New API for GPAW?

Input-parameter naming and grouping

• initial guess:

  • random

  • basis

  • hund

  • magmoms

• environment (SolvationGPAW, SJM (solvated jellium model), CDFT):

  • background_charge

  • external

• wave-function representation (basis set or grid-spacing):

  • mode

• real-space wave-function grid:

  • h

  • gpts

• more things:

  • kpts

  • nbands

  • setups

  • soc

  • xc

  • charge

  • symmetry

  • poissonsolver

• can be changed on the fly:

  • occupations

  • convergence

  • eigensolver

  • mixer

• not really parameters:

  • parallel

  • txt

DFT-calculation objects and ASE-calculators

atoms.calc = GPAW(...)  # ASE-calculator
e = atoms.get_potential_energy()
dft = atoms.calc.dft  # DFT-calculation (new-GPAW only)

from gpaw.dft import DFT, FermiDirac, MonkhorstPack, ...

dft = DFT(
    atoms=...
    mode=PW(600),
    xc='PBE',
    occupations=FermiDirac(0.2),
    # occupations={'name': 'fermi-dirac', 'width': 0.2},
    kpts=MonkhorstPack((4, 4, 1), include_gamma=True),
    txt='...')

e = dft.calculate_energy()

# Use ASE to do structure optimization:
atoms.calc = dft.create_ase_calculator()

# Magic tricks:
dft.converge(steps=4, mixer=..., eigensolver=...)
for ctx in dft.iconverge():
    dft.occupations = ...

Breaking backwards compatibility

• Blöchl Poisson-solver?

• ecut defaults?

• GGA gradients with FFT?

• …