Problem in optimisation of H on a tmdc/metal hybrid.

[ashita_jose]

Dear Admin,
I have been trying to optimise H on Au/monolayer(ML)-MoS2 hybrid for a while but after optimisation sulphur atoms are moving apart. Before adding hydrogen the hybrid system (Au/ML-MoS2) is properly optimised to force tolerance of -0.01 eV/Å. van der Waals correction (IVDW =1) was included in the calculation. After optimisation of the hybrid, H was kept at 1.7 Å above the top sulphur atom and performed optimisation. It reached the required accuracy on optimisation but resulted in a distorted structure. Please help me figure out the issue.


INCAR
Global Parameters
SYSTEM = Au-MoS2

ISTART = 0
ENCUT = 525
PREC = Accurate
LWAVE = .FALSE.
LCHARG = .FALSE.

Electronic Relaxation
ISMEAR = 0
EDIFF = 1E-06
NELM = 200

Ionic Relaxation
NSW = 500
IBRION = 2
ISIF = 2
EDIFFG = -0.01

IVDW = 1
ADDGRID = .TRUE.

[ferenc_karsai]

There could be many reasons, but it seems you have not enough atoms in x,y direction (and maybe in more atoms in the bulk or z direction). Have you tested the effect if you increase the number of atoms in these directions? Also try the effect of calculational parameters like k-mesh grid in the x and y direction and IBRION=1. Try to change these things and see if they have a positive effect for the relaxation.

[ashita_jose]

Thank you for your response.
I have tried performing calculations on a 3x3x1 supercell of the same hybrid structure but the same unusual displacement of atoms is observed. Also, I used ibrion =1 but it didn't work in this case. The k mesh used for the calculation is 5 x 5 x 1.
The initial structure is created by transforming the MoS2 to a rectangular lattice such that the lattice mismatch between Au (111) and ML-MoS2 is minimum. I then optimised the distance between ML-MoS2 and Au slab. On the optimised structure, H was kept and further optimisation led to the distorted structure. If you have any further suggestions on the problem, kindly let me know.

[ferenc_karsai]

So I asked a colleague and he noticed that you are using

Code: Select all

IVDW=1

, which corresponds to the more or less obsolete DFT-D2. In his experience it can sometimes lead to strange results.

He suggested to try DFT-D3 (

Code: Select all

IVDW=11

or

Code: Select all

IVDW=12

) in combination with

Code: Select all

GGA=PE

.

[hszhao.cn@gmail.com]

Dear ashita_jose,

I've recently compiled VASP.6.4.2 with DFT-D4 support. Can you send me all the input files except potcar so I can check it using DFT-D4?

Regards,
Zhao

[hszhao.cn@gmail.com]

Dear ashita_jose,

According to your description, I used the following script to build all possible heterostructures using OgreInterface and hetbuilder, which is based on coincidence lattice theory:

Code: Select all

# Please use the revised versions of the following packages by me to execute this script:
# https://github.com/hongyi-zhao/OgreInterface
# https://github.com/hongyi-zhao/hetbuilder/


#Problem in optimisation of H on a tmdc/metal hybrid.
#https://www.vasp.at/forum/viewtopic.php?t=19400


#https://github.com/DerekDardzinski/OgreInterface/issues/6#issuecomment-1959468808
from OgreInterface.generate import SurfaceGenerator

# Create a surface generator instance from a CIF file
subs = SurfaceGenerator.from_file(
    "Au_mp-81.cif",
    miller_index=[1, 1, 1],  # Specify the Miller index for the surface
    layers=1,                # Number of layers in the generated surface
    vacuum=15                # Vacuum spacing around the surface
    # refine_structure=False  # Don't refine the initial structure; uncomment if needed
)


subs._slabs[0].write_file(output="Au_mp-81_POSCAR_111", orthogonal=True)

from hetbuilder.algorithm import CoincidenceAlgorithm
from hetbuilder.plotting import InteractivePlot
import ase.io

# Read in the structure files via ASE
bottom = ase.io.read("Au_mp-81_POSCAR_111")
top = ase.io.read("MoS2_2H_1L.xyz")

# Set up the algorithm class
alg = CoincidenceAlgorithm(bottom, top)

# Run the algorithm for a choice of parameters
results = alg.run(tolerance=0.15)

for i, res in enumerate(results):
    res.stack.write(str(res.stack.symbols), format='vasp')
    
#https://hetbuilder.readthedocs.io/en/latest/python_interface.html
iplot = InteractivePlot(bottom=bottom, top=top, results=results, weight=0.5)
iplot.plot_results()


# if True:
    # pass

However, among all the results I obtained, there were no supercells of the rectangular lattice type. The attachment includes the above script, all input files, and the generated supercells in vasp poscar format.

Any comments or further discussion are welcome.

Regards,
Zhao

[ashita_jose]

I attached the input files herewith.

[hszhao.cn@gmail.com]

Dear ashita_jose,

Based on your input files, I changed the INCAR to the following:

Code: Select all

$ cat INCAR 
Global Parameters
SYSTEM = Au-MoS2

ISTART =  0             (Start job: "begin from scratch")      
ENCUT  =  525           (Cut-off energy for plane wave basis set, in eV, 130% of ENMAX value of POTCAR file)
PREC   =  Accurate      (Precision level)
LWAVE  = .FALSE.         (Write WAVECAR or not)
LCHARG = .TRUE.        (Write CHGCAR or not)

Electronic Relaxation
ISMEAR =  0             (Gaussian smearing; metals:1)
EDIFF  =  1E-06         (SCF energy convergence; in eV)
NELM   = 200            (NELM sets the maximum number of electronic SC steps

Ionic Relaxation
NSW    =  500           (Max ionic steps)
IBRION =  2             (Algorithm: 0-MD; 1-Quasi-New; 2-CG)
ISIF   =  2             (Stress/relaxation: 2-Ions, 3-Shape/Ions/V, 4-Shape/Ions)
EDIFFG = -0.01          (Ionic convergence; eV/Å)

!IVDW = 1                (DFT-D2 method of Grimme)
IVDW = 13                DFT-D4 method (available as of VASP.6.2 as external package)
GGA=PE
ADDGRID = .TRUE.

