The ubiquitous defects in crystalline solids play a crucial role in determining their behavior. These defects can be categorized according to their topological dimensionality from 0D, namely point defects such as vacancies, 1D linear defects (dislocations, triple lines), to 2D surfaces, grain boundaries, or stacking faults. Dislocations are essential carriers of plastic deformations and can create or annihilate additional defects when moving, thus mediating the macroscopic mechanical response. Impurities and alloying elements often segregate to these defects, altering the defects’ energetic stability, geometric structure, and mobility. Understanding the interactions between alloying elements and crystal defects is crucial for property engineering. Atom probe tomography (APT) is a characterization routine capable of discerning these interactions.
Due to the limited spatial resolution of APT, crystallographic defects cannot be directly extracted from atomic positions in the APT reconstructed data. Solute decorated defects, however, are routinely analyzed in APT owing to its high chemical resolution. Dislocations in the reconstructed data from APT can have linear morphology (elongated in one direction or along an arc) or can appear as loops [1]. A typical APT analysis of dislocations comprises calculating 1D composition profiles along and perpendicular to the dislocation by manually placing primitives of, for example, cylindrical shapes [2]. Further local proxigrams are also used to quantify radial composition profiles from the dislocation core [3]. If dislocations are curved or form complex networks, as seen frequently in materials, such manual analysis becomes tedious and cumbersome.
In this paper, we introduce a novel workflow for automatically extracting and analyzing line-like microstructural features, particularly dislocations, in APT reconstructed data, significantly enhancing existing methodologies. The workflow comprises the following key steps: extracting and filtering iso-composition surface (iso-surface) meshes delineating dislocations using principal component analysis (PCA), computing mesh skeletons through mean curvature flow (MCF) algorithm [4], isolating dislocation segments, fitting B-splines for precise representation of each dislocation segment and placing regions of interest (ROIs) on the splines for automated composition analysis. Additionally, it integrates crystallographic information from APT data for orientation analysis of dislocations in the crystal coordinate system. This kind of analysis is traditionally performed manually; for example, Dubosq et al. [5] had to manually find out the crystallographic planes/directions and dislocation line direction in the APT data to find the orientation of the dislocation in the crystal coordinate system. However, the developed workflow completely automates these various cumbersome steps.
The workflow is applied to isolate and analyze individual dislocations from a dislocation network in a Ni-based superalloy. Fig. 1(A) shows 6 at. % Cr iso-surface in a Ni-based single crystal superalloy [6]. A dislocation network extends from the γ phase into a γ' precipitate. MCF curve skeleton nodes for the iso-surface mesh are shown in Fig. 1(B). Using the segmentation scheme based on the nearest neighbor analysis, skeleton nodes are segmented into individual linear segments as shown in Fig. 1(B). Then based on the underlying segmented skeleton and the local morphology, linear features are segmented from the iso-surface mesh, shown in Fig. 1(C). For each segmented linear mesh, ROIs are defined using the underlying skeleton, shown in Figure 2(A), and a detailed composition analysis is performed, for example, as shown in Fig. 2 (B and C). Furthermore, the workflow incorporates crystallographic data from APT to ascertain the orientations of dislocations within the crystal coordinate system. This aspect of the workflow is demonstrated through the application on an APT dataset of a Fe-Mn alloy. Both case studies showed the robustness and automation features of the developed pipeline while dealing with complex linear features. By enabling automatic segmentation and composition analysis, the workflow has the potential to provide insights into the role of dislocations in affecting the properties of a material. This approach not only decreases the human effort but also opens new avenues for high throughput study analyzing property-structure relationships [7].
(A) 6 at. % Cr iso-surface mesh in a Ni-based superalloy. (B) Segmented MCF skeleton for the mesh. (C) Morphology-based segmentation of the mesh to extract linear features.
(A) Segmented mesh corresponding to a dislocation with underlying skeleton. Arrows correspond to the cylindrical ROIs perpendicular to the dislocation. (B, C) 1D composition profiles for Cr along the red and green cylindrical ROIs.
