The morphology of primary alpha-Be dendrites during the solidification process directly determines the porosity defect and thus the mechanical properties of beryllium-aluminum alloys. However, the features of the threedimensional microstructure remain unclear, and how the structure evolves under the complex multiple physical fields still requires to be explored. Here, X-ray microtomography characterisations were conducted to uncover the three-dimensional dendritic structure of the hexagonal close-packed (HCP) alpha-Be dendrite with 12 preference growth directions. A thermal-fluid-solute-microstructure fully-coupled model was developed to describe dendritic morphology evolutions under the solute transport driven by the natural convection. The results reveal that the Al element significantly enriched at the solid-liquid interface drove an intense downward natural convection, accelerating outward solute transport from the roots of the dendritic branches. Furthermore, due to the in-phase arrangement of growth directions on both sides of {0001}, the unique symmetrical structure of alpha-Be dendrites induces solute distribution and thus the asymmetric melt flow. Such asymmetry inhibits the secondary dendrite arm growth in some directions, ultimately resulting in the hard-to-identify asymmetric threedimensional dendritic structure. Accelerated solute transport also leads to the increase in the secondary dendrite arm spacing and the decrease in the interfacial area density. As the temperature approaching to 1431 K (the temperature at which the liquidus slope approaches to zero), a rapid increase in solid fraction occurs and thereby induces frequent grain coalescences, causing an additional increase in the melt flow velocity by compressing the space and enhancing the solute enrichment.

The structures of metallic melts are of utmost significance for understanding liquid properties, atomic dynamics, and solidification behaviors, including the formation of solidified microstructures. In recent years, with the continuous advancement of experimental techniques and numerical simulation methods, researchers have obtained deeper insights into the microscopic details of the liquid structure, the structure evolution, and the correlations between the liquid structure and the solidified microstructure. This article reviews the main experimental techniques and simulation methods employed in the study of metallic melt structures, as well as the pioneering findings in the field. High-energy synchrotron X-ray diffraction and in situ X-ray imaging techniques and their applications in elucidating the liquid structure and its evolution during solidification are introduced. Special attention is given to the development of synchrotron equipment. Simulation methods for analyzing melt structures include classical molecular dynamics (MD), ab initio molecular dynamics (AIMD), reversed Monte Carlo (RMC), and machine learning (ML), all of which have been applied to predict the atomic structures of metallic melts. The most recent progress in machine learning potentials or force fields is also introduced. The article also discusses future research directions, including the integration of high-resolution imaging with high-energy X-ray diffraction techniques, the application of artificial intelligence-assisted simulations to reduce computational costs, and the investigation of external factors such as pressure and cooling rates on solidification behavior. By combining advanced experimental and computational approaches, research on metallic melt structures is expected to move toward a more comprehensive and in-depth understanding, opening new opportunities for breakthroughs in metallurgy and materials science.

Rare earth elements are commonly incorporated into aluminum alloys to refine grains and enhance mechanical properties through forming the intermetallic compounds. However, the atomic-level mechanism governing the formation of these compounds during solidification remains insufficiently understood. In this work, Al-Cu and Al-Cu-La liquid alloys were investigated using synchrotron high-energy X-ray diffraction, ab initio molecular dynamics and reverse Monte Carlo simulations. The solute-solvent interactions and atomic structure of both liquid alloys were analyzed and compared. The results reveal a strong affinity between Al and La atoms, and minor La addition significantly enhances icosahedral order in the Al-Cu liquid. Furthermore, the local atomic arrangement in the Al-Cu-La liquid closely resembles that of the Al11La3 crystal, providing favorable structural conditions for the nucleation of Al11La3 phase. These findings demonstrate the structural role of La in Al-Cu liquids and provide deeper insights into the formation of Al11La3 intermetallic compound.

