Introduction to Backscatter Kikuchi Diffraction

Pattern Formation
Backscatter Kikuchi patterns ( BKP), also known as Electron BackScattering Patterns (EBSD) are produced by incoherent wide-angle scattering of a stationary beam of high-energy electrons from a virtually perfect volume of crystal. A small fraction of these electrons are channeled along low-index lattice planes, leave the crystal and form a Backscatter Kikuchi pattern on a phosphor screen placed close to the specimen. It consists of straight Kikuchi bands whose widths are, according to the channeling conditions, proportional to the Bragg's angles. The center lines of the bands correspond to the (imaginary) section lines of the lattice planes with the screen, the star-like band crossings to zone axes of the crystal lattice, and the angles between the bands to interplanar angles. Therefore, the crystal orientation of the grain under the beam can be determined with high accuracy by simply measuring the widths and the positions of several bands in a pattern. In the SEM, the patterns are recorded with a low-light level CCD camera, digitized, corrected for background and transmitted to a PC for indexing.

Advanced BKD Systems
Texture analysis and characterization of microstructure require a large number of grain orientations to be measured on a selected area on the specimen without the interaction of the operator (Automated Crystal Orientation Measurement/Microscopy = ACOM) or in short Orientation Microscopy (OM). For indexing a pattern and determining the grain orientation, the positions of at least three bands in the pattern are required. The band positions are extracted automatically by applying a Radon or Hough transform (see below).
In advanced BKD systems, preference is given to digital beam scan, due to higher speed and precision, over a mechanical stage scan. To make allowance for the strong forward scattering of fast electrons and to obtain sufficiently intense patterns, the specimen is steeply inclined to make a shallow angle of typically 20° with the primary beam.
The specimen tilt, however, has an unwanted side effect on systems with digital beam scan in that the diffraction geometry changes from one measured point to the next and that the beam spot runs out of focus when scanning down the specimen one line after the other. Therefore, the specimen-to-screen distance and the pattern center have to be calibrated automatically from point to point, as well as the probe-forming lens has to be focused dynamically. The microscope is under full control of the ACOM program during automated measurement. When automated calibration is implemented in the BKD system, the operator need not bother about keeping a fixed working distance.
The band geometry is extracted on-line in a fast pattern recognition subroutine by applying a Radon transform. The higher the diffracting crystal volume is plastically deformed, the more diffuse the Kikuchi pattern appears to be. Pattern quality, as a measure of local plastic deformation, is hence determined by applying a 1D FFT on the Radon transformed pattern and by weighting the high spatial frequencies of the Radon peaks of the most prominent bands.

Performance
Spatial resolution is better than a tenth of a micron, depending on the spot size of the beam, the accelerating voltage and the density of the material. Resolution does in principle not depend neither on the beam current nor on the actual microscope magnification. The specimen area accessible to OM measurement is thus only limited by the lowest microscope magnification provided that the beam is focused dynamically. Since some EBSD systems are equipped with a camera of poor sensitivity, the spot size of the SEM has then to be widened unduly in order to obtain sufficient beam current, and hence spatial resolution gets worse.
Depth of information beneath the surface is in the range of the mean free path of the backscattered electrons (estimated at some 10 nm only). BKD is thus a surface sensitive method. Special care is required in preparing clean specimen surfaces free from artifacts. The accuracy of grain orientation measurement is < 0.5°. A test of spatial resolution and accuracy can be performed by mapping a specimen that contains fine deformation twins. Dynamic system calibration can be checked by scanning across a large single crystal (e.g. silicon) at low SEM magnification and verifying the uniformity of orientation data. Speed of advanced BKD systems exceeds several ten to several hundred thousand measured orientations per hour (see Fast EBSD).

(Typical shortcomings of out-of-date EBSD systems are:

• Neither automated system calibration nor pattern calibration "on-the-fly" is available. =>The pattern center, as one reference direction for calculating the grain orientation, as well as the specimen-to-screen distance (= pattern magnification) is determined experimentally for one working distance, e.g. with a reference crystal or by pulling back the camera (zoom center = pattern center). The operator is well advised to place the sample at the fixed working distance and to avoid digital beam scan at medium or low microscope magnifications. Insufficient calibration may lead to false indexing.
• Dynamic focusing is not provided, or the standard dynamic focusing appliance of the SEM is employed. Due to the steep specimen tilt and the high sensitivity of the quality of BKD on spot size, precise focusing is essential not only at low magnifications, but also for studying fine grain materials.
• Camera sensitivity is insufficient. => The pattern is either integrated, on the expense of speed, on the camera chip or by software to reduce noise, or the beam current has to be increased excessively. In the latter case, spatial resolution is lost since the probe-forming lenses are not operated under optimal conditions. In addition, contamination rate increases with beam current density, in particular when using a FE gun.
• Indexing is performed by only considering the center lines of Kikuchi bands rather than the center lines and the band widths, i.e. the interplanar angles and the Bragg angles => indexing is less reliable.
• Indexing is done with only a few bands since the Hough transform and the band tracing algorithm are inadequate to provide a sufficient number of located bands => propensity to false indexing.
  (N.B.: Typically 7 or more bands should be available for cubic symmetry, and about 10 bands or more are desirable for lower symmetries.)