Sonu,

What Andrei refers to by displacement amplitude is not the eggbox effect,
it's the displacement applied by siesta to a given atom when you do the
phonon run. As you may know, siesta calculates the force constant matrix by
applying a displacement plus or minus delta to a given atom on each of the
cartesian directions. Sometimes, siesta's default displacement can be too
large because of the anharmonicities he mentioned, and then you'll get wrong
results. So one thing you can do is to check which flga controls the
amplitude of atomic displacements for a phonon calculation, make it smaler,
re-run your phonon calculation and see if anything changes in your results.

Another thing to be careful about is the degree of relaxation of your
system, even though I'm not so sure it would be a major source of error for
you. You are calculating forces on other atoms due to the displacement on a
pretty small distance scale, so on the relaxation of your system prior to
the force constant matrix calculation, you should have the force tolerance
set to a very small value, much smaller than the default 0.04 eV/Ang.  0.001
would be better already, although depending on your availability of time and
patience, you could set it to even lower values :)

Cheers,

Marcos

On Sat, Dec 12, 2009 at 6:06 PM, sonu kumar <[email protected]> wrote:

> Dear Sir,
>
> Effect of displacement amplitude i suppose  is eggbox effect. I have not
> checked it, will CHECK it out, although  for reasonable values of
> meshcutoff, the effect of  eggbox on the total energy or on the relaxed
> structure is normally unimportant as has been written in SIESTA manual.
>
> Sorry, right now information of La(+3)  PP is not with me.
>
> Also,now, i will do rest of the calculations as have been asked by you.
>
> Thank you very much
>
> 1009u
>
>
>
>
> On Sat, Dec 12, 2009 at 8:58 PM, <[email protected]> wrote:
>
>> Dear 1009u,
>>
>> the things to look at are:
>> 1. Your acoustic modes are close enough to zero, that means that,
>> most probably, an insufficient MeshCutoff is not your problem.
>> 2. The frequencies which are supposed to be degenerate are reasonably
>> close, that means that numerical errors due to k-mesh etc.,
>> are not obviously to blame.
>> 3. You say that not all frequencies are bad from which I conclude
>> that the rest (out of 7, or so? I do not have the symmetry decomposition
>> at hand) modes, are OK. This is not bad.
>> For the modes which worry you, you should have a more attentive look at
>> - the effect of the displacement magnitude; some modes are strongly
>> anharmonic and depend on it.
>>
>> If the difference persists, it can be either
>> - the general DFT problem (do you have other calculations to compare?), or
>> - the experiment problem (are the data reliable? Are other experiments
>> available? Do these frequencies strongly depend on temperature?).
>>
>> After eliminating all these source of difference, whatever rests is
>> a purely Siesta problem - pseudos and bases. It does not mean, though,
>> that you can tune it easily. How good is your La ? Semicores etc.
>>
>> Do the elastic properties come out right? To test them might be
>> not so easy for a hexagonal structure. Take a simple cubic compound
>> with your La and check the bulk modulus. Compare it against WIEN2k
>> or other all-electron calculation.
>>
>> Good luck
>>
>> Andrei Postnikov
>>
>>
>> > Thank you Sir,
>> >
>> > I have done few calculations of zone center phonons  using primitive
>> unit.
>> >
>> > But i got wrong zone center phonon frequencies ( not all ) in LaB6
>> system.
>> > Similarly in carbide systems.
>> >
>> > e.g. For LaB6 system
>> >
>> >                                                           Expt. freq
>> >                    Freq (by SIESTA)
>> >
>> > F1u mode ( translation)                    210
>> > 181.12,182.30, 182.97
>> >
>> > F1u mode (bending)                        1095
>> > 877.12, 877.60, 879.85
>> >
>> >
>> > and also
>> >
>> > Accoustical
>> > modes
>> 0.0310,
>> > 0.0396, 0.0766,
>> >
>> >
>> > Many times  i changed the meshcutoff (kept more than 400 Ryd) and kgrid
>> (
>> > kept more than 25  Bohr) to get the required frequiences, But only the
>> > digits after decimal place changes i.e meshcutoff  and kgrid doesn't
>> > change
>> > the frequencies appreciably.
>> >
>> > Is the problem because of Pseudo-potential and basis.?
>> >
>> > I also checked the frequencies using GGA approximation to particle
>> density
>> > functional, but no success?
>> >
>> > hope that the information given is sufficient.
>> >
>> > With Regards
>> >
>> > 1009u
>> >
>> >
>> >
>>
>>
>
>
> --
> 1009U
>

Responder a