Mechanics Of Quantum Mechanical Behaviors And Machine Learning Model For Predicting High-entropy Alloy

Through the ages,many classical theories have been developed to describe the macroscopic objects'mechanical behaviors e.g.motion,force,deformation,friction,wear and manufacture.But,with the resolutions of characterization techniques increasing,it's found that the mechanical behaviors of materials are strongly depended on the motion of electrons in many situations,which must be considered through quantum mechanics.When applied external physical fields e.g.stress,electric or magnetic field,the behaviors of electrons moving in the local fields will change,affecting materials'mechanical behaviors and leading to macroscopically measurable multi-field coupling phenomena.Especially when a certain dimension of a material goes down to nanoscale,the multi-field coupling effects are more significant.The discovery of graphene has aroused worldwide research interests in two-dimensional materials.In recent years,the family of two-dimensional materials is expanding and contains many novel mechanical,electrical,magnetic and optical properties,which provides an idea platform for studying multi-field coupling phenomena in nanoscale and for developing functional devices.High-entropy alloys,a unique class of multi-principal element alloys,show many unusual mechanical and electronic properties,attributed to the single-phase microstructures,including high strength,high fracture toughness,resistance to oxidation,corrosion and high temperature,thermoelectricity,magnetism and superconductivity.Through first-principles calculations based on density functional theory,classic equations of state for solids,classic dynamic equations and machine learning models,we investigate the charge-doping induced electromechanical coupling behaviors,spin friction between layers of two-dimensional magnetic materials,phase classifications of multi-principal element alloys and identification of high-entropy alloys.Our findings are briefly concluded as below.(1)Equation of state for charge doping induced deformation and hardening in cubic crystals.Charge doping has been extensively applied in electronic and optoelectronic devices to modulate the band alignment,electrical resistance and magnetic properties of semiconductor.But deformation is always concurrent with charge doping and seriously influences the performance and safety of devices,while difficult to be predicted.By combing the concept of quantum electronic stress with Birch-Murnaghan equation of state and the modified Tait equation,we present a set of equations of state to describe charge-doping induced deformation and hardening in cubic crystals.The parameters used in the equations of state are basic physical quantities of materials like deformation potential,bulk modulus and lattice constant,which can be derived experimentally or theoretically without charge doping.Through density functional calculations,the equations are proved to be efficient for a variety of cubic crystals(Si,Ge,diamond,Ga As,Al and Zr O₂).Besides,.the equations can be used to consider the effects of hydrostatic pressure and charge doping simultaneously,allowing for the direct assessment of the deformation in the applications of electronic and optoelectronic devices.(2)Spin degree dominant interlayer friction and stick-slip behaviors in two-dimensional antiferromagnetic crystals.Using Mn₂C,a two-dimensional antiferromagnetic crystal,as a prototype,we show spin-dominant frictional behaviors by comprehensive first-principles calculations and Ising model:while the frictional force in spin-unpolarized state is isotropic and independent with forward and backward moving directions,the interlayer exchange interactions between the antiferromagnetic orderings not only reduces the rotational symmetries of potential energy surfaces,but also induces anisotropic,nonsinusoidal and even asymmetrical energy landscapes on the energetically preferred sliding paths,leading to anisotropic and even direction-dependent frictional force.Employing the Tomlinson model,we also find a novel stick-slip behavior with slips across a fractional number of lattice sites in sharp contrast with the prevailing concept of integer stick-slip behaviors,and the transition conditions from this fractional slip to multiple slip are predicted.(3)We have predicted hundreds of new high-entropy alloy compositions using machine learning model.Owing to the complex mechanism of phase formations of multi-principal element alloys,high-entropy alloys were mainly discovered through time-consuming trial-and-error approaches and there is no robust or reliable model capable to predict which combinations of elements will form high-entropy alloys.We have proposed and demonstrated a support vector machine model as an efficient approach to identify BCC and FCC high-entropy alloys from multi-principal element alloys.Firstly,we built a comprehensive data set containing 322 as-cast samples of multi-principal element alloys with microstructure characterizations for training and testing.After features selection and parameters optimization,the optimal model has five features and manages to achieve an extremely high cross-validation accuracy of 90.66%.With the trained model,we predicted 369 FCC and 267 BCC equiatomic high-entropy alloys from the composition space of 16 different elements.Through the assessments of melting temperatures and densities,dozens of refractory high-entropy alloys have been further screened out,eleven of which are validated by recent experimental results.The model and results can serve as a guide for experimental design and realization of new high entropy alloys with desired properties.