CHGCAR will be kept constant for ICHARG=11?

[yong_han]

In the VASP manual, I can find the statement: "non-selfconsistent calculations: Adding 10 to the value of ICHARG, e.g., ICHARG=11 or 12 (or the less convenient value 10) means that the charge density will be kept constant during the entire electronic minimization."

I made a test when calculating DOS for a large slab system by setting in INCAR file:
ISTART = 1
ICHARG = 11
LORBIT = 11
LCHARG = T
NSW = 0
IBRION = -1

where "LCHARG = T" means that initial CHGCAR (from SCF relaxation calculation) will be updated after the above NSCF calculation (only one ionic step).

I checked and compared the initial CHGCAR and the final CHGCAR (updated after this NSCF calculation) and found that these two CHGCAR files are not identical (i.e., not constant) with the differences between -9.45330E-05 to +9.45926E-05. I am not sure such differences are just the computational errors or from other reasons. It looks like a difference close to +0.0001 or -0.0001 is not negligible. Anyone can comment on this? Thank you!

[manuel_engel1]

I could not reproduce your issue in a simple test system. Both CHGCAR files are identical on my end. To investigate this issue further, please provide all the relevant files according to the Forum Posting Guidelines.

A potential difference could come from an insufficient LMAXMIX in this context. From the ICHARG page on the VASP wiki:

If ICHARG is set to 11 or 12, it is strongly recommended to set LMAXMIX to twice the maximum l-quantum number in the pseudopotentials. Thus, for s and p elements LMAXMIX should be set to 2, for d elements LMAXMIX should be set to 4, and for f elements LMAXMIX should be set to 6.

[yong_han]

Thank you for your information, manuel_engel1.

I re-ran the NSCF calculation with LMAXMIX=4 due to the system including Au element (i.e., d-electrons), as attached (notes: I initially obtained CHGCAR with a k mesh of 1x1x1 from a SCF run, and the attached INCAR and KPOINTS are changed into the setting for calculating DOS from a NSCF run). CHGCAR files are too large and cannot be provided.

I, again, checked and compared the initial CHGCAR (from the initial SCF run) and the final CHGCAR (updated after this new NSCF calculation) and found that these two CHGCAR files are still not identical (i.e., not constant) with the differences between -3.15309E-05 and +3.15209E-05, i.e., the fluctuation is reduced to one-third of that (-9.45330E-05 to +9.45926E-05) with LMAXMIX=2.

Any comments on this?

[manuel_engel1]

Thanks for the additional files and information. I'm happy to help. Unfortunately, I am currently unable to reproduce your problem. I have a few questions that might help us resolve the problem:

1. Could you please provide small excerpts of the CHGCAR files that show this difference so I can look at it on my end?
2. If possible, could you please also provide the OUTCAR files of both runs (at least the beginning of the file).
3. Can you reproduce this difference in a smaller/test system or does it only appear in this particular case?

[yong_han]

>> 1. Could you please provide small excerpts of the CHGCAR files that show this difference so I can look at it on my end?
>> 2. If possible, could you please also provide the OUTCAR files of both runs (at least the beginning of the file).

See the attached "LargeSystem.rar", including INCAR, POSCAR, KPOINTS, POTCAR, out (stdout file), and CHGCAR for two runs, where any CHGCAR is huge, and then only first 1500 lines are provided. As you can see, two CHAGCAR are different:

1496,1500c1496,1500
< 0.15440113625E-02 0.15504970866E-02 0.15577650506E-02 0.15567825362E-02 0.15473540470E-02
< 0.15368922304E-02 0.15300943396E-02 0.15236107920E-02 0.15122826626E-02 0.14977818091E-02
< 0.14880555436E-02 0.14886620265E-02 0.14969618514E-02 0.15055777167E-02 0.15103782753E-02
< 0.15141256463E-02 0.15226117466E-02 0.15379102260E-02 0.15556590697E-02 0.15687489649E-02
< 0.15731848107E-02 0.15701016455E-02 0.15630276939E-02 0.15550364988E-02 0.15488836070E-02
---
> 0.15477169718E-02 0.15551355227E-02 0.15591660908E-02 0.15525076547E-02 0.15359188556E-02
> 0.15231298212E-02 0.15210476022E-02 0.15288366849E-02 0.15233057748E-02 0.14985087300E-02
> 0.14772000586E-02 0.14950643163E-02 0.14970058292E-02 0.15027595800E-02 0.15142334632E-02
> 0.15108874031E-02 0.15219631168E-02 0.15395420378E-02 0.15591580159E-02 0.15682556004E-02
> 0.15699758696E-02 0.15650489440E-02 0.15578342791E-02 0.15501641901E-02 0.15439185117E-02


>> 3. Can you reproduce this difference in a smaller/test system or does it only appear in this particular case?

I tested a small system, and attached two runs as:
SmallSystem_SCF.rar
and
SmallSystem_NSCF.rar

As you can check, for this small system, two CHGCAR file are also not identical with a fluctuation between -4.26173E-06 and +4.27663E-06, although this fluctuation is smaller than those in the above large system.

Thank you for your help!