make precise and accurate strain measurements. It is recommended to use clean samples and anti-contamination devices such as a liquid nitrogen cold finger.
5. Number of Pixels in the Pattern
If a pattern is taken with a camera’s highest resolution the shift measurements will be more precise and accurate. However, patterns taken at a higher resolution are proportionally longer to expose and this means that issues such as sample drift and beam contamination become more apparent. Therefore, it may be necessary to compromise between quality and speed with compromises tolerable in cases where an experiment’s focus is not elastic strain but rotations caused by plastic strain.
6. Calibration
To interpret EBSD patterns it is necessary to know the EBSD pattern centre and specimen to film distance i.e. the standard calibration data. This information is stored in the datasets’ project files or, if no project file exists, must be entered manually – the calibration routines provided by the EBSD system manufacturers’ can be used.
The accuracy of the measurements is depends on the accuracy of the calibration data. Calibration parameters are required to enable conversion of the directly measured shifts of the EBSD pattern within an ROI, which are in units of pixels, to an angular measure made in relation to the EBSD pattern centre; this calibration is essential to calculate the strain tensor.
Villert et al, see reference below, show that errors in the pattern centre cancel out to a first order approximation if all patterns are taken from the same scan, i.e. when the systematic error of the pattern centre is the same for all patterns.
Villert, S., Maurice, C., Wyon, C. and Fortunier, R. (2009), “Accuracy assessment of elastic strain measurement by EBSD.” Journal of Microscopy, 233: 290–301. doi: 10.1111/j.1365-2818.2009.03120.x
7. Pattern Centre Position
The value of Y*, i.e. the height of the pattern centre on the phosphor is important. If the sample is too low in relation to the phosphor screen, the spread of the spray of back scattered electrons will be centred low down the phosphor screen; this will mean the signal is weak at the top of the screen and the results from ROI centred here very noisy and consequently degrade the overall performance of the calculations.
You can try and adjust the stage height in order centre the spray of electrons is centred the phosphor screen. This adjustment requires setting the pattern centre at a height of 2/3rds the screen diameter from the bottom of the screen. The exact Y* position will vary with the atomic number of the specimen.
8. Choosing the Reference Pattern
The reference EBSD pattern should be taken from a region which is known to be strain free. As all other patterns will be compared to this reference pattern it should be of the highest possible quality. If it is not known precisely if the selected region for the reference pattern is actually strain free, the strain values measured will not be accurate even though the technique’s precision remains unchanged. All relative measures will have an accuracy and precision at the stated 2 parts in 10000.
For manual input of the reference point, you may wish to choose a point far away from a known point of strain, e.g. an indentation point, crack or grain boundary triple point in a polycrystalline material. The selection can be based on the quality of the EBSD pattern, for example, how sharp it is. The same criteria can be used for automated reference pattern detection, CrossCourt will search the data provided in the project files of the EBSD pattern indexing routines of EDAX, Oxford Instruments or Bruker, whichever, relevant, and search for the best pattern(s) as reported using their internal measures. Often, the reference patterns found by these automated routines are not at the centre of grains, where the strain might be thought to be least, but closer to a grain boundary. Where this is the case, you can opt to override the automated choice and manually place the reference pattern.
Strain maps are presented once strain measurement is conducted. Inspecting these strain maps may then reveal a more likely location for the reference pattern. For example, at the centre of a region showing no strain gradients over several microns for all strain components.
9. Correction for beam movement
When the electron beam of the SEM is moved across the sample there is an inherent and identical shift of the EBSD pattern across the phosphor screen. If large enough, generally more than 3 microns, the software will interpret this shift, unless it is removed, as a rigid body rotation; because of this the reference pattern should be as close as possible to the area of interest. If beam movements of more than 3 microns are made, for example, in a line scan or a map, a beam movement correction procedure should then be applied by finding the effective pixel size in microns – see Section 6 above on calibration for further details.
Calibration values calculated with CrossCourt3 should still be valid for the same EBSD equipment.
Where a large area scan is taken, a secondary problem will almost certainly arise. SEM scans are designed to be perfectly rectangular for a horizontal sample, but in the case of EBSD where the sample is highly tilted, this rectangular scan becomes distorted into a trapezium due to the difference in working distance between the top and bottom of the scan. Therefore, this will mean the actual step size will vary across the scan making calibration impossible and introducing phantom strains. Due to this likely problem, we recommend making sure the reference pattern is no more than 150µm from the test pattern.
Please note, phantom strains will appear if large scans, > 150 µm, of single crystal materials are examined.
10. Keep experimental conditions the same for all Patterns
All patterns must have the same image size and be collected under the same microscope conditions.
11. Specimen to Screen Distance
For taking measurements the best experimental arrangement is with the phosphor screen close to the sample collecting a wide solid angle of the diffraction pattern.
12. Data Scan Collection
We recommend using your EBSD collection software’s automated area scan for collecting the data set to be analysed. Beforehand, the recording of patterns at each point in the video page needs to be initialised. If possible, always select 12 bit tiff images when selecting the format of the image. If recording an area image, select the square grid option on the data collection page.
The quality of the data collected is improved significantly by, if available, collecting and averaging more than a single frame at each point in the data set; this improves the signal to noise ratio in the patterns collected and leads to less noise in the strains calculated.
13. A note about Euler angle settings
Use your EBSD system’s default Euler Angle settings.
Each EBSD manufacturer uses a different default frame of reference for the Euler Angles. Crosscourt automatically rotates these from the manufacturer’s standard set into the frame of reference used by Crosscourt. If the Euler data is collected in a non-standard frame of reference, Crosscourt will still apply the default correction leading to the mixing up of the normal strains. This happens because the Euler angles are only used in the rotation of the elasticity coefficients which are used to set the traction free condition, i.e. S33 the stress normal to the surface is zero, in order to be able to calculate all 9 tensor components. If the wrong frame of reference is used, the stress is set to zero in a different plane and the normal strains are not calculated correctly.
14. Special instructions for Oxford Instruments’ AZtec Software Users
Oxford Instruments have produced a document explaining how to set up AZtec to collect the best data for use with CrossCourt and we highly recommend that you read this carefully.
We also recommend you have at least Version 3 of AZtec installed, this will allow CrossCourt to perform background removal using the static background stored by AZtec from Version 3 onwards.