Abstract:Conventional phase diagram generation involves experimentation to provide an initial estimate of the set of thermodynamically accessible phases and their boundaries, followed by use of phenomenological models to interpolate between the available experimental data points and extrapolate to experimentally inaccessible regions. Such an approach, combined with high throughput first-principles calculations and data-mining techniques, has led to exhaustive thermodynamic databases (e.g. compatible with the CALPHAD method), albeit focused on the reduced set of phases observed at distinct thermodynamic equilibria. In contrast, materials during their synthesis, operation, or processing, may not reach their thermodynamic equilibrium state but, instead, remain trapped in a local (metastable) free energy minimum, which may exhibit desirable properties. Here, we introduce an automated workflow that integrates first-principles physics and atomistic simulations with machine learning (ML), and high-performance computing to allow rapid exploration of the metastable phases to construct "metastable" phase diagrams for materials far-from-equilibrium. Using carbon as a prototypical system, we demonstrate automated metastable phase diagram construction to map hundreds of metastable states ranging from near equilibrium to far-from-equilibrium (400 meV/atom). We incorporate the free energy calculations into a neural-network-based learning of the equations of state that allows for efficient construction of metastable phase diagrams. We use the metastable phase diagram and identify domains of relative stability and synthesizability of metastable materials. High temperature high pressure experiments using a diamond anvil cell on graphite sample coupled with high-resolution transmission electron microscopy (HRTEM) confirm our metastable phase predictions. In particular, we identify the previously ambiguous structure of n-diamond as a cubic-analog of diaphite-like lonsdaelite phase.

Machine-Learning-Based phase diagram construction for high-throughput batch experiments

Machine Learning the Metastable Phase Diagram of Materials

Determining Multi‐Component Phase Diagrams with Desired Characteristics Using Active Learning

Machine Learning the Metastable Phase Diagram of Covalently Bonded Carbon.

Machine learning assisted predictions of multi-component phase diagrams and fine boundary information

Efficient Phase Diagram Sampling by Active Learning

$\textit{Ab initio}$ construction of full phase diagram of MgO-CaO eutectic system using neural network interatomic potentials

Accelerating CALPHAD-based Phase Diagram Predictions in Complex Alloys Using Universal Machine Learning Potentials: Opportunities and Challenges

Efficient mapping of phase diagrams with conditional Boltzmann Generators

High-throughput analysis of magnetic phase transition by combining table-top sputtering, photoemission electron microscopy, and Landau theory

Using Machine Learning and Feature Engineering to Characterize Limited Material Datasets of High-Entropy Alloys

Uncertainty quantification of phase boundary in a composition-phase map via Bayesian strategies

Accelerating Multicomponent Phase-Coexistence Calculations with Physics-informed Neural Networks

Accelerating Phase Prediction of Refractory High Entropy Alloys Via Machine Learning

Human-In-the-Loop for Bayesian Autonomous Materials Phase Mapping

Machine-Learning Surrogate Model for Accelerating the Search of Stable Ternary Alloys

Phase Selection Rules of Multi-Principal Element Alloys

Calculating Phase Diagrams Using PANDAT and Panengine

Bayesian inference of composition-dependent phase diagrams

Machine learning, phase stability, and disorder with the Automatic Flow Framework for Materials Discovery

From electrons to phase diagrams with machine learning potentials using pyiron based automated workflows