The Stokes-Einstein relation (SER) links diffusion and viscosity, but its applicability is challenged in systems with complex structure and chemical properties. Medium-entropy alloys (MEAs), with their inherent chemical complexity and propensity for diverse local ordering, offer a compelling platform to study such coupling phenomena in the liquid state. However, how element-specific structural preferences influence SER breakdown remains unexplored. Here, we investigate the microscopic mechanisms underlying SER breakdown in liquid equiatomic NiCoCr MEA using high-energy x-ray scattering and molecular dynamics simulations. A dynamic transition is observed near 1700 K (melting point 1682.9 K). While the self-diffusion coefficients of all elements retain Arrhenius behavior down to the deeply supercooled regime, the viscosity and structural relaxation times exhibit a clear cross-over between two Arrhenius regimes below this temperature, leading to the breakdown of the inverse scaling between diffusion and viscosity. This decoupling is primarily governed by the anomalous temperature dependence of the viscosity. Fractional SER analysis reveals element-specific decoupling, with Cr showing a fundamentally distinct departure from classical scaling compared to Ni and Co. Concurrently, non-Gaussian parameters reveal growing dynamic heterogeneity upon cooling. Structurally, short-range order strengthens with significant increases in icosahedral-like and mixed clusters, especially those centered on Cr, accompanied by an enhancement of local five-fold symmetry. The formation of these rigid, Cr-centered ordered domains amplifies local geometric constraints, which severely hinder cooperative atomic rearrangements, while leaving single-atom diffusion less affected. These results connect element-specific ordering to the viscosity-driven breakdown of SER in NiCoCr, providing a structural perspective on diffusion-viscosity decoupling in MEA liquids.

Conventional solidification theory primarily describes steady-state growth under a constant temperature gradient. However, in practical casting, the thermal field evolves dynamically due to the complex interactions between latent heat release and non-uniform cooling. In this study, the transient solidification behavior of an Al-20 wt pctCu alloy was investigated during natural cooling with varying initial temperature gradients (G0: 3.7–12.1 K/mm). By coupling microstructural evolution and thermal field variation with in situ synchrotron observation, we quantitatively analyzed the dynamic relationship between the advancing velocity of the liquidus isotherms (VLiquidus) and the solidification front (VSolidification). The results reveal three distinct solidification regimes governed by morphological evolution within the field of view. At the lowest gradient (G0 = 3.7 K/mm), the system exhibits a volumetric solidification characterized by dispersed, synchronous burst-like nucleation. Increasing gradients shift the mode to sequential solidification, where an ordered spatial sequence of nucleation is established. At the highest gradient (G0 = 12.1 K/mm), extreme spatial constraint triggers a morphological transition to elongated dendrites, with the grain elongation factor surging from 1.5 to 5.0. This evolution is driven by non-linear coupling under heat flux constraints. Higher gradients severely restrict VLiquidus, leading to a drastic reduction in the active nucleation zone (LANZ). This microscopic spatial compression limits available space for newly formed nuclei, activating a probabilistic selection mechanism. Under this mechanism, only a small fraction of grains grow preferentially along the steep gradient, evolving into elongated shape through competitive growth. These findings establish a quantitative framework for microstructure control in non-steady-state industrial casting processes.

The structural stability of metallic glasses (MGs), a key requirement for their practical application, can be significantly enhanced by accessing low-energy configurational states. The reentrant glass transition offers a promising route toward achieving such ultrastable glassy configurations. However, the atomic-scale topological mechanisms underlying this anomalous macroscopic behavior remain poorly understood. In the present work, we systematically investigate the composition-dependent transition behaviors of a series of Pd-Ni-P metallic glasses (Pd40Ni40P20, Pd41.25Ni41.25P17.5, Pd42.5Ni42.5P15, and Pd43Ni20Cu27P10) to elucidate the microscopic physical origin of the reentrant phenomenon and its role in the formation of ultrastable glasses. Samples were prepared via arc-melting and characterized through differential scanning calorimetry (DSC) and in situ high-energy X-ray diffraction (HE-XRD) at a synchrotron radiation facility. DSC results show that Pd41.25Ni41.25P17.5 and Pd42.5Ni42.5P15 exhibit distinct anomalous exothermic peaks within the temperature range between the glass transition temperature (Tg) and crystallization temperature (Tx), which are located at 611 K and 601 K, respectively. The associated configurational enthalpy changes account for only 10%-20% of the crystallization enthalpy, thereby ruling out the occurrence of phase separation or macroscopic crystallization. In contrast, no such intermediate events were observed for Pd40Ni40P20 and Pd43Ni20Cu27P10. In situ HE-XRD measurements confirmed the absence of crystallization during these exothermic processes. Instead, profound structural evolution was observed: the intensity of the first peak in the structure factor S(Q) increases by up to 29% (for Pd42.5Ni42.5P15), while its full width at half maximum (FWHM) decreases significantly by 38%, indicating highly enhanced short-to-medium-range order. Furthermore, analysis of the reduced pair distribution function G(r) revealed a dramatic restructuring of atomic cluster connectivity during the transition. Specifically, an increase in 1-atom (vertex-sharing) connections and a concurrent decrease in 2-atom (edge-sharing) connections were confirmed, which point toward the formation of a more ordered medium-range topological network. The key findings of this work are as follows: 1) Composition-specific macroscopic response, the reentrant glass transition exhibits a strong composition dependence, occurring specifically in Pd41.25Ni41.25P17.5 and Pd42.5Ni42.5P15 while remaining absent in Pd40Ni40P20 and Pd43Ni20Cu27P10; 2) microscopic structural origin: the anomalous exotherm fundamentally corresponds to an intrinsic amorphous-to-amorphous polymorphic ordering process within the glassy state; 3) topological restructuring mechanism: the transition is governed by the evolution of atomic cluster connectivity modes, which drives the amorphous network toward an ultrastable low-energy configuration. These findings provide atomic-level insights into the reentrant transition mechanism and offer valuable guidelines for the rational design of metallic glasses with tailored stability.

