Modern Materials Characterization

Comparison of an experimental 20kV EBSD pattern for silicon (right) and the corresponding simulated pattern (left).

In the mid 1980s, when I was a physics graduate student in the Materials Department at the K.U. Leuven (Belgium), it was relatively rare to find a computer that was hooked up to a scientific instrument; computers were mostly used for word processing and spreadsheets, not for data acquisition. Scanning electron microscopes (SEMs) had a primitive single-color CRT screen and a device to transfer images to Polaroid prints.

Other microscopes used negatives to record images, meaning that every materials department had an extensive darkroom to develop the plates and print them. Energy-dispersive spectroscopy (EDS) was just becoming available as an add-on to electron microscopes. These units were controlled by a very slow but large computer (in size) with very little memory (measured in kilobytes) and no internal storage.

Fast forward three decades to today’s electronics (and labs without darkrooms). A simple smartphone has more capability than a 1980s desktop computer, and it is now possible to remotely operate and monitor scientific equipment using simple apps.

Modern electron microscopes can only be operated by computer. Depending on the level of instrument sophistication, some SEMs have simple graphical user interfaces that make them accessible to middle-school students.

Other instruments, such as transmission electron microscopes (TEMs), are also operated via a graphical user interface, but the complexity of these instruments requires significant user training. The amounts of data that can be produced by a modern microscope in a single session are astounding and are often measured in hundreds of gigabytes or more.

In addition to running my research group, I am also co-director of the Materials Characterization Facility (MCF) at Carnegie Mellon University in Pittsburgh, Pa. The MCF is a user facility, accessible to students and researchers from any college, as well as external users from other schools and local industry.

We have about 250 unique users each year, and our instrument pool includes three TEMs, four SEMs, two focused ion-beam SEMs that allow for the controlled removal of thin slices of material, two X-ray diffractometers, a nano-CT system for X-ray-computed tomography, several probe microscopy instruments for surface analysis, and a centralized classroom equipped with hardware to enable remote operation of most of these instruments. The staff members in our facility are not only responsible for user training and keeping the machines running, but they must also have a decent knowledge of computer systems, networks and data storage to handle the large volume generated by our instruments.

The photographs accompanying this column online show several of our instruments as well as the central remote-control cluster. In upcoming quarterly columns, I will highlight several state-of-the-art research projects that make heavy use of our facility, including: 3D serial-sectioning reconstruction of material microstructures; tomographic analysis of solid-oxide fuel cells; and the modeling work that we carry out to support, quantify and explain our materials-characterization results.

A Deeper Dive into Professor De Graef’s Research

Electron backscatter diffraction (EBSD) is one of several tools that can be used to determine the orientations of grains in polycrystalline and/or multiphase microstructures. To improve the capabilities of this technique, in particular for studying heavily deformed microstructures, Prof. De Graef’s research group has developed a simulation technique that relies on basic physics to model the scattering of electrons in the material and the subsequent formation of the EBSD pattern.

The simulated patterns are very realistic and can be used to create a new pattern-indexing approach that is far-more robust against noise than commercially available techniques. Ongoing work considers machine-learning approaches to speed up the new technique and turn it into a near real-time indexing engine.

Dr. Marc De Graef is Professor of Materials Science and Engineering at Carnegie Mellon University. Over the past two decades, Prof. De Graef’s research group has been at the forefront of developing the numerical tools that allow us to convert raw experimental data from serial sectioning experiments into accurate 3D reconstructions. The group has developed a simulation technique that relies on basic physics to model the scattering of electrons in materials and the subsequent formation of an EBSD pattern. Ongoing work involves machine-learning approaches to speed up the new technique and turn it into a near-real-time indexing engine. De Graef’s research group has also contributed extensively to the field of Lorentz microscopy.