Visualization and validation of twin nucleation and early-stage growth in magnesium

Magnesium, a lightweight HCP metal, relies on deformation twinning due to limited easy slip systems, but twinning induces plastic anisotropy and low ductility. A central unresolved question is how {10-12}<10-11> deformation twins nucleate and grow at early stages: via conventional shear-shuffle mechanisms (twinning dislocations gliding on the twin plane) or via a pure-shuffle mechanism involving prismatic-to-basal (P→B) and basal-to-prismatic (B→P) transformations. Direct visualization has been challenging because twin nucleation is localized, ultrafast, and stochastic. This study aims to isolate and observe twin nucleation and early-stage growth in Mg by engineering stress fields through pillar geometry, and to validate the operative mechanism using in-situ TEM, finite element analysis, and atomistic simulations.

Prior work has extensively investigated conditions and microstructural and alloying effects governing twinning in Mg and other HCP metals. The community has debated whether {10-12} twinning nucleates via shear-shuffle (involving twinning dislocations/disconnections on the twin plane) or via a pure-shuffle, phase-transformation-like process. In-situ TEM on conventional pillars typically captures only rapidly formed, mature twins, with twinning events completing in ~0.04 s, precluding observation of nucleation and earliest growth. Twin formation near stress concentrators (e.g., nanoindentation) has been predicted and simulated, but the sequence from embryo to visible twin in Mg has not been experimentally tracked. Atomistic simulations and theoretical models have proposed roles for PB/BP interfaces and pure-shuffle mechanisms, but experimental confirmation of nucleation pathways has been lacking.

- Specimens: Single-crystal Mg (0001) (99.999%) pillars (750 nm thickness, 1200 nm height) prepared via FIB from a 3 mm diameter crystal. Multiple geometries: conventional rectangular pillar (750 nm width) and truncated wedge-shaped pillars (TWP) with top widths of 400, 250, and 100 nm, all with 45° inclined sides and 750 nm bottom width. - FIB preparation minimized damage using cross-section cutting and 2 keV low-energy cleaning; affected layer limited to a few nm. Post-mortem thinning by FIB and Nano Mill for atomic-resolution TEM. - In-situ TEM compression: Hysitron PicoIndenter (P195) inside JEOL 3010FEG (300 kV) and a double-aberration-corrected Titan with K2-IS camera. A 1 µm-wide flat compression platen contacted pillar tops; loading direction parallel to basal plane. Indentation speeds: 0.5 nm/s (rectangular, 400 nm, 250 nm TWPs); 0.1 nm/s (100 nm TWP) to capture early events. Alignment ensured to avoid misalignment artifacts. Dark-field and bright-field TEM used to track twin boundary/tip motion. Selected area diffraction and HRTEM characterized boundary structures. - Finite Element Analysis (Abaqus/CAE): Modeled stress fields under axial compression for each geometry; Mg anisotropic elastic properties with axes x:[0001], y:[10-10], z:[1-210]. Applied 5 nm displacement to top, bottom fixed. Extracted normal stresses (σx, σy) and shear on 45° planes corresponding to twin planes. Normalized experimental stress differences between pillar top and bottom were also computed from load divided by respective cross-sections to quantify stress localization. - Atomistic simulations (LAMMPS): Embedded-atom method potential for Mg (Liu et al.) chosen for accurate surface/interface energetics. Quasi-2D TWP model (top width 12 nm, bottom 40 nm, height 30 nm), x:[0001], y:[10-10], z:[1-210]; periodic in z (3.0 nm), bottom 1 nm fixed. Indentation-like compression applied to top at 10 m/s at 300 K. Tracked twin embryo formation and growth, PB/BP interfaces, and emission of basal <a> dislocations. - Interface energetics and mechanics analysis: Molecular statics calculated formation energies of coherent twin boundary (CTB) vs coherent PB (CPB) and coherent BP (CBP) interfaces, finding lower energies for CPB/CBP. A simplified mechanics model compared work to form nuclei bounded by CPB/CBP (pure-shuffle) vs CTB (shear-shuffle), as a function of twin dimension L, using Mg anisotropic stiffness constants.

