Dear Carlos,
Sorry to interrupt you again. I have 3 more questions which might be related to
each other:In the output file of Siesta I see the "Maximum dynamic memory
allocated" increases for every cycle (starting with about 4MB and going up to
about 200MB). Would be thankful to help me why this happen?
The second question is that in the continuation job I see the mentioned memory
optionstart again from 4MB which is a little strange to me. If it is a real
continuation job why allocatedmemory decreases suddenly?
According to your last comment:"I see you are using a large Pulay number, i.e.
6. I guess you find that number necessary for convergence. This means that the
new density matrix is constructed using the input and output matrices of the
DM.NumberPulay (=6 in your case) former SCF cycles. Only the last density
matrix is stored in the output files, while the rest is stored in memory and so
lost if TranSIESTA stops."the last question is that is there any way to keep
all density matrices and helping to have a true continuation run? How about
using "DM.PulayOnFile .true." option?
Thanks in advance for your help,Zeynab
On Monday, July 4, 2016 8:44 AM, Carlos Sanz <[email protected]> wrote:
Zeynab,
just a last check. Make sure you copy all the necessary restarting files in
the directory where you rerun TranSIESTA. In my previous messages I mainly
focus on those files that are specific for a TranSIESTA run, but you should
also have the typical SIESTA files for restarting too (e.g.: *.DM, *.XV, ....).
Mainly those you require to use in the fdf file (e.g.. DM.UseSaveDM,
MD.UseSaveXV, MD.UseSaveCG,..), but also there might be other important
restarting files read by SIESTA by default in a rerun. Usually I copy the whole
directory of the previous run into the new directory to avoid missing any file
needed for restarting. This should work in your case unless any file was not
completed before the previous TranSIESTA run stopped.
If still you see the same poor SCF convergence, then you should consider
seriously what Nick is suggesting.
But, if after you play with the convergence parameters for a while, you still
see those large oscillations in the first SCF steps after restarting, then
perhaps the main problem is the N-dimensional energy well close to the KS
minimum. I see you are using a large Pulay number, i.e. 6. I guess you find
that number necessary for convergence. This means that the new density matrix
is constructed using the input and output matrices of the DM.NumberPulay (=6 in
your case) former SCF cycles. Only the last density matrix is stored in the
output files, while the rest is stored in memory and so lost if TranSIESTA
stops.
best regards,
SIMUNE support team
El 04/07/16 a las 08:00, zeynab mohamad hoseyni escribió:
Dear Nick,
Thanks a lot for your prompt reply and your guidance..
All the Best, Zeynab
On Monday, July 4, 2016 6:58 AM, Nick Papior <[email protected]> wrote:
Sometimes restarting from a not-so-well converged density matrix ends up
going out of the convergence basin. This seems to be what happens here.
Secondly, it seems you should change some of your calculation parameters. It
converges with very low rate. Regarding transiesta convergence, please see
previous mails on the mailing list.
2016-07-04 7:02 GMT+02:00 zeynab mohamad hoseyni <[email protected]>:
Hi Carlos,
Thanks a lot for the answer. As you said I had a look at the last lines of
the OUTPUT.fdf for the first run and the first lines of that of continuation
run. I think the energies aren't that much close. Would be thankful to let me
know your idea about it:
