Texture-dependent Measurement of Recrystallization Kinetics Using Electron Backscatter Diffraction

Partially recrystallized samples of 5352 aluminum alloy and pure titanium were sectioned and characterized with automated electron backscatter diffraction (EBSD) in a scanning electron microscope. Pixels were partitioned into recrystallized (annealed state) and unrecrystallized (deformed state) using a criterion based on the confidence index (CI) and the diffraction pattern quality (IQ). Unrecrystallized pixels were linked to Low C I and IQ. Errors in indexing were minimized by incorporating isolated unrecrystallized pixels into surrounding recrystallized regions. Good agreement was obtained with conventional methods of determining fraction recrystallized. INTRODUCTION: The traditional method of measuring the fraction recrystallized, in deformed and partially annealed metals, uses optical micrographs and the operator’s judgment based on the appearance of each region. Uniform gray level is typically associated with recrystallized and speckle is associated with unrecrystallized. These qualitative criteria suggest that local perfection of the lattice should lead to a quantitative criterion using EBSD maps. In recent years, this has been attempted with digital imaging using a grayscale threshold on the image quality (IQ) values, which is a measure of the diffraction quality at any point (Wright[2000]). Given a suitable choice of threshold, the fraction recrystallized is then equal to the area fraction of high IQ points. Also of interest is observation of the evolution of any preferred orientation (or texture) in the material as annealing proceeds. This is typically measured in a separate step using X-ray diffraction. While this procedure remains useful as a measure of the average texture, it cannot be used to obtain detailed information on individual components. In addition to the need for information on spatial and texture-dependent information on the kinetics of recrystallization, it is evident that the cutoff threshold used to determine the recrystallization state is very subjective as it is heavily operator dependent. These considerations motivated a desire to create and test a new approach to this problem. This work is allied to other efforts to characterize the deformed state at a smaller scale in which the geometrically necessary remnant dislocation structure is derived from the orientation field. PROCEDURE, RESULTS AND DISCUSSION: Our approach is based on TexSEM Labs’ Orientation Imaging Microscopy (OIM) system. In this technique, a focused electron beam is used to interrogate a sample surface on a regular grid of points (see Published in the Proceedings of Plasticity’02, the 9 Intl. Symposium on Plasticity and its Current Applications, edited by A.S. Khan and O. Lopez-Pamies, Aruba, NEAT Press, Fulton MD, p.42 (2002). Adams[1993]). An Electron Backscatter Diffraction Pattern (EBSP) is captured for each point, and this pattern is subsequently indexed to determine the orientation of the crystal at that grid position. The IQ value, which represents the contrast present in the pattern, and the CI value, representing how certain the software is of the indexing are also recorded. However, despite an observed relationship between the unrecrystallized portions and low IQ and CI values, testing with thresholds on the IQ or CI yielded inadequate results. Specifically, it was found to be too aggressive and yielded unrealistic microstructures, as it eliminated too many points around recrystallized grains (including grain boundaries), while occasionally leaving points surrounded on all sides by unrecrystallized material, as can be seen in the change from Figure 1(a) to 1(b). In order to create a more physically based and justifiable algorithm, we set a CI threshold of 0.1. This was justified since experience indicated that, in general, the transition between welland poorly-indexed points occurred at this value. Next, an IQ threshold was chosen to be the average IQ of all points with a CI value less than 0.1. This value was not only found to be in many cases, surprisingly, close to the cutoff that would have been chosen by the operator, it also completely eliminated the need for a subjective operator decision. Lastly, a neighbor correlation was applied in order to ensure that the final microstructure was physically reasonable. This allowed a point to be considered recrystallized if it has at least four recrystallized nearest neighbors, or three of its nearest neighbors are within 3° of the same orientation. We can represent all of this as follows: