Page 1 of 1

Questions in relation to dimer output

Posted: Fri Aug 12, 2005 1:10 am
by Kwansun Yoon
The out.dat contains the data for each iteration in the six columns.

What's meaning of the third column, U?

In the website, it was explained that the third column is the energy of the image of the dimer.

But, I can't make sense of this sentence exactly. Please give me an explanation in detail.

If the value is the energy of the image of the dimer, I think that the value will be increased.

The reason is that the energy of the saddle point is higher than that of local minimum state.

But, what's the reason of decrease?

And if the value is positive, is this a sign of problem?

Does the positive value relate with the "WARNING in EDDRMM: call to ZHEGV failed, returncode = x x x"?

Posted: Fri Aug 12, 2005 5:24 pm
by andri
Hi,

Could you email me or post here your out.dat file and maybe the INCAR as well.

There are several different energy values floating around in VASP. The one printed in the third column of the out.dat is NOT the true energy. It is because some artificial temperature (http://cms.mpi.univie.ac.at/vasp/vasp/node117.html) has been introduced for the electrons to aid with the electron structure minimization. The true energy can be obtained using our vef.pl script or by grep-ing "energy without entropy" from the OUTCAR. The true energy is the one labeled as "energy(sigma->0) =". To get the energy for the dimer, the average from the two images as to be taken.

Positive value for the energy usually indicates some type of a problem. How did you choose the initial configuration? I have never seen that error message you mention before. It could have something to do with your positive energy values?

Andri

Posted: Tue Aug 16, 2005 2:49 pm
by graeme
One possibility is that the initial geometry is something crazy. You can check this by running one normal vasp iteration with the POSCAR files in 01 and 02 to make sure the energies make sense.

Another possibility is that your initial images in 01 and 02 are far apart. When the dimer routine interpolates between them, perhaps the geometry no longer makes sense. You can check this by interpolating half way between images 01 and 02 and checking the energy, or by looking at your OUTCAR. If the first energy calculation is reasonable, but the second is positive, the interpolation could be the problem.

The energy issue that Andri mentions is a somewhat subtle issue. The difference between the energy printed in the out.dat file, and the extrapolated energy without entropy that you should use is usually on the order of meV/Atom. So it is important to get the correct energy, but it does not sound like this is the source of the problem.

The dimer energy can go either up or down during the optimization. It is true that the saddle point energy will be higher than your initial minimum, but during a saddle point search, the dimer can be either higher or lower than the true saddle. In fact, in a high dimensional system, the dimer is minimizing along many more degrees of freedom than the one it is maximizing along. This means that the energy tends to drop as the saddle point is found.