- Geometric stress localization enables isolation of nucleation and early-stage twin growth. FEA and experimental normalization show increasing stress concentration at the pillar top as TWP top width decreases (400 → 250 → 100 nm). - Conventional pillars (750 nm width, near-uniform stress): Twins appear abruptly and propagate rapidly across the pillar; smallest captured twin at 4.85% strain was ~310 nm (equivalent diameter), indicating only mature twins are observed. Growth proceeds initially by fast shearing along the {10-12} twinning plane, followed by thickening along the basal plane; boundaries show serrated PB/BP terraces. - TWP 400 nm top: Two twins nucleate from top corners (stress concentration points) and rapidly shear along twin planes to intersect; initial sizes ~200 nm at 1.7% strain; co-zone twins subsequently thicken via basal–prismatic transformation. - TWP 250 nm top: A triangular twin nucleus first identified at 0.7% strain with size ~90 nm. Early-stage growth occurs by boundary migration along the basal plane (compression direction), not by lateral shearing along the twinning plane. Final twin (~250 nm) shows primary boundaries parallel to the matrix basal plane; HRTEM reveals BP and PB terraced interfaces. - TWP 100 nm top: Twin nucleation is directly visualized from the top contact surface at only 0.2% strain; initial triangular embryo ~20 nm grows to ~90 nm. Diffraction and HRTEM confirm boundaries comprised of BP and PB interfaces; no coherent {10-12} twin boundary present at this stage. The measured stress gap at the nucleation point was ~320 MPa (top–bottom difference). For the 250 nm TWP, the stress gap at nucleation was ~104 MPa. - Pure-shuffle nucleation validated: All observed embryos are bounded by PB/BP interfaces, indicating nucleation via P→B and B→P transformations without net shear along the twinning plane. MD reproduces embryo formation as few-layer PB/BP-faceted nuclei at corners, layer-by-layer growth via PB/BP migration, coalescence into larger triangle-like embryos, and emission of basal <a> dislocations from PB interfaces that provide misfit dislocations to relax coherency strains. - Energetics and mechanics: Molecular statics shows CPB/CBP interfaces have lower formation energy (~105 mJ/m²) than CTB (~125 mJ/m²), favoring PB/BP-bounded embryos. Mechanics analysis indicates nuclei bounded by CPB/CBP require less work than CTB-bounded (shear) nuclei for twin dimensions L < ~2.5 nm. - Transition to mature twins: As embryos grow, misfit dislocations accumulate and pin PB/BP facets, making slower-growing coherent {10-12} twin planes bound the twin, evolving toward lenticular, mature twins with CTBs and the classic 86.22° twin/matrix orientation relation. - Plastic flow stability: Geometry-controlled early-stage growth via slower PB/BP interface migration suppresses strain bursts: conventional pillars show abrupt bursts >1.2%, 400 nm TWP ~0.4%, 250/100 nm TWPs ~0.2% at nucleation. This suggests grain-geometry design can mitigate twinning-induced plastic instability in Mg.

The study resolves a longstanding controversy by directly visualizing that {10-12} twin nucleation in Mg proceeds via a pure-shuffle mechanism involving prismatic–basal and basal–prismatic transformations, rather than initial shear along the twin plane. By engineering stress gradients with truncated wedge geometries, nucleation locations and early growth were confined and observable. The observed PB/BP-faceted embryos and their growth along the basal plane address how twins first form and advance before coherent twin boundaries develop. Atomistic simulations corroborate experimental observations, reproducing embryo formation, PB/BP facet migration, and basal <a> dislocation emission that transforms initially coherent PB/BP interfaces into semi-coherent ones, enabling continued growth. As embryos expand, the kinetics and pinning of PB/BP facets favor subsequent development of coherent {10-12} twin planes, explaining the transition to mature twins commonly observed post-mortem. The geometry-dependent growth pathway impacts macroscopic plasticity: confining early-stage growth to slower PB/BP migration reduces strain bursts and stabilizes plastic flow. These insights are relevant for designing microstructures (e.g., grain shapes/stress concentrators) to control twinning activity in Mg and, by extension, other HCP metals.

By combining in-situ TEM compression on geometry-tailored Mg pillars, finite element stress analysis, and atomistic simulations, the work demonstrates that {10-12} twin nucleation occurs via a pure-shuffle mechanism requiring prismatic–basal and basal–prismatic transformations. Early-stage growth is geometry dependent: conventional pillars enable rapid glide-shuffle along the twin plane, while truncated wedge geometries confine growth to slower PB/BP interface migration along the basal plane, reducing plastic instability. Embryos are PB/BP-faceted without coherent twin boundaries; with growth, misfit dislocations pin these facets and coherent twin planes emerge, producing mature twins. These findings provide a mechanistic foundation for microstructure and geometry design to tune twinning and plasticity in Mg and other HCP metals. Future work could explore translation to polycrystalline Mg via grain shape engineering, interactions with alloying and precipitates, and real-time observation of transitions from PB/BP-faceted embryos to CTB-bounded twins under varied loading paths and temperatures.

- Atomic-resolution dynamics of PB/BP transformations during in-situ loading could not be directly recorded due to the required specimen thickness for twin activity; atomic structures were characterized post-mortem after thinning. - Sample size and geometry effects: Nanoscale pillars (<~100 nm thick) favored dislocation plasticity over twinning; study focused on submicron (750 nm thick) pillars to mitigate size effects, but constraints differ from bulk. - MD simulations used quasi-2D periodic boundaries and high loading rates typical for MD; while capturing key mechanisms, absolute timescales and some kinetics may differ from experiments. - Interface energy and mechanics analyses employed simplified models (e.g., neglected corner/excess energies, used specific interatomic potential); quantitative values may vary with potential and conditions, though qualitative trends are robust. - FIB preparation, though minimized, may introduce residual damage or stresses; alignment and cleaning steps were used to mitigate artifacts.