My Community

Hi, everyone,
I have a question about Scanning Tunneling Microscopy (STM) simulation using VASP. I set up my parameters and did the calculations by following the instruction in the VASP manual (Band decomposed chargedensity).
The problem is that when I set k-points as 1x1x3 (which is the one I am using), the STM image is not what I expect. If I set k-points as 1x1x2 which does not include T point, the image seems right.
Does anyone have such experience? any suggestions? How can I generate correct STM image by using 1x1x3 k-point mesh?

INCAR file:
Keywords for generating STM information
LPARD=.TRUE.
NBMOD= -3
EINT= -0.5

Startparameter for this Run:
ISTART = 2

NBANDS=304
NPAR = 8
LPLANE = .TRUE.

PREC = Normal
ISPIN = 2

Electronic Relaxation

EDIFF = 1E-04
NELMIN = 3
LREAL =auto
IALGO = 48

Ionic Relaxation

NSW = 0
IBRION = 2
ISIF = 2
ISYM = 1
POTIM = 0.5
EDIFFG = -3E-02


Write flags
LWAVE = .TRUE.
LCHARG = .TRUE.
LVTOT = .FALSE.
NWRITE = 1

DOS related values:
ISMEAR = 1
SIGMA = 0.05
# LORBIT = 2

Thanks in advance.


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

the STM image should be generated with a k-mesh that covers the IBZ as completely as possible. of course the k-points (and hence the Fermi level and the states integrated for that E-interval) of a 1x1x2 and 1x1x3 mesh may differ significantly. Have you tested for k-point convergence before you did the scf calculations?
Also, the precision you choose is rather low (NORMAL,EDIFF =1e-4) this may result in wrong images as well.

Yes, I tested the k-point convergence. 1x1x3 is the correct one. I will change EDIFF to see what will happen. Thanks.

I used EDIFF = 1E-07 later. That did not change the results. Any other suggestions?

1x1x3 kpoints correspond to a linear chain of atoms along z. Are you sure that STM images can be taken from such a probe?

The system I am calculating is a single-wall carbon nanotube with periodic condition adding to z direction. As we know that you can "see" SWNTs through STM. Any suggestion on simulating STM images of SWNTs? Thank you.

just a silly suggestion: are you sure your WAVECAR file from the selfconsistent calculation is read correctly by VASP for the chargedensity caculations?
Check if it explicitly says WAVECAR read succesfully(or something similar)...using different amounts of CPU's to do the scc and the stm calculation could be the cause(even though you define the number of bands)

Also keep in mand that STM images depend on
* distance of the tip to the surface(or whatever your scanning)
* the simulated bias
--> so always check for a range of biases and distances(they are not always exactly the same as the experimantal parameters)

--> also check if your bandstructure/DOS of your swcnt is as it is expected experimentally...here it looks like the extra kpoint(0,0,0) might not have the structure you should expect.


Danny
PS: a program to simulate STM images you can find here
http://cmsmaster.tnw.utwente.nl/~danny/ ... ml&content
(just mail me about it )