Mz287

12/5/10:
Run with a walltime of 24 hours, density barely missed convergence. Died on step number 25, previous diff etot = -2.55e-7. Rerunning for 36 hours instead, will report full parameters after that run.

12/9/10:
Job got to the top of the queue, however was aborted by the PBS server. Most likely due to scratch file system error that is currently effecting much of NERSC. Job resubmitted to the queue.

12/11/10:
In light of evidence from mz83, I'm setting prtelf = 1 in the input file. I think I should be able to print the ELF immediately after converging the density.

12/14/10:
Density successfully converged after 31 SCF steps and ELF printed.
Final etot = -1003.3689762576 Hartrees.
Final diff in etot = -5.062E-09.
nkpt = 258.
natom = 160.
ngfft = 180 64 120.
ecut = 25 Hartrees.
Run on 32 nodes, 8 ppn.
It took 107925.0 seconds per processor = 29 hours 59 minutes
Size of abinit ELF output file: 11.7 mb
Size of ELF vesta file: 31.6 mb

3/7/11:

Elf maximum found:

nGrid = < 120 , 64 , 180 >
max ELF = 9.99380455945891E-01
location = < 10 , 22 , 108 >

Therefore fractional coordinates of the maximum = 10/120, 22/64, 108/180 = 0.083333333, 0.34375, 0.6

3/29/11:
Attempted to calculate I-Hirshfeld charges, however separate_atomtypes.r returned a segmentation fault from the DEN file, likely owing to its size (12MB).

3/29/11:
Talked to Josh, and he made a new version of his program with a larger limit on input file size (as I understand it). Calculated charges successfully after fix. For more information, see Hirschfeld-I analysis.

3/30/11:
Put in large-core Te psp calculation in order to allow for I-H charges to be calculated that can be compared to those for 318 and 358 whereby the large-core psp was used.

4/13/11:
Attempted to reconverge the density with the large-core Te psp, however the calculation ran out of time. I'll have to break it up into 25 SCF steps and then 25 more. Putting in the first leg now.

5/14/11:
Finally getting back to these calcs. The first leg went smoothly. Here are the specs:
Density NOT converged after 28 SCF steps.
Final etot = -1651.42684592877 Hartrees.
Final diff in etot = E-03
nkpt = 258.
natom = 160.
ngfft = 180 64 120.
ecut = 25 Hartrees.
Run on 32 nodes, 8 ppn.
It took 116234.6 seconds per processor = 32 hours 17 minutes

I don't have the exact final diff in etot because I accidentally wrote over the log file while preparing the input file for the next iteration. Oh well. I'm upping the # of iterations to 30 steps, because I have a walltime of 36 hours, and thus two more should be safe. It's back in the queue.

5/26/11:
This is the problem with forgetting to update my wiki… sometimes things get a little messed up. Let's review:

4/15 - Working on converging the density of mz287 with the large core Te psp.

4/16 - Alex and Josh realized that the disorder present in the mz287 CIF would yield non-physical charges. Alex generated a new CIF which has four times the atoms but has no disorder. I set up a calculation with the new CIF and put it into the queue, but I did not record it here.

4/18 - The faked order calculation finishes, but I wasn't checking on it and thus it went unknown for some time. Here are the specs:

Density successfully converged after 19 SCF steps and ELF printed.
Final etot = -978.80727124057 Hartrees.
Final diff in etot = -4.228E-09.
nkpt = 432.
natom = 140.
ngfft = 180 64 120.
ecut = 25 Hartrees.
Run on 54 nodes, 8 ppn.
It took 28978.2 seconds per processor = 8 hours 3 minutes

This was actually all that I needed to do for 287 at this point.

5/12 - Alex e-mailed me to check up on the status of mz287. Because I hadn't recorded anything here about the new structure, I basically completely forgot and thought that he was talking about the old (bad) structure but with large core Te psps. Thus I spent some of my time and some computational time on these calculations, which we didn't actually need.

5/25 - I hadn't even finished what I believed to be the calculations of interest when I saw the directory new287. That led to me remembering / figuring out exactly wtf was going on, thus I killed the bad jobs in the queue and calculated the IH charges of the new287 density. Seems my ride on the struggle bus has finally come to an end.

6/2/11:
Large core new287 finished. Here are the specs:
Density successfully converged after 21 SCF steps.
Final etot = -1627.6920393136 Hartrees.
Final diff in etot = -5.886E-09.
nkpt = 432.
natom = 140.
ngfft = 180 64 120.
ecut = 25 Hartrees.
Run on 54 nodes, 8 ppn.
It took 107925.0 seconds per processor = 10 hours 36 minutes

Unless otherwise stated, the content of this page is licensed under Creative Commons Attribution-ShareAlike 3.0 License