Machine-learned multi-system surrogate models for materials prediction

Abstract

Surrogate machine-learning models are transforming computational materials science by predicting properties of materials with the accuracy of ab initio methods at a fraction of the computational cost. We demonstrate surrogate models that simultaneously interpolate energies of different materials on a dataset of 10 binary alloys (AgCu, AlFe, AlMg, AlNi, AlTi, CoNi, CuFe, CuNi, FeV, NbNi) with 10 different species and all possible fcc, bcc and hcp structures up to 8 atoms in the unit cell, 15 950 structures in total. We find that the deviation of prediction errors when increasing the number of simultaneously modeled alloys is less than 1 meV/atom. Several state-of-the-art materials representations and learning algorithms were found to qualitatively agree on the prediction errors of formation enthalpy with relative errors of <2.5% for all systems.

Identifying domains of applicability of machine learning models for materials science

“…(3)); the resulting quantity we refer to as the mean absolute error, MAE) of 14.2, 14.1, and 14.7 meV/cation, respectively. This confirms previously reported virtually indistinguishable accuracies for MBTR and SOAP in the prediction of formation energies of alloys 40 . However, using the proposed method, key differences can be observed in the MAEs of their respective DAs (see Table 2 and Fig.…”

Efficient machine-learning based interatomic potentialsfor exploring thermal conductivity in two-dimensional materials

“…The training procedure is thus a nonlinear optimization which is solved using the Broyden-Fletcher-Goldfarb-Shanno (BFGS) algorithm. MTPs have been tested on a number of atomic modeling applications [51], where they demonstrate excellent characteristics for reproducing the properties of both single-component [41,52] and multicomponent materials [40,43].…”

Accelerated design and characterization of non-uniform cellular materials via a machine-learning based framework

“…First, it only needs a relatively simple ML model for the response prediction, based on which the pattern generation can be directly realized, thus the training is not computationally expensive compared with similar studies 30,69,70 . Second, while the architecture of the ML model is not complicated, it is capable of predicting the entire nonlinear response history with plasticity and large deformation other than individual properties 10,18,28 . Third, from the geometric perspective, it is able to predict the response for the majority of the patterns although the ML model alone as a surrogate method run into difficulty to provide accurate response for patterns with a larger infill number.…”

“…Thanks to recent advances in computational power and robust algorithms, we have witnessed the rise of an interdisciplinary field in computational material science [6][7][8][9] . Specifically, researchers have already shown success in utilizing ML-based or other AI methods in two major categories: (1) to accelerate the prediction of material properties for specific applications [10][11][12][13][14][15][16][17][18][19][20][21][22][23][24] , and (2) to accelerate the on-demand design and the optimization of material microstructure and composition for targeted properties [25][26][27][28][29][30][31][32][33][34][35][36][37][38][39][40][41][42] . These promising studies have shown superior effectiveness of ML techniques compared to traditional computational modeling or experimental measurements on a variety of materials.…”