Ryan J. Larsen and Dr. Brent L. Adams, Mechanical Engineering

A growing focus of material science research is relating material properties to grain boundary geometry. The purpose of my research has been to improve current methods of determining grain boundary geometry and to relate interface geometry to sensitization of AISI type 304 stainless steel. Sensitization is the formation of chromium carbides that occurs in a temperature range of 500-800 °C, and makes the material more susceptible to intergranular stress corrosion cracking and stress corrosion cracking.

It has been observed that adjacent grains which have a low angle of misorientation between their crystal lattices often have improved resistance to sensitization. However, certain high-angle boundaries have been shown to exhibit special properties as well. The low energy levels of these special boundaries can be explained in terms of the Coincident Site Lattice (CSL) model. The CSL model is a geometric construction which only considers the misorientation between the crystal lattices. If the lattices were allowed to overlap in the same space, the ratio of coincident sites would be given by the reciprocal of ∑. If the boundary plane passes through the coincident sites, the plane is coherent and the boundary is classified as a coherent CSL, otherwise it is a non-coherent CSL. It has been shown in various studies that non-coherent ∑3’s are more likely to sensitize than coherent ∑3’s, but less likely to sensitize than random high-angle boundaries.

Although it is not difficult to identify CSLs using existing technology and software, it is considerably more difficult to determine whether or not the grain boundary plane is coherent. Determining the orientation of the grain boundary plane requires a three dimensional data set, which is difficult to construct with a Scanning Electron Microscope. In two-dimensional scans of metal surfaces, the boundary plane shows as a single line, the grain boundary trace.

The idea behind my research project was simply to express the grain boundary trace as a vector in reference frames embedded in the adjoining crystal lattices. If this vector does not lie within a given crystallographic plane, it can be concluded that the grain boundary cannot lie in that plane. By using this method I hoped to identify grain boundaries that could not lie along the {111} crystal plane (the coherent plane for ∑3 boundaries). Thus, certain boundaries could be identified as non-coherent ∑3’s. As a measure of the success of this technique, I would like to produce evidence supporting the claims made by other researchers that non-coherent ∑3’s are more likely to sensitize than coherent ∑3’s.

The original program that I wrote analyzed data exported from Orientation Imaging Microscopy (OIM) Analysis software. This software package interprets data generated by the Electron Back Scatter Patterns (EBSP) from a Scanning Electron Microscope. It constructs maps, pole figures, and performs a wide variety of calculations. Within OIM Analysis, the user clicks on points that define the grain boundary trace, and on points within the two adjacent grains, and outputs the necessary information about these points to my program. The program calculates the minimum angle that the boundary trace makes with a {111} plane in each of the adjacent grains.

Although the initial results from the original program yielded some interesting insights, it had severe limitations. Repeated runs on the same boundary typically yielded a range of results of 10 degrees. This is caused by variations in the points the user chooses to describe the boundary endpoints, as well as variation in the points in the grain the user chooses to describe the orientation of the grain. Another problem of this method was the considerable time required to analyze a large number of boundaries. Because of these limitations, no extensive effort was made to use this program to label non-coherent and study their sensitization information.

Since that time, developments have made it possible to perform the analysis described above on a large number of boundaries in an automated fashion. Elements of the my original program have recently been incorporated into the OIM Analysis program. By working with the original creators of OIM technology and software, namely, my faculty mentor, Dr. Brent Adams, and one of his former students, Dr. Stuart Wright of TSL, Inc., I have been able to make a number of additions to the program to assist with my research. The new additions can be used to automatically extract the needed information about all grain boundaries in a scan, as well as the needed information on any number of selected boundaries in an automated way. The results are more accurate than the “manual” method because the program determines the grain boundary trace more precisely than a user could. Furthermore, the modified program is able to average a number of lattice orientations near the boundary to give the grain orientations. The addition of this feature to the OIM Analysis was an outgrowth of a more extensive project, stereology, which will be discussed later.

Although this improved method of analysis eliminates error due to user variation, it still has inherent sources of error. Most of the error stems from the limited resolution due to the discrete nature EBSP scanning. Because of the grid pattern used by OIM software to reconstruct boundaries, a boundary must be large with respect to the stepsize of the scan. It is difficult for the scans to show the microscopic step-like nature of many of the boundaries, caused by interchanging coherent and non-coherent segments. The overall slope of the boundary would appear to be several degrees off the coherent plane, even though large segments of the boundary plane are coherent.

It is our feeling that the resolution problems inherent in the OIM method are offset by the ability of OIM to collect information about a large number of boundaries in an automated fashion in order to perform statistical analysis. The next level of analysis, stereology, involves creating a statistical representation of how the planes are orientated in space. This is done by taking measurements of the slopes and lengths of thousands of grain boundaries from various oblique section planes. The automation of the calculations described earlier was made possible by modifications I made in OIM Analysis to accommodate the information extraction necessary for stereology. By comparing the results of the two methods I hope to make considerable advances in relating the orientation of the grain boundary plane to sensitization.