Hello users:
I am running a molecular dynamics simulation with SIESTA and the job keep
running into situations where the forces grow large all of a sudden and
then it will go back to normal in the next time step.
For example: take a look below at the maximum atomic force printed for the
MD steps [this is not the entire trajaectory; I am just showing a tiny
segment to illustrate this problem].
Max 8.169457 constrained
Max 7.725134 constrained
Max 7.231393 constrained
Max 6.698717 constrained
Max 6.434356 constrained
Max 180.612893 constrained <--------- large force of 180.61 eV/Ang
here
Max 6.351384 constrained
Max 6.500315 constrained
Max 6.616196 constrained
Max 6.705574 constrained
Max 7.375029 constrained
Max 7.971425 constrained
Max 8.440149 constrained
Max 8.768627 constrained
Now take a look at the SCF loop for the MD step where the large force
was generated:
iscf Eharris(eV) E_KS(eV) FreeEng(eV) dDmax
Ef(eV) dHmax(eV)
scf: 1 -217188.442539 -217204.666859 -217206.302454 0.064243
-1.492048 5.165406
scf: 2 -146020.001861 -193579.811302 -193580.989056 1.710060
-0.338743357.676349
scf: 3 -217800.695121 -217208.716553 -217210.351765 1.701213
-1.495349 6.087548
scf: 4 -217213.535597 -217211.812703 -217213.449302 0.032944
-1.494384 2.570971
scf: 5 -217212.451595 -217212.518659 -217214.156524 0.027437
-1.493541 0.659157
scf: 6 -217212.666366 -217212.617779 -217214.255836 0.005069
-1.493661 0.297730
scf: 7 -217212.671543 -217212.651068 -217214.289179 0.002601
-1.493708 0.171617
scf: 8 -217212.670437 -217212.664758 -217214.302857 0.002015
-1.493723 0.141605
scf: 9 -217212.676500 -217212.671768 -217214.309790 0.001025
-1.493792 0.096767
scf: 10 -217212.679444 -217212.676034 -217214.314015 0.000846
-1.493810 0.083722
scf: 11 -217212.682014 -217212.679321 -217214.317264 0.000677
-1.493834 0.059076
scf: 12 -217212.684219 -217212.682043 -217214.319956 0.000602
-1.493864 0.037595
scf: 13 -217212.684718 -217212.683469 -217214.321359 0.000377
-1.493896 0.032728
scf: 14 -217212.685645 -217212.684620 -217214.322474 0.000429
-1.493929 0.029768
scf: 15 -217212.686201 -217212.685444 -217214.323249 0.000401
-1.493955 0.027512
scf: 16 -217212.687024 -217212.686259 -217214.323994 0.000466
-1.493982 0.024955
scf: 17 -217212.687903 -217212.687112 -217214.324756 0.000566
-1.494013 0.021878
scf: 18 -217212.688413 -217212.687788 -217214.325350 0.000508
-1.494041 0.018827
scf: 19 -217212.689186 -217212.688519 -217214.325984 0.000660
-1.494070 0.014994
scf: 20 -217212.689633 -217212.689113 -217214.326468 0.000768
-1.494101 0.010350
scf: 21 -217212.689427 -217212.689275 -217214.326592 0.000290
-1.494111 0.008499
scf: 22 -217212.689461 -217212.689369 -217214.326663 0.000185
-1.494118 0.006871
scf: 23 -217212.689521 -217212.689448 -217214.326723 0.000161
-1.494124 0.006698
scf: 24 -217212.689528 -217212.689492 -217214.326755 0.000075
-1.494126 0.006319
scf: 25 -217212.689561 -217212.689528 -217214.326778 0.000063
-1.494128 0.005641
scf: 26 -217212.689564 -217212.689544 -217214.326788 0.000047
-1.494127 0.004544
scf: 27 -217212.689564 -217212.689554 -217214.326799 0.000053
-1.494124 0.003604
scf: 28 -217212.689567 -217212.689561 -217214.326808 0.000053
-1.494119 0.002869
scf: 29 -217212.689568 -217212.689565 -217214.326817 0.000059
-1.494113 0.002762
scf: 30 -217212.689569 -217212.689567 -217214.326822 0.000037
-1.494108 0.002618
scf: 31 -217212.689566 -217212.689567 -217214.326826 0.000028
-1.494103 0.002409
scf: 32 -217212.689566 -217212.689567 -217214.326831 0.000032
-1.494099 0.002048
scf: 33 -217212.689569 -217212.689568 -217214.326835 0.000038
-1.494095 0.001620
scf: 34 -217212.689576 -217212.689572 -217214.326839 0.000037
-1.494093 0.001183
scf: 35 -217212.689582 -217212.689577 -217214.326841 0.000025
-1.494093 0.000872
SCF Convergence by DM+H criterion
max |DM_out - DM_in| : 0.0000250455
max |H_out - H_in| (eV) : 0.0008717786
SCF cycle converged after 35 iterations
Using DM_out to compute the final energy and forces
No. of atoms with KB's overlaping orbs in proc 0. Max # of overlaps:
222 160
siesta: E_KS(eV) = -217212.6890
siesta: Atomic forces (eV/Ang):
----------------------------------------
Tot -87.051127 -112.933365 -273.703901
----------------------------------------
Max 180.612893
Res 19.109629 sqrt( Sum f_i^2 / 3N )
----------------------------------------
Max 180.612893 constrained
As far as I can tell, there is nothing wrong in the SCF convergence but yet
still the forces shoot up suddenly and then goes back to normal in the next
step. This causes the temperature to shoot up and throws the system out of
equilibrium for a while, resulting in an increase in the total energy
(energy conservation is destroyed). It also occurs a couple of times. Could
this be an eggbox effect?
Thank you.
PS: I am using the latest version of SIESTA
