Convergence of ionic relaxation

[MBaeker]

Dear all,
I'm running a fairly large (136 atom) supercell of a nickel-titanium alloy with a two-phase boundary.
I have created an initial structure where the two phases are correctly aligned crystallographically and am running a ionic relaxation step.
Here is the INCAR I'm using:

## --- ENERGY ---
ENMAX=400
ENCUT=520
#
## --- Simulation precision ---
PREC=Normal
LREAL=Auto
ROPT=1.e-4 1.e-4
ISMEAR=-5
#
## --- Spin definition ---
ISPIN=2
MAGMOM= 136*1.0
#
## --- KSPACING
KSPACING=0.2
KGAMMA=.TRUE.
## Relaxation
ISIF=2
NSW=50
IBRION=2
## 14/10/16: try POTIM=0.1 instead of default for better convergence
POTIM=0.05
#
NCORE=24
KPAR=16
##
IALGO=38

From the OSZICAR file, I find the following energies for each relaxation step:
1 F= -.82830818E+03 E0= -.82830818E+03 d E =-.828308E+03 mag= 83.7098
2 F= -.82851271E+03 E0= -.82851271E+03 d E =-.204530E+00 mag= 46.4913
3 F= -.82847059E+03 E0= -.82847059E+03 d E =-.162404E+00 mag= 44.7249
4 F= -.82849065E+03 E0= -.82849065E+03 d E =-.182468E+00 mag= 44.4651
5 F= -.82852931E+03 E0= -.82852931E+03 d E =-.221130E+00 mag= 44.2885
6 F= -.82858519E+03 E0= -.82858519E+03 d E =-.277005E+00 mag= 44.1764
7 F= -.82861591E+03 E0= -.82861591E+03 d E =-.307733E+00 mag= 44.1448
8 F= -.82870979E+03 E0= -.82870979E+03 d E =-.938747E-01 mag= 44.0847
9 F= -.82895450E+03 E0= -.82895450E+03 d E =-.338585E+00 mag= 44.0194
10 F= -.82969005E+03 E0= -.82969005E+03 d E =-.107413E+01 mag= 43.6396
11 F= -.99282963E+08 E0= -.99282963E+08 d E =-.992821E+08 mag= 41.1850
12 F= -.82925503E+03 E0= -.82925503E+03 d E =-.639120E+00 mag= 44.8134
As you can see, in step 11, the energy is totally off, which seems worrying to me. Is this normal behaviour for the algorithm (possibly due to too much shifting of the ions?).
Varying the POTIM-parameter does not change this behaviour, I tried the default value, a value of 0.1 and 0.05. In all cases, there is one relaxation step where the energy has increased by several orders of magnitude and the calculation stops after the next step.

When I run the same simulation but with IALGO=48, this does not happen, but I end up with a slightly different value of the energy.

So my questions is how to choose the correct algorithm and what to do about the strange increase in energy. The manual seems rather terse to me at this point and I find it difficult to decide this on my own.

Any help is greatly appreciated,

Martin.

[support_vasp]

Hi,

We're sorry that we didn’t answer your question. This does not live up to the quality of support that we aim to provide. The team has since expanded. If we can still help with your problem, please ask again in a new post, linking to this one, and we will answer as quickly as possible.

Best wishes,

VASP