PhaseForge

PhaseForge is a versatile tool designed to assist in the creation and management of thermodynamic database (TDB) files, particularly for materials science applications. It provides a user-friendly interface for generating TDB files, making it an essential resource for researchers and engineers in the field.

Report a Bug | Request a Feature

---

Features

• Integrates MLIPs (e.g., GRACE, ORB, CHGNet, SevenNet) into CALPHAD workflows
• Interfaces seamlessly with ATAT and MaterialsFramework
• Supports automated structure relaxation, phonon and MD-based free energy calculations
• Capable of fitting TDB files for use in CALPHAD-based software (e.g., Pandat, Thermo-Calc)
• Benchmarking framework for evaluating MLIP accuracy in predicting phase stability

---

Installation

Prerequisites

• CMake ≥ 3.10
• C++14 compatible compiler (e.g., GCC, Clang)
• Git
• Python 3 with dependencies (see MaterialsFramework)
• Alloy Theoretic Automated Toolkit (ATAT)
• (Optional) LAMMPS with MLIP integration

Build Instructions

git clone https://github.com/dogusariturk/PhaseForge.git
cd PhaseForge
mkdir build && cd build
cmake .. [-DCMAKE_INSTALL_PREFIX=/your/desired/path]
make
make install

---

Quick Start

Use the sqscal command to generate a phase diagram with minimal setup:

sqscal -e Ni,Re -l FCC_A1,HCP_A3 -lv 2 -mlip Grace -model GRACE-2L-OMAT

This command performs:

• Structure sampling using SQS
• MLIP-based energy calculations
• Optional MD and vibrational contributions
• CALPHAD model fitting (TDB output)

---

Commands

sqscal

 sqscal -e [Element1, Element2, ...] -l [Lattice1,Lattice2] -lv [Level] -mlip [MLIP] -model [MLIP model version] [-vib] [-sro]

A simple in line command for a quick start.

MLIPrelax

MLIPrelax -mlip=[MLIP] -model=[model]

Relax the structure with MLIP-model & single point calculation. fmax=0.001, step=1000

MLIPcalc

MLIPcalc -mlip=[MLIP] -model=[model]

Single point calculation using MLIP

extract_MLIP

extract_MLIP

Extract the MLIP calculation results by MLIP to the ATAT form. CONTCAR, force_temp.out → force, stress_temp.out → stress

runstruct_MLIP

runstruct_MLIP -mlip=[MLIP] -model=[model] [-static]

runstruct_vasp -nr, MLIPrelax, then extract_MLIP. If -static, runs MLIPcalc instead of MLIPrelax.

MLIPmd

MLIPmd -mlip=[MLIP] -model=[model] -temp=[temperature] [-nr]

MLIP MD with NVT ensemble for 2000 steps. Average energy → energy. If -nr, only generate the MLIPmd.py but not run.

MLIPliquid

MLIPliquid -mlip=[MLIP] -model=[model] -dt=[temperature_offset] [-nr] [-lammps]

Calculate the average of melting temperature in atomlabel.tmp and add “dt”. Performing MLIPmd with that temperature. If -nr, only generate the MLIPmd.py but not run. If -lammps, runs with MLIPlammps.

MLIPlammps

MLIPlammps -mlip=[MLIP] -model=[model] -temp=[temperature] [-nr]

Perform MD calculations with LAMMPS and MLIP. If -nr, only generate the lammps.in but not run.

ternary_search

ternary_search -ll=[lower limit] -ul=[upper limit] -eps=[epsilon] -c=[command for energy]

Perform ternary search for the volume. Stop when upper limit - lower limit < epsilon.

robustrelax_MLIP

 robustrelax_MLIP -mlip=[MLIP] -model=[model] [other options]

Perform robustrelax_vasp (command in ATAT) with MLIP and the options

---

Citation

If you use PhaseForge in your research, please cite the following:

Siya Zhu, Doğuhan Sarıtürk, Raymundo Arróyave. Machine Learning Potentials for Alloys: A Detailed Workflow to Predict Phase Diagrams and Benchmark Accuracy, npj Computational Materials, 11, 340 (2025). https://doi.org/10.1038/s41524-025-01814-z

Siya Zhu, Doğuhan Sarıtürk, Raymundo Arróyave. Accelerating CALPHAD-based phase diagram predictions in complex alloys using universal machine learning potentials: Opportunities and challenges, Acta Materialia, 286, 120747 (2025). https://doi.org/10.1016/j.actamat.2025.120747

Sarıtürk, D., Zhu, S., & Arróyave, R. (2025). PhaseForge (v1.0.0). Zenodo. https://doi.org/10.5281/zenodo.15730912

BibTeX:

@article{zhu2025machinelearningpotentials,
  author       = {Zhu, Siya and Sarıtürk, Doğuhan and Arróyave, Raymundo},
  title        = {Machine Learning Potentials for Alloys: A Detailed Workflow to Predict Phase Diagrams and Benchmark Accuracy},
  journal      = {npj Computational Materials},
  year         = 2025,
  volume       = {11},
  pages        = {340},
  doi          = {10.1038/s41524-025-01814-z},
  url          = {https://doi.org/10.1038/s41524-025-01814-z},
}

@article{zhu2025accelerating,
  author       = {Zhu, Siya and Sarıtürk, Doğuhan and Arróyave, Raymundo},
  title        = {Accelerating {CALPHAD}-based phase diagram predictions in complex alloys using universal machine learning potentials: Opportunities and challenges},
  journal      = {Acta Materialia},
  year         = 2025,
  volume       = {286},
  pages        = {120747},
  doi          = {10.1016/j.actamat.2025.120747},
  url          = {https://doi.org/10.1016/j.actamat.2025.120747},
}

@software{sariturk_2025_15730912,
  author       = {Sarıtürk, Doğuhan and Zhu, Siya and Arróyave, Raymundo},
  title        = {PhaseForge},
  month        = jun,
  year         = 2025,
  publisher    = {Zenodo},
  doi          = {10.5281/zenodo.15730911},
  url          = {https://doi.org/10.5281/zenodo.15730911},
}

---

License

This project is licensed under the GNU GPLv3 License. See the LICENSE file for details.