Benchmarking WSe2 relaxation

163 views
Skip to first unread message

Gayani Pallewela

unread,
Dec 26, 2023, 6:04:13 PM12/26/23
to Twister Help
Hi Indrajith,

I need to benchmark my structural relaxation (in-plane and out of plane) of WSe2 against one of your studies. Which paper you suggest.

Reconstruction of moir´e lattices in twisted transition metal dichalcogenide bilayers?
I found this.
I need data for WSe2

Please advice me

Best,
Gayani

Indrajit Maity

unread,
Jan 2, 2024, 2:51:21 AM1/2/24
to Gayani Pallewela, Twister Help
Hi Gayani,

Happy New Year! 

Yeah, the paper you referred would be a good starting point. I am afraid I didn't retain the the data files. You should be able to quickly generate the input files by following the MoS2 example. 

Best wishes,
Indrajit 
--
You received this message because you are subscribed to the Google Groups "Twister Help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to twister-help+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/twister-help/CACC_f2y5o%2B_6QyYFEu_v6%2B-M6SF7yZphpFX41Fo%2BVDQJiFj7sg%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.

Gayani Pallewela

unread,
Jun 4, 2024, 3:27:40 AM6/4/24
to Twister Help
Hi Indrajith,

Referring to the paper "Reconstruction of moiré lattices in twisted transition metal dichalcogenide bilayers"

I have tried to reproduce the SA results of this paper. This my input file. Can you please point out if there is something I have done wrong when do the simulating annealing. Is the 1K temperature is higher enough? I don't see any difference between SA and SR according to my simulations.


# Initialize
# General settings
units           metal
dimension       3
box tilt        large
atom_style      atomic
newton          on

# Structure
boundary        p p p
read_data       WSe2_n21_theta58_46newcode.lmp

# Potentials
pair_style hybrid/overlay sw/mod sw/mod kolmogorov/crespi/z 14.0 kolmogorov/crespi/z 14.0 kolmogorov/crespi/z 14.0 kolmogorov/crespi/z 14.0 lj/cut 10.0

# Intralayer Interaction
pair_coeff * * sw/mod 1 tmd.sw W Se Se NULL NULL NULL
pair_coeff * * sw/mod 2 tmd.sw NULL NULL NULL W Se Se

# Interlayer Interaction
# W-W interactions
pair_coeff 1 4 kolmogorov/crespi/z 1 WSe.KC W NULL NULL W NULL NULL
# Se-Se interaction
pair_coeff 2 6 kolmogorov/crespi/z 2 WSe.KC NULL Se NULL NULL NULL Se
# W-Se interaction
pair_coeff 1 6 kolmogorov/crespi/z 3 WSe.KC W NULL NULL NULL NULL Se
pair_coeff 2 4 kolmogorov/crespi/z 4 WSe.KC NULL Se NULL W NULL NULL

pair_coeff * * lj/cut 0.0 3.0

# Neighbor settings
neighbor        2.0 bin
neigh_modify    every 1 delay 0 check yes

# Output settings
thermo          1000
thermo_style    custom step pe press

# NVT ensemble at 1 K
timestep        0.001
fix             1 all nvt temp 1.0 1.0 0.1
run             10000  # Run MD for a certain number of steps at 1 K

# Cool down to 0 K
unfix           1
fix             2 all nvt temp 1.0 0.1 0.1
run             10000  # Run MD for a certain number of steps to cool down to 0 K

# Energy minimization
min_style       cg
minimize        0.0 1.0e-4 1000000 1000000

# Write the final structure
write_dump all custom  WSe2_n21_theta58_46SA.xyz id type x y z

# End of script

On Tuesday, January 2, 2024 at 3:51:21 PM UTC+8 indrajit...@gmail.com wrote:
Hi Gayani,

Happy New Year! 

Yeah, the paper you referred would be a good starting point. I am afraid I didn't retain the the data files. You should be able to quickly generate the input files by following the MoS2 example. 

Best wishes,
Indrajit 


On Tuesday, December 26, 2023, Gayani Pallewela <gayanipa...@gmail.com> wrote:
Hi Indrajith,

I need to benchmark my structural relaxation (in-plane and out of plane) of WSe2 against one of your studies. Which paper you suggest.

Reconstruction of moir´e lattices in twisted transition metal dichalcogenide bilayers?
I found this.
I need data for WSe2

Please advice me

Best,
Gayani

--
You received this message because you are subscribed to the Google Groups "Twister Help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to twister-help...@googlegroups.com.

Indrajit Maity

unread,
Jun 4, 2024, 11:24:06 AM6/4/24
to Gayani Pallewela, Twister Help
Hi Gayani,

The temperature scale very much depends on the lattice constant chosen for simulations. For example, to get annealed reconstructed structures if you use a lattice constant of MoS2 to be 3.15 Angstrom as opposed to the LAMMPS optimized lattice constant with the same SW potential, which is 3.11 Angstrom, you may need either higher temperature or much smaller twist angle to see the reconstructed structures.

