4.4. DFT calculations

This example is about the SCF function and how to set it up.

This function is responsible for the DFT calculation and the respective SCF cycle.

---

from eminus import Atoms, SCF

Start by creating an Atoms object for helium

Use a very small ecut for a fast calculation

atoms = Atoms("He", [0, 0, 0], ecut=5)

Optional parameters with examples are listed as follows

Dictionary to set the maximum amount of steps per minimization method and their order

Set it to a very small value for a short output

opt = {"pccg": 5}

The SCF class only needs an Atoms object, but only calculate 5 steps for less output

print("First calculation:")
SCF(atoms, opt=opt).run()

Exchange-correlation functional description (case insensitive), separated by a comma

xc = "lda,pw"

The Libxc interface can be used by adding libxc: before a functional (you can also neglect the libxc part)

Names and numbers can be used, and mixed with the internal functionals as well

# xc = "libxc:LDA_X,:LDA_C_PW"
# xc = "libxc:1,pw"

Type of potential (case insensitive)

pot = "gth"

Initial guess method for the basis functions (case insensitive)

Adding sym to the string will use the same guess for all spin channels

An unsymmetric guess will be used by default

guess = "random"

Convergence tolerance of the total energy

etol = 1e-8

Convergence tolerance of the gradient norm

gradtol = 1e-7

Calculate a self-interaction correction from the Kohn-Sham orbitals after an SCF calculation.

sic = True

Calculate a dispersion correction after an SCF calculation (need the dispersion extra).

disp = False

The amount of output can be controlled with the verbosity level

By default the verbosity level of the Atoms object will be used

verbose = 4

Start a new calculation with new parameters

print("\nSecond calculation with more output:")
scf = SCF(
    atoms=atoms,
    xc=xc,
    pot=pot,
    guess=guess,
    etol=etol,
    gradtol=gradtol,
    opt=opt,
    sic=sic,
    disp=disp,
    verbose=verbose,
)

Arguments for the minimizer can be passed through via the run function, e.g., for the conjugated-gradient form

etot = scf.run(cgform=2)

The total energy is a return value of the SCF.run function, but it is saved in the SCF object as well with all energy contributions

print(f"\nEnergy from SCF function = {etot} Eh")
print(f"\nEnergy in Atoms object:\n{scf.energies}")

Download 04_dft_calculations.py