The examination of the surface and cross-sections of cast or processed metals to assess the presence of particles, inclusions, or analyze grain size is a fundamental practice to metallurgy. The microstructures and particles embedded within a metal matrix influence its mechanical properties, such as strength, ductility, and fracture toughness. 
Consequently, microscopy and image analysis have played pivotal roles in the development of metallurgy, progressing from qualitative observation to quantitative analysis due to advances in technology. Today, microscopes are indispensable in applications like failure analysis, quality assurance, inspection, and research. 

This article explores the use of optical and scanning electron microscopy in the steelmaking and metal casting industry, the role of energy dispersive spectroscopy (EDS) and electron backscatter diffraction (EBSD) techniques, automated feature analysis for non-metallic inclusions (NMI), and recent innovations in digital segmentation and image characterization.

Optical Microscopy and SEM

Historically, light microscopy has been commonly employed for examining casting specimens and their cross-sections. These methods provided foundational insights, yet the emergence of advanced electron microscopy techniques has radically expanded analysis capabilities in the metalcasting and steelmaking industry. Modern microscopy offers a variety of methods for evaluating steel’s microstructural properties with unparalleled precision and versatility. Scanning Electron Microscopy (SEM) has become a widely adopted tool in the industry for generating high-resolution images of specimens, capable of magnifying up to 100,000X or more. SEM images are produced by scanning the specimen’s surface with a focused electron beam, providing a detailed view of surface structure and topography.

An SEM is an electron microscope that generates images by scanning a sample’s surface with a focused electron beam. Initially, tungsten filament-based SEMs (W- SEM) were common for imaging, but over the last 20–30 years, field-emission SEMs (FESEM) have emerged. 

The main distinction between W-SEM and FESEM lies in their electron emitters. W-SEMs use a thermionic emitter that heats a tungsten filament to release electrons, while FESEMs employ a field emission gun (FEG) with a sharp tungsten tip, emitting electrons through a potential gradient. 

FESEMs are preferred for their ability to produce high-resolution images with enhanced brightness and stability, vital for detailed analyses. SEM imaging utilizes signals generated by electron beam interactions at various sample depths, including secondary electrons (SE) for surface detail and back-scattered electrons (BSE) for compositional insights. The BSE signal is valuable as its intensity correlates with the atomic number of sample elements.

Characteristic x-rays, detected with energy dispersive spectroscopy, provide elemental analysis, aiding in evaluating carbides, nitrides, oxide scales and serving as a key tool for failure analysis and assessing sample cleanliness.

The combination of the SEM and EDS techniques gave rise to automated feature analysis (AFA). AFA offers statistical data on the number, elemental composition, morphology, and size of inclusions, facilitating more detailed assessment of non-metallic inclusions (NMIs). AFA employs a threshold-based method for NMI identification within a continuous matrix, with rule and vector files that define the composition and characteristics of different inclusions. The smallest detectable feature typically ranges between 0.7 µm and 5 µm, depending on the settings for the search grid and magnification. Each inclusion is measured for geometric characteristics using the rotating chords method, while EDS provides a detailed elemental analysis across each chord, offering an area-averaged elemental profile for each feature.

Electron Backscatter Diffraction 

Electron backscatter diffraction (EBSD) is an SEM-based imaging technique that visualizes, quantifies, and analyzes the microstructure of materials. For an EBSD experiment, a sample must be highly flat and polished, and the electron beam must be directed at a grazing angle requiring a stage tilt of 70 degrees. Using acceleration voltages ranging from 10–30 kV and an incident current between 1–50nA, EBSD produces diffraction patterns through interaction of the electron beam with the sample’s crystal lattice.

EBSD data reveal the orientation of crystals, crystallographic differences, grain boundaries, and levels of local crystalline integrity within the sample. By scanning the electron beam across a polycrystalline sample and measuring orientation at each location, a detailed map of the grain morphology, orientation, and boundaries can be generated. 
This mapping process also identifies the sample’s preferred crystal orientation, or texture, providing a comprehensive view of the microstructure. In EBSD analysis, orientation data is typically displayed using either Euler maps or inverse pole figure (IPF) maps. 

IPF maps use colors from an IPF color key, assigning color based on the measured orientation in a chosen viewing direction. This makes IPF maps particularly useful for showing preferred orientation, or texture, which appears as regions of similar or uniform colors. By displaying orientation data in this way, IPF maps make it easy to visualize and understand the spatial distribution of specific textures. 

Software from providers like EDAX and Oxford Instruments facilitates EBSD analysis and are widely used across the industry. Figure 1 shows the EBSD inverse pole figure maps for an austenitized press-hardened steel (PHS) grade (ULTRALUME® 1500), showing martensitic structure.

Image Segmentation and Analysis

When it comes to post-processing of images obtained on the optical microscopes or SEM, ImageJ has been the industry leader for many decades. It’s a Java-based image processing program developed at the National Institutes of Health and the Laboratory for Optical and Computational Instrumentation released in 1977. Most of the foundries and steelmaking industries that house a microscope would most probably have a version of this used in day-to-day analysis of their images. 

But in the last decade, with advances in machine learning, artificial intelligence, and scientific computing, image segmentation and characterization have made huge leaps. While there are currently software products from some microscope manufacturers that can perform live analysis of an image during observation, another emerging area is the post-processing of these images. Software such as MIPAR use a host of contrast enhancement and noise reduction filters for image alignment and pre-processing, offer classical and AI- based computer vision tools for segmentation, and provide an environment for segmentation visualization and quantification for any digital image, including those acquired by SEM and optical microscopy.

Modern image analysis software has provided multiple advantages over earlier versions, such as faster processing times, improved accuracy, and a higher capacity for calculating various features from a single image, including area fraction, thickness, roundness, and aspect ratio. Although these programs were not developed solely for metallurgical applications, many of their capabilities are highly beneficial for metallurgists. 

Applications in the steel industry include grain diameter calculations, thickness measurements of alloy layers and coatings, determining the area and area fraction of multiple phases, calculating roundness and aspect ratio of grain etc. in images generated from both optical and scanning electron microscopes. 

Figure 2 shows the coating thickness for an interdiffusion layer of a press-hardened steel grade (ULTRALUME 1500), where the thickness is calculated by drawing lines across the coating in regular intervals. The length of these lines provide the mean coating thickness while individual data points are also generated. 

Figure 3 shows the area fraction of an iron aluminide phase and Si-rich phase of a press- hardened steel grade, generated using a recipe on software utilizing segmentation techniques. These insights directly contribute to improving steel quality and minimizing failure risks by optimizing production processes and quality control measures.

Conclusion

In the last few decades, image acquisition and characterization in the steel industry have undergone tremendous transformation. From basic magnification tools like simple microscopes to advanced field emission electron microscopes, and from manual methods for grain size calculation to software driven by artificial intelligence, these changes have collectively redefined the field. 

The integration of technologies like SEM, EDS, EBSD, and advanced image segmentation software has led to an enhanced ability to inspect, analyze, and interpret microstructural characteristics with unprecedented precision. As technology continues to evolve at a rapid pace, adapting and embracing these advances will be crucial for the industry to stay competitive, improve productivity, and maintain stringent quality standards.  

Koushik Balasubramanian is a research engineer at the Research and Innovation Center of Cleveland-Cliffs Inc., working on materials characterization, inclusion analysis, and product development for carbon and electrical steel. He can be reached at koushik.balasubramanian@clevelandcliffs.com