Please note that we can't comment on the validity of your molecular dynamics simulations at this time. LAMMPS documentation/resource is a better place for that. 
Please also read more details of the calculations which is freely available in the Supplementary Information of the paper. 

You may be missing an important step of the relaxation(https://journals.aps.org/prb/supplemental/10.1103/PhysRevB.103.L121102/si.pdf): "The standard relaxation (SR) is performed with the target pressure of P = 0 bar. For the simulated annealing (SA), the simulation box dimensions were kept the same as in SR. We use the Nosé-Hoover thermostat while performing simulations with the canonical ensemble" We specify all the steps to achieve the relaxation, and dynamics."

Best wishes,
Indrajit

Gayani Pallewela

unread,
Jun 11, 2024, 2:33:15 AM6/11/24
to Twister Help
Hi Indrajith,

May I know
Which force field you used? kolmogorov/crespi/full or kolmogorov/crespi/z?. 

"We use the Stillinger-Weber and Kolmogorov-Crespi (KC) potential to capture the intralayer and interlayer interaction of tBLTMDs, respectively [32,33]. The used KC parameters have been shown to accurately capture the interlayer van der Waals interaction present in the TMDs [33]."

Thanks
Gayani

Indrajit Maity

unread,
Jun 11, 2024, 6:42:55 AM6/11/24
to Gayani Pallewela, mi...@berkeley.edu, Twister Help
Hi Gayani,

We have used Kolmogorov/Crespi/Z. At that time, @mi...@berkeley.edu  had written a Kolmogorov/Crespi/Full version as well, which we used to validate the use of Kolmogorov/Crespi/Z. I am not sure if the Kolmogorov/Crespi/Full is available for TMDs in the stable version of LAMMPS.

Best wishes,
Indrajit


Gayani Pallewela

unread,
Jun 11, 2024, 7:08:39 AM6/11/24
to Twister Help
Hi Indarajith,

Thank you very much for your quick reply. 
Still I couldn't reproduce the antiparallel interlayer distance profile (twist=58.46.). There is no problem with the parallel configuration. 

The standard relaxation (SR) is performed with the target pressure of P = 0 bar. I have added this as well. Could you please comment on this input script. Is there any obvious error?

units           metal

dimension       3

box tilt        large

atom_style      atomic

newton            on

# structure

boundary        p p p

read_data       WSe2_n21_theta58_46newcode.lmp

# potentials

pair_style hybrid/overlay sw/mod sw/mod kolmogorov/crespi/z 14.0 kolmogorov/crespi/z 14.0 kolmogorov/crespi/z 14.0 kolmogorov/crespi/z 14.0 lj/cut 10.0

# Intralayer Interaction

pair_coeff * * sw/mod 1 tmd.sw W Se Se NULL NULL NULL

pair_coeff * * sw/mod 2 tmd.sw NULL NULL NULL W Se Se

# Interlayer Interaction

# W-W interactions

pair_coeff 1 4 kolmogorov/crespi/z 1 WSe.KC W NULL NULL  W NULL NULL

#Se-Se interaction

pair_coeff 2 6 kolmogorov/crespi/z 2 WSe.KC NULL Se NULL  NULL NULL Se

#W-Se interaction

pair_coeff 1 6 kolmogorov/crespi/z 3 WSe.KC W NULL NULL  NULL NULL Se

pair_coeff 2 4 kolmogorov/crespi/z 4 WSe.KC NULL Se NULL  W NULL NULL


pair_coeff * * lj/cut 0.0 3.0

neighbor        2.0 bin

neigh_modify every 1 delay 0 check yes


#optimize at 0 K

#dump 1 all custom 100  WSe2out.xyz

#dump            1 all custom 100 dump.initial id type x y z

thermo          1000

thermo_style    custom step pe press

#undump          1


fix             1 all box/relax iso 0.0


min_style       cg

minimize        0.0 1.0e-4 1000000 1000000

unfix           1


write_dump all custom    WSe2_n21_theta58_46relaxed.xyz id type x y z

~                                                                                                                                                                                                                     

~                                                                           



Screenshot 2024-06-11 at 19.01.17.png

Indrajit Maity

unread,
Jun 11, 2024, 8:06:07 AM6/11/24
to Gayani Pallewela, Twister Help
Hi,

I am assuming you are plotting interlayer spacing. If so, why do they look discontinuous?

Here are some thoughts:
1. Try molecular dynamics for long times.
2. Maybe use a larger supercell; Something like at least 3x3x1 of the moiré unit cell.
3. If nothing happens, increase the temperature and see what the atoms are doing.
4. Try simulating smaller twist angles.

Hope this helps.

Best wishes,
Indrajit

Gayani Pallewela

unread,
Jun 11, 2024, 6:12:18 PM6/11/24
to Twister Help
Hi Indrajith,

I will check your suggestions today.
Do you assign pressure=0 when you relax the system always? (I have not seen this in your example scripts)
I believe this is doing to avoid the conflicts from periodic boundary condition, close to experiments and thereby to get physically realistic configuration.

