Periodic PySCF Interface

The ciderpress.pyscf.pbc.dft module decorates periodic Kohn–Sham objects from pyscf.pbc.dft. Its make_cider_calc() function supports the semilocal and SDMX features used for methodological reproduction of the CIDER24X work. [1] Packaged periodic NLDF calculations use the classic GPAW/PAW interface.

NOTE: Periodic PySCF calculations must use pseudopotentials and uniform integration grids. All-electron calculations and atom-centered grids are not supported.

Numerical implementation

ciderpress.pyscf.pbc.numint connects the mapped evaluator to PySCF’s periodic boundary condition implementation. It evaluates the semilocal and SDMX blocks, applies their adjoint matrix contributions, and supports the k-point layouts accepted by the periodic Kohn–Sham object.

ciderpress.pyscf.pbc.sdmx_fft constructs periodic SDMX quantities from real- and reciprocal-space orbital representations. Its Gaussian smoothing and k-point conventions implement the CIDER24X periodic descriptor path.

ciderpress.pyscf.pbc.util supplies the FFT interpolation used when the XC evaluation mesh is denser than the cell’s base mesh. Forward feature evaluation and the returned matrix potential use the corresponding pair of mesh transfers.

ciderpress.pyscf.pbc.dft.make_cider_calc(ks, mlfunc, xmix=1.0, xc=None, xkernel=None, ckernel=None, mlfunc_format=None, nlc_coeff=None, nldf_init=None, sdmx_init=None, dense_mesh=None, rhocut=None)

Same as ciderpress.pyscf.dft.make_cider_calc(), but for periodic systems. Note that only semilocal and SDMX features are supported. The ks object must use a uniform XC integration grid and pseudopotentials.

[1]

Kyle Bystrom, Stefano Falletta, and Boris Kozinsky. Training machine-learned density functionals on band gaps. Journal of Chemical Theory and Computation, 20(17):7516–7532, 2024. PMID: 39178337. URL: https://doi.org/10.1021/acs.jctc.4c00999, doi:10.1021/acs.jctc.4c00999.