----- I. Camps <ica...@gmail.com> a écrit : >> Dear Andrei, >> Thank you very much for your email. >> What I mean by convergence study is: > - I choose the accuracy for my energy calculations > - Do several calculations with different values of meshcutoff energies > - Plot the system energy (converged within the accuracy selected previously) > - Select the value of meshcutoff from were the system energy is stable >> Is this ok? > []'s, > Camps
Dear Camps,This is OK up to the last point. What is your definition of "stable"?Does the energy have to be stable to within 0.01 eV or 0.1 meV ? This would of coursedepend on your problem under study. Moreover, the "system energy" as such(the absolute value as it comes out of the calculation) has no big meaning;what really interests you is typically the difference of this energy from someother energy (that of different phase, different magnetic state, etc.).This physically meaningful difference may converge to physically meaningful accuracy faster (or, on the contrary, slower)than you might have guessed while looking at your [system energy vs cutoff] plot. Best regardsAndrei > On Mon, Dec 19, 2016 at 8:59 PM, Andrei Postnikov <andrei.postni...@univ-lorraine.fr> wrote: A small comment to the latest remarks,sorry as it must be obvious:it hardly makes sense to run the convergence test on a calculation of just one molecule(or, in any standalone calculation) because you wont't have a cluewhat exactly, and to which accuracy, should be converged.Typically, you are after the energy difference between two (or more) cases:atoms vs molecule, free molecule + surface vs adsorbed molecule, etc.The convergence to be checkedis that of the meaningful energy difference, to the accuracy you need. Best regards Andrei Postnikov ----- I. Camps <ica...@gmail.com> a écrit : >Dear Riya, Its always good that, instead selecting those parameters (cutoff energy, cell parameters, numbers of k points, etc.) manually, you run convergence test. In this way, maybe you get smaller values for them and then less computational resources involved. []'s, Camps > On Sat, Dec 17, 2016 at 8:23 PM, Riya Rogers <rogers.riy...@gmail.com> wrote: Dear Dr Camps >Thank u so much. I m taking very high cutoff greater than 340Ry. N put the >Molecule in middle of big lattice. > My bond lengths are lil away frm exam abt 1.2 to 1.4%. >Thanking you >Regards > Riya > On 18-Dec-2016 2:40 am, "I. Camps" <ica...@gmail.com> wrote: >Can anyone tell me how to calculate energy of one molecule of gas. >Should I take molecule inside lattice? > Or > Should I follow the following example1: >http://personales.unican.es/junqueraj/JavierJunquera_files/Metodos/Basic/Basic-exercises.html SIESTA is designed for periodic calculation, so, you have to put your molecule inside a fictional lattice. The lattice vectors should be big enough to not get interaction between two neighborhood molecules. Take a look in the second example. Also, you can check how SIESTA is treating your system (molecular, surface, slab, etc.) in the output file. >Does k points n meshcutoff and basis set info req?Yes. K vectors has meaning only for periodic system, so, in case of a molecule, only one is needed (the gamma point (0,0,0). For the meshcutoff, you should do the convergence study as usually. The basis set should be selected according the accuracy you want or need (as in any SIESTA calculation). Regards, -- []`s >Camps > > >
