The Investigations On The Correlation Between The Intrinsic Properties,Microscopic Structures And Glass-forming Ability In Metallic Glasses

The glass-forming ability(GFA)and macroscopic properties of metallic glasses(MGs)are determined by their microscopic atomic-level structures.Although it has been known that the structure of MGs has the characteristics of showing order in the short range but disorder in the long range,the geometrical packing of short-range order(SRO)and medium-range order(MRO)is still unclear.As a result,the structure of MGs and the relationship between the structure and GFA as well as properties need further to be clarified.In addition,obtaining MGs from metallic liquids by rapid quenching involves the phase transition process and the in-depth and comprehensive study on this process can provide necessary theoretical guidance on the design and further fabricating of bulk MGs.However,the investigations on the change of the liquid miscroscopic structure involved in the glass-forming process are still lacking in the scientific community.Consequently,the structural mechanism of the significant change in dynamic properties during the phase transition process in MGs is in urgent need to be revealed.Based on this,this thesis utilized the conventional data analysis,classical molecular dynamics(MD)simulations,ab initio molecular dynamics(AIMD)simulations,and machine learning methods based on neural networks to investigate the microscopic atomic-level structures,GFA and glass transition process in Cu-Zr,Ni-Zr,Ni-Nb,and Pd-Si systems.The results clarified the influence of the microscopic atomic-level structures on the dynamic properties and GFA,established the correlation between the dynamic properties and microscopic atomic-level structures,explored the relationship between the crystal growth rate and GFA,and developed the neural network potential(NNP)for use in the structural analysis of MGs and materials discovery of crystal structures.The main results are as follows:(1)A new index to effectively indicate the GFA was proposed.By starting from the classical nucleation rate and crystal growth rate equations as well as the experimental data for various types of MGs,a new index to effectively characterize the GFA was proposed by using the methods of theoretical derivation and data analysis:G(0.15)=(T_g-T₀)T_l²/(T_g(T_l-T₀)²)(T_x/(T_l-T_x))^0.15,T_g is the glass transition temperature,T₀ is the ideal glass transition temperature,T_l is the liquidus temperature,and T_x is the onset crystallization temperature.For23 various types of MGs,the new index can well represent and predict the GFA.Through theoretical analyses,the new index correlates with the viscosity at the liquidus temperature,which provides effective theoretical guidance on developing bulk MGs in experiments.(2)The correlation between the microscopic structures and the dynamic properties as well as the atomic energy was revealed.For Cu_64.5Zr_35.5 MG system,the rapid increase of the dynamic relaxation time and dynamic heterogeneity at 1100 K is resulted from the rapid increase of the fraction of the SROs.In addition,when the atomic clusters are more similar to the SROs,the energy and dynamic displacement of the center atom decrease linearly.For Cu-Zr-Al MG system,when the MROs formed by the SROs connect more firmly with each other,the dynamics in MROs is in close relation with the dynamics slowdown required by the glass-forming process.(3)Novel SROs in Ni-Zr MG system were determined and the factors contributing to its marginal GFA were revealed.For Ni₅₀Zr₅₀ MG,the major SROs are"Z11","mixed",and"Ni-B33"motifs.Among them,the"Z11"and"mixed"motifs are composed of the B2 phase and icosahedron(ICO)structures,while the chemical order of"Ni-B33"motif is completely the same as that of the B33 crystal structure.For Ni_64.5Zr_35.5 MG,the major SROs are"mixed","intertwined-cube"and"icosahedron-like"motifs.Among them,the"intertwined-cube"motif shows the similar chemical order as that in the crystal structures.In this case,the low fraction of ICO and the existed partial crystal structures contribute to the marginal GFA in Ni-Zr MG system.In addition,in the MD simulation process,the competition between the major motifs in glass("Z11","mixed",and"icosahedron-like"motifs)and the motifs in crystal structures("Ni-B33"and"intertwined-cube"motifs)was observed and the total fraction of the major SROs increased during the annealing process.(4)The SROs and their packing characteristics in different compositions of Ni-Nb MG system were clarified and the reasons for the best GFA around the eutectic point were revealed.In the Ni₆₂Nb₃₈ MG system,Ni₄₈Nb₅₂,Ni_59.5Nb_40.5,and Ni₇₅Nb₂₅ metallic liquids,the major SROs centered on Ni are the distorted ICO and ICO structures,while the major SROs centered on Nb are the Frank-kasper Z14,Z15,and Z16 motifs.Among them,the chemical order of the distorted positions in distorted ICO is quite different from that at other positions and the spatial packing efficiency of distorted ICO is lower than that of ICO.In addition,the GFA of Ni-Nb system is simultaneously determined by the fraction of SROs,the degree of five-fold symmetry and the networks of the MROs.(5)A novel interatomic potential was developed to investigate the solidification process in Ni-Nb system.By including the formation energies of Ni-Nb compounds in the potential fitting process,a novel interatomic potential which can well describe the energies of the Ni-Nb crystal structures was developed.In addition,with the help of the newly developed potential,the solidification processes of Ni₃Nb and Ni₆Nb₇ phases were studied and the good GFA of Ni₆₂Nb₃₈ MG was explained from the perspective of the slow crystal growth.(6)A novel NNP for Pd-Si system was developed by the machine learning method.By considering the ab initio density functional theory energies and atomic forces of the liquids and crystals in the potential fitting process,a novel NNP for Pd-Si system was developed by the machine learning neural-network method.The novel NNP can well describe the structures of both liquids and crystals,the energy orders of the crystal structures,and the morphology and migration of the solid-liquid interface.The results demonstrated the applications of machine learning methods in the potential fitting process and clarified the reasons for the good GFA of Pd₈₂Si₁₈ system.