nwpw optimization wrong

From NWChem

Viewed 2707 times, With a total of 5 Posts
Jump to: navigation, search

Gets Around
Threads 27
Posts 51
I am having multiple problems optimizing geometry in nwpw at gamma point. My unit cell collapses and atoms become coordinated in an unreasonable fashion. Also, it's taking forever (now already 100 cycles) to optimize and it's still far away from being over. Also, I am getting bunch of strange warnings, such as

Error reading psi - bad unitcell
- Ignored if the following parameter is set
- set nwpw:psi_nolattice .true.

and, usual, pseudopotential incorrectly formatted. What a mess...

It's probably mostly since I am familiar with other periodic boundary condition codes, not NWChem, but logically it shouldn't collapse the unit cell and shouldn't take 100 cycles to optimize crystal taken from the experimental coordinates.

I uploaded two files to the links below, would appreciate any guidance

Jonas

[1]

[2]
Edited On 2:23:20 AM PST - Sat, Nov 10th 2012 by Jbaltrus

Gets Around
Threads 27
Posts 51
both optimizations simply ran out of 400 cycles

  • Bylaska Forum:Admin, Forum:Mod, NWChemDeveloper, bureaucrat, sysop
    Profile
    Send PM
Gets Around
Threads 0
Posts 46
The problem is that you're using SIC with the wrong type of optimizer. You can see the problem in your first iteration.

       ============ Grassmann lmbfgs iteration ============
    >>>  ITERATION STARTED AT Fri Nov  9 16:31:36 2012  <<<
iter. Energy DeltaE DeltaRho
------------------------------------------------------
- 15 steepest descent iterations performed
10 -0.2933164569E+03 -0.28251E+01 0.25142E+02
- 10 steepest descent iterations performed
20 -0.3044814762E+03 -0.38190E+00 0.37921E+00
- 10 steepest descent iterations performed
30 -0.3069735714E+03 -0.12021E+00 0.13592E+00
- 10 steepest descent iterations performed
40 -0.3079887118E+03 -0.37873E-01 0.35121E-01
- 10 steepest descent iterations performed
50 -0.3083151316E+03 -0.94546E-02 0.14078E-01
60 -0.3083580553E+03 0.56843E-13 0.14995E-01
*** energy going up. iteration not terminated
- 10 steepest descent iterations performed
70 -0.3086062132E+03 0.00000E+00 0.53551E-02
80 -0.3086062132E+03 0.28422E-13 0.11288E-27
*** energy going up. iteration not terminated
*** tolerance ok. iteration terminated
>>> ITERATION ENDED AT Fri Nov 9 16:36:53 2012 <<<



My suggestion is to first run it with pbe96 and then turn on SIC. You'll first need to localize the orbitals for an initial guess and set the minimizer to the following

set nwpw:minimizer 4


A strategy would be

nwpw
  xc pbe96
end
task pspw optimize ignore

nwpw
  xc pbe96-0.4SIC
end
set nwpw:minimizer 4
task pspw wannier
task pspw wannier
task pspw optimize ignore

Gets Around
Threads 27
Posts 51
thank you, I will adjust. Is there any effect of the cutoff (I am using 36 from one of yoru papers) on optimization? I will test cutoff WRT bandgap properties but not sure whether there is effect on the optimization

Gets Around
Threads 27
Posts 51
Much better. It still runs out of cycles when optimiznig with bpe, though and I am not sure why. WHat I can see is this, as if there is some numerical inaccuracy that makes the optimization simply oscillate near the minimum and it never converges.

Are there tricks to using some tighter settign (large grid?) to finsih this optimization?

 Step       Energy      Delta E   Gmax     Grms     Xrms     Xmax   Walltime
---- ---------------- -------- -------- -------- -------- -------- --------
@ 304 -312.94981443 5.8D-05 0.02416 0.00798 0.00514 0.03107 9844.8
                                                                   

Taking step in negative mode    1  eval=-1.1D-02 grad= 8.1D-04 step=-7.2D-02
Restricting overall step due to uphill motion. alpha= 0.50

Gets Around
Threads 27
Posts 51
My other structures managed to converge PBE but in PBE-SIC it has difficulties during SCF cycles, what conveger aids should I use to rectify that?

