Atomic-Scale Quantification of Grain Boundary Segregation in Nanocrystalline Material

Grain boundary segregation leads to nanoscale chemical variations that can alter a material’s performance by orders of magnitude (e.g., embrittlement). To understand this phenomenon, a large number of grain boundaries must be characterized in terms of both their five crystallographic interface parameters and their atomic-scale chemical composition. We demonstrate how this can be achieved using an approach that combines the accuracy of structural characterization in transmission electron microscopy with the 3D chemical sensitivity of atom probe tomography. We find a linear trend between carbon segregation and the misorientation angle $\omega$ for low-angle grain boundaries in ferrite, which indicates that $\omega$ is the most influential crystallographic parameter in this regime. However, there are significant deviations from this linear trend indicating an additional strong influence of other crystallographic parameters (grain boundary plane, rotation axis). For high-angle grain boundaries, no general trend between carbon excess and $\omega$ is observed; i.e., the grain boundary plane and rotation axis have an even higher influence on the segregation behavior in this regime. Slight deviations from special grain boundary configurations are shown to lead to unexpectedly high levels of segregation.

Figure 1

Experimental setup and output data. After a sample fabrication by focused ion beam milling (a), transmission electron microscopy (b) is used to measure grain boundary orientations in STEM mode (d) and grain orientations by scanning nanobeam diffraction (e). Finally, atom probe tomography (c) is used to reconstruct a 3D atom map (f).

Figure 2

Overlay of projected 3D atom map and bright-field STEM micrograph of a pearlitic steel atom probe specimen. Grain orientations are determined by nanobeam diffraction in the TEM. GB2, a $\mathrm{\Sigma }3$ coherent twin, shows significantly lower carbon segregation than the average grain boundary [see, also, Fig. 4].

Figure 3

(a) Bright-field STEM micrograph of a pearlitic steel atom probe specimen. Red arrows mark dislocation lines along low-angle grain boundaries. (b) Projection of 3D atom map. Red dots mark carbon atoms. Carbides are highlighted by green envelopes (12 at. % isoconcentration surfaces). (c) Overlay of (a) and (b). (d) Magnification of (a). Yellow arrows mark the intersections of grain boundaries with upper/lower sample surface.

Figure 4

Plot of carbon excess in grain boundary space. For reasons of visualization the $n$-dimensional grain boundary segregation space (a) is projected into a 1D plot showing carbon excess, $\mathrm{\Gamma }$, over misorientation angle, $\omega$, for 121 grain boundaries in ferrite (b). Error bars estimated by determining the upper and lower bounds for manual peak fitting of $\mathrm{\Gamma }$ are marked in gray. A linear fit corresponding to the increase in dislocation density was performed for the low-angle regime (red line). The strong spread of data points for $\omega >14°$ is attributed to the large influence of the four crystallographic parameters in the high-angle regime that are not taken into account in this diagram. GB1-3 are the $\mathrm{\Sigma }5$, $\mathrm{\Sigma }3$, and $\mathrm{\Sigma }3$ boundaries, respectively. The difference in carbon segregation between GB2 and 3 can be understood with the help of (c). The high deviation from the ideal $\mathrm{\Sigma }3$ misorientation of GB3 means the presence of misfit dislocations which are responsible for the measured carbon segregation.