I am trying to compare the density of states for a few 1st row metals and partically the d-band density of states. I am using the Ni example that came with the software to determine a good starting point for my analysis however I am having some difficulty calculating what I am expecting for the Ni d-band. In particular, the d-band drops quickly after the fermi level but then reaches a small value of about 10 states/atom and stays at about this value for the rest of the window analyzed. Does anyone have a suggestion of what tags should be used in the exciting.in file for improving the calculation of the d-band density of states for this Ni example so that the d-band does not "ooze" above the fermi level and rather just drops to zero? I have tried to increase the k-points up to 12x12x12 as well as increasing the kmesh? Thank you very much for your help.
q365
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Due to the itinerant nature of the d-electrons at the Fermi level in the 3d transition metals you may find a tail in the d-projected dos above the Fermi level. With the following parameters exciting-0.9.139 produces a projected dos in good agreement with other calculations from literature.
In general you are right Kay, but in the case of fcc Ni, with only 2 empty d-states (in an atomic picture) per unit cell I get almost identical d-projected DOS with nempty=20 and nempty=5 (default).
Cheers,
/fredrik
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I am still learning the details of this method so I could be way off, but is it possible that the tail in the energy range above the fermi level is due to non-orthogonality issues for semi-core states? Would it help if I were to promote some core states to valence states and add local orbitals?
Thanks you,
q365
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
The only core electrons that you could add to the valence would then be the 3s states with an energy of about -3.5 hartree. I doubt they would change the density of states at the fermi level much. The semicore p states are well taken care of by the extra local orbital we discussed in the other thread.
Why do you want the d-projected DOS to drop to zero above the fermi level? I'm unable to find any reference that suggest that it should.
Thank you all for a lot of great information. I think I am almost there but I am still doing something slightly wrong. I have been trying to recreate some "simple" numbers from litearture but I am off. For example, I am trying to get the numbers that O.K. Anderson got for Ni, Fe, or Cr as reported in his paper:
Anderson, O.K. Jepsen, O., and Glotzel, D., in "Highlights of Condensed Matter Theory," LXXXIX, p.59 Corso Soc. Italiana di Fisica, Bologna, 1985.
In that paper, he reports that Ni, Fe, and Cr have 8.551, 6.528, and 4.518 d electrons, respectively. However, when I integrate the DOS that I get from exciting LAPW in the PDOS file, I am missing approx. 1 d electron. Is there a way to get angular resolved DOS for the interstital region? That seems to be where my missing d electron is.
Happy holidays!!!
q365
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
After looking at the dos.f90 source code, it dawned on me that I was incorrect in asking my previous question. I forgot that previous planewave DFT programs that I used created DOS by using projected atomic orbitals on atom centers and integrating within a given cutoff radius. Therefore, if I understand this correctly, there isn't a way to get angular momentum resolved DOS for the interstitial region because this area is described by planewaves and the atomic orbitals for the LAPW method only reach up to the radius of the muffin around each atom, right?
Instead, is there a standard method for incorporating the interstitial electrons in band diagrams? For example, I would like to calculate the average energy of the d-band (or the d-band center). Since some of the d-band DOS is outside of the radius of the muffin, should I increase the size of the radius of the muffin to try and incorporate as much of the interstitial electrons back into the atomic orbitals when I calculate the DOS? Or is it common to just determine the d-band center without incorporating information from the interstitial region?
My concern is that if I do not incorporate the interstitial region that my d-band center will be incorrect as well as I will not have a definite ending point for my integration when determining the d-band center. I would like to integrate from below the d-band up until I reach a total of 10 d states. If I do not include the interstitial region, does anyone have a suggestion for what the clear end point should be for this integration?
Thank you very much for any advice that can be provided,
q365
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
you will get the orbital character of each eigenstate at every k-point. Then you will be able to tell which bands have enough d-character to be called d-bands.
with
autormt
.true.
the MT radius of all atoms will be as large as possible and in this way a larger part of space will end up inside the spheres.
The interstitial density has no orbital character and can not be included. Therefore the total number of d-states (filled and empty) will be smaller than 10.
happy computing!
/fredrik
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I am trying to compare the density of states for a few 1st row metals and partically the d-band density of states. I am using the Ni example that came with the software to determine a good starting point for my analysis however I am having some difficulty calculating what I am expecting for the Ni d-band. In particular, the d-band drops quickly after the fermi level but then reaches a small value of about 10 states/atom and stays at about this value for the rest of the window analyzed. Does anyone have a suggestion of what tags should be used in the exciting.in file for improving the calculation of the d-band density of states for this Ni example so that the d-band does not "ooze" above the fermi level and rather just drops to zero? I have tried to increase the k-points up to 12x12x12 as well as increasing the kmesh? Thank you very much for your help.
q365
Due to the itinerant nature of the d-electrons at the Fermi level in the 3d transition metals you may find a tail in the d-projected dos above the Fermi level. With the following parameters exciting-0.9.139 produces a projected dos in good agreement with other calculations from literature.
