6
$\begingroup$

Why is my plot of 'Total Energy vs Energy cut off' is jagged. Previously ecutwfc was kept at 40, 60 and now 80. But still I am unable to figure why isn't it converging. Where as 'Total Energy vs K points' show convergent behavior.

 &CONTROL
                 calculation = 'scf' ,
                restart_mode = 'from_scratch' ,
                      outdir = '../../tmp' ,
                  pseudo_dir = '../../pseudo' ,
                      prefix = 'Pt' ,
 /
 &SYSTEM
                       ibrav = 2,
                   celldm(1) = 7.5828,
                         nat = 1,
                        ntyp = 1,
                     ecutwfc = 80 ,
                 occupations = 'smearing' ,
                     degauss = 0.02 ,
                    smearing = 'mp' ,
 /
 &ELECTRONS
 /
 &IONS
 /
 &CELL
               cell_dynamics = 'bfgs' ,
 /
ATOMIC_SPECIES
   Pt  195.07800  Pt.pbe-nd-rrkjus.UPF
ATOMIC_POSITIONS alat
   Pt    0.000000000    0.000000000    0.000000000
K_POINTS automatic
  8 8 8  0 0 0

enter image description here

enter image description here

Edit 1:
Changing 'ecutwfc' ranging between 6-40 in the intervals of 2 gives very good convergence.

Why doesn't this happen in other plot where it ranges from 20-80 in the intervals of 1?

enter image description here

$\endgroup$
5
  • 1
    $\begingroup$ +1 Something to note is that the scale of the k-point convergence plot is 350 mRy, while the scale of the energy cutoff convergence is over an order of magnitude smaller at 20 mRy. Perhaps your numerical parameters are not tight enough to resolve an energy of the order of 1 mRy in the first plot, hence the erratic shape of the curve? $\endgroup$
    – ProfM
    Oct 2, 2021 at 11:25
  • $\begingroup$ @ProfM May be seems right. When I took ecutwfc ranging from 6 - 40 in the intervals of 2, I got excellent convergence. Please look at the edit. I believe it has something to do with precision? $\endgroup$
    – 147875
    Oct 2, 2021 at 12:39
  • $\begingroup$ The increments on your first plot (.002) are 25x smaller than on the second plot (.05). Try redoing the second plot starting with 6 k-points or the change the scale on the first plot and you should find they look fairly similar. $\endgroup$
    – Tyberius
    Oct 2, 2021 at 12:51
  • $\begingroup$ @Tyberius I didn't understand why you're comparing energy convergence to k-point ocnvergence? $\endgroup$
    – 147875
    Oct 2, 2021 at 13:24
  • 1
    $\begingroup$ You have to look at the scale. On your new plot, each increment is 1 Ry, the original plot has increments of .002 Ry, 500 times smaller. The same oscillations are still there, you just can't see them on that large a scale. Whether energy differences of that size are converged enough depends on the application. $\endgroup$
    – Tyberius
    Oct 2, 2021 at 14:24

1 Answer 1

2
$\begingroup$

The energy should be variational with regard to encut, but this is not something to expect for kpoints. However, normally this results in the opposite of what you are observing with the kpoints showing some sort of oscillation.

The result you see is due to numerical noise in the encut convergence calculations. Try to tighten the electronic cutoff and you will see less oscillation, however also keep in mind the scale of the graph. You have an oscillation of about 0.005 Ry from 40 Ry - 60 Ry. This is about 0.06 eV in terms I am more used to seeing. This seems like more than it should be. This says to me that there might still be a problem, but this can be solved by raising the convergence. However, you should never expect this behavior to totally go away.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .