DIIS turns off after iteration 300


Clicked A Few Times
Hello all,

I have trouble with convergence of a quite big system. The main problem is that after the 300th iteration (after setting Maxiter to 2000) the DIIS is disabled after the 300th iteration, which causes the energy to spike and oscillate. Is there any way to elongate application of DIIS?

d= 0,ls=0.5,diis 295 -5652.6397539560 -4.48D-05 1.38D-05 6.70D-05 6540.9
                                                    1.80D-05  6.59D-05
d= 0,ls=0.5,diis 296 -5652.6398117909 -5.78D-05 7.97D-06 7.01D-05 6559.7
1.12D-05 6.58D-05
d= 0,ls=0.5,diis 297 -5652.6398430119 -3.12D-05 2.54D-05 7.82D-05 6578.3
3.15D-05 7.19D-05
d= 0,ls=0.5,diis 298 -5652.6399542571 -1.11D-04 4.34D-05 5.29D-05 6596.5
5.34D-05 5.61D-05
d= 0,ls=0.5,diis 299 -5652.6401356067 -1.81D-04 1.20D-05 3.37D-05 6614.6
1.82D-05 4.58D-05
d= 0,ls=0.5,diis 300 -5652.6401826528 -4.70D-05 9.41D-06 3.64D-05 6632.7
1.56D-05 4.89D-05
d= 0,ls=0.0 301 -5652.6402221707 -3.95D-05 1.06D-04 6650.6
1.69D-04
d= 0,ls=0.0 302 -5652.6406961749 -4.74D-04 9.50D-05 6668.4
1.74D-04
d= 0,ls=0.0 303 -5652.6411178178 -4.22D-04 2.26D-04 6686.3
1.41D-04
d= 0,ls=0.0 304 -5652.6399116946 1.21D-03 9.58D-04 6704.1
1.13D-03
d= 0,ls=0.0 305 -5652.5672689413 7.26D-02 5.01D-03 6722.2
7.66D-03
d= 0,ls=0.0 306 -5645.2062333062 7.36D+00 2.12D-02 6740.3
1.95D-02
d= 0,ls=0.0 307 -5564.7959031021 8.04D+01 1.03D-01 6758.2
9.97D-02
d= 0,ls=0.0 308 -5127.6086332086 4.37D+02 1.20D-01 6777.0
1.19D-01
d= 0,ls=0.0 309 -4571.6537443236 5.56D+02 1.08D-01 6795.4
1.10D-01
d= 0,ls=0.0 310 -4014.5514583279 5.57D+02 1.27D-01 6813.8
1.27D-01
d= 0,ls=0.0 311 -3708.1652773820 3.06D+02 1.40D-01 6832.9
1.40D-01
d= 0,ls=0.0 312 -3629.1232425122 7.90D+01 1.32D-01 6851.4
1.32D-01
d= 0,ls=0.0 313 -3717.8020511332 -8.87D+01 1.31D-01 6869.4
1.31D-01
d= 0,ls=0.0 314 -3628.5756631809 8.92D+01 1.31D-01 6887.6
1.31D-01
d= 0,ls=0.0 315 -3718.4657086714 -8.99D+01 1.31D-01 6905.6
1.31D-01
d= 0,ls=0.0 316 -3628.5107692061 9.00D+01 1.31D-01 6923.7
1.31D-01
d= 0,ls=0.0 317 -3718.5210035995 -9.00D+01 1.31D-01 6941.8
1.31D-01
d= 0,ls=0.0 318 -3628.5049027627 9.00D+01 1.31D-01 6960.9
1.31D-01
d= 0,ls=0.0 319 -3718.5258294774 -9.00D+01 1.31D-01 6979.6
1.31D-01
d= 0,ls=0.0 320 -3628.5043865491 9.00D+01 1.31D-01 6997.7
1.31D-01
d= 0,ls=0.0 321 -3718.5262524879 -9.00D+01 1.31D-01 7016.3
1.31D-01
d= 0,ls=0.0 322 -3628.5043412685 9.00D+01 1.31D-01 7035.3
1.31D-01
d= 0,ls=0.0 323 -3718.5262895825 -9.00D+01 1.31D-01 7054.1
1.31D-01
d= 0,ls=0.0 324 -3628.5043372975 9.00D+01 1.31D-01 7072.8
1.31D-01
d= 0,ls=0.0 325 -3718.5262928355 -9.00D+01 1.31D-01 7091.4
1.31D-01
d= 0,ls=0.0 326 -3628.5043369518 9.00D+01 1.31D-01 7109.4
1.31D-01
d= 0,ls=0.0 327 -3718.5262931223 -9.00D+01 1.31D-01 7127.6

