Non-self-consistent calculations with istart=2 & icharg=11

[youngsam_kim]

Hello,

I have a question about the tag combination of ISTART = 2 and ICHARG = 11.

To clarify the question, the calculation procedure that I did is summarized below.

1) The system was calculated along with an insufficient electric minimization loop value.
(NELM=10 in this case and note that this is a non-converged calculation!)
From the OUTCAR of 1) calculation,
Hartree energy = -495.12291229
free energy = -13.18860079 eV

2) Non-selfconsistent calculation was performed with the WAVECAR and CHGCAR from 1).
(ISTART = 2 and ICHARG = 11)
From the OUTCAR of 2) calculation,
Hartree energy = -495.18085910
free energy = -13.52057273

If I understood correctly, the Hartree energies should be equal to each other in 1) and 2).
Because the same charge density is used in 1) and 2)!

3) Further calculation helped to understand the cause.
The 1) calculation was the result of the 10th cycle, and the result of the 11th cycle was as follows.
Hartree energy = -495.18057932
free energy = -13.52056611

It seems that ICHARG = 11 calculation implies one more continuous loop of electronic minimization.
And as a result, it means that the density changes and Hartree energy also changes.

Because general calculations with "ICHARG = 11" tag are performed for the band structure or DOS calculations after convergence of previously given Brillouin zone sampling, the density does not change. (Due to the variation principle.)

If so, is there a way to perform the calculation while maintaining non-converged density?

Thank you for reading this long question!

Sincerely,

Youngsam Kim

details of the input files are as follows. (Gamma-point only calculation)
------------------------------
POSCAR
water_mon
1.00000
15.00000 0.00000 0.00000
0.00000 15.00000 0.00000
0.00000 0.00000 15.00000
O H
1 2
Cart
5.00111 3.60865 5.00000
5.49057 3.28121 5.76087
5.49057 3.28121 4.23913
----------------------------------
INCAR of 1)
nwrite = 3
encut = 400
symprec = 1e-5

# spin polarization parameters
ispin = 2 ! 2(spin polarized), 1(non-polarized)

# ionic relaxation
nsw = 0 ! static

# electronic relaxation
ediff = 1e-6
nelm = 10

ismear = 0
sigma = 0.0001

kpar = 1
npar = 1

lwave = .true.
lcharg = .true.
----------------------------------
INCAR of 2)
nwrite = 3
istart = 2
icharg = 11
encut = 400
symprec = 1e-5

# spin polarization parameters
ispin = 2 ! 2(spin polarized), 1(non-polarized)

# ionic relaxation
nsw = 0 ! static

# electronic relaxation
ediff = 1e-6
nelm = 1

ismear = 0
sigma = 0.0001

kpar = 1
npar = 1

lwave = .false.
lcharg = .false.

[alexey.tal]

Dear youngsam_kim,

The energy difference in 1) and 2) is due to the fact that the density written in CHGCAR corresponds to the next iteration, i.e. density after mixing. Hence, 1) calculates the energies for iteration #10 and 2) calculates the energies for iteration #11. This behavior cannot be changed without modifying the code.

If the density is converged, the energies in 1) and 2) should be the same. However, in your case the density is not converged.
Could you please explain what properties you are trying to calculate and why you would like to do a non-selfconsistent calculation with an unconverged density?

[youngsam_kim]

First of all, thank you for the rapid and detailed explanation!

I would like to calculate the energy in "A" potential with the pre-converged density on "B" potential.

Note that the pre-converged density(B) is no longer the converged density on the other potential(A).
Therefore, in the previous question, non-converged density was used to understand the combination of tags.

For example, after calculating the density using PBE functional, then I want to calculate the energy on the potential of RPBE functional with the pre-converged PBE density.
When one iteration is run, the density is only the meaning of the trivial density for a given potential.

What I want to do is find out how much the "XC energy" changes as the "XC potential" changes for a given wavefunction and the corresponding density.
(The kinetic energy, Hartree energy, and Coulombic energy between the nucleus and the electron will be the same.)

Thanks a lot for your kindness.

Sincerely,

Youngsam Kim

[alexey.tal]

I assume that for such comparison you would like to use exactly the same density. Probably the easiest way to do it would be to converge the charge density with functional 1) A and then perform non-selfconsistent calculations with functionals 2) A and 3) B reading the same WAVECAR and CHGCAR produced in 1).

[youngsam_kim]

Yes, I agree very much!

Therefore, I did the 3 calculations described in the first post.
However, the non-self-consistent calculation seems to be performed with one more iteration.

Is the combination of ISTART=2 and ICHARG = 11 wrong?
or, did I make another mistake?

Thanks for your kindness!

Youngsam Kim

[alexey.tal]

If I understood correctly, in your first post you are comparing the Hartree energy in 1) and 2) which are different because the charge density is different. In the scheme that I described calculations 2) and 3) should produces the same Hartree energy if they read the same charge density. Do you get different Hartree energies in two non-selfconsistent calculations with different xc functionals?