Hello to APW guru.
I got a couple of questions about the linearization energies of APW orbitals:
1. What is the magic number 0.15 for the choice of default apwe0 value? Why it's the same for all orbital quantum nubers?
2. Why in the case of apword=1 (default) the linearization energies remain fixed at the input value 0.15 even when I switch apwve to .true.
Cheers,
Anton.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I am not an APW Guru but as far as I understand it
0.15 H is a guess for the Fermi level in a solid. Of
course that is subject to be shifted with the potential
if usual (L)APW convetions, e.g. interstitial potential
is zero, are applied. But as far as I can see a good
first guess. Of course it is then assumed that the
"interesting" regin where linearization errors should
be minimized is around the Fermi level.
The WIEN2k web pages contain a remark that for iron additional
local orbitals for the _conduction_ bands should be used in
order to get smooth volume-versus-energy curves for the bulk.
Please note that 0.15 H is only for the valence orbitals.
The linearization energies for local orbitals representing
the core electrons move in fact during calculation (depending
on your system of course), I am not sure which quantum number
you were refering to.
Dear Christof!
Thanks for the reply. Still I'm puzzled by the fact that the linearization energy is the same 0.15HA for (let's say) Hydrogen and Uranium and it is not adjusted during the SCF loop. BTW, if I introduce LAPW orbitals for the valence states by putting apword=2, then the linearization energies will vary if apwve is set to .true.
Cheers,
Anton.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
the linearization energies are a guess for the upper valence region/
Fermi level in a solid if the mean interstitial potential is shifted
to zero. The important part is here the shift in potential which will
shift the absolute value of the eigenvalues. If I remember correctly
somewhere in the literature (Andersen paper ?) there are estimates
that the linearization energies are good (for some measure of good)
for LAPWs around 1 H of the linearization energy ?
Of course adjusting them should in principle lead to smaller
linearization errors.
For a "free" atom, e.g. an isolated atom in a hughe supercell, I am
myselves not sure if the default linearization energies are good.
Best Regards
Christof
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I would add that from my (although not too extensive) experience the errors due to the not-quite-optimal linearization energies are indeed rather small. For example, adjusting them by 0.3Ha in silicon improves the total energy by only 0.1mHa/atom. By the way, there is a good primer on LAPW and optimizing the calculations at http://www.fys.kuleuven.ac.be/iks/nvsf/publications/DFT_and_LAPW.pdf
andrei
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hello to APW guru.
I got a couple of questions about the linearization energies of APW orbitals:
1. What is the magic number 0.15 for the choice of default apwe0 value? Why it's the same for all orbital quantum nubers?
2. Why in the case of apword=1 (default) the linearization energies remain fixed at the input value 0.15 even when I switch apwve to .true.
Cheers,
Anton.
Dear toxa81,
I am not an APW Guru but as far as I understand it
0.15 H is a guess for the Fermi level in a solid. Of
course that is subject to be shifted with the potential
if usual (L)APW convetions, e.g. interstitial potential
is zero, are applied. But as far as I can see a good
first guess. Of course it is then assumed that the
"interesting" regin where linearization errors should
be minimized is around the Fermi level.
The WIEN2k web pages contain a remark that for iron additional
local orbitals for the _conduction_ bands should be used in
order to get smooth volume-versus-energy curves for the bulk.
Please note that 0.15 H is only for the valence orbitals.
The linearization energies for local orbitals representing
the core electrons move in fact during calculation (depending
on your system of course), I am not sure which quantum number
you were refering to.
For the species file you might be interested in
http://exciting.wiki.sourceforge.net/SpeciesFileSyntax
Best Regards
Christof
Dear Christof!
Thanks for the reply. Still I'm puzzled by the fact that the linearization energy is the same 0.15HA for (let's say) Hydrogen and Uranium and it is not adjusted during the SCF loop. BTW, if I introduce LAPW orbitals for the valence states by putting apword=2, then the linearization energies will vary if apwve is set to .true.
Cheers,
Anton.
Dear Anton,
the linearization energies are a guess for the upper valence region/
Fermi level in a solid if the mean interstitial potential is shifted
to zero. The important part is here the shift in potential which will
shift the absolute value of the eigenvalues. If I remember correctly
somewhere in the literature (Andersen paper ?) there are estimates
that the linearization energies are good (for some measure of good)
for LAPWs around 1 H of the linearization energy ?
Of course adjusting them should in principle lead to smaller
linearization errors.
For a "free" atom, e.g. an isolated atom in a hughe supercell, I am
myselves not sure if the default linearization energies are good.
Best Regards
Christof
Dear Anton,
I would add that from my (although not too extensive) experience the errors due to the not-quite-optimal linearization energies are indeed rather small. For example, adjusting them by 0.3Ha in silicon improves the total energy by only 0.1mHa/atom. By the way, there is a good primer on LAPW and optimizing the calculations at http://www.fys.kuleuven.ac.be/iks/nvsf/publications/DFT_and_LAPW.pdf
andrei
Daer Andrei,
thanks for the comment and a useful link.
Anton.