Here I copy the last lines of the first run:
transiesta: TSiscf Eharris(eV) E_KS(eV) FreeEng(eV) dDmax Ef(eV)
transiesta: 59 -27749.7793 -27802.2693 -27802.2693 0.6819 -3.5816
TranSiesta: Qsol,Qtot: 715.511 713.000 1.000 transiesta: 60 -27829.2053
-27817.4318 -27817.4318 0.7644 -3.5816 TranSiesta: Qsol,Qtot: 715.936
713.000 1.000 transiesta: 61 -27805.3324 -27808.0481 -27808.0481
0.9509 -3.5816 TranSiesta: Qsol,Qtot: 715.196 713.000 1.000 transiesta:
62 -27809.7837 -27812.0000 -27812.0000 0.7693 -3.5816 TranSiesta:
Qsol,Qtot: 715.049 713.000 1.000 transiesta: 63 -27820.2730
-27817.4024 -27817.4024 0.7280 -3.5816 TranSiesta: Qsol,Qtot: 714.691
713.000 1.000
And here is the one of continuation run:
transiesta: TSiscf Eharris(eV) E_KS(eV) FreeEng(eV) dDmax Ef(eV)
transiesta: 1 -26177.9898 -26389.7478 -26389.7478 9.5642 -3.5816
timer: Routine,Calls,Time,% = IterSCF 1 0.000 0.00 elaps:
Routine,Calls,Wall,% = IterSCF 1 0.000 0.00 timer:
Routine,Calls,Time,% = TS 1 6109.618 15.80 elaps:
Routine,Calls,Wall,% = TS 1 127.847 15.83 timer:
Routine,Calls,Time,% = IterSCF 1 38369.594 99.21 elaps:
Routine,Calls,Wall,% = IterSCF 1 801.301 99.21 TranSiesta:
Qsol,Qtot: 868.854 713.000 1.000 transiesta: 2 4588.1213
-1130.8881 -1130.8881 22.0851 -3.5816 TranSiesta: Qsol,Qtot: 767.213
713.000 1.000 transiesta: 3 -33998.9321 -24075.0873 -24075.0873
11.8589 -3.5816 TranSiesta: Qsol,Qtot: 720.054 713.000 1.000 transiesta:
4 -28365.1093 -27515.3187 -27515.3187 4.6533 -3.5816 TranSiesta:
Qsol,Qtot: 727.959 713.000 1.000 transiesta: 5 -27507.8708
-27520.8220 -27520.8220 4.0404 -3.5816 TranSiesta: Qsol,Qtot: 737.516
713.000 1.000 transiesta: 6 -26131.8016 -25891.0284 -25891.0284
16.7552 -3.5816 TranSiesta: Qsol,Qtot: 713.770 713.000 1.000 transiesta:
7 -26949.2847 -27274.1379 -27274.1379 6.1382 -3.5816 TranSiesta:
Qsol,Qtot: 717.130 713.000 1.000 transiesta: 8 -27697.9093
-27604.8120 -27604.8120 1.7657 -3.5816 TranSiesta: Qsol,Qtot: 712.777
713.000 1.000 transiesta: 9 -27595.7559 -27600.6101 -27600.6101
1.9578 -3.5816 TranSiesta: Qsol,Qtot: 744.675 713.000 1.000 transiesta:
10 -24264.7612 -23823.9262 -23823.9262 24.9160 -3.5816 TranSiesta:
Qsol,Qtot: 727.493 713.000 1.000 transiesta: 11 -27053.7951
-27406.1813 -27406.1813 6.4667 -3.5816 TranSiesta: Qsol,Qtot: 720.304
713.000 1.000 transiesta: 12 -27569.4696 -27643.9851 -27643.9851
1.7944 -3.5816 TranSiesta: Qsol,Qtot: 714.457 713.000 1.000
I also see the following options in the OUTPUT.fdf of Continuation run,
which I guess they confirm it is really a continuation job:
New_DM. Step: 1 Initializing Density Matrix...
ts_iodm: Reading Density Matrix from files TRANSIESTA: Continuation run
All the Best, Zeynab
On Sunday, July 3, 2016 4:55 PM, Carlos Sanz <[email protected]> wrote:
Zeynab,
and one more thing, in order to check that the TranSIESTA calculation
restarts properly, you should look at whether the energies, convergence of
the density matrix and Fermi energy for the first SCF steps are very close to
those of the last SCF steps of your previous run. The SCF counter simply starts
from 1 again as Nick points out.
regards,
SIMUNE support team
El 03/07/16 a las 16:50, Carlos Sanz escribió:
Hi Zeynab,
the *.TSHS files are built in a previous stage, during the Electrodes
calculations (which is actually a pure SIESTA calculation with a high number
of k-points in the transport direction), and then used when the Scattering
Region calculation starts to describe the electrodes, but it is true that
they do not change anymore at this last stage.
regards,
Carlos
El 03/07/16 a las 11:21, zeynab mohamad hoseyni escribió:
Hi Nick!
