Analysis of recrystallized volume fractions in uranium using electron backscatter diffraction


Rodney J. McCabe. Tel: 505 606 1649; fax: 505 667 8021; e-mail:


Electron backscatter diffraction was used to examine the recrystallization behaviour of warm, clock-rolled uranium. A new uranium preparation method was developed, resulting in acceptable specimen surface finishes nearly every time, even for as-rolled specimens. Recrystallized fractions were differentiated from unrecrystallized fractions using differences in the grain average misorientation, a measure of the internal level of misorientation within a grain. This new approach better estimates the recrystallized fraction than hardness measurements, and has the advantage over standard metallographic techniques of providing texture information.


Like those of most structural metals, the mechanical properties of cast uranium can be improved by working and recrystallizing the material. This type of processing generally results in a more favourable, smaller grain size, but also results in texturing that may be important because of the strong anisotropic nature of uranium (Foote et al., 1956; Vandermeer, 1982; Sherby et al., 1976). The orientation and magnitude of texture is strongly dependent on the thermomechanical process used to work the metal. For example, clock-rolling (a series of rolling passes with a 45° rotation between passes) produces less in-plane texture than unidirectional rolling (Mueller et al., 1955; Choi & Staker, 1996). It is important to understand how the microstructure evolves during the processing steps in order to explain, predict and/or tailor processing parameters for a given application. This article details an electron backscatter diffraction (EBSD)-based analysis of the evolution of the microstructure of warm, clock-rolled uranium during recrystallization.

EBSD analysis is a good tool for characterizing microstructures at the grain level that works by correlating an EBSD pattern, similar to Kikuchi patterns observed in transmission electron microscopy, with a crystal orientation for each point of a user-defined grid in a scanning electron microscope. EBSD has the advantage that grain size, grain orientation, texture and other grain and boundary properties can be obtained from a scan, whereas most other diffraction techniques are limited to measuring bulk texture properties, and standard metallographic techniques are limited to measuring grain dimensions.

EBSD has the potential to directly measure various recrystallization parameters while requiring a minimum of human subjectivity. For scans of a partially recrystallized material, there are often specific differences in the characteristics of the EBSD patterns that depend on whether or not a grain is recrystallized. Many of these differences are based on the fact that recrystallized grains have a significantly lower density of deformation defects, such as dislocations, than worked material. Therefore, the image quality (IQ) (a quantitative measure of the quality of an EBSD pattern) is generally better for a recrystallized grain. Similarly, the larger degree of accumulated defects in a previously worked metal generally results in larger degrees of misorientation within individual grains. Using EBSD line scans, Woldt & Juul Jensen (1995) first used a combination of qualitative changes in IQ and variations in pattern orientation to characterize the recrystallization of copper. Several recent studies have successfully differentiated recrystallized grains using quantitative differences in IQ (Black & Higginson, 1999; Tarasiuk et al., 2002; Caleyo et al., 2002) or a combination of IQ and observations of resolvable cell or subgrain boundaries (Bocos et al., 2003). Similarly, groups have been studying recrystallization using various measures of misorientation, including the kernel average misorientation (Dingley, 2004), a local measure of misorientation, and grain orientation spread (GOS) (Alvi et al., 2004; Rollett et al., 2004), one measure of the internal level of misorientation within a grain. One advantage of using a measure of misorientation vs. IQ is that the relative value of IQ is affected by many factors, including specimen preparation and EBSD scan parameters, whereas the measured orientation, from which the misorientation values are based, should be relatively unaffected.

To date, successful preparation of uranium samples for EBSD has proved to be difficult. There are currently two published studies of uranium microstructures determined by EBSD (Bingert et al., 2004; Carpenter & Bullock, 2000), one of as-cast and one of fully recrystallized samples, where a low density of deformation defects is expected. In addition, the success rate of preparing samples resulting in satisfactory EBSD patterns was very low in these studies. Because uranium has such a large Z number, the escape depth of backscattered electrons is very small. Therefore, the quality of the EBSD pattern strongly depends upon the surface condition of the sample (e.g. the amount of deformation remaining due to grinding and polishing or the presence of an oxide film). One difficulty in preparing a decent uranium EBSD sample is the propensity for uranium to rapidly grow an oxide that greatly diminishes the quality of the EBSD pattern.