The substrate temperature plays a pivotal role in regulating the orientation of metal single crystals (SCs), serving as a valuable complement to substrate control strategies. However, due to the interplay of multiple temperaturesensitive factors, including lattice parameters, cooling rate, and undercooling, its regulatory mechanism remains incompletely understood. In this work, sessile drop experiments combined with numerical simulations were employed to investigate the formation of SCs' orientation on substrates at various temperatures, taking Zn droplets solidified on alpha-Al2O3 substrates as a model system. The results indicate that nucleation energy barriers imposed by the Al2O3 substrate constrain the orientations of Zn grains successfully nucleated on the substrate within a specific range: 8 and 9 distinct orientations are formed on the Al2O3 (1010) and (0001) substrates, respectively. Notably, substrate temperature-induced variations in lattice parameters do not significantly alter the nucleation energy barriers of Zn grains, thus failing to change grain orientations. On low-temperature substrates (320 degrees C and 370 degrees C), all orientations within the constrained range are viable for nucleation. In contrast, on high-temperature substrates (420 degrees C, 470 degrees C and 520 degrees C), the undercooling of the Zn melt decreases sharply from 12.89 K to 0.27 K. This weak nucleation driving force is insufficient to overcome the nucleation energy barrier of certain orientations within the constrained range, thereby narrowing the number of viable orientations to 5 and 4. These findings significantly advance the comprehension of how substrate temperature modulates the orientation of metal SCs.

Sliver is a casting defect formed during solidification of nickel-based single-crystal superalloys, which degrades the high temperature properties of superalloys. However, the formation mechanism of sliver cannot be fully deduced from the solidified microstructure alone and thus requires clarification. In this work, a multi-scale model of thermo-mechanical deformation was developed to investigate the formation of sliver. Results reveal that casting contraction-driven dendrite deformation induces significant plastic strain localisation, which is inferred as a key step to dendrite fragmentation and formation of sliver defects. Two distinct dendrite deformation mechanisms during sliver formation are identified: The first mechanism is contraction-driven plastic deformation of dendritic trunks along 45 degrees, and the second mechanism arises from constraint from mould protuberances causing large-angle bending of dendrites. This work sheds light on the cross-scale mechanisms of dendrite deformation and fragmentation for sliver defect formation, providing a basis for defect mitigation.

Accurate prediction of the atomic structure at metal-oxide interfaces is of great significance for developing highperformance semiconductors, catalysis, coating materials, etc. However, the existing models fall short of accurately and rapidly predicting metal-oxide's interfacial defects and crystal orientation. In this study, we propose a novel substrate-directed atomic stacking model (SASM) tailored to metal-oxide interfaces in heterogeneous nucleation, based on the minimum interfacial energy principle. The SASM demonstrates reliable prediction of defects and crystal orientations in the Cu/alpha-Al2O3 system. Specifically, it forecasts dislocation formation at Cu/ alpha-Al2O3 (1011) and (0001) interfaces, while predicting micro-twin occurrence uniquely at the (1120) interface. The predictions align closely with experimental observations. The strain analysis further reveals that the atomic displacement caused by interfacial strain is the primary cause of defect formation: the lateral displacement destroys the continuity of the crystal surface and generates dislocations, whereas the longitudinal displacement disrupts the atomic packing order and results in micro-twins. Compared with the existing models and molecular dynamics simulations, the SASM can extend the applicability and significantly accelerate computational efficiency, laying a foundation for rapidly developing high-performance metal-oxide systems.

