Conventional alloys mainly consist of one principal element with minor property modifications made by adding relatively small proportions of other elements. While as many as a dozen additional elements may be included in conventional alloys, one base element possesses the majority atom fraction in the material.
Multi-principal element alloys or MPEAs consist of several base elements, usually in significant proportions. These MPEAs possess unique and sometimes better physical and mechanical properties than the conventional alloys. Nathan Linton, who recently joined the PhD program in Materials Science and Engineering at Clemson University, is hoping to find out the physical properties of novel MPEAs in an unusually short span of time using machine learning techniques.
To keep up with an ever-increasing demand of material performance required for various high-temperature applications, a significant amount of work needs to be done to further the development of novel materials. Old-school trial-and-error based experimental methods would take plenty of time to develop novel materials and to consequently predict material properties. First principles methods such as Density Functional Theory (DFT) can predict material properties of unknown systems computationally, without any experimental input.
Nathan, who works with Prof. Dilpuneet S. Aidhy, seeks to reduce the computational time and resources needed even further, to obtain the mechanical properties of these MPEAs by using machine learning models.
First, a computational simulation of two, three, four and five element alloys containing Ni, Cu, Au, Pd, and/or Pt is performed to obtain material properties such as elastic constants, Young’s modulus, bulk and shear moduli and Poisson’s ratio. Next, a machine learning model is trained on the two element alloys which is then used to predict the mechanical properties of the three, four and five element alloys. The entire computational work is done using Clemson’s high-performance computer cluster, Palmetto.
“This is done in a matter of seconds for all the alloys compared to the 40+ hours per alloy using conventional computations using density functional theory”, Nathan said.
In describing the relevance of his work in today’s life, Nathan believes that machine learning and computational simulations can drastically reduce the amount of time it takes to engineer a new material with desired qualities. “You reduce the amount of experimental work needed by cutting down the number of alloys to create and test”, Nathan said.
Nathan, who is originally from Wyoming, did his undergraduate in Mechanical Engineering at the University of Wyoming and obtained his BS degree in the December of 2020. During his undergraduate studies, he took Prof. Aidhy’s course on “Properties of materials”, which got him hooked on working towards designing novel materials with better mechanical properties.
He has already helped with the publication of the article, “Machine Learning Based Methodology to Predict Point Defect Energies in Multi-Principal Element Alloys” on point defect energies and machine learning. Currently, he is working on publishing his current work and will be attending the Materials Research Society’s (MRS) Conference in Boston in the Fall of 2022 to present his findings.
Nathan holds a good relationship with Prof. Aidhy, with whom he communicates research ideas and enjoys progress as well as understands his expectations as a graduate student. Currently, his work is funded through the Department of Energy’s Basic Energy Sciences program. He hopes that this work will allow him to join a National Laboratory to design and develop new materials along with becoming a professor to teach Materials Science courses at a university.