The present study represents the first successful study using EBSD to examine worked and partially recrystallized uranium. It is a study of the recrystallization of α-uranium that has been warm clock-rolled and then annealed at 450 °C for various times. The EBSD specimens were prepared by a two-step electropolishing process that was successful nearly every time in producing an acceptable EBSD surface finish. In this study, the recrystallized grains were quantitatively distinguished from the unrecrystallized grains through differences in the grain average misorientation (GAM), a measure of the internal level of misorientation within a grain.

Materials and methods

The uranium used for this study contained a relatively low impurity content, as shown in Table 1. A series of processing steps was used to produce the uranium plate, including vacuum induction casting, hot upset forging (625 °C), and warm clock-rolling (300 °C, eight passes with rotations of 0°, 90°, 135°, 225°, 270°, 360°, 45° and 135°), with the clock-rolling resulting in a final reduction of approximately 50%. Cylindrical samples (3 mm in diameter, 10 mm in length) were machined from the plates for isothermal heat treatments using a quench dilatometer. The specimens were heated to 450 °C under a vacuum of 10−5 torr at a rate of 7.5 °C s−1. After the specimen had been held at 450 °C for a specified period of time, it was quenched at a rate of 10 °C s−1 to room temperature.

Table 1.  Impurities (wt. p.p.m.).

Metallographic mounting consisted of a vacuum impregnation and slow pressure curing process that minimizes heating of the sample (Kelly et al., 2006). For microhardness testing, specimens were ground and polished to a 1-µm diamond finish using standard uranium metallographic preparation techniques (Kelly et al., 2006). Ten Vickers microhardness indents were made across each specimen using an indentation load of 25 g.

The EBSD preparation technique described in this article has been used to successfully prepare specimens of uranium in the as-cast, warm-worked and partially recrystallized forms, and consists of polishing to a fresh 1-µm diamond finish followed by two electropolishing steps. The purpose of the first electropolishing step is to remove the damaged surface layer that remains after mechanical grinding and polishing. The first electrolyte used is a solution of 45% ethanol, 27% ethylene glycol, and 27% phosphoric acid (ASM International Handbook Committee, 1985). The partially recrystallized samples were electropolished at 10–15 V for 5–10 min at room temperature with the electrolyte being stirred. For larger specimen geometries and different microstructures (i.e. as-cast vs. worked), the best voltage may be as high as 30 V, and is determined on the basis of whether the sample is being etched, polished, or pitted. Following the electropolish, the specimens were rinsed in tap water, rinsed in propanol, warm air dried, and re-immersed in the unbiased electrolyte, which removes any residual reaction product from the surface. The samples were then re-rinsed and dried with unheated, forced air. When the samples are viewed under a light microscope following this first electropolishing step, primary inclusions are seen to stand out from the surface, evidence that uranium matrix material has been removed.

Although the first electropolishing step removes the damaged layer and leaves a nice mirror-like finish, the finish results in unsatisfactory EBSD patterns, probably because of the propensity for an oxide layer to rapidly form on the uranium surface. It is believed that the benefit of the second electropolishing step is that it helps to passivate the surface, at least long enough to get the sample into the scanning electron microscope. Other evidence of this passivating effect can be seen in the significantly slower rate at which a visual oxide forms on specimens, depending on whether it has had this electropolish or received just a mechanical polish. The second electrolyte used is a solution of 5% phosphoric acid and 95% de-ionized water (Kelly et al., 2006). Samples were electropolished at 5 V for 1–2 s at room temperature in an unstirred bath. Following the second electropolish, the specimens were rinsed with water, rinsed with propanol, and dried with unheated, forced air. The specimens were then well grounded with colloidal graphite and copper tape and expeditiously transferred into the scanning electron microscope.

Automated EBSD scans were performed at 25 kV in an FEI XL30 scanning electron microscope equipped with TSL data acquisition software. Regions roughly at the centre of the 3.175-mm-diameter mounted specimens were orientation mapped with a step size of 1 µm. The orientation data were analysed using TSL orientation imaging microscopy analysis 4 software. In order to obtain decent IQ and confidence indices from unrecrystallized regions, it was necessary to process the EBSD patterns with the autocontrast function of the acquisition software.