Macrosegregation is a critical defect in direct-chill (DC) casting of aluminum alloys, significantly affecting the working performance of the final products. This work employed a diffusion-governed three-phase Eulerian model, integrating melt convection, the growth of columnar trunks, the nucleation, growth, and sedimentation of equiaxed grains, to study macrosegregation patterns in an Al-4.34 wt.% Cu billet and to quantify how casting speed controls the transition of central macrosegregation via competition between thermal-solutal convection and grain sedimentation. The simulations identified a critical casting speed (ti 42 mm & sdot;min-1) governing the transition between positive and negative central macrosegregation. Below this threshold, shallow sump conditions suppress melt flow intensity (maximum liquid speed ti 8 mm & sdot;s-1 at 35 mm & sdot;min-1), allowing buoyancy-driven accumulation of solute-enriched melt near the sump bottom and generating central positive segregation. Above this threshold, deeper sumps with intensified convection (maximum liquid speed ti15 mm & sdot;s-1 at 60 mm & sdot;min-1) enhance solute dilution and promote equiaxed grain settling, establishing negative segregation. A comparison between 2D-axisymmetric and 3D simulations shows that the computationally efficient 2D simulation can capture consistent overall macrosegregation trends, but it systematically overestimates the central segregation due to the omission of tangential flow.

Given the limitations of current mainstream laser-based additive manufacturing (AM) technologies, metal jet printing is regarded as a promising low-cost AM technology in the printing of metals with high laser reflectivity. However, the full understanding of its capacity to produce copper alloy parts with a refined microstructure with uniform chemical composition remains elusive. A numerical model is developed to uncover the formation mechanism of the printed part shape and the evolution of solute elements and microstructure. The simulated and experimental results consistently indicate that the solute-diluted zones are located at the junctions and at the center of the droplets. The former is attributed to the flow-induced solute dilution in the solute partition after remelting, while the latter corresponds to a gradual weakening of the solute transport by continuous momentum decline of droplet oscillations. The tracked thermal history during the remelting and solidification process governs the equiaxed-columnar-equiaxed transition occurring in each droplet with the reduced morphology factor (from 107 to 105 (K s)/m2 ) at the solid-liquid interface. Meanwhile, the reduced cooling rate (from 104 to 103 K/s) contributes to the increase in grain size from bottom to top within each droplet. Despite the requirement of metallurgical bonding, excessive heat accumulation also contributes to the poor resolution and solute segregation. The understanding of the mechanisms in the shape morphology, microstructure, and solute transport provides insights for the property control and further advancing applications of the metal jet printing process. (c) 2026 Published by Elsevier Ltd on behalf of The editorial office of Journal of Materials Science & Technology.

In-situ alloying is a promising material processing method for the direct fabrication of alloys with arbitrary elemental content. In-situ alloying of biomedical titanium-niobium alloys using laser powder bed fusion (LPBF) necessitates a fundamental understanding of the underlying physics governing solute transport, which remains a challenge in addressing the 'porosity-segregation dilemma'. This work developed a thermal-fluid-solute model integrated with actual solidification paths for the LPBF process to advance the in-depth understanding of melt pool dynamics, solute mixing, and elemental segregation mechanisms during in-situ alloying of Ti-Nb alloys. It reveals the influence of solute content and energy density on the melt pool dynamics. Higher Nb fraction up to 41 wt.% results in a deeper and shorter melt pool with a stronger flow, and higher energy density leads to a larger melt pool with a liquid metal flow circuit. Moreover, Nb-rich bands parallel to the melt pool boundaries are formed due to solute partition and the strong flow through the mushy zone. Additionally, Nb-rich ridges are formed between tracks due to insufficient heat at the track edges and the fluid buildup effect. During the LPBF of the second layer, a 'wide-narrow-wide' melt pool is formed with vertical stratification of Nb-rich and Nb-lean.

Recent studies have established a research paradigm for decoupling thermoelectric parameters through Hf doping in half-Heusler (HH) compounds with high structural symmetry. However, the high cost of Hf significantly hinders commercialization, thereby driving the urgent need to develop low-cost and efficient alternatives for performance regulation. Here, we propose a simple and effective modulation strategy, tailoring Zr vacancies to achieve a notable thermoelectric performance for Hf-free ZrCoSb-based HH compounds. We demonstrate the multifunctional roles of Zr vacancies in regulating power factor and thermal conductivity, including local electronic states, in-gap states, and vacancy collapse. As a result, a high figure of merit of similar to 0.81 is achieved in Zr0.9CoSb0.8Sn0.2 at 973 K by tailoring Zr vacancies, which is nearly the highest value reported for Hf-free ZrCoSb-based HH compounds. This work not only presents a promising pathway to optimize thermoelectric performance in Hf-free HH compounds but also provides new insights into vacancy engineering for HH compounds.

