Density successfully converged after 16 SCF cycles.
Final etot = -514.50481032421 Hartrees.
Final diff in etot = 1.151E-09.
nkpt = 258.
natom = 78.
ngfft = 54 180 54.
ecut = 25 Hartrees.
Run on 32 nodes, 8 ppn.
It took 7711.1 seconds per processor = 2 hours 8 minutes
Iterative Hirschfield charges successfully calculated.
V charge range: Between 1 and -2
Te charge range: Between 0 and -5
O charge range: Between -1 and 2
N charge range: Between -1 and 1
C charge range: Between -1 and 1
H charge range: Between -1 and 1
Unfortunately, the results do not seem correct. All aspects of the process have been checked, however no mistakes were found. Will wait until school starts up again to meet with Josh and see if he can find any problems.
Met with Josh, however no problems were found. We did come up with some sanity checks, however:
Make an image from the _DEN file, check it against the CIF. This should check the abinit calculation.
Check that qtot = dintt from 0 to inf of r^2 rho of r 4 pi dr equals the expected total charge. This should check the .ae file generation.
Run hirschfield on another compound to see if it also gives screwy charges. This should check the hirschfield code and Josh's iterative addition.
These will all be done shortly.
Tried to check the qtot integral, but I ran into a few problems. I did, however, successfully run the iterative hirschfield on mz133 and got screwy charges, therefore it appears that the problems are with the method as a whole, rather than problems with this compound specifically.
Met with Josh again, determined that the problem was in the .ae file generation by integrating charge density on mathematica. We realized that we had never successfully done a spin polarized calculation, and this had been our solution for negative ions. In fact, only half of the electrons were being read in, therefore there wasn't actually a negative ion. Unfortunately we also determined that GGA / LDA codes do not yield stable cations, therefore we are switching to an atomic HF code. For more information, see Hirschfeld-I analysis.
Missed a few updates, but I'll give an overview:
- The new atomic HF code successfully produces reliable .ae files for negative, positive, and neutral species. See Hirschfeld-I analysis for exactly how to do this.
- After successfully creating all of the .ae files we were still having problems, therefore we thought it might have to do with not having the d-orbitals present in the tellurium pseudopotential. I generated a new pseudopotential with these orbitals present (for more information, see Generating Pseudopotentials) and then reran the abinit calculations to generate new density files. These still did not yield good iterative charges, sending us back to the drawing board. For reference, the calculations with the new pseudopotential took 9399.1 seconds = 2 hours 37 minutes. This corresponds to about a 20% increase in computational time.
- While doing a little extra digging, I found an extra negative sign in the iterative hirshfield code that was originally attributed to the negative sign of the electron's charge back when its present made our charges reasonable (when everything else was wrong, like my abinit code and all of our .ae files). It turns out it was responsible for the failures of the iterative hirchfield, so by taking it out we were able to calculate accurate charges. Hurrah!
Now the goal is to also compute the AIM charges for comparison. I'll be running these calculations tonight. For all updates on this, please refer to AIM/bader