Thanks a lot for your attention.. So much happy.. ;-)
All the Best, Zeynab
On Sunday, July 3, 2016 10:14 AM, Nick Papior <[email protected]>
wrote:
Hi zeynab,
See comments in line. -- Kind regards Nick Papior
On 3 Jul 2016 10:13, "zeynab mohamad hoseyni" <[email protected]> wrote:
>
> Hi Carlos,
>
> Thanks a lot for the reply. Nobody answered my question and I really didn't
> know what should I do.
> I progressed well and don't want to switch to another software due to this
> simple problem..
>
> I've enabled all the options you mentioned in my input.fdf (I copy it in
> the following) file.
> I had a look into my directory where I ran previous TranSiesta job. When I
> list *.TSDE I find "scat.TSDE" file,
> but when I list *.TSHS I can't find such file except for the one I already
> put there as the input for previous
> TranSiesta file (it is called "elec.TSHS").
In fact, TSHS file isn't used for a continuation run.
>
> Another point is that when I run the continuation job, inside the produced
> OUTPUT.fdf file it is written "Continuation run".
> But as soon as the section "transiesta: TSiscf" is printed out, it starts
> from number 1 while the previous run
> stops at number 63 and I expect the continuation run start from 64 and not
> 63..
A continuation run does not know what the previous runs Scf count is, so it
starts over.
If it says it has read in the TSDE file, then it is a continuation run.
>
> Can you guess what is wrong?
So I would guess nothing is wrong and it actually behaves as expected. :)
>
> All the Best and thanks again for your time,
> Zeynab
>
> P.S.
> INPUT.fdf file :
> **********************************
> SystemName scat
>
> SystemLabel scat
>
> %block kgrid_Monkhorst_Pack
> 1 0 0 0.0
> 0 1 0 0.0
> 0 0 5 0.0
> %endblockkgrid_Monkhorst_Pack
>
>
> xc.functional GGA
> xc.authors PBE
>
> MeshCutoff 1000.00000000 Ry
> SolutionMethod Transiesta
> OccupationFunction MP
> ElectronicTemperature 300 K
>
> SpinPolarized F
> FixSpin F
>
> MaxSCFIterations 1000
>
> DM.NumberPulay 6
> DM.NumberBroyden 0
> DM.MixingWeight 0.05
> DM.OccupancyTolerance 0.1000000000E-11
> DM.NumberKick 0
> DM.KickMixingWeight 0.5000000000
> DM.Tolerance 1.000000000E-04
> UseSaveData T
>
> MD.NumCGsteps 0
> MD.TypeOfRun CG
> MD.VariableCell F
> MD.MaxCGDispl 0.2000000000 Bohr
> MD.MaxForceTol 0.05 eV/Ang
> #MD.MaxStressTol 0.0001 eV/Ang**3
> MD.UseSaveXV .true.
> MD.UseSaveCG T
> DM.UseSaveDM true
>
> Diag.ParallelOverK F
> PAO.EnergyShift 50 meV
> PAO.SplitNorm 0.1500000000
> PAO.BasisType split
> PAO.BasisSize DZP
>
> WriteMullikenPop 0
> WriteBands F
> SaveRho .true.
> SaveDeltaRho .true.
> SaveHS .true.
> SaveElectrostaticPotential .true.
> SaveTotalPotential yes
> WriteCoorXmol .true.
> WriteMDXmol .true.
> WriteMDhistory .false.
> WriteEigenvalues yes
> WriteDM T
>
>
>
> %include POSITIONS.fdf
>
>
> # Transiesta information
> SolutionMethod Transiesta
> # GENGF OPTIONS
> TS.ComplexContour.Emin -24 eV
> TS.ComplexContour.NPoles 16
> TS.ComplexContour.NCircle 16
> TS.ComplexContour.NLine 10
> # BIAS OPTIONS
> TS.biasContour.NumPoints 10
>
>
> # TS OPTIONS
> TS.Voltage 0.00 eV
>
> # TBT OPTIONS
> TS.TBT.Emin -3 eV
> TS.TBT.Emax +3 eV
> TS.TBT.NPoints 500
> TS.TBT.NEigen 3
> TS.TBT.Eta 0.000001 Ry
>
> # Write electrode hamiltonian
> TS.SaveHS .true.
> TS.SaveLead .true.