The orientation relationship (OR) at the interface is a critical feature for heterogeneous nucleation, typically characterized by parallel lattice planes and directions. However, under conditions of high-misfit OR, nucleated crystals often exhibit distinct tilting behavior. This study systematically investigates the ORs of Al/Al2O3 heterogeneous nucleation interfaces under high misfit conditions using the “Equator Mismatch Figure”, which enables comprehensive 3D characterization of ORs by projecting extensive lattice planes and directions from both phases into a single map. A serious of nucleation experiments were conducted on the identical substrate orientations, revealing three types of ORs: parallel matching, simple tilting, and complex tilting. The parallel matching corresponds to the conventional description and allows for the calculation of the lattice misfit through the classical 2D misfit equation. In contrast, neither type of tilting matching permits an evaluation of 2D misfit due to the fact that Al crystals in tilting configurations do not terminate with low-index planes at the interface and may lack low-index directions within the interface plane-exceeding the description of classical theories that necessitate parallel lattice planes and directions. These findings illustrate that heterogeneous nucleation interfaces can adopt various mismatch pathways even on the substrates with identical orientations, which is attributed to the energy contributed from substantial melt superheating. The misfits of parallel matching align with previous studies, while those related to tilting cannot be evaluate their misfits. This work underscores the necessity for further development of interfacial mismatch theories in heterogeneous nucleation to elucidate the mechanisms underlying such complex tilting ORs.

Nickel-based single-crystal (SX) superalloys are the key metallic materials of aeroengines. However, thermomechanical deformation always occurs during the directional solidification of SX superalloys, negatively influencing the SX structure. Casting deformation is simulated in most of the previous studies, whereas the direct simulation of dendritic thermomechanical deformation has been largely ignored, resulting in a lack of comprehensive understanding of this process. In this study, we systematically investigate dendritic thermomechanical deformation with a model coupled with dendrite growth, fluid flow, and thermomechanical deformation behavior. Results reveal that the dendritic thermomechanical deformation-induced dendrite bending is not randomly distributed but is mainly concentrated on the casting surface. The dendritic thermal stress increases as dendrite grows and accumulates after dendrite bridging. Transverse thermal contraction mainly occurs at the edge of casting in the corner, and axial thermal contraction is larger than transverse contraction. The high-stress region of the primary dendrite trunk is mainly distributed below the dendrite bridging near the solidified part, and the stress along the transverse direction reaches its maximum value on the casting surface. Stress concentrated on the casting surface is mainly attributed to variations in transverse temperature gradients caused by heat dissipation on the lateral mold wall, and inconsistent constraints in the lateral mold walls.

AgSbSe2 is regarded as a promising p-type I-V-VI2 thermoelectric material owing to the intrinsically low thermal conductivity and high Seebeck coefficient. However, the intrinsic low electrical conductivity impedes the further enhancement of the thermoelectric performance of AgSbSe2. Here, a novel approach is initiated to enhance the thermoelectric properties of AgSbSe2 by combining atomic off-centering with grain boundary engineering. This work simultaneously promotes the grain growth and amplifies off-centering behavior for Ag1+ ySb1-x-yBixSe2 by the precise adjustment of the Ag/Sb ratio based on Bi-doping. The enlarged grain induces the increasing room-temperature carrier mobility from 2.49 cm2 V-1 s-1 for pristine AgSbSe2 to 11.16 cm2 V-1 s-1 for Ag1.01Sb0.9Bi0.09Se2, and a high power factor of ≈7.2 µW cm-1 K-2 is achieved in Ag1.01Sb0.9Bi0.09Se2. Simultaneously, the amplified localized off-centering behavior drives a low lattice thermal conductivity of ≈0.37 W m-1 K-1 for Ag1.01Sb0.9Bi0.09Se2 at 673 K, representing a 20% reduction in lattice thermal conductivity than that of pristine AgSbSe2. As a result, Ag1.01Sb0.9Bi0.09Se2 obtains an excellent figure-of-merit zT of ≈1.11 at 673 K. The synergistic optimization of cation modulation simultaneously promotes the grain growth and amplifies off-centering behavior, which provides a new optimization paradigm for designing high-performance AgSbSe2 thermoelectric materials.

