This framework provides methods for solving kinetic theory formulations uptil 5-dimensional phase space, and has been throughly unit tested. The framework consists of a linear as well as a non-linear solver.
The non-linear solver can make use of the following solver methods:
- Conservative Finite Volume Method
- Reconstruction Methods: minmod, PPM, WENO5
- Riemann Solvers: Local Lax Friedrichs Flux, 1st order Upwind-Flux
- Advective semi-Lagrangian solver based on the method proposed in Cheng & Knorr, 1976.
The framework has been written with ease of use and extensibility in mind, and can be used to obtain solution for any equation of the following form:
Where , , , , and are terms that need to be coded in by the user.
The generalized structure that the framework uses can be found in lib/
. The code has been thoroughly documented, with most folders containing their own README which explains the components of the folder and all functions have been provided docstrings which are indicative of their usage. Additionally, we have validated the solvers by solving the Boltzmann-Equation:
The functions that have been used for solving the above equation may be referred to from src/nonrelativistic_boltzmann
. The test problems we have solved can be found under example_problems/nonrelativistic_boltzmann
. A README has been provided under each of the test folders for appropriate context.
The solver makes use of ArrayFire for shared memory parallelism, and PETSc(Built with HDF5 file writing support) for distributed memory parallelism and require those packages to be built and installed on the system of usage in addition to their python interfaces(arrayfire-python and petsc4py). Additionally, following python libraries are also necessary:
- numpy
- h5py
- matplotlib
- pytest
- mpi4py
- tqdm
- prettytable
The documentation is built using sphinx, and requires the following dependencies to be built locally:
- sphinx
- sphinx_rtd_theme
- sphinx-autobuild
- numpydoc
- Shyam Sankaran - GitHub Profile
- Mani Chandra - GitHub Profile