Conference Tools for 2027 TMS Annual Meeting & Exhibition

Login

Register as a New User

Help

Submit An Abstract

Propose A Symposium

Presenter/Author Tools

Organizer/Editor Tools

About this Abstract

Meeting

2027 TMS Annual Meeting & Exhibition

Symposium

Computational Discovery and Design of Materials

Presentation Title

PhaseForge: Automated, Uncertainty-Quantified Construction of CALPHAD Thermodynamic Databases via Machine-Learning Interatomic Potentials and Bayesian Parameter Refinement

Author(s)

Raymundo Arroyave

On-Site Speaker (Planned)

Raymundo Arroyave

Abstract Scope

The predictive design of multicomponent materials hinges on thermodynamic databases that encode phase stability and equilibrium properties across composition and temperature. Despite decades of progress in first-principles electronic-structure methods, constructing these databases remains a labor-intensive craft: ab initio energetics of competing phases are hand-curated, fit to CALPHAD models, and calibrated against fragmented experimental data through cycles that occupy specialists for months. PhaseForge aims to close this loop autonomously, by treating machine-learning interatomic potentials as scalable surrogates for first-principles physics, embedding them within a Bayesian framework that learns CALPHAD parameters with calibrated uncertainty, and coupling the resulting models to active-learning agents that direct targeted experiments. The aim is to make uncertainty-aware phase-stability prediction as routine for any new multicomponent system as DFT calculations are today, opening the design space of functional, energy, and reliability-critical materials at a scale conventional CALPHAD efforts cannot reach.

Proceedings Inclusion?

Planned:

Keywords

Computational Materials Science & Engineering, Machine Learning, ICME

OTHER PAPERS PLANNED FOR THIS SYMPOSIUM

A Multi-Fidelity Machine-Learning Workflow for Accelerated Lithium-Ion Cathode Materials Development

Accelerating Quantum Materials Discovery with AI

Agapi-agents: An open-access agentic ai platform for accelerated materials design on atomgpt. org

Assessing, training, and fine-tuning foundation potentials for plasticity-relevant properties of metals

Beyond First Principles: Convergent and Thermodynamically Self-Consistent Atomistic Phase Diagrams from Interatomic Potentials

Capturing Spin-Disorder in Paramagnetic Austenitic Steels: A First-Principles Approach

Characterization of the Atomic-Scale Strengthening Mechanisms in Oxide Dispersion Strengthened Alloys at Elevated Temperatures

Closing the Loop on Superconductor Discovery

Collective motion of screw dislocations in α-Fe and its effect on Peierls stress

Computational discovery of synthesizable and hard multicomponent transition metal carbides

Crystal Structure Exploration Under Extreme Conditions

Defect-Controlled Carbon Transport in Zirconium Carbide under Charge, Temperature, and Pressure Extremes.

Digital material design guidelines for mitigation of alloy embrittlement

Effects of Chemical Fluctuations and Short-Range Order on Dislocation Behavior in Multi-Principal-Element Alloys

Enhancing of the Thermal Conductivity in Refractory-based High Entropy Alloys via Mott s-d Scattering Mechanism

ExactBO: GPU-Accelerated Certified Bayesian Optimization for Continuous and Constrained Materials Design

First-Principles and Graph Learning Framework for Discovery of Li-Ion Transport Pathways in Solid Electrolyte Interphases

First-principles mapping of defect energetics across the 211 MAX phase chemical space

First-Principles Transport in AlGaN/GaN HEMT-like Interfaces

First Principles Data Assisted Investigation of Spin Ordering in Spinel Ni1−xZnxFe2O4 Crystal

From DFT Defect Mechanisms to Transport-Scale Modeling of Hydrogen Isotopes in Li2O

Geometric Design of d- and i-wave Altermagnetism in Graphene Nanohole Superlattices

Humans-(less)-in-the-loop: moving toward autonomous materials design for metal-organic frameworks

Hysteresis in silicon-dominated anodes in lithium-ion batteries.

Integrating Physical Models, Specialized Databases, AI/ML Tools, and Self-Driving Labs to Understand the Synthesis of Inorganic Materials

Intermetallic Composition as a Variable in Alloy Design

Machine Learning-Assisted Design of High-Performance HOR Catalysts for Proton-Conducting Fuel Cells

Machine Learning Screening of Oxide Semiconductors for Energy-Efficient Back-End Transistors

Metadynamics-Accelerated Active Learning of Interatomic Potentials for InP Magic-Sized Cluster Nucleation

PhaseForge: Automated, Uncertainty-Quantified Construction of CALPHAD Thermodynamic Databases via Machine-Learning Interatomic Potentials and Bayesian Parameter Refinement

Physics-Grounded AI for Materials Discovery

Physics-Informed Discovery of Hydride Superconductors

Predicting plateau slope in metal hydrides from first-principles: hydrogen energy and charge distributions in TiMn₂-based alloys

Prediction and Screening of Transition Metal Nitrides for Ammonia Synthesis

Quantifying the Contributions of Strengthening Mechanisms in Refractory Complex Concentrated Alloys through Physics-Informed Statistical Learning.

Role of a Ni₁₃ Cocatalyst in Promoting CO₂ Activation and Hydrogenation on CaCu₃Ti₄O₁₂: A First-Principles Study

Size Effects on Chemical Ordering in High-Entropy MXenes: A First-Principles Validated Atomistic Study

Stabilizing Titanium Carbide MXenes in Water with Montmorillonite Heterostructures: Ab Initio Molecular Dynamics Study

Thermodynamic models for the epitaxial growth of oxide solid solutions

Understanding Point Defects in Semiconductors using First Principles Simulations Accelerated by ML Potentials

Questions about ProgramMaster? Contact programming@programmaster.org | TMS Privacy Policy | Accessibility Statement