Gayani Pallewela

unread,
Jun 26, 2024, 10:44:23 PM6/26/24
to Twister Help
Hi Indrajith,

I Have tried to do the simulations as per your comments.
Please find the analysis I have done.
1. Parallel simulations are totally fine.
2. Antiparallel inplane displacements are reasonable however out of plane displacements and interlayer distances are not. As per your suggestion I have extended the moire cell 3x3x1. Still the results are same.
3. Then I tried setting P=0 as per the supporting information in the manuscript.
Then the results again are not what I expected.

Can you please help me to reproduce the figure of WSe2.
This is not yet simulated annealing. I'm still struggling to do Antiparallel SR.

Thanks
Gayani

WSe2_n21_theta58_46fullnewextended.lmp
WSe2_Antiparallel.docx

Gayani Pallewela

unread,
Jun 26, 2024, 10:49:38 PM6/26/24
to Twister Help
I have attached script and relaxed structure

WSe2_n21_theta58_46fullnewexteninputlammpstestrelaxed.dat
lammpsWSe2newton.in

Indrajit Maity

unread,
Jun 27, 2024, 6:50:22 AM6/27/24
to Gayani Pallewela, Twister Help

Hi Gayani,

While the near 0-degree structure looks okay to me, the near 60-degree structure is not correct. For instance, the discontinuity of the interlayer spacing on a 3x3x1 and the weird pattern on a 1x1x1 suggest that the relaxed structure is not producing the right results.

I suggest you carefully examine the basis_* files for near 0 and near 60 degrees when creating the structure. It is very likely that the ordering or numbering of the layers is incorrect. The atom types are absolutely central to getting the right Kolmogorov-Crespi interactions.


Best,
Indrajit

On Thu, Jun 27, 2024 at 3:49 AM Gayani Pallewela <gayanipa...@gmail.com> wrote:
I have attached script and relaxed structure

--
You received this message because you are subscribed to the Google Groups "Twister Help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to twister-help...@googlegroups.com.

Gayani Pallewela

unread,
Jul 11, 2024, 6:23:51 AM7/11/24
to Twister Help

Hi Indrajith,

Thank you very much for your suggestions on reproducing your results from the paper. I believe I have finally achieved that. However, I wonder if you also observe significant corrugation in your simulations, as shown in the attachment (e.g., at 59.27 degrees).

Please comment on this. I also conducted the simulation for 3x3x1 and still observed the same results.

BenckmarkingagainstJain.docx

Indrajit Maity

unread,
Jul 11, 2024, 3:27:30 PM7/11/24
to Gayani Pallewela, Twister Help
Hi Gayani,

Well done!

Indeed, we did. There are physics reasons behind it.

Best wishes,
Indrajit

Gayani Pallewela

unread,
Aug 16, 2024, 3:02:40 AM8/16/24
to Twister Help

Dear Indrajith,

I have successfully reproduced the simulated annealing (SA) results for WSe2. Thank you very much for your guidance and suggestions on reproducing the results from your paper. I increased the temperature up to 10 K and used the extended system as discussed.

However, I have a few questions, and I would greatly appreciate your insights. I followed the steps in your paper closely, and I’ve attached the input script for your reference.

In your paper, you mentioned that some of the experimental results match exactly with the structures observed from simulated annealing. Given that experiments are typically conducted at 300 K, I am curious about the best approach for a theorist to study relaxed structures at this temperature.

Would it be advisable to heat the system up to 300 K, perform NVT for a few nanoseconds to obtain the equilibrated structure, or should we first heat the system to 300 K, cool it down to zero, and then perform minimization?

Your advice on this matter would be invaluable.

Best regards,
Gayani

fix 1 all box/relax iso 0.0

min_style       cg

minimize        0.0 1.0e-10 1000000 1000000

unfix 1

# NVT ensemble at 1 K

timestep        0.001

fix             2 all nvt temp 10.0 10.0 0.1

run             1000000  # Run MD for a certain number of steps at 1 K

# Cool down to 0 K

unfix           2

fix             3 all nvt temp 10.0 0.1 0.1

run             10000  # Run MD for a certain number of steps to cool down to 0 K

# Energy minimization

min_style       cg

minimize        0.0 1.0e-10 1000000 1000000


SimulatedAnnealing.pdf

Indrajit Maity

unread,
Sep 3, 2024, 4:53:44 PM9/3/24
to Gayani Pallewela, Twister Help

I have successfully reproduced the simulated annealing (SA) results for WSe2.
* Well done!

Would it be advisable to heat the system up to 300 K, perform NVT for a few nanoseconds to obtain the equilibrated structure, or should we first heat the system to 300 K, cool it down to zero, and then perform minimization?
* I would say it entirely depends on what you wish to study. This is very much beyond the scope of the discussion on TWISTER.

Best wishes,
Indrajit

Reply all
Reply to author
Forward
0 new messages