The FeOs package provides Rust implementations of different equation of state and Helmholtz energy functional models and corresponding Python bindings.
importfeosimportsi_unitsassi# PC-SAFT parameters for methanol (Gross and Sadowski 2002)record=feos.PureRecord( feos.Identifier(name="methanol"), molarweight=32.04, m=1.5255, sigma=3.23, epsilon_k=188.9, association_sites=[{ "kappa_ab": 0.035176, "epsilon_k_ab": 2899.5, "na": 1, "nb": 1, }] ) # Build an equation of stateparameters=feos.Parameters.new_pure(record) eos=feos.EquationOfState.pcsaft(parameters) # Define thermodynamic conditionscritical_point=feos.State.critical_point(eos) # Compute propertiesp=critical_point.pressure() t=critical_point.temperatureprint(f"Critical point for methanol: T={t/si.KELVIN:.1f} K, p={p/si.BAR:.1f} bar.")Critical point for methanol: T=531.5 K, p=106.5 bar. Models
The following models are currently published as part of the FeOs framework
namedescriptioneosdftpcsaftperturbed-chain (polar) statistical associating fluid theory✓✓epcsaftelectrolyte PC-SAFT✓gc-pcsaft(heterosegmented) group contribution PC-SAFT✓✓petsperturbed truncated and shifted Lennard-Jones mixtures✓✓uvtheoryequation of state for Mie fluids and mixtures✓saftvrqmieequation of state for quantum fluids and mixtures✓✓saftvrmiestatistical associating fluid theory for variable range interactions of Mie form✓multiparameterMultiparameter Helmholtz energy equations of state for common pure components✓The list is being expanded continuously. Currently under development are implementations of Helmholtz energy functionals for the UV theory and for SAFT-VR Mie.
Other public repositories that implement models within the FeOs framework, but are currently not part of the feos Python package, are
namedescriptioneosdft
heterosegmented fused-sphere chain functional✓Parameters
In addition to the source code for the Rust and Python packages, this repository contains JSON files with previously published
for the different models including group contribution methods. The parameter files can be read directly from Rust or Python.
Warning
The format for parameter files changed between releases 0.8.0 and 0.9.0. You find parameters for new versions in the
directory and for versions up to 0.8.0 in the
directory.
If you maintain your own parameter files, there are two adjustments you need to make when upgrading to feos 0.9.0:
Flatten the contents of the model_record field directly into the pure or segment record.
Move association parameters into a list called association_sites.
The structure of the JSON files and the PureRecord and SegmentRecord classes in Python is identical.
Properties and phase equilibria
The crate makes use of
generalized (hyper-) dual numbers
to generically calculate exact partial derivatives from Helmholtz energy equations of state. The derivatives are used to calculate
equilibrium properties (pressure, heat capacity, fugacity, and many more),
transport properties (viscosity, thermal conductivity, diffusion coefficients) using the entropy scaling approach
critical points and phase equilibria for pure components and mixtures.
In addition to that, utilities are provided to assist in the handling of parameters for both molecular equations of state and (homosegmented) group contribution methods and for the generation of phase diagrams for pure components and binary mixtures.
Classical density functional theory
FeOs uses efficient numerical methods to calculate density profiles in inhomogeneous systems. Highlights include:
Fast calculation of convolution integrals in cartesian (1D, 2D and 3D), polar, cylindrical, and spherical coordinate systems using FFT and related algorithms.
Automatic calculation of partial derivatives of Helmholtz energy densities (including temperature derivatives) using automatic differentiation with
generalized (hyper-) dual numbers
.
Modeling of heterosegmented molecules, including branched molecules.
Functionalities for calculating surface tensions, adsorption isotherms, pair correlation functions, and solvation free energies.
Cargo features
Without additional features activated, the command
cargo test --release will only build and test the core functionalities of the crate. To run unit and integration tests for specific models, run
cargo test --release --features pcsaft to test, e.g., the implementation of PC-SAFT or
cargo test --release --features all_models to run tests on all implemented models.
Python package
FeOs uses the
framework to provide Python bindings. The Python package can be installed via pip and runs on Windows, Linux and macOS:
pip install feos If there is no compiled package for your system available from PyPI and you have a Rust compiler installed, you can instead build the python package from source using
pip install git+https://github.com/feos-org/feos This command builds the package without link-time optimization (LTO) that can be used to increase the performance further. See the Building from source section for information about building the wheel including LTO.
Building from source
To compile the code you need the Rust compiler and maturin installed. To install the package directly into the active environment (virtualenv or conda), use
cd py-feos maturin develop --release which uses the wheel feature collection specified in the pyproject.toml file.
Alternatively, you can specify the models or features that you want to include in the python package explicitly, e.g.
maturin develop --release --features "pcsaft dft" for the PC-SAFT equation of state and Helmholtz energy functional.
To build wheels including link-time optimization (LTO), use
maturin build --profile="release-lto" which will use the wheel feature collection specified in the pyproject.toml file. The emscripten_wheel collection provides the same models and AD support for Pyodide while omitting incompatible features (such as Rayon, SQLite, and the Rayon-dependent 3D DFT classes). Use the following command to build a wheel with specific features:
maturin build --profile="release-lto" --features "pcsaft ..." LTO increases compile times measurably but the resulting wheel is more performant and has a smaller size. For development however, we recommend using the --release flag.
Documentation
For a documentation of the Python API, Python examples, and a guide to the underlying Rust framework check out the
.
Benchmarks
Check out the
directory for information about provided Rust benchmarks and how to run them.
Developers
This software is currently maintained by members of the groups of
Prof. Philipp Rehner,
Molecular Engineering Thermodynamics (MET), ETH Zurich
Prof. Joachim Gross,
Institute of Thermodynamics and Thermal Process Engineering (ITT), University of Stuttgart
Contributing
FeOs grew from the need to maintain a common codebase used within the scientific work done in our groups. We share the code publicly as a platform to publish our own research but also encourage other researchers and developers to contribute their own models or implementations of existing equations of state.
If you want to contribute to FeOs, there are several ways to go: improving the documentation and helping with language issues, testing the code on your systems to find bugs, adding new models or algorithms, or providing feature requests. Feel free to message us if you have questions or open an issue to discuss improvements.
Cite us
If you find FeOs useful for your own scientific studies, consider citing our
accompanying this library.
@article{rehner2023feos, author = {Rehner, Philipp and Bauer, Gernot and Gross, Joachim}, title = {FeOs: An Open-Source Framework for Equations of State and Classical Density Functional Theory}, journal = {Industrial \& Engineering Chemistry Research}, volume = {62}, number = {12}, pages = {5347-5357}, year = {2023}, }