Dear All,
I am finding some issues in doing anneal calculations.
I am heating some VO2 in the T range 100-120K within 1500 steps and with a
timestep of 0.5 fs and with +U correction for Ti d orbitals.
I submit the job which ends in 24 hrs (walltime of the machine) after ~1000
steps.
Then I restart it, expecting ~500 steps more to accomplish the duty.
But actually, even if the job restarts properly (from .XV and other saved
data) it runs a lot longer (it actually redoes the same 1000 steps in 24
hrs). Where am I wrong?
I am using the following parallel version with ELSI implementation with
("+U" for metallic ions)
Whatever the value of " MD.FinalTimeStep" I specify in the fdf, restarting
the job I notice that the steps calculated in the previous run are totally
neglected ( the number of MD.FinalTimeStep turns to be pretty larger than
that specified in the fdf file)
Siesta Version : MaX-1.2.1
Architecture : Master-template
Compiler version: GNU Fortran (GCC) 8.4.0
Compiler flags : mpif90 -O2
This the fdf section related to Anneal
MD.TypeOfRun Anneal
MD.AnnealOption TemperatureAndPressure
MD.LengthTimeStep 0.5 fs
MD.FinalTimeStep 1500
MD.InitialTemperature 100 K
MD.TargetTemperature 120 K
MD.TargetPressure 0. GPa
MD.TauRelax 200 fs
This other tags
solution-method elsi
elsi-solver elpa
number-of-eigenstates -50
elsi-elpa-gpu 1
elsi-elpa-flavor 2 # Either 1 or 2; 2 is better
elsi-output-level 3 # (Optional) for verbosity
WriteDM .true.
MD.UseSaveXV .true.
PAO.BasisSize DZP
PAO.EnergyShift 0.01 Ry
%block LDAU.proj
V 1 # number of shells of projectors
n=3 2 # n, l
0.75 0.00 # U(eV), J(eV)
0.000 0.000 # rc, \omega (if zero default values are taken)
%endblock LDAU.proj
this the .MDE file
# Step T (K) E_KS (eV) E_tot (eV) Vol (A^3) P (kBar)
1 250.02 -116947.36381 -116936.92542 3301.123 14.302
2 249.64 -116947.35021 -116936.92769 3301.833 13.943
...
...
...
1020 261.50 -116928.47588 -116917.55842 3322.508 -1.376
1021 262.22 -116928.46900 -116917.52153 3322.441 -1.225
# Step T (K) E_KS (eV) E_tot (eV) Vol (A^3) P (kBar)
1 262.91 -116928.45950 -116917.48331 3322.382 -1.056
2 263.55 -116928.45138 -116917.44823 3322.330 -0.922
...
...
...
684 298.91 -116876.87470 -116864.39549 3516.631 14.232
685 298.57 -116876.85251 -116864.38742 3517.383 14.166
....
....
that is
1021+681=1702 which is pretty larger than the initially specified
MD.FinalTimeStep (1500)
I am really unable to figure out the origin of this behaviour and in this
sense I would be extremely grateful if anyone could help me.
Thanks in advance.
Very best,
Giacomo
--
SIESTA is supported by the Spanish Research Agency (AEI) and by the European
H2020 MaX Centre of Excellence (http://www.max-centre.eu/)
