Very high energies, no convergence, perhaps NBAND issue?

[jasius]

Hi all, I am experiencing a very strange issue which I can't diagnose. The system is ionic insulator, rather simple for calculations. But when I cut the slab and run various terminations or k-point convergence tests, the energies during SCF are just off by 3 orders of magnitude and they never converge. I tried changing ALGO, SYMM, ADDGRID - nothing worked *except* if I run it twice. I run it from scratch, it does not converge, I simply restart and then it does. I am really in need of somebody to diagnose why this thing happens. Thank you very much, see the attached files

This particular job signals about NBANDS but I has NELECT=352 and NBAND=256, I only need 176 bands in this case so I do not understand the problem.

DAV: 1 -0.242756585037E+06 -0.24276E+06 -0.36683E+06 5536 0.134E+04
DAV: 2 -0.284071665246E+06 -0.41315E+05 -0.37254E+05 8992 0.422E+03
DAV: 3 -0.291763815303E+06 -0.76922E+04 -0.70980E+04 9216 0.146E+03
DAV: 4 -0.294590670420E+06 -0.28269E+04 -0.26473E+04 9216 0.884E+02
DAV: 5 -0.295940239943E+06 -0.13496E+04 -0.12825E+04 9216 0.624E+02 0.396E+02
RMM: 6 -0.177642471836E+06 0.11830E+06 -0.12684E+04 8846 0.435E+02 0.260E+02
RMM: 7 -0.318466977241E+05 0.14580E+06 -0.47415E+04 9216 0.629E+02 0.203E+02
RMM: 8 0.214351789183E+05 0.53282E+05 -0.12858E+05 9098 0.124E+03 0.143E+02
RMM: 9 0.822280458195E+05 0.60793E+05 -0.42640E+05 6127 0.150E+03 0.274E+02
RMM: 10 0.392913516274E+05 -0.42937E+05 -0.42912E+05 7545 0.280E+03 0.367E+02
RMM: 11 -0.142384867914E+05 -0.53530E+05 -0.39589E+05 8785 0.287E+03 0.405E+02
RMM: 12 -0.334059940770E+05 -0.19168E+05 -0.26796E+05 9145 0.233E+03 0.385E+02
RMM: 13 -0.574058842071E+05 -0.24000E+05 -0.25062E+05 9137 0.191E+03 0.383E+02
RMM: 14 -0.770726553066E+05 -0.19667E+05 -0.18090E+05 9054 0.142E+03 0.530E+02
RMM: 15 -0.894132013729E+05 -0.12341E+05 -0.16028E+05 9149 0.109E+03 0.904E+02
RMM: 16 -0.118317999948E+06 -0.28905E+05 -0.10169E+05 9178 0.101E+03 0.119E+03
RMM: 17 -0.176172453039E+06 -0.57854E+05 0.31582E+04 9052 0.202E+03 0.122E+03
RMM: 18 -0.193997691386E+06 -0.17825E+05 -0.19679E+05 8538 0.159E+03 0.112E+03
RMM: 19 -0.179342687148E+06 0.14655E+05 0.78150E+03 7933 0.248E+03 0.110E+03
RMM: 20 -0.169754809221E+06 0.95879E+04 -0.53248E+04 8824 0.111E+03 0.101E+03
RMM: 21 -0.169129618466E+06 0.62519E+03 -0.45072E+04 8002 0.824E+02 0.950E+02
RMM: 22 -0.227377827347E+06 -0.58248E+05 -0.95137E+04 9110 0.378E+03 0.830E+02
RMM: 23 -0.192586764051E+06 0.34791E+05 -0.27055E+04 8190 0.140E+03 0.777E+02
RMM: 24 -0.189446460111E+06 0.31403E+04 -0.38728E+04 7766 0.123E+03 0.750E+02
RMM: 25 -0.202926647212E+06 -0.13480E+05 -0.20712E+04 7922 0.708E+02 0.739E+02
RMM: 26 -0.194412036969E+06 0.85146E+04 0.12020E+04 8262 0.110E+03 0.697E+02
RMM: 27 -0.196457236159E+06 -0.20452E+04 0.96485E+04 7457 0.165E+03 0.733E+02
RMM: 28 -0.200197580513E+06 -0.37403E+04 -0.35358E+04 8469 0.991E+02 0.660E+02
RMM: 29 -0.184183334252E+06 0.16014E+05 -0.16195E+06 8987 0.526E+03 0.698E+02
RMM: 30 -0.197484586839E+06 -0.13301E+05 0.71757E+04 8963 0.112E+03 0.685E+02

[christopher_sheldon1]

Hi Jasius,

Thank you for your question. I've repeated your calculation in VASP 6.5.1 and found the same issue. I do not think that NBANDS is the issue.

By setting ICHARG = 0, the charge density is calculated from the initial wavefunctions. In your case, these are not good approximations. I fixed this by changing ICHARG from 0 to 2. When ICHARG = 2 (the default), the initial charge density is taken as a superposition of atomic charge densities. The electronic structure then converges smoothly. You may also need to increase

Code: Select all

NELM      =   55

so that there are enough SCF steps to achieve convergence. When I used ICHARG = 2, it did not converge within 55 steps.

Did this answer your question?

Best wishes,

Chris

[jasius]

I am very grateful for your timely response, it worked!

[christopher_sheldon1]

Good, glad that I could help.

If your question is answered, I'll close the topic on Friday afternoon.

Best,

Chris

[jasius]

please close