Hi Bach Tran Tung. If it is not a problem for you, can you share your input file and the specifications of your current hardware, too?

Berkay Sungur Sorry for the missing information. I have updated the input file. I'm using Quantum Espresso version 7.4.1 running on Ubuntu version 24.04.2 Linux system and workstation PC with 48 cores CPU and 128GB RAM.

Bach Tran Tung, your input looks good. However, I noticed a few things.

• You chose ecutwfc = 80. If you did a convergence test with respect to the threshold and selected ecutwfc as 80 Ry, it is good practice. However, as I inspect your pseudopotentials used in your input, the maximum recommended ecutwfc is around 47 Ry. You may use 50 Ry as ecutwfc for fast and reliable results. This will help you to increase calculation speed.
• Another thing is the k-point mesh. You selected 4 4 1 as your k-points. I understood the reason for the z-axis because of the planar structure. However, why did you choose 4 for the x- and y-axis? If you did a convergence test with respect to the k-point mesh and selected 4, it is understandable. If it is an arbitrary number, I think it is overkill for adsorption studies. Optimization of the unit cell and relaxation of electronic structures requires larger k-points, but you can use lower k-points for adsorption studies. You may check the link for choosing k-points for your structure, https://qeinputgenerator.materialscloud.io/. I suggest a fine option for k-points distance in the Advanced Settings section for unit cell optimization, and coarse or medium for adsorption studies.
• For faster optimization cycles, you may fix the position of the lower-layer atoms. In most cases, they mostly do not change position so much, and omitting the relaxation of these atoms from the calculation may accelerate your simulation.

However, if you still want to use your original input parameters, I have one more suggestion for you. Divide your simulation into chunks. First, use a gamma point or a lower k-point mesh, like 2 2 1, for initial calculation. These lower k-point meshes will be calculated faster and give you approximately the optimum structure. After that, you can use coarsely optimized positions as input for the second calculation with higher k-point meshes like 4 4 1 or even 8 8 1. Dividing the calculation into manageable steps helps you to detect the problems that may arise early and gain valuable insight faster. If you have any more questions, I am happy to answer. Good luck.

Dear Dr. Berkay Sungur , thank you very much for your response, it was extremely helpful. Firstly, let me clarify some of your questions.

• I first performed convergence tests for the plane-wave cutoff energy (ecutwfc) and k-point mesh, finding that 80 Ry and a 7×7×1 grid were sufficient. Then I ran a vc-relax calculation on the pristine unit cell with vacuum adding. Afterward, I built a 4×4×1 supercell for the adsorption study and, as you said, I used a arbitrary 4×4×1 k-point grid which refer to values reported in other literatures for this system.

I have a few more questions—could you please clarify them for me?

Finally, your suggestion to break the simulations into two or three smaller steps is very appealing and new to me. I think I will try this approach first.

Once again, thank you for your invaluable guidance.

Dear Bach Tran Tung, it is good to know that you did converge test respect to ecutwfc and k-points. I can suggest doing converge test for the size of the supercell too. Sometimes, molecule can alter charge density of the surface drastically and periodic image interactions effect your result. You can start 3x3x1 supercell for adsorption studies and increase it step by step to be sure that periodic image interaction is negligible.

In my experience, there is no strictly correct methodology for DFT calculations whether you use QE or VASP or others. In our laboratory, we first create simple methodology to fastly get results. Afterwards, we try to compare theoretical findings with experimental data. If the deviation from real world results is negligible, we continue to calculations. However, if deviations are large, we optimize our methods.

If the experimental data comparison is not applicable for your case you may use reference materials calculations. For example, whenever I conduct DFT studies on doped graphene, I use graphene and graphite for my reference materials. I conduct DOS and phonon calculations on my reference materials and compare with literature values. For example, as in the literature, pristine graphene exhibits zero band gap. As for your questions:

• You can start with moderate k-points and relatively high cut off potentials like 80 Ry and 7x7x1 for vc-relax of unit cell. However, I do not recommend you to change cut off potential in any further studies if it is not required. Moderate k-point values ensure your unit cell has relatively free from strain and further DOS studies are accurate. You can conduct SCF with higher k-points like 13×13×1 and use it for post process calculations. For adsorption studies, you may start lower k-points like gamma or 2x2x1 and increase it later if it is needed.
• If your converge test suggest that you should use 80 Ry, use it. Each pseudopotential has recommend Ry potential which translates to do not use lower than recommend value but higher values are up to you.
• Unfortunately, I am not familiar to your material and can not say anything about multi- or single-layer differences. What I suggested on my previous message is that you may fix the bottom sulfur layer in your single-layer system. Most of the bottom layer sulfur atoms may not change positions except the atoms near to active center. You may freeze first and unfreeze later by dividing the calculation into multiple steps.

Breaking the simulation into manageable steps are not shown in any tutorials or webinars. After I understood the mathematics and algorithm behind the DFT with many wasted hours and failed simulations, I use multistep approaches on long and hardware demanding calculations. You may mesage me for your further question on Researchgate. I am happy to help. Good luck on your study.

Dear Dr. Berkay Sungur, thank you so much for taking the time to share your experience and insights, your detailed explanations and practical tips are incredibly valuable and I really appreciate them. I now have several ideas to pursue and will apply them in my calculations. Thanks again, and best of luck with your research!