What is materialdft-demo?

MaterialDFT is an experimental scientific computing library that performs quantum mechanical calculations for materials research, specifically using a method called Density Functional Theory (DFT). DFT is a widely used computational approach in physics and chemistry that predicts how electrons are arranged in a material, which in turn reveals properties like energy, stability, and electronic structure. Researchers use it to study metals, semiconductors, and other materials without needing to run physical experiments for every scenario. The distinctive feature of this project is that it is built to accept the same input files and produce the same output files as VASP, a widely used commercial DFT software package. This means a researcher who already has VASP input files (the INCAR configuration file, POSCAR crystal structure file, and KPOINTS file) can run them through MaterialDFT without changing anything. The outputs land in familiar formats like OUTCAR, EIGENVAL, and DOSCAR, making comparison straightforward. VASP itself is closed-source and requires a paid license, so this project is an independently written alternative built from public documentation, academic papers, and other open-source DFT implementations. The current scope is narrow by design. The library only covers one specific calculation path: static, non-spin-polarized PBE calculations on a single CPU thread. PBE is a particular flavor of the exchange-correlation functional that DFT calculations depend on. More complex calculation types, parallel execution, and relaxation runs are not yet supported targets. Benchmark results show the library reaches agreement with VASP in the range of 0.00001 to 0.001 electron-volts of error for simpler systems like small metal unit cells and basic semiconductors, but errors can reach around 0.01 eV on more complex or low-symmetry structures. The code is organized as a C++ core with a Python interface layer built using pybind11. Building it requires a C++17 compiler, CMake, FFTW3 for fast Fourier transforms, LAPACK and BLAS for linear algebra, and the Libxc library for the exchange-correlation functional. The repository does not include POTCAR files, which are the pseudopotential data files that VASP uses and distributes under a separate commercial license.