Best regards,
Fredrik Bultmark
exciting.in:
tasks
10
spinpol
.true.
bfieldc
0.0 0.0 0.01
avec
0.5 0.5 0.0
0.5 0.0 0.5
0.0 0.5 0.5
scale
6.66
sppath
'../../species/'
atoms
1 : nspecies
'Ni.in' : spfname
1 : natoms
0.0 0.0 0.0 0.0 0.0 0.0 : atposl, bfcmt
ngridk
8 8 8
vkloff
0.5 0.5 0.5
dos
500 200 0
-0.5 0.5
You have to increase the number of empty states to get the DOS for higher energies. Try
nempty
15
Cheers, Kay.
In general you are right Kay, but in the case of fcc Ni, with only 2 empty d-states (in an atomic picture) per unit cell I get almost identical d-projected DOS with nempty=20 and nempty=5 (default).
Cheers,
/fredrik
I am still learning the details of this method so I could be way off, but is it possible that the tail in the energy range above the fermi level is due to non-orthogonality issues for semi-core states? Would it help if I were to promote some core states to valence states and add local orbitals?
Thanks you,
q365
The only core electrons that you could add to the valence would then be the 3s states with an energy of about -3.5 hartree. I doubt they would change the density of states at the fermi level much. The semicore p states are well taken care of by the extra local orbital we discussed in the other thread.
Why do you want the d-projected DOS to drop to zero above the fermi level? I'm unable to find any reference that suggest that it should.
Suggested reading on the subject:
V. Heine Phys. Rev. 153, p673
J.W.D. Connolly Phys. Rev. 159, p415
And study the band structure with angular character of the bands using the exciting.in below.
Have fun!
/fredrik
tasks
0
21
spinpol
.true.
bfieldc
0.0 0.0 0.01
xctype
22
avec
0.5 0.5 0.0
0.5 0.0 0.5
0.0 0.5 0.5
scale
6.66
sppath
'../../species/'
atoms
1 : nspecies
'Ni.in' : spfname
1 : natoms
0.0 0.0 0.0 0.0 0.0 0.0 : atposl, bfcmt
ngridk
12 12 12
vkloff
0.05 0.05 0.05
plot1d
7 200 : nvp1d, npp1d
0.0 0.0 1.0 : vlvp1d
0.5 0.5 1.0
0.0 0.0 0.0
0.5 0.0 0.0
0.5 0.5 0.0
0.5 0.25 -0.25
0.5 0.0 0.0
Thank you all for a lot of great information. I think I am almost there but I am still doing something slightly wrong. I have been trying to recreate some "simple" numbers from litearture but I am off. For example, I am trying to get the numbers that O.K. Anderson got for Ni, Fe, or Cr as reported in his paper:
Anderson, O.K. Jepsen, O., and Glotzel, D., in "Highlights of Condensed Matter Theory," LXXXIX, p.59 Corso Soc. Italiana di Fisica, Bologna, 1985.
In that paper, he reports that Ni, Fe, and Cr have 8.551, 6.528, and 4.518 d electrons, respectively. However, when I integrate the DOS that I get from exciting LAPW in the PDOS file, I am missing approx. 1 d electron. Is there a way to get angular resolved DOS for the interstital region? That seems to be where my missing d electron is.
Happy holidays!!!
q365
After looking at the dos.f90 source code, it dawned on me that I was incorrect in asking my previous question. I forgot that previous planewave DFT programs that I used created DOS by using projected atomic orbitals on atom centers and integrating within a given cutoff radius. Therefore, if I understand this correctly, there isn't a way to get angular momentum resolved DOS for the interstitial region because this area is described by planewaves and the atomic orbitals for the LAPW method only reach up to the radius of the muffin around each atom, right?
Instead, is there a standard method for incorporating the interstitial electrons in band diagrams? For example, I would like to calculate the average energy of the d-band (or the d-band center). Since some of the d-band DOS is outside of the radius of the muffin, should I increase the size of the radius of the muffin to try and incorporate as much of the interstitial electrons back into the atomic orbitals when I calculate the DOS? Or is it common to just determine the d-band center without incorporating information from the interstitial region?
My concern is that if I do not incorporate the interstitial region that my d-band center will be incorrect as well as I will not have a definite ending point for my integration when determining the d-band center. I would like to integrate from below the d-band up until I reach a total of 10 d states. If I do not include the interstitial region, does anyone have a suggestion for what the clear end point should be for this integration?
Thank you very much for any advice that can be provided,
q365
With
tasks
21
you will get the orbital character of each eigenstate at every k-point. Then you will be able to tell which bands have enough d-character to be called d-bands.
with
autormt
.true.
the MT radius of all atoms will be as large as possible and in this way a larger part of space will end up inside the spheres.
The interstitial density has no orbital character and can not be included. Therefore the total number of d-states (filled and empty) will be smaller than 10.
happy computing!
/fredrik