CASTEP Core Charge + Grid Convergence
Posted: Sat Feb 13, 2021 11:01 pm
Dear all,
I'm using CASTEP to get the charge density in .den_fmt format, and then using the den2vasp tool to produce a CHGCAR file that I then give to bader. I'm using ultrasoft pseudopotentials in CASTEP. This whole process seems to work well, at least in the sense that no errors are produced and the numbers seem sensical.
After reading these forums I keep seeing:
"It is very important to do the Bader partitioning based upon the total charge and the integration based upon the valence charge."
The solution for VASP calculations seems to be to output the core and valence charge separately, and then do:
bader CHGCAR -ref CHGCAR_sum
With CHGCAR being the valence charge and CHGCAR_sum being the total charge. However, from my CASTEP calculation, I get only one density and convert this using den2vasp. Given that the bader output adds up perfectly to the number of valence electrons, I assume this is the valence charge only. The reasoning for the above total charge Vs valence procedure seems to be that:
"The Bader analysis assumes that charge density maxima are located at atomic centers"
When I visualise the density using VESTA or similar, I can quite clearly see that the charge density is nicely maximised at the atomic cores, even though it would appear that the total charge density is not being included. Is this, therefore, sufficient for bader to work properly?
My second question relates to converging the FFT grid. My cell is approximately of size 4x4x6 (Ang) and contains 32 atoms (it's a highly compressed cell), and I started with a grid of 50x50x75. However, even after steadily increasing the density of the grid all the way up to 120x120x180, the bader charge on each atom is still not converged. Is there a characteristic grid spacing that provides a good starting point?
Many thanks.
I'm using CASTEP to get the charge density in .den_fmt format, and then using the den2vasp tool to produce a CHGCAR file that I then give to bader. I'm using ultrasoft pseudopotentials in CASTEP. This whole process seems to work well, at least in the sense that no errors are produced and the numbers seem sensical.
After reading these forums I keep seeing:
"It is very important to do the Bader partitioning based upon the total charge and the integration based upon the valence charge."
The solution for VASP calculations seems to be to output the core and valence charge separately, and then do:
bader CHGCAR -ref CHGCAR_sum
With CHGCAR being the valence charge and CHGCAR_sum being the total charge. However, from my CASTEP calculation, I get only one density and convert this using den2vasp. Given that the bader output adds up perfectly to the number of valence electrons, I assume this is the valence charge only. The reasoning for the above total charge Vs valence procedure seems to be that:
"The Bader analysis assumes that charge density maxima are located at atomic centers"
When I visualise the density using VESTA or similar, I can quite clearly see that the charge density is nicely maximised at the atomic cores, even though it would appear that the total charge density is not being included. Is this, therefore, sufficient for bader to work properly?
My second question relates to converging the FFT grid. My cell is approximately of size 4x4x6 (Ang) and contains 32 atoms (it's a highly compressed cell), and I started with a grid of 50x50x75. However, even after steadily increasing the density of the grid all the way up to 120x120x180, the bader charge on each atom is still not converged. Is there a characteristic grid spacing that provides a good starting point?
Many thanks.