Forum Vet
Could you post the input file?
Thanks

Clicked A Few Times
start mnclust

geometry all noautosym noautoz nocenter
Mn 0.227857745 2.049454583 -4.137926720
O -1.717479181 2.814620406 -4.160164219
O -0.719178031 -0.159425481 -3.803350720
H -0.274045579 -0.901793091 -3.388203237
Mn -2.436101375 1.186131699 -3.137954517
Mn -2.221619559 -2.065635695 -2.750351158
O -2.948011924 -2.036021005 -4.454183046
O -3.250899365 -0.658668686 -2.173843645
Mn -4.455985850 -0.705241422 -3.928295851
O -3.371414816 0.936508680 -4.807576964
Ca -1.746398934 -0.624164101 -5.957745584
end

geometry f1 noautosym noautoz nocenter
O -1.717479181 2.814620406 -4.160164219
O -0.719178031 -0.159425481 -3.803350720
H -0.274045579 -0.901793091 -3.388203237
O -2.948011924 -2.036021005 -4.454183046
O -3.250899365 -0.658668686 -2.173843645
O -3.371414816 0.936508680 -4.807576964
Ca -1.746398934 -0.624164101 -5.957745584
end

geometry f2 noautosym noautoz nocenter
Mn -2.221619559 -2.065635695 -2.750351158
end

geometry f3 noautosym noautoz nocenter
Mn -4.455985850 -0.705241422 -3.928295851
end

geometry f4 noautosym noautoz nocenter
Mn -2.436101375 1.186131699 -3.137954517
end

geometry f5 noautosym noautoz nocenter
Mn 0.227857745 2.049454583 -4.137926720
end

basis spherical
* library 6-31g*
end

title "first fragment: the rest of the system"

charge -7
set geometry f1
dft
odft
mult 1
xc b3lyp
direct
vectors input atomic output f1.mos
maxiter 300
end

task dft ignore

title "MnIV spin up"

charge 4
set geometry f2
dft
odft
mult 4
xc b3lyp
direct
vectors input atomic output f2.mos
maxiter 300
end

task dft ignore

title "MnIII spin up"

charge 3
set geometry f3
dft
odft
mult 5
xc b3lyp
direct
vectors input atomic output f3.mos
maxiter 300
end

task dft ignore

title "MnIV spin down"

charge 4
set geometry f4
dft
odft
mult -4
xc b3lyp
direct
vectors input atomic output f4.mos
maxiter 300
end

task dft ignore

title "MnIII spin down"

charge 3
set geometry f5
dft
odft
mult -5
xc b3lyp
vectors input atomic output f5.mos
maxiter 300
end

task dft ignore

title "preopt"

charge 7
set geometry all
dft
mulliken
odft
mult 1
xc b3lyp
vectors input fragment f1.mos f2.mos f3.mos f4.mos f5.mos output pre_all.mos
maxiter 2000
end

task dft optimize

Clicked A Few Times
Is there a way to keep it as long as the maxiter? it should be a keyword, but I cannot find it anywhere in the documentation or the forums.

Clicked A Few Times
I am aware that DIIS is not compatible entirely with open shell and UHF, but convergence in this case is key.

Forum Vet
Try
convergence ncyds 2001

https://github.com/nwchemgit/nwchem/wiki/Density-Functional-Theory-for-Molecules#convergen...

Clicked A Few Times
Thank You a lot! that should do it!


Forum >> NWChem's corner >> Running NWChem