hBN-phosphorene heterostructure.pdf (9.9 MB)
Hello everyone,
I am trying to reproduce the molecular dynamics simulations reported in the paper:
“Twist-angle dependent anisotropic thermal transport across phosphorene/hBN van der Waals heterostructures.”
Before proceeding to the heterostructure simulations, I am first trying to reproduce the validation results presented in Table 2, where the thermal conductivity of monolayer black phosphorene is compared with previously published values.
The paper states that the phosphorene structure was initially constructed using Material Studio before being imported into LAMMPS.
To use an open-source workflow, I generated the phosphorene structure using Python ASE. Although I can successfully build the structure and perform the MD simulations, I have been unable to reproduce the thermal conductivity values reported in Table 2.
This makes me wonder whether the discrepancy originates from the initial atomic structure rather than the simulation itself. I suspect there may be subtle differences in the generated phosphorene, such as:
-
the unit cell definition,
-
lattice constants,
-
armchair/zigzag orientation,
-
replication procedure,
-
or other structural details introduced during construction in Material Studio.
I would appreciate any advice on the following questions:
-
Is there a recommended open-source workflow for generating monolayer black phosphorene equivalent to Material Studio?
-
Are there any verified ASE scripts or repositories that generate the phosphorene structure commonly used for LAMMPS thermal conductivity simulations?
-
Can small differences in the initial phosphorene structure significantly affect the thermal conductivity obtained from RNEMD/NEMD simulations?
-
Has anyone successfully reproduced the phosphorene validation results reported in studies such as Hong et al. (Nanoscale, 2015) using ASE or another open-source workflow?
Any suggestions or recommendations would be greatly appreciated.
Thank you!
# For reference, my validation results are:(material BP)
Paper:
Zigzag: 6.36 W/m·K
Armchair: 2.10 W/m·K
My results:
Zigzag: 5.60 W/m·K
Armchair: 3.80 W/m·K
The zigzag value is reasonably close, but the armchair value is significantly higher than the reported value. Since the anisotropy is not reproduced correctly, I suspect the discrepancy may originate from the initial phosphorene structure or its orientation rather than the simulation procedure itself.
First of all, I don’t know why you are so strongly believing that the problem is in the initial structure. Why the problem is not in the MD setup? The “anisotropy not reproduced correctly” thing could have many reasons.
The paper you mentioned did not provide their structure files and LAMMPS input files, and you are not providing them either. So it’s impossible to tell where the difference originates.
Anyway, if you really believe that the problem is the difference between ASE and Materials Studio, then you should go to create the structure with Materials Studio (or other software) and try again. It’s just logical reasoning.
You should also discuss this matter with your supervisor (if possible).
You can also try to send emails to the corresponding author of that paper and ask for their input files, and compare them with yours. They may not respond though.
1 Like
thanks for your valuable insight.
- Is there a recommended open-source workflow for generating monolayer black phosphorene equivalent to Material Studio?
You should get the crystal structure and the there are several tools (including LAMMPS) that can generate geometries from that. Those are more-or-less all not that great but a starting point. What matters is that you then carefully relax and equilibrate the structure and compare the results with experimental findings. There is no simple “do this, not that” approach that is guaranteed to work. You have to do that iteratively.
- Are there any verified ASE scripts or repositories that generate the phosphorene structure commonly used for LAMMPS thermal conductivity simulations?
You are on the wrong track here, it is not the software you use, but how you build your systems and how you validate the results and determine errors and error bars.
- Can small differences in the initial phosphorene structure significantly affect the thermal conductivity obtained from RNEMD/NEMD simulations?
If you equilibrate your system carefully, it does not matter how you generate the initial structure.
- Has anyone successfully reproduced the phosphorene validation results reported in studies such as Hong et al. (Nanoscale, 2015) using ASE or another open-source workflow?
Any suggestions or recommendations would be greatly appreciated.
Without knowing any details of the workflow for determining the results it is not possible to make meaningful assessments. I’ve recently checked some tutorial material for determining thermal conductivity and a key element there was that there are significant size effects when simulating molecular cystals and converging kappa on its own would require systems far too large so the only way would be to collect a table with 1/kappa versus 1/length and then extrapolate to infinity and determine for each value the error bars so that you know what the error bars on your results are.
In short, it is unlikely that the initial geometry matters a lot when you do your simulations carefully and the results mismatch is more likely the result from having data that has very large error bars (you didn’t mention any even though for these kinds of properties they are essential).
1 Like
Thank you very much for your thoughtful and detailed response. Your explanation helped me understand that careful relaxation, equilibration, validation, and finite-size analysis are far more important than the specific software or initial structure generation method. I sincerely appreciate your time and valuable advice, and I will incorporate these considerations into my workflow.