Dear Andrei, Thanks for the fast respond. You are comeletely right, I use siesta-3.2. Last midnight, when I ran the relax calculation with diagon method for a system of 730 atoms, which its fdf details are brought at the end, I faced these errors: ***An error occurred in MPI_Bcast, then ***on communicator MPI_COMM_WORLD, next line ***MPI_ERR_COMM: invalid communicator, and finally my MPI job was aborted!
I don't know if that's because of the solution method or not, but then, I hopelessly desided to try OrderN. Unfortunately, while deleteing last output files, I carelessly deleted the Au.psf file as well!! Oops!! But that's what I've noticed now when trying to send the details for you! But the good news is that the program is now running. So sorry for such a mistake! Anyway, I send the details of the input file for further consideration. Best Regards B. Fakhraei PAO.EnergyShift 250 meV PAO.SplitNorm 0.15 PAO.BasisSize SZ MeshCutoff 125.0 Ry MaxSCFIteration 50 DM.MixingWeight 0.4 DM.NumberPulay 2 SolutionMethod OrderN MD.TypeOfRun CG MD.NumCGsteps 50 MD.MaxCGDispl 0.1 Bohr MD.MaxForceTol 0.04d0 eV/Ang On Tuesday, August 2, 2016, Andrei Postnikov < [email protected]> wrote: > Dear Bahareh: > > > *But I recieve error at the very beginnig of the calculation * > - what kind of error? (What was the error message)? > > *and the job terminates badly!* > - how badly? > It is difficult to help you as you give no details, but let me guess. > You are using some old version of Siesta > where the SolutionMethod by default was diagon for less than 100 atoms > and OrderN for beyond 100 atoms. That's OK, just define > SolutionMethod diagon > and go ahead. Yes you can. If the calculation is too heavy that's another > story; > you can always try OrderN later (if your system has a gap, etc. etc.) > > Good luck > > Andrei Postnikov > > > ----- Bahareh Fakhraei <[email protected] > <javascript:_e(%7B%7D,'cvml','[email protected]');>> a écrit : > > Thak you so much dear Andria, > > So you suggest that I can use diagon for a system consisted of about 730 > atoms? > But I recieve error at the very beginnig of the calculation and the job > terminates badly! > I thought it might be because of that. But if that's ok with diagon, I > must go over it again and give it another shot. > > Sincerely yours, > > B. Fakhraei > > > On Tuesday, August 2, 2016, Andrei Postnikov < > [email protected] > <javascript:_e(%7B%7D,'cvml','[email protected]');>> > wrote: > >> Hi, >> >> > I think there must be some misunderstanding: >> >> > the Siesta rule of thumb about ~100 atoms implies only that >> >> > for large enough system, O(N) might ultimately provide >> >> > better scaling than diagon, and ~100 is a rough estimate >> >> > of when the crossing of ~N^3 and ~N tendencies happens in practice. >> >> > However, O(N) only works with large enough band gap, and is otherwise >> fragile. >> >> > There is nothing wrong about using diagon to systems of any size, >> >> > and 100 atoms are not that many to make it markedly unpractical. >> >> > I'd say, go ahead with diagon as long as the calculation time >> >> > is not a problem. If it starts to become an issue >> >> > and you think you can gain by switching to O(N), >> >> > do some test on whether it works correctly and is indeed faster. >> >> > >> >> > Best regards >> >> > >> >> > Andrei >> >> > >> >> > >> >> > ----- RCP <[email protected] <http://JAVASCRIPT-BLOCKED;>> a écrit : >> >> > > Hi, >> >> > > >> >> > > I'm not an expert here, but the naive answer is "use order N >> >> > > method". However, that depends on the physics of your system, >> >> > > for instance, O(N) cannot be applied to metals. >> >> > > >> >> > > Regards, >> >> > > >> >> > > Roberto >> >> > > >> >> > > >> >> > > On 08/01/2016 05:42 PM, Bahareh Fakhraei wrote: >> >> > > > Hello every one, >> >> > > > >> >> > > > For relaxing the large systems (more than 100 atoms) in SIESTA, >> seemingly >> >> > > > *diagon* is not a good option for solution method. Why and What >> should be >> >> > > > used instead? >> >> > > > >> >> > > > yours, >> >> > > > >> >> > > > B. Fakhraei >> >> > > > >> >> > > > A beginner to SIESTA program >> >> > >> >> > -- >> >> > > > > >
