ABINIT Discussion Forums

Posted: **Thu May 05, 2011 5:24 pm**

Dear all

I am performing response function calculations on a rocksalt structure, and have this warning in the abinit LOG:

Code: Select all

 ITER STEP NUMBER    11
-P-0000  cgwf3: problem of minimisation (likely metallic), set resid to -2
-P-0000  cgwf3: converged with theta=      0.0000E+00
-P-0000  cgwf3: problem of minimisation (likely metallic), set resid to -2
-P-0000  cgwf3: converged with theta=      0.0000E+00
-P-0000  cgwf3: problem of minimisation (likely metallic), set resid to -2
-P-0000  cgwf3: converged with theta=      0.0000E+00
-P-0000  leave_test : synchronization done...
 vtorho3: loop on k-points and spins done in parallel
 vtorho3 : MPI_ALLREDUCE, buffer of size26543160 bytes
 ETOT 11   5.0261545528265     -1.113E-05 6.769E+00 2.071E-03

I couldn't find the 'resid' keyword on the abinit website. My material is metallic indeed but I don't understand what the warning means.

Could someone please share their experience about this?

Thank you

Boris

Posted: **Fri May 06, 2011 2:29 am**

This warning is coming from the cgwf3 routine, which does the actual band-by-band conjugate gradient search for the minimal energy at a given k point. "resid" is an internal variable, it is the current wavefunction residual. It is being set to -2 internally as a flag for the current situation (see Phys Rev B 55, 10337 (1997) Eq. 31--it's a line minimization algorithm). As this is a warning it doesn't necessarily signal calculation failure, more that the algorithm is having trouble. I think the best thing to try is to increase the number of bands you are giving to your system. The default is a pretty small number of initially unoccupied bands, try increasing it by 20% or so.

Posted: **Mon May 09, 2011 5:21 pm**

Thank you very much for your answer, Prof. Zwanziger

I will try to increase the number of bands.

ABINIT Discussion Forums

Problem of minimisation (likely metallic)

Problem of minimisation (likely metallic)

Re: Problem of minimisation (likely metallic)

Re: Problem of minimisation (likely metallic)