Machine learning guided appraisal and exploration of phase design for high entropy alloys

Ziqing Zhou,Zhaoyi Ding,Yeju Zhou,Yong Yang,Fucheng Li,Quanfeng He

doi:10.1038/s41524-019-0265-1

Abstract

High entropy alloys (HEAs) and compositionally complex alloys (CCAs) have recently attracted great research interest because of their remarkable mechanical and physical properties. Although many useful HEAs or CCAs were reported, the rules of phase design, if there are any, which could guide alloy screening are still an open issue. In this work, we made a critical appraisal of the existing design rules commonly used by the academic community with different machine learning (ML) algorithms. Based on the artificial neural network algorithm, we were able to derive and extract a sensitivity matrix from the ML modeling, which enabled the quantitative assessment of how to tune a design parameter for the formation of a certain phase, such as solid solution, intermetallic, or amorphous phase. Furthermore, we explored the use of an extended set of new design parameters, which had not been considered before, for phase design in HEAs or CCAs with the ML modeling. To verify our ML-guided design rule, we performed various experiments and designed a series of alloys out of the Fe-Cr-Ni-Zr-Cu system. The outcomes of our experiments agree reasonably well with our predictions, which suggests that the ML-based techniques could be a useful tool in the future design of HEAs or CCAs.

Highlights

Since their advent in 2004,1 high entropy alloys (HEAs) have been attracting tremendous research interest because of their remarkable mechanical and physical properties.[2,3,4] Compared with traditional alloys, HEAs usually contain more than five elements mixed with a similar atomic fraction, thereby known as multiprincipal element alloys[1] or compositionally concentrated alloys.[2]
These rules were developed mainly based on whether there was a correlation between the observed phases and several empirical parameters that could be calculated readily from a HEA composition, such as the ideal mixing entropy Sid,[1] the parameters of atomic size difference δ and the average mixing enthalpy ΔHmix proposed by Zhang et al.,[8] the mean valance electron concentration (VEC) by Guo et al.,[13] the packing misfit parameter γ by Wang et al.,[14] the root mean square residual strain parameter by Ye et al.,[15] the electronegativity difference parameter by Guo et al.,[13] the dimensionless parameters
Our machine learning (ML) modeling is based on three algorithms, including the artificial neural network (ANN), the one-dimensional convolutional neural network (CNN), and the support vector machine (SVM), to assess the efficiency of the existing phase design rules and to explore new ones

Summary

Introduction

Since their advent in 2004,1 high entropy alloys (HEAs) have been attracting tremendous research interest because of their remarkable mechanical and physical properties.[2,3,4] Compared with traditional alloys, HEAs usually contain more than five elements mixed with a similar atomic fraction, thereby known as multiprincipal element alloys[1] or compositionally concentrated alloys.[2]. According to the Hume–Rothery rules,[6] the misfit in the physical and electronic properties of constituent elements strongly affects the formation of SS in binary alloys and possibly multicomponent alloys as well, such as HEAs. Aside from SS, a variety of other phases, such as IM7,8 or even amorphous phase (AM),[9,10,11,12] were observed in the as-cast HEAs. Aside from SS, a variety of other phases, such as IM7,8 or even amorphous phase (AM),[9,10,11,12] were observed in the as-cast HEAs To rationalize this phenomenon, a number of empirical or semiempirical rules were proposed for phase selection in HEAs. To rationalize this phenomenon, a number of empirical or semiempirical rules were proposed for phase selection in HEAs These rules were developed mainly based on whether there was a correlation between the observed phases and several empirical parameters that could be calculated readily from a HEA composition, such as the ideal mixing entropy Sid,[1] the parameters of atomic size difference δ and the average mixing enthalpy ΔHmix proposed by Zhang et al.,[8] the mean valance electron concentration (VEC) by Guo et al.,[13] the packing misfit parameter γ by Wang et al.,[14] the root mean square residual strain parameter by Ye et al.,[15] the electronegativity difference parameter by Guo et al.,[13] the dimensionless parameters

Methods

Results

Conclusion