Results and Discussion

As-rolled and partially recrystallized uranium microstructures were characterized using EBSD and scanning electron microscopy. Figure 1(a) and (b) show transverse direction (with respect to the rolled plate normal) inverse pole figures overlying IQ maps of the microstructures of samples annealed at 450 °C for 600 s and 1800 s, respectively. The rolling normal and in-plane directions for the cylindrical specimens were estimated using texture asymmetries and based upon previous measurements of texture on this plate where we knew the rolling normal and in-plane directions. The directions obtained in this manner correlate well with the shapes of the unrecrystallized grains. The raw EBSD data were cleaned using a standard TSL cleanup function (neighbour orientation correlation). The black spots on these images, areas where patterns were not indexed as α-U, are largely due to the presence of uranium carbonitride primary inclusions (Kelly et al., 2006). Another interesting aspect of the microstructures is the presence of {130}〈310〉 twins in some unrecrystallized grains. Twin boundaries were identified with the EBSD data on the basis of a 69° rotation about the 〈001〉 axis (Bingert et al., 2004). Figure 1(c) and (d) distinguish the unrecrystallized (yellow) and recrystallized (blue) grains in the microstructure, with grain boundaries shown in black. Recrystallized grains were defined as grains with a GAM less than 0.95° and 1.15° for the 600-s and 1800-s anneals, respectively (grain size minimum was five pixels (∼4.33 µm2), with a grain tolerance angle of 5°). Determination of the GAM cutoff values was based on curve fitting to the GAM distribution and is discussed in more detail later. Another characteristic of the microstructure is the presence of local misorientations, particularly in the grains with GAM greater than the cutoff values (point-to-point misorientations between 1° and 5° are mapped in red). These misorientations are probably related to dislocation structures (cell walls or boundaries) within the grains.

Figure 1.

Transverse direction inverse pole figures overlying IQ maps for annealing times of (a) 600 s and (b) 1800 s. Rolling normal (ND) and in plane (IP) directions were determined on the basis of asymmetries in textures. (c) and (d) show recrystallized grains in blue and unrecrystallized grains in yellow.

On the basis of the EBSD data, there were several possible approaches for determining the recrystallized fraction. In order to obtain reasonable EBSD patterns from worked uranium, it was necessary to use the EBSD pattern autocontrast function of the acquisition software. This had the advantage of improving both the IQ and the confidence indices of the orientation solution, particularly for the unrecrystallized grains. However, using autocontrast had the drawback of normalizing the IQ such that IQ could not be used for discerning the recrystallized fraction. The distributions of grain average IQs for annealing times of 600 s and 1800 s are shown in Fig. 2(a) and (d), respectively, where the recrystallized distribution in each plot of Fig. 2 was determined with a GAM cutoff based on curve fitting to the GAM distribution. Although the average IQ for recrystallized grains is greater than that for unrecrystallized grains, the distribution differs greatly from the bimodal distributions of the previous studies that used IQ to quantify recrystallization (Black & Higginson, 1999; Tarasiuk et al., 2002; Caleyo et al., 2002), and it is clear that differentiating with IQ is not possible for our case.

Figure 2.

Comparison of methods for determining the recrystallized fraction in uranium for the scans shown in Fig. 1. The IQ method is shown in (a) and (d), the GOS method in (b) and (e), and the GAM method in (c) and (f). Specimens were annealed at 450 °C for 600 s (a, b, c) and 1800 s (d, e, f).

Using a measure of the internal grain misorientation physically seems like a reasonable approach for examining recrystallization. Deformed grains have high densities of dislocations, often organized into dislocation structures such as cell boundaries, whereas recrystallized grains tend to have much lower dislocation densities. Dislocations and dislocation structures generally result in local misorientations as high as several degrees, and measures of internal grain misorientation may be related directly to dislocation density and, thus, recrystallization. Although IQ is also related to defect density, it also depends on other factors, such as grain orientation, specimen preparation, indexing parameters, and video processing (autocontrast).

There are several measures of internal grain misorientation that might be used to differentiate recrystallized from unrecrystallized grains. GOS, having been used in previous studies (Alvi et al., 2004; Rollett et al., 2004), is the average deviation between the orientation of each point in a grain and the average orientation of the grain, and GOS distributions are shown in Fig. 2(b) and (e). GAM is the average misorientation between all neighbouring pairs of points in a grain, and GAM distributions are shown in Fig. 2(c) and (f). Using a cutoff value to separate the misorientations characteristic of recrystallized and unrecrystallized grains can be used for either GOS or GAM to estimate the recrystallized fraction. However, for both GOS and GAM, the appropriate cutoff does seem to vary with annealing time.

