Setup in mbx.json concerning Ewald grid
50 views
Skip to first unread message
Wenjie LIU
Oct 23, 2024, 3:01:19 AM10/23/24
to MBX-users
Dear MBX users and developers:
I want to conduct a comparative (with classical SPC/E) Mb-pol simulation of a system composed of water vapor and a CNT which confines some water molecules inside. Before I start the simulation, I have some questions concerning the way mb-pol meshing is used for long-range coulomb interactions.
In my previous SPC/E simulation, the long-range coulomb interactions were dealt with by the pppm method. Since my simulation box is very "empty" except for the Q1D core part, I introduced a very large cutoff distance for the lj/long/coul/long interactions to reduce the grid numbers when the same tolerance is used for the pppm.
If mb-pol is used, the setup concerning long-range interactions is in mbx.json, where I found something concerning the cutoff of 2B/3B interactions and some other parameters like the "grid_density_elec" with a default value of 2.5. Given that my simulation cell is very large (~800 angstroms in a dimension), if this value is adopted, the number of grids will be crazy. How do you suggest setting up my case? Can I try to increase the 2B/3B cutoff and adjust some other parameters to achieve a similar way I simulate the SPC/E system?
Thank you for all your replies!
I want to conduct a comparative (with classical SPC/E) Mb-pol simulation of a system composed of water vapor and a CNT which confines some water molecules inside. Before I start the simulation, I have some questions concerning the way mb-pol meshing is used for long-range coulomb interactions.
In my previous SPC/E simulation, the long-range coulomb interactions were dealt with by the pppm method. Since my simulation box is very "empty" except for the Q1D core part, I introduced a very large cutoff distance for the lj/long/coul/long interactions to reduce the grid numbers when the same tolerance is used for the pppm.
If mb-pol is used, the setup concerning long-range interactions is in mbx.json, where I found something concerning the cutoff of 2B/3B interactions and some other parameters like the "grid_density_elec" with a default value of 2.5. Given that my simulation cell is very large (~800 angstroms in a dimension), if this value is adopted, the number of grids will be crazy. How do you suggest setting up my case? Can I try to increase the 2B/3B cutoff and adjust some other parameters to achieve a similar way I simulate the SPC/E system?
Thank you for all your replies!
Francesco Paesani
Oct 24, 2024, 1:19:03 AM10/24/24
to MBX-users
Hi Wenjie,
Thank you for your interest in MBX! For setting up your water simulations in a CNT with MBX and MB-pol, you can follow a similar approach to what you did with SPC/E. One key adjustment is to increase the 2-body cutoff to match the value you used for SPC/E. However, there's no need to change the 3-body cutoff, as that term is short-ranged and doesn’t contribute beyond the cutoff specified in the json file (increasing the 3-body cutoff won’t affect the 3-body energy).
You can also try reducing the "grid_density_elec" value, since you're using a large 2-body cutoff and calculating a significant portion of the long-range electrostatic interactions in real space. Running a few benchmarks should help you find the best balance between accuracy and computational efficiency.
Just a quick question: are you using a cubic box with sides of approximately 800 Å? And along which direction is the CNT oriented?
If you have any further questions or need help using MBX for your system, feel free to reach out!
Best regards,
- The MBX Team
Thank you for your interest in MBX! For setting up your water simulations in a CNT with MBX and MB-pol, you can follow a similar approach to what you did with SPC/E. One key adjustment is to increase the 2-body cutoff to match the value you used for SPC/E. However, there's no need to change the 3-body cutoff, as that term is short-ranged and doesn’t contribute beyond the cutoff specified in the json file (increasing the 3-body cutoff won’t affect the 3-body energy).
You can also try reducing the "grid_density_elec" value, since you're using a large 2-body cutoff and calculating a significant portion of the long-range electrostatic interactions in real space. Running a few benchmarks should help you find the best balance between accuracy and computational efficiency.
Just a quick question: are you using a cubic box with sides of approximately 800 Å? And along which direction is the CNT oriented?
If you have any further questions or need help using MBX for your system, feel free to reach out!
Best regards,
- The MBX Team
Reply all
Reply to author
Forward
0 new messages