Multi-activity derivative automated labeling and margin-sampling active learning for efficient determination of phase boundaries

Abstract

To realize the exploration of unknown diagrams directly from measurable physical/chemical properties, we improve the property derivatives-based machine learning (ML) framework of phase boundary prediction to overcome prior limitations in signal enhancement, labeling efficiency, uncertainty quantification, and potential applicability to multicomponent composition spaces. An automated phase-region labeling approach combining multiple activity derivatives is introduced to amplify the differentiability signal near phase boundaries, enabling precise identification of distinct phase regions. Three uncertainty sampling (US) strategies, i.e., least confidence (LC), margin sampling (MS), and entropy-based acquisition (EA), are employed to guide query selection in active learning. MS generally promoted a more uniform distribution of queries along phase boundaries and reduced redundant sampling near triple junctions, thereby facilitating efficient convergence. The truncated ML phase boundaries based on multi-layer perceptron + MS achieved Macro-F1 scores of 0.9994 and 0.9814 using only 120 and 186 labeled samples for the Pd-Pt and Au-Ag-Ge systems, respectively. The mean boundary displacements of ML-predicted phase boundaries against the Thermo-Calc-calculated results were 0.0059 for the Pd-Pt system and 0.0778 for the Au-Ag-Ge system.

Keywords

Labeling / sampling / activity derivatives / uncertainty / phase boundary

References

[1] Liu Z.Thermodynamics and its prediction and CALPHAD modeling: review, state of the art, and perspectives.Calphad2023;82:102580

[2] Arróyave R.Phase stability through machine learning.J Phase Equilib Diffus2022;43:606-28

[3] Chew PY.Phase diagrams-Why they matter and how to predict them.J Chem Phys2023;158:030902

[4] Shen C.The synergy of machine learning and CALPHAD: revitalizing traditional approaches.Comput Mater Sci2025;258:113970

[5] Kruglov IA,Oganov AR.Phase diagram of uranium from ab initio calculations and machine learning.Phys Rev B2019;100:174104

[6] Rosenbrock CW,Shapeev AV.Machine-learned interatomic potentials for alloys and alloy phase diagrams.npj Comput Mater2021;7:477

[7] Zhu S,Arróyave R.Accelerating CALPHAD-based phase diagram predictions in complex alloys using universal machine learning potentials: opportunities and challenges.Acta Mater2025;286:120747

[8] Xie JZ,Jin B.Machine learning force field-aided cluster expansion approach to phase diagram of alloyed materials.J Chem Theory Comput2024;20:6207-17

[9] Bocklund B,Egorov A,Roslyakova I.ESPEI for efficient thermodynamic database development, modification, and uncertainty quantification: application to Cu–Mg.MRS Commun2019;9:618-27

[10] Zhang H,Zhang L.Efficient thermodynamic modelling with uncertainty quantification of the Ag–Cu–Co system from its sub-binary systems.Mater Chem Phys2023;308:128276

[11] Deffrennes G,Abe T.Data-driven study of the enthalpy of mixing in the liquid phase.Calphad2024;87:102745

[12] Wu B,Zhang L.Estimating the temperature dependent zero-phase-fraction features in ternary phase diagram via Bayesian approach.Scripta Mater2023;235:115615

[13] Lund J,Braatz RD.Machine learning of phase diagrams.Mater Adv2022;3:8485-97

[14] Deffrennes G,Abe T.A machine learning–based classification approach for phase diagram prediction.Mater Design2022;215:110497

[15] He J,Wang C.Machine learning assisted predictions of multi-component phase diagrams and fine boundary information.Acta Mater2022;240:118341

[16] Terayama K,Katsube R.Acceleration of phase diagram construction by machine learning incorporating Gibbs’ phase rule.Scripta Mater2022;208:114335

[17] Li H,Chen F.Estimating phase boundaries using sampling and machine learning of derivatives of thermochemistry properties.Calphad2026;92:102920

[18] Hao L,Xiong W.CALPHAD modeling based on Gibbs energy functions from zero kevin and improved magnetic model: a case study on the Cr–Ni system.Calphad2021;73:102268

[19] Liu ZK.Computational thermodynamics of materials. Cambridge University Press: Cambridge, 2016.

[20] Koizumi A,Terayama K.Performance of uncertainty-based active learning for efficient approximation of black-box functions in materials science.Sci Rep2024;14:27019 PMCID:PMC11541781

[21] Terayama K,Nose Y.Efficient construction method for phase diagrams using uncertainty sampling.Phys Rev Mater2019;3:033802

[22] Wu B,Zhou Y,Wang H.Uncertainty quantification of phase boundary in a composition-phase map via Bayesian strategies.Phys Rev Mater2023;7:025201

[23] Liu P,Qin H.Construction of material phase diagram using different uncertainty estimation strategies.Mater Lett2025;384:138036

[24] Varivoda D,Omee SS.Materials property prediction with uncertainty quantification: a benchmark study.Appl Phys Rev2023;10:021409

[25] Tian Y,Xue D.Role of uncertainty estimation in accelerating materials development via active learning.J Appl Phys2020;128:014103

[26] Tian Y,Xue D.Determining multi-component phase diagrams with desired characteristics using active learning.Adv Sci2020;8:2003165 PMCID:PMC7788591

[27] Dai C.Efficient phase diagram sampling by active learning.J Phys Chem B2020;124:1275-84

[28] Lookman T,Xue D.Active learning in materials science with emphasis on adaptive sampling using uncertainties for targeted design.npj Comput Mater2019;5:153

[29] Kanopoulos N,Baker R.Design of an image edge detection filter using the Sobel operator.IEEE J Solid State Circuits23:358-67

[30] Turchi PEA,Kudrnovský J.Stability and ordering properties of fcc alloys based on Rh, Ir, Pd, and Pt.Phys Rev B2006;74:064202

[31] Wang J,Tang C,Zhou H.Thermodynamic description of the Au–Ag–Ge ternary system.Thermochim Acta2011;512:240-6

[32] Deringer VL,Bernstein N,Ceriotti M.Gaussian process regression for materials and molecules.Chem Rev2021;121:10073-141 PMCID:PMC8391963

[33] Rumelhart DE,Williams RJ.Learning representations by back-propagating errors.Nature1986;323:533-6

[34] Wang X,Yan C.A multi-candidate batch mode active learning approach. In Proceedings of the 2023 15th International Conference on Machine Learning and Computing. Association for Computing Machinery; 2023. pp 528-33.

[35] Yacouby R.Probabilistic extension of Precision, Recall, and F1 score for more thorough evaluation of classification models. In Proceedings of the First Workshop on Evaluation and Comparison of NLP Systems. Association for Computational Linguistics; 2020. pp 79-91.

[36] Heese R,Walczak M.Calibrated simplex-mapping classification.PLoS One2023;18:e0279876 PMCID:PMC9844900

[37] Krajewski AM,Reinhart WF.Efficient generation of grids and traversal graphs in compositional spaces towards exploration and path planning.npj Unconv Comput2024;1:12

[38] Greenacre M,Bacon-Shone J,Quinn T.Aitchison’s compositional data analysis 40 years on: a reappraisal.Statist Sci2023;38:386-410

[39] Egozcue JJ,Mateu-Figueras G.Isometric logratio transformations for compositional data analysis.Math Geol2003;35:279-300