In the face of the contradiction arising from the insatiable demand for natural resources and their limitations, a new mode of resource circulation is urgently needed. The "urban mines" formed by the accumulation of scrap metal are growing, and the environmental, energy, and economic benefits of reusing them far exceed those of raw ore. The traditional recycling process is too passive and focuses on end-of-life treatment after scrap generation, neglecting front-end information and resulting in low-value application of recycled materials and resource wastage. Herein, the "Metal 2.0" mode is proposed. This mode advocates the active control of metal circulation at every stage. It encompasses source design to ensure easy recyclability and outstanding service performance, process repair for damaged metal components, and categorized end-treatment of scrap metal with impurities. The barriers limiting the current circular application of metals and the key breakthrough directions of Metal 2.0 are pointed out. Potential challenges are also discussed. Through the application and promotion of the Metal 2.0 mode, a circular scheme with economic and environmental benefits is expected, fostering the harmonious coexistence of humans and nature and contributing to global sustainable development.

Slag entrainment in the continuous casting mold has always been the focus of surface quality control of the slab for the steel strips. However, the invisible and nonlinear molten steel flow that determines the defect formation make online measurement and evaluation particularly difficult. We developed a fast prediction method visualising the characteristics of the melt flow and estimating the slag entrainment online by combining numerical simulation and machine learning. The data-driven surrogate model was constructed using the proper orthogonal decomposition method and a fully connected neural network based on the numerical simulation data. The model showed a high hit rate of over 91% with an absolute error of less than 0.05, and exhibited millisecond-scale real-time responsiveness. With the ultra-high computational efficiency, a high-resolution parameter-defect index map was established to clarify the effect of the operating parameters at multiple dimensions, and to locate the low-risk range, the argon flow rate within 3–8 L/min, the casting speed within 1.0–1.7 m/min, and the nozzle immersion depth within 180–220 mm. This approach provides a promising technical route for the design of an efficient online slag entrainment monitoring system.

VO2/Al composites were fabricated via spark plasma sintering (SPS) under diverse temperatures and pressures for the purpose of tuning phase transition behaviours. The experimental results suggest that appropriately lowering the sintering temperature is favourable to generate pure VO2/Al composites, on the basis of which the pressure is increased to make the material dense. At the same time, it has been found that high pressures in the sintering process have a negative impact on the latent heat of the composites. This adverse effect is attributed to the introduction of residual stress and the formation of oxygen vacancies, which are recognized as destabilizing the insulating phase of VO2 and suppressing the sharpness of the phase transition. Nevertheless, the bulk VO2/Al composites still possess tunable emissivity properties. At the actual temperature of 105 degrees C, the composite sintered at 500 MPa has a thermal radiation temperature as low as 46.6 degrees C, which provides good infrared stealth performance. The findings provide crucial insights for optimizing the sintering conditions to balance the density and phase transition performance of VO2/Al composites, which are pertinent to their application in adaptive infrared stealth and other temperature-sensitive applications.

Low-angle grain boundaries (LAGBs) are one of the solidification defects in single-crystal nickel-based superalloys and are detrimental to the mechanical properties. The formation of LAGBs is related to dendrite deformation, while the mechanism has not been fully understood at the mesoscale. In this work, a model coupling dendrite growth, thermal-solutal-fluid flow, thermal stress and flow-induced dendrite deformation via cellular automaton-finite volume method and finite element method is developed to study the formation of LAGBs in single crystal superalloys. Results reveal that the bending of dendrites is primarily attributed to the thermal-solutal convection-induced dendrite deformation. The mechanical stress of dendrite deformation develops and stabilises as solidification proceeds. As the width of the mushy zone gets stable, stresses are built up and then dendritic elastoplastic bending occurs at some thin primary dendrites with the wider inter-dendritic space. There are three characteristic zones of stress distribution along the solidification direction: (i) no stress concentration in the fully solidified regions; (ii) stress developing in the primary dendrite bridging region, and (iii) stress decrease in the inter-dendritic uncontacted zone. The stresses reach maximum near the initial dendrite bridging position. The lower temperature gradients, the finer primary dendritic trunks and sudden reductions in local dendritic trunk radius jointly promote the elastoplastic deformation of the dendrites. Corresponding measures are suggested to reduce LAGBs.