Discrete "Coulp.-recip" values for different conformations of the same system

User discussions
1

GROMACS version: 2023.2
GROMACS modification: No
Hello everyone,

I calculated the average energy terms for XG in water across 25 different conformations. All systems are identical, sharing the same water molecules; the only difference is the conformation of XG. Specifically, I analyzed the “Coulp.-recip” term, as shown below. The x‑axis values, such as “60000–65000,” represent the conformation names, while the y‑axis values correspond to the difference between each system and the lowest value.

image
image1319×587 18.2 KB

However, the “Coulp.-recip” values appear discrete. It should be random. In my MDP setup, the coulombtype in the .mdp file was set to PME. I also tried adjusting the parameters:

fourierspacing = 0.08  
pme_order = 6  
ewald_rtol = 1e-6

…but the values remain discrete.

Do you have any suggestions or insights on this? Thank you very much!

2

This is very strange. Is the unit completely filled with a uniform water density? Are the box sizes the same?

3

Yes, the box sizes are totally the same. The water was filled automatically by Gromacs program.

4

One thing to check: PME tuning could explain this, if it is active. The split between Coul.-recip and Coul SR depends on the Ewald splitting parameter. mdrun -notunepme disactivates tuning.

5

Thanks a lot. I will try.

6

Thank you very much for this suggestion！Disabling PME tuning with -notunepme resolves the issue. You were absolutely right！This was very helpful！

7

Note that this is not really an issue, as the decomposition of the Coulomb energy is artificial. You should only look a the total Coulomb energy.

8

Yeah. Thanks a lot for the clarification! It makes sense.

9

Hi again,

Thank you very much for your previous help. After test, I expected no significant difference in total energy, since -tunepme should only affect load balancing between PP and PME, not the physics. However, I observed a difference of about 100 kJ/mol in total energy. It shouldn’t be like this, right? Thank you again for your time!

10

But is that systematic? It takes an enormous amount of sampling to get accuracy potential energy differences. 100 kJ/mol difference can be an expected “random” difference, depending on the size of the system.

11

Thanks! There are about 50000 atoms in the system. The total energy is around ~10⁶ kJ/mol. The 100 kJ/mol difference refers to the time-averaged total energy. Even so, this still corresponds to a very small relative variation, right? I think I should futher check whether the difference is systematic. Thanks again for your time and help!