The properties of most engineering alloys and ceramics are strongly affected by the grain size and morphology. Various standards exist for measurement of grain size by light microscopy e.g. ASTM E112 or by other methods including electron back-scatter diffraction such as ASTM E2627. There is a fundamental factor common to all methods – differentiation of one grain from its neighbors. For light micrographs, this is facilitated by appropriate etching – either to highlight the grain boundary (typical in steels and nickel alloys) or by coloring each grain differently from its neighbor (e.g. some aluminium alloys under polarised light). Once individual grains have been identified, measuring their size/shape distribution is trivial. Figures 1 and 2 show examples of grain boundary detection in metals and ceramics using light and field emission scanning electron microscopy. The metal (Alloy 600) was polished to a 0.25µm finish and then electro-etched in a dilute sulfuric acid.
The grain boundaries are clearly visible, as are the twinning lines within the grains. However, the twinning lines are lighter than the grain boundaries, and grain boundary detection was straightforward using machine learning. The zirconia sample was more challenging – it was examined in the as-received (un-polished and un-coated) condition in a ZEISS Sigma 300 FE-SEM using secondary electron imaging at 1kV.
Figure 1 (Top) Nickel Alloy 600 after metallographic preparation and electro etching. Brightfield imaging on a ZEISS Axio Imager Z2.m. (Bottom) ZEISS ZEN Intellesis segmentation of this image, showing grains in red and grain boundaries in green.
Grain boundaries are visible, but there are significant variations in contrast across the sample, as well as several pores. Using machine learning in ZEISS ZEN Intellesis, it was possible to directly segment grain boundaries to permit determination of grain size/shape, while simultaneously detecting and measuring pore size/shape/distribution.