3D reconstruction and characterization of polycrystalline microstructures using a FIB–SEM system

Abstract

A novel methodology is described in this paper which is a step towards three-dimensional representation of grain structures for microstructure characterization and processing microstructural data for subsequent computational analysis. It facilitates evaluation of stereological parameters of grain structures from a series of two-dimensional (2D) electron backscatter diffraction (EBSD) maps. Crystallographic orientation maps of consecutive serial sections of a micron-size specimen are collected in an automated manner using a dual-beam focused ion beam–scanning electron microscope (FIB–SEM) outfitted with an EBSD system. Analysis of the serial-sectioning data is accomplished using a special purpose software program called “Micro-Imager”. Micro-Imager is able to output characterization parameters such as the distribution of grain size, number of neighboring grains, and grain orientation and misorientation for every 2D section. Some of these data can be compared with results from stereological exercises. Stacking the 2D statistical information obtained from the analysis of the serial-sectioning data provides a means to quantify the variability of grain structure in 3D.

Introduction

The ability to characterize microstructure is an important tool for materials scientists, because it allows one to quantify microstructure–property relations and anticipate the capability of a material to perform in a given application as a function of process history. For example, it is well known that the grain-size of a material has a strong effect on mechanical properties; therefore, an accurate measure of the grain size distribution is desirable to predict material performance. Classical methods for characterizing microstructure usually involve viewing an image from a sectioned surface, where the area of interest is mechanically polished and viewed in an optical or scanning electron microscope (SEM) [1]. With adequate resolution in this image-analysis process, grain boundaries and second-phase particles can be delineated, and stereological methods can subsequently be used to infer three-dimensional (3D) statistical attributes from the 2D microstructural images.

However, there are a number of microstructural parameters such as feature connectivity, true feature size, and true feature shape that cannot be inferred from 2D sections [2]. For the example of grain-size measurements, a common stereological measurement, e.g. mean linear intercept, can be used to determine the average grain size, but only for certain assumptions of the grain morphology. By comparison, this measurement can be computed without any bias if a full 3D data set is available for the microstructure. In addition, many stereological parameters yield only average scalar quantities to describe microstructural features. Recognizing the fact that many properties (especially those associated with failure) require extreme values of the microstructure [3], it is evident that characterizing the full distribution of these features may be more appropriate for some predictive models [4], [5].

The need to more completely characterize the 3D microstructure has led to the development of methods that allow one to directly obtain 3D microstructural data [6], [7]. One methodology that has been successfully used to perform this task is serial sectioning [6], [7]. However, this technique can be time-consuming and, if performed manually, can be prone to errors related to maintaining a constant sectioning thickness. This paper discusses a newly developed technique to characterize grain structure via serial sectioning utilizing a new type of electron microscope; a dual-beam focused ion beam–scanning electron microscope (FIB–SEM). This microscope is capable of highly localized micro-machining and ion imaging using the FIB column, and non-destructive high-resolution imaging or other analytical methods such as electron backscatter diffraction (EBSD) using the electron column. For this study, the FIB is used to serial-section specimens and the EBSD system is used to obtain an orientation map for each section. The dual-beam FIB–SEM shows great potential because it can be automated to perform this analysis without user interaction.

Focused ion beam or FIB has become very popular in the recent days and has been used by a number of researchers to serial section and visualize microstructure in 3D [8], [9], [10], [11], [12], [13], [14], [15]. This technique has brought about a tremendous advancement in the ability to view the true 3D microstructure of metallic materials with complex microstructural morphology and crystallographic orientation. Many of the published literature (e.g. [8], [9], [10], [11], [12], [13], [14], [15]) provide only “slice and view” data. Sample repositioning in these studies was generally performed manually. Most of these experiments conduct fine-scale sectioning using a single beam FIB. Sakamoto et al. [8] and Fraser et al. [14] have used dual-beam FIB–SEM techniques to acquire morphological data for microanalysis. Much of this study, while sufficient for 3D reconstruction and visualization, does not contain adequate quantitative orientation information or allow for a completely automated method for grain segmentation to be incorporated. Additionally, manual sectioning is simply not feasible, even with a FIB, when collecting a large-scale high-resolution 3D data, a necessary ingredient for high fidelity statistical characterization.

This paper presents three significant advancements over the previous work, through (i) automated sectioning and scanning, (ii) collecting orientation data on each section for providing quantitative 3D crystallographic data, and (iii) automated segmenting and characterizing microstructural features. Customized scripts are developed in this work to completely control the FIB–SEM system during the sectioning and EBSD collection. The collection of EBSD maps on every section yields a completely automated path to grain reconstruction and segmentation. In addition to visualization, this work introduces a new methodology for the collection of 3D data, as well as a framework for the handling of the 3D data set. This paper discusses a novel procedure for obtaining data from the FIB–SEM, starting from sample preparation to the post-processing of the data. Post-processing of the data is performed using a customized in-house program called “Micro-Imager”. This program has the capability for processing the raw data characterization and eventual finite element modeling. Starting from 2D EBSD maps, Micro-Imager automatically defines grains and grain boundary segments, and calculates various statistical features of the microstructure. By utilizing the statistical data for a sequence of 2D EBSD maps from a serial-sectioning experiment, statistical correlations and 3D characteristics of a polycrystalline microstructure can be understood. The focus of the present paper is on developing improved data collection and subsequent statistical analysis techniques. The methods presented in this paper are an integral part of a larger characterization and modeling framework. An outline of this overarching framework is given in Fig. 1. While the ultimate goal of the overall research is to obtain microstructure–property relationships, the present paper only discusses an essential stepping stone in that direction. The results of this study are particularly useful in constructing truly representative material microstructures as input for modeling and simulation programs.