Energy Calculation



         ======= Stiefel conjugate gradient iteration =======
    >>>  ITERATION STARTED AT Tue Nov 13 00:13:13 2012  <<<
iter. Energy DeltaE DeltaRho
------------------------------------------------------
10 -0.3129688105E+03 -0.54783E-03 0.85538E-05
20 -0.3129729364E+03 -0.56193E-03 0.30289E-04
- 10 steepest descent iterations performed
30 -0.3129552999E+03 -0.29230E-05 0.69584E-09
40 -0.3129553282E+03 -0.27634E-05 0.61649E-09
50 -0.3129553551E+03 -0.26250E-05 0.54839E-09
60 -0.3129553806E+03 -0.25040E-05 0.48960E-09
70 -0.3129554051E+03 -0.23972E-05 0.43857E-09
80 -0.3129554285E+03 -0.23023E-05 0.39405E-09
90 -0.3129554511E+03 -0.22173E-05 0.35504E-09
100 -0.3129554728E+03 -0.21407E-05 0.32072E-09
110 -0.3129554938E+03 -0.20714E-05 0.29040E-09
120 -0.3129555142E+03 -0.20082E-05 0.26353E-09
130 -0.3129555340E+03 -0.19505E-05 0.23963E-09
140 -0.3129555532E+03 -0.18975E-05 0.21832E-09
150 -0.3129555719E+03 -0.18487E-05 0.19926E-09
160 -0.3129555901E+03 -0.18035E-05 0.18218E-09
170 -0.3129556079E+03 -0.17616E-05 0.16683E-09
180 -0.3129556253E+03 -0.17226E-05 0.15300E-09
190 -0.3129556423E+03 -0.16861E-05 0.14052E-09
200 -0.3129556590E+03 -0.16520E-05 0.12924E-09
210 -0.3129556754E+03 -0.16200E-05 0.11902E-09
220 -0.3129556914E+03 -0.15900E-05 0.10974E-09
230 -0.3129557071E+03 -0.15616E-05 0.10132E-09
240 -0.3129557226E+03 -0.15348E-05 0.93645E-10
250 -0.3129557378E+03 -0.15095E-05 0.86653E-10
260 -0.3129557528E+03 -0.14855E-05 0.80272E-10
270 -0.3129557675E+03 -0.14627E-05 0.74440E-10
280 -0.3129557820E+03 -0.14410E-05 0.69105E-10
290 -0.3129557963E+03 -0.14204E-05 0.64217E-10
300 -0.3129558104E+03 -0.14007E-05 0.59734E-10
310 -0.3129558243E+03 -0.13819E-05 0.55619E-10
320 -0.3129558380E+03 -0.13639E-05 0.51837E-10
330 -0.3129558516E+03 -0.13466E-05 0.48359E-10
340 -0.3129558649E+03 -0.13301E-05 0.45156E-10
350 -0.3129558781E+03 -0.13142E-05 0.42204E-10
360 -0.3129558912E+03 -0.12989E-05 0.39481E-10
370 -0.3129559041E+03 -0.12843E-05 0.36967E-10
380 -0.3129559169E+03 -0.12701E-05 0.34644E-10
390 -0.3129559295E+03 -0.12565E-05 0.32496E-10
400 -0.3129559420E+03 -0.12433E-05 0.30508E-10
410 -0.3129559544E+03 -0.12306E-05 0.28668E-10
420 -0.3129559666E+03 -0.12183E-05 0.26961E-10
430 -0.3129559787E+03 -0.12064E-05 0.25379E-10
440 -0.3129559907E+03 -0.11949E-05 0.23910E-10
450 -0.3129560026E+03 -0.11838E-05 0.22546E-10
460 -0.3129560144E+03 -0.11730E-05 0.21277E-10
470 -0.3129560261E+03 -0.11625E-05 0.20098E-10
480 -0.3129560376E+03 -0.11523E-05 0.19000E-10
490 -0.3129560491E+03 -0.11424E-05 0.17978E-10
500 -0.3129560605E+03 -0.11328E-05 0.17025E-10
510 -0.3129560717E+03 -0.11235E-05 0.16137E-10
520 -0.3129560829E+03 -0.11143E-05 0.15308E-10
530 -0.3129560940E+03 -0.11055E-05 0.14534E-10
540 -0.3129561050E+03 -0.10968E-05 0.13810E-10
550 -0.3129561159E+03 -0.10884E-05 0.13134E-10


Forum >> NWChem's corner >> Running NWChem



Who's here now Members 0 Guests 1 Bots/Crawler 0


AWC's: 2.5.10 MediaWiki - Stand Alone Forum Extension
Forum theme style by: AWC