Equilipy

Equilipy is an open-source python package that offers multicomponent multiphase equilibrium calculations based on the CALPHAD (CALculation of PHAse Diagram) approach. With a set of Gibbs energy description (Thermochemical database) and input conditions (Composition, temperature, pressure), equilibrium phase configureation, amount, composition, and thermochemical properties can be obtained. Equilipy uses the Gibbs energy descriptions furnished by THERMOCHIMICA with the modified Gibbs energy minimization algorithm initially proposed by de Capitani, C. and Brown, T.H. (1987).

Dependencies

Dependency	Version	Required	Libraries
Fortran	-	Yes	-
Python	3.9+	Yes	numpy, wheel, meson, ninja

Before installation

Single computing nodes (desktop/laptop)

Equilipy requires a Fortran compiler in the local environment. To install gfortran using conda, for Linux:

conda install -c conda-forge gfortran_linux-64

for MacOS:

conda install -c conda-forge gfortran_osx-64

for Windows:

conda install -c conda-forge fortran-compiler

Alternatively, gfortran can be install for Ubuntu and Debian,

sudo apt-get update
sudo apt-get upgrade
sudo apt-get install gfortran 
sudo apt-get install libopenmpi-dev

for MacOS,

brew install gcc open-mpi

To install gfortran on Windows,

1. Download the latest MinGW-w64 and unzip.
2. Copy the unzipped folder to C-drive and rename the folder/directory as mingw in C-drive C:\mingw\
3. Click the Windows button and type "environment variables" to access Edit the system environment variables.
4. Click the Environment Variables at the bottom right corner
5. Click Path in System variables dialog to display Edit environment variable window
6. Click New and add C:\mingw\bin to the path

Equilipy also requires Python version 3.9 and above. The Fortran backend needs to be compiled through the f2py module in numpy which requires meson and ninja. The wheel library is used for packaging. These can both be installed through pip.

Multiple computing nodes (HPC) on Linux

Equilipy uses mpi4py to interface with MPI tools. To install OpenMPI, mpi4py, and gfortran without using sudo privilage, we recommand install gfortran, OpenMPI, and mpi4py using conda:

conda install -c conda-forge gfortran_linux-64 openmpi mpi4py

Alternatively, users with sudo privilage may install without conda: for Debian-based (Debian, Ubuntu, Mint, etc..)

sudo apt-get install gfortran 
sudo apt-get install libopenmpi-dev

for CentOS (Red Hat Enterprise Linux, CentOS,Fedora, openSUSE),

sudo yum install gcc-gfortran 
sudo yum install openmpi openmpi-devel

for Fedora and Red Hat Enterprise Linux

sudo dnf install gcc-fortran
sudo dnf install openmpi openmpi-devel

Installation

Pypi

Installation using pip is available for Equilipy.

pip install equilipy

For HPC environment, use equilipy-hpc

pip install equilipy-hpc

Install from the source

Here is the procedure to install Equilipy from the source:

1. Clone the repository

git clone https://github.com/ORNL/Equilipy.git

2. Install dependant packages using pip.

pip install numpy wheel meson ninja

3. Create wheels

python setup.py bdist_wheel --dist-dir=./wheelhouse

Note that on macOS it may be necessary to explicitly use GNU gcc instead of the Apple clang gcc. For example, with a brew installation, export CC before calling setup.py:

export CC=/usr/local/Cellar/gcc/*/bin/gcc-*

4. Install from .whl file

pip install wheelhouse/equilipy-*.whl

Features and example

The following features are currently available.

• Single condition equilibrium calculations
• Batch equilibrium calculations
• Scheil-Gulliver solidification
• Phase selection

For details, check out the example directory.

Additional note

Examples in Equilipy uses polars dataframe for fast data processing. If you are using old CPUs, install

pip install polars-lts-cpu

If you are using large dataset (> 4billion), install

pip install polars-u64-idx

For details, check out polars.