>
> TS.MixH T
> TS.UpdateDMCROnly T
> SCFMustConverge T
>
> # LEFT ELECTRODE
> TS.HSFileLeft ./elec.TSHS
> TS.NumUsedAtomsLeft 12
> TS.BufferAtomsLeft 0
>
> # RIGHT ELECTRODE
> TS.HSFileRight ./elec.TSHS
> TS.NumUsedAtomsRight 12
> TS.BufferAtomsRight 0
>
> DM.UseSaveDM T
>
>
>
>
> ==================================================
> ==================================================
>
>
> On Saturday, July 2, 2016 11:24 AM, Carlos Sanz <[email protected]> wrote:
>
>
>
> Dear Zeynab,
>
> as in any SIESTA run, you will have to set up the typical restarting
> parameters in the fdf file such as
>
> DM.UseSaveDM .true. (during a TranSIESTA run, the .DM values are used
> for the density matrix in the buffer (if used)
> and electrode regions)
>
> DM.UseSaveXV .true.
>
> etc... In addition, and what really makes a specific TranSIESTA
> restarting, you must check that you have the *.TSHS and *.TSDE files from
> the previous run in the directory where you are going to restart
> TranSIESTA. By the way, TS.SaveHS is true by default, so you should find
> the *.TSHS from the previous run unless you set up TS.SaveHS false
> previously.
>
> - The *.TSHS file contains the information (e.g.: Hamiltonian matrix once
> self consistency is achieved and overlap matrix) necessary for the
> scattering region calculation. If both electrodes are identical
> structures, the same .*.TSHS can be used to describe both. In general,
> however, the electrodes can be different and therefore you will have two
> different *.TSHS files; one for the left electrode and and another for the
> right electrode.
>
> - The *.TSDE file contains the density matrix for the TranSIESTA
> calculations. It is equivalent to the *.DM file for SIESTA (i.e., density
> matrix with periodic boundary conditions in all directions and no
> voltage), but the density and energy matrix stored in *.TSDE has open
> boundary conditions and bias potential.
>
> best regards,
>
> SIMUNE support team
>
> El 02/07/16 a las 07:08, zeynab mohamad hoseyni escribió:
>>
>> Dear all,
>>
>> I would really appreciate if someone kindly answer to my question:
>> How I can restart a TranSiesta job from the previous run? Is there any
>> option
>> that I have to enable in the .fdf file so that continuing a job from
>> previous job
>> which has been stopped due to lack of wall-time?
>>
>> Would be so much thankful to help me on that.
>>
>> All the Best,
>> Zeynab
>>
>
> --
> Support
> Simune Atomistics
> Tolosa Hiribidea 76 | 20018 Donostia - San Sebastián | Spain
> T.: +34 943 57 40 42 | [email protected] | www.simune.eu
>
>
--
| Support
Simune Atomistics
Tolosa Hiribidea 76 | 20018 Donostia - San Sebastián | Spain
T.: +34 943 57 40 42 | [email protected] | www.simune.eu | |
--
| Support
Simune Atomistics
Tolosa Hiribidea 76 | 20018 Donostia - San Sebastián | Spain
T.: +34 943 57 40 42 | [email protected] | www.simune.eu | |
--
Kind regards Nick
--
SIMUNE Atomistic Simulations#yiv1058254989 body{font-family:Arial, Helvetica,
sans-serif;font-size:11px;color:#4D4D4D;line-height:18px;}#yiv1058254989
.yiv1058254989horia{color:#E7731F!important;font-size:13px!important;font-weight:bold;line-height:20px;}#yiv1058254989
.yiv1058254989titulo{font-weight:bold!important;font-size:13px!important;}#yiv1058254989
td{font-size:12px;}#yiv1058254989 .yiv1058254989borde-arriba{}#yiv1058254989
.yiv1058254989borde-abajo{}#yiv1058254989
a{cursor:pointer;color:#4a4a4a!important;text-decoration:none!important;}#yiv1058254989
a:hover{text-decoration:underline!important;}#yiv1058254989 img{border:none;}
| Support
Simune Atomistics
Tolosa Hiribidea 76 | 20018 Donostia - San Sebastián | Spain
T.: +34 943 57 40 42 | [email protected] | www.simune.eu | |
