Hi Sonu, So what was the change in the results with the stronger relaxation and smaller displacement for the force constant matrix calculation? If nothing has changed appreciably, then indeed you should test for semicore in La, as Andrei says. Again, if nothing has changed appreciably in your results, at least you have ruled out the easiest possible elements to be checked for (since generating a new pseudo is often boring work)... And most likely you can do the calculations with semicore using the same higher (0.01 eV/Ang) force tolerance as before, because your calculations will become a bit heavier with semicore.
Cheers, Marcos On Mon, Dec 14, 2009 at 9:07 AM, <[email protected]> wrote: > > Dear Andrei and Marcos, > > > > 1) As a DFT problem, i used La in its +3 state along with non-linear core > > correction with pseudocore, Not semocore > > states. > > La(+3) PP gen configuration is : > > 5s2 1.83 > > 5p6 2.19 > > 5d0 3.00 > > 4f0 1.25 with pseudocore radius 1.50 bohr. > > > > I think i need to change pseudocore ? Please give suggestions over > > this why i should increase or decrease the pseudocore? > > I did not spoke about pseudocore, but about semicore, meaning, 5p. > You include 5s and 5p in the valence; this is probably OK. > I calculated something containing La with 5s in the core, > being somehow afraid of a too much ionized +3 configuration, > but this was not a priori better. You should test. > > > how it can affect the frequencies? > > Who knows? In principle, the semicore (and also pseudocore) stuff > often has effect on elastic properties, even if you'll see no changes > in the electronic structure. > > > 2) I used FC Displ= 0.005Bohr, hopefully within harmonic approximation. > > This is not really the point. For some modes this number won't affect > anything at all, and for others, where anharmonicity is important, it will > change quite a lot. But for such anharmonic modes, also the experimental > data may be ambiguous, and at least depend on temperature. > You objective is not to get the best agreement with SOME experiment, > but provide a consistent and reliable und understandable result. > If at the end your frequency comes 200 cm-1 off, but you are sure > that it is the "DFT truth" and not due to some stupid mistake, - > so is life... Someone else can be able to explain it... > > > 3) I Relaxed the system with MaxForceTol= 0.0001 ev/Ang. > > This parameter is rarely a reason to worry, once you are safe on what > regards the MeshCutoff. Even if your starting point for phonon > calculations is not exactly at the eauilibrium, but close enough to it, > it won't be a problem to get good force constants (because of the way > they are evaluated). > > > 4) Can SIESTA calculates bulk modulus directly? > > No > > > or do i need to calculate it through fitting process using Murnaghan > > equation of state? > > Yes (or so I think) > > Best regards > > Andrei Postnikov > > >
