Accurately Modeling STM dI/dV Spectra

[mrm]

Dear All,

I perform surface calculations on relatively large supercells/slabs (200 - 500 atoms) in order to support scanning tunneling microscopy (STM) data. I often need LDOS from VASP since it can be directly compared to dI/dV spectra from the STM. However, the STM measures primarily around the Gamma point. But in order to obtain good quality LDOS from VASP one usually needs to generate a mesh in k-space (using SCF, followed by non-SCF run) which covers more than just the Gamma point. But this is not really what STM measures so it is often difficult to make comparisons.

My questions are:
(1) For large supercells/slabs, is it valid to perform an SCF calculation using a reasonable k-points mesh (e.g.5x5x1), and then to do a non-SCF run at just the Gamma point (1x1x1) to produce an LDOS that is more like what the STM is measuring? I am concerned about the validity of this and also the numerical accuracy.

(2) If the method above is not valid, then can I do an SCF followed by non-SCF runs to get the LDOS (using several kpoints) and then somehow find the LDOS projected onto certain kpoints (specifically the Gamma point)?

Thank you in advance for your help! I believe this is a really important question since I see a lot of STM/DFT papers and none of them seem to make mention of this fact that STM is looking at roughly just the Gamma point.

<span class='smallblacktext'>[ Edited ]</span>

[admin]

if you just take Gamma, you will obtain a line spectrum, not a DOS.
However, it is of course possible to calculate the scf-potential for the k-mesh covering the complete BZ, and then do a non-scf run with a small sample of k-points at and very close to Gamma. Please choose the KPOINTS file format from which the k-points are read explicitely for this approach, but proceed from the scf-CHGCAR file.

[mrm]

Dear Head Admin,

Thank you for your help in confirming this. I appreciate it.

The idea of doing a non-scf run with a kmesh centered near Gamma was something we had considered before I posted the question but it was not clear if it was valid to this. Thanks again for confirming.

<span class='smallblacktext'>[ Edited Wed Jan 13 2010, 03:36PM ]</span>