So far, I have run 138 ionic steps and obtained the following CONTAR. I am not sure if this is reasonable:

Code: Select all

$ cat CONTCAR 
Au/MoS2-hybrid                          
   1.00000000000000     
     5.3020000457999998    0.0000000000000000    0.0000000000000000
     0.0000000000000000    3.0611109733999999    0.0000000000000000
     0.0000000000000000    0.0000000000000000   36.0000000000000000
  Au/d0044ae04e2  Mo/e91eef5e768  S/0fc7481fb069  H/030f79b5d3ab
               8               2               4               1
Selective dynamics
Direct
  0.0000000000000000  0.0000000000000000  0.0000000000000000   F   F   F
  0.0245874243258706  0.0000000000000000  0.1892913193244395   T   T   T
  0.6781577302208242  0.0000000000000000  0.1282735964532198   T   T   T
  0.3333332538161713  0.0000000000000000  0.0666341409166691   F   F   F
  0.1801742958626597  0.4999998511324790  0.1280207215340731   T   T   T
  0.8333334326354844  0.4999998511324790  0.0666341409166691   F   F   F
  0.5000000000188578  0.4999998511324790  0.0000000000000000   F   F   F
  0.5276413025860066  0.4999998511324790  0.1902132595731924   T   T   T
  0.8667115829955319  0.4999998511324790  0.2990145796400474   T   T   T
  0.3630344732546424  0.0000000000000000  0.3043363868287595   T   T   T
  0.5014066999404796  0.4999998511324790  0.3533268674665762   T   T   T
  0.0104545656569426 -0.0000000000000000  0.3455586072213017   T   T   T
  0.5279113530752755  0.4999998511324790  0.2585605742127685   T   T   T
  0.0420074722815525  0.0000000000000000  0.2571006405572085   T   T   T
  0.2923981564384938  0.4999998511324790  0.3773228418929618   T   T   T
 
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00

Regards,
Zhao

[ashita_jose]

So I asked a colleague and he noticed that you are using

Code: Select all

IVDW=1

, which corresponds to the more or less obsolete DFT-D2. In his experience it can sometimes lead to strange results.

He suggested to try DFT-D3 (

Code: Select all

IVDW=11

or

Code: Select all

IVDW=12

) in combination with

Code: Select all

GGA=PE

.

Dear Ferenc_Karsai,
I tried calculating with DFT-D3 (IVDW =12) but I am facing a similar problem.

[ashita_jose]

Dear Zhao,
Thank you for your response. I will try H adsorption on the heterostructures that you have suggested.

The optimisation using DFT-D4 seems reasonable compared to the other vdW corrections I have tried. I will be grateful if you share the CONTCAR once the optimisation is done.

Regards,
Ashita

[ferenc_karsai]

Let's wait what Zhao gets out...

[hszhao.cn@gmail.com]

Dear ferenc_karsai,

See below:

Code: Select all

$ cat CONTCAR 
Au/MoS2-hybrid                          
   1.00000000000000     
     5.3020000457999998    0.0000000000000000    0.0000000000000000
     0.0000000000000000    3.0611109733999999    0.0000000000000000
     0.0000000000000000    0.0000000000000000   36.0000000000000000
  Au/d0044ae04e2  Mo/e91eef5e768  S/0fc7481fb069  H/030f79b5d3ab
               8               2               4               1
Selective dynamics
Direct
  0.0000000000000000  0.0000000000000000  0.0000000000000000   F   F   F
  0.0111444173979812  0.0000000000000000  0.1894067910919785   T   T   T
  0.6713453890915535  0.0000000000000000  0.1283499684370994   T   T   T
  0.3333332538161713  0.0000000000000000  0.0666341409166691   F   F   F
  0.1734655091879417  0.4999998511324790  0.1280119386503783   T   T   T
  0.8333334326354844  0.4999998511324790  0.0666341409166691   F   F   F
  0.5000000000188578  0.4999998511324790  0.0000000000000000   F   F   F
  0.5134433187983907  0.4999998511324790  0.1903285898680275   T   T   T
  0.8425875289823674  0.4999998511324790  0.2991752562347844   T   T   T
  0.3385643192327905  0.0000000000000000  0.3045655678175372   T   T   T
  0.4770604499280395  0.4999998511324790  0.3536526219781888   T   T   T
 -0.0143407373236675  0.0000000000000000  0.3457420097468190   T   T   T
  0.5038297635072209  0.4999998511324790  0.2588001175540732   T   T   T
  0.0183297971624707  0.0000000000000000  0.2572738450181339   T   T   T
  0.2670688559228859  0.4999998511324790  0.3774147619028350   T   T   T
 
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00
  0.00000000E+00  0.00000000E+00  0.00000000E+00

P.S. Dear ashita_jose, if you want, I can privately send you the complete results of all the calculation files.

Regards,
Zhao

[hszhao.cn@gmail.com]

Dear ashita_jose,

As I have described here, based on coincidence lattice theory, I cannot obtain the supercell of the rectangular lattice type you have used. So, I'm still not sure if your modeling method makes sense.

Currently, I've put this example with the script to generate all supercells here for the reference of interested researchers

Regards,
Zhao

[ashita_jose]

Thank you for sharing.