Investigation Of Stress Relaxation And Correlation Of Structure And Rearrangement In Metallic Glasses

Metallic glass?MG?,a new amorphous material,has many novel mechanical properties such as high elastic limit and strength and so on compared to the corresponding crystalline materials.These features give metallic glasses a broad future of application.However,under ambient temperature,the deformation is mainly localized into thin shear bands,thus metallic glasses lack macroscopic tensile ductility,which are dreadfully limited in the engineering application.Consequently,a deeply understanding of the deformation mechanism of metallic glasses is needed.Due to the nature of disordered structure,the mechanism like dislocation mediated plasticity of crystalline materials is not working in disordered systems,the deformation mechanism of transformation events of disordered systems is still unclear until now.Therefore,MGs offer a model system for studying these fundamental issues.Based on the molecular dynamics simulations,we investigate the correlation between the activation of transformation events and macroscopic physical quantities under the stress relaxation processes.Furthermore,we study the correlation between rearrangements and the local structural parameter under quiescent system and AQS deformation by machine-learning methods.It is hard to identify the deformation events of MG in static state,we impose a constant strain on Zr₅₀Cu₅₀0 metallic glass?MG?to investigate the signature and movement behavior of deformation events.The stress relaxation processes of strained MG are investigated by molecular dynamics simulations.We provide the direct evidence for the strain-accelerated relaxation which can be attributed to the activation of deformation events as flow units in MG.Moreover,we observe that stress attenuation degree and relaxation time reach saturation corresponding to a critical applied strain,which is related to the crossover from stochastic activation to percolation of flow units.In the quiescent systems,we successfully apply machine-learning methods on local structure to identify liquid-like particles prone to structural rearrangements in simulated Zr₅₀Cu₅₀ metallic glassy systems both above and below its glass transition temperature.We find that the structural differences between liquid-like particles and the rest of the system are beyond short-range order.We further redefine the structural entropy by manipulating the cutoff distance,and confirm that this structural indicator is reasonable only if it is defined beyond the first nearest neighbor.Our results show it is possible to identify subtle structural features responsible for structural rearrangements of metallic glassy systems and the structural descriptors to predict rearrangements should be composed of atoms beyond short-range order.Under AQS deformation,we use machine learning method to quantify the prediction efficiency of local structural parameters to rearrangements to determine the reliability of local structural properties for predicting rearrangements.We find that such structural descriptors containing dynamic information as participation fraction and vibrational MSD give the most predictive information of rearrangements.And machine-learned structural quantity?softness?,a pure structural quantity,outperforms most structural indicators.