One advantage of GAM is the bimodal nature of the distribution. The recrystallized fraction can be determined by fitting a double-peaked curve to the GAM distribution, and taking the area under the low-misorientation peak. Similarly, a cutoff value based on the curve fit can be used to separate (partition) the recrystallized fraction in order to do further EBSD analysis. Because the distribution peaks are nonsymmetrical, double-peaked curves, each peak an asymmetrical double sigmoidal of the form


were fitted to the distributions, where A is related to the peak amplitude, xc is the peak position, and w1, w2 and w3 are related to the width and asymmetry of the peak distribution. Single peaked curves of this form fit the as-rolled and 100 000-s annealing time distributions well, with R-values of 0.933 and 0.997, respectively. Curve fits for samples annealed at 450 °C for 600 s and 1800 s are shown in Fig. 2(c) and (f) with the GAM distributions. Based on the curve fits, GAM cutoff values vary from around 0.85° for a 30-s annealing time to 1.15° for a 6000-s annealing time. Plots of the recrystallization based on the curve fitting along with Vickers hardness measurements as a function of annealing time at 450 °C are shown in Fig. 3. The EBSD-measured recrystallized fraction follows a sigmoidal curve expected for recrystallization.

Figure 3.

Vickers hardness and volume fraction of recrystallized grains vs. annealing times at 450 °C determined by curve fitting the GAM distribution for each temperature. As-annealed data are represented at 1 s. (Note that the hardness axis is reversed so that it can be more directly compared to the recrystallized fraction.)

When compared to other methods for measuring recrystallization, the present method has many advantages. The Vickers hardness measurements shown in Fig. 3 do not correlate well with the EBSD measurements, and the scatter in the hardness data is quite large compared to the differences in hardness for the different annealing times. The more rapid decrease in hardness at early annealing times may be indicative that recovery or other processes may be occurring, affecting hardness. The scatter in Vickers hardness is not unexpected, as microhardness measurements often sample only one grain, which may be either recrystallized or unrecrystallized. Even in fully recrystallized samples, the anisotropy in mechanical properties with grain orientation for uranium is enough to result in significant degrees of scatter.

It is expected that metallographic techniques should give similar results for recrystallized fractions. A primary advantage of the present EBSD technique over metallographic techniques is that EBSD largely removes human subjectivity from the analysis, whereas common metallographic techniques require a person to decide whether a given grain is recrystallized or not. The other main benefit of EBSD is the inherent orientation data, from which misorientation and textures can be measured, in particular, texture evolution during recrystallization. The as-rolled and fully recrystallized textures are shown in Fig. 4. The uncertainty in the as-rolled texture is higher than that for the recrystallized structure, because the unrecrystallized grain size is much larger than the recrystallized grain size, so the number of grains sampled is much smaller. The distribution of texture peaks is very similar throughout the recrystallization process, with all specimens having a strong c-axis texture within several degrees of the normal direction. The texture trend is a general weakening of the texture magnitude with recrystallization from something higher than 7 to around 5.

Figure 4.

Pole figures showing the texture in the (a) as-rolled and (b) 100 000 s at 450 °C annealed conditions (nearly fully recrystallized). The preferred orientation does not appear to change with recrystallization. The general trend is for the maximum intensity to decrease from greater than 7 to around 5.


EBSD was used to study recrystallization in uranium. A new uranium preparation method was developed that allowed acceptable EBSD surface finishes nearly every time, even for as-rolled specimens. The recrystallized grains were differentiated from the unrecrystallized grains through differences in the grain average misorientation, GAM, by fitting a two-peaked distribution function to the GAM distribution. This resulted in a sigmoidal isothermal recrystallization curve, as expected for recrystallization. A cutoff GAM based on the curve fit was used to partition the recrystallized from the unrecrystallized grains for further EBSD analysis. The general texture trend is for the distribution of texture peaks to stay very similar throughout the recrystallization process with a general weakening of the texture magnitude.


The authors gratefully acknowledge several individuals in MST-6 at Los Alamos for their contributions in casting, forging, rolling, machining, annealing, metallographic preparation, and hardness testing of the alloy and specimens, particularly Chastity Vigil and Ann Kelly. We also thank Carl Necker for valuable discussions of EBSD and recrystallization.