- Posts: 105
Inhomogeneity (lumping) of very large CG simulation
- visvaldas
-
Topic Author
- Visitor
6 years 11 months ago - 6 years 11 months ago #7239
by visvaldas
Inhomogeneity (lumping) of very large CG simulation was created by visvaldas
Dear All,
I am running a very large simulation at 372 K of DNA and some dendrimers in PW. I get very strange artifact(s), which I was trying to half-systematically explore (see below). I use the martini_v2.x_new-rf.mdp as a template adapted to my needs and running it using the neat and fast 2016.3 version of GROMACS.
After the energy minimization the system looks normal (homogeneous):
This image is hidden for guests.
However, when I run equilibration using V-rescale, the particles strangely lump together, forming "channels", forming nearly vacuum in the inter-channel regions:
This image is hidden for guests.
At first I though something's wrong with the pressure control, but I get the same picture using no pressure control (NVT conditions). I tried to change to other temperature control means, but berendsen or nose-hoover yield similar picture. Only andersen-massive temperature control gave somewhat "better" picture (I wasn't able to set up andersen temperature control because of some strange input error which I will not delve into here):
This image is hidden for guests.
However, even here there are "bubbles" of ions clumped together, and PW clustered around those bubbles and around the DNA.
If I completely switch off temperature and pressure control (NVE conditions) I get somewhat better picture:
This image is hidden for guests.
but it's a bit weird because it's (almost) good only if I get this when I use epsilon_r=15; with 2.5 (recommended for PW) I get the same artifacts again.
If I try to look at the energies, I see that the LJ interaction tends to increase quite a bit while coulomb interaction decreases somewhat. However, the pressure is negative (without pressure control or even with berendsen pressure control!).
It would seem to me that polarized waters don't stick enough to each other, or perhaps charged particles interact too much and hence lump together, and PW just follow them.
I am out of ideas, so I will probably try "normal" water (although I saw some strange artifacts there too; I'll probably talk about them in a couple of weeks if I can't resolve them).
Vis
PS. An afterthought: Is it possible that the density of PW is too low for some reason? That could explain it.
I am running a very large simulation at 372 K of DNA and some dendrimers in PW. I get very strange artifact(s), which I was trying to half-systematically explore (see below). I use the martini_v2.x_new-rf.mdp as a template adapted to my needs and running it using the neat and fast 2016.3 version of GROMACS.
After the energy minimization the system looks normal (homogeneous):
This image is hidden for guests.
Please log in or register to see it.
However, when I run equilibration using V-rescale, the particles strangely lump together, forming "channels", forming nearly vacuum in the inter-channel regions:
This image is hidden for guests.
Please log in or register to see it.
At first I though something's wrong with the pressure control, but I get the same picture using no pressure control (NVT conditions). I tried to change to other temperature control means, but berendsen or nose-hoover yield similar picture. Only andersen-massive temperature control gave somewhat "better" picture (I wasn't able to set up andersen temperature control because of some strange input error which I will not delve into here):
This image is hidden for guests.
Please log in or register to see it.
However, even here there are "bubbles" of ions clumped together, and PW clustered around those bubbles and around the DNA.
If I completely switch off temperature and pressure control (NVE conditions) I get somewhat better picture:
This image is hidden for guests.
Please log in or register to see it.
but it's a bit weird because it's (almost) good only if I get this when I use epsilon_r=15; with 2.5 (recommended for PW) I get the same artifacts again.
If I try to look at the energies, I see that the LJ interaction tends to increase quite a bit while coulomb interaction decreases somewhat. However, the pressure is negative (without pressure control or even with berendsen pressure control!).
It would seem to me that polarized waters don't stick enough to each other, or perhaps charged particles interact too much and hence lump together, and PW just follow them.
I am out of ideas, so I will probably try "normal" water (although I saw some strange artifacts there too; I'll probably talk about them in a couple of weeks if I can't resolve them).
Vis
PS. An afterthought: Is it possible that the density of PW is too low for some reason? That could explain it.
Last edit: 6 years 11 months ago by visvaldas. Reason: fix temperature
Please Log in or Create an account to join the conversation.
- Pim
-
- Offline
- Expert Boarder
Less
More
6 years 11 months ago #7242
by Pim
Replied by Pim on topic Inhomogeneity (lumping) of very large CG simulation
Hey, I'm not sure what's going on but here's 2 possible options:
1. Your temperature is below the freezing of water. That seems dangerous. The freezing point of Martini PW water is ~282 K and for normal Martini water ~290K.
2. I think you might just not have enough PW particles in the system and the p-coupling can't catch up. How did you solvate your system? Did you use normal, single-bead water and then use the triple-w.py or did you solvate with PW directly? In the latter case you often don't get enough water in the box (you can somewhat counter this by setting the -radius flag lower than the for martini recommended 0.21(because PW contains dummy particles ~0.1 works but it is more error-prone to use this, I recommend the triple-w.py script from the website).
I hope it's either of these, but maybe others have better ideas!
1. Your temperature is below the freezing of water. That seems dangerous. The freezing point of Martini PW water is ~282 K and for normal Martini water ~290K.
2. I think you might just not have enough PW particles in the system and the p-coupling can't catch up. How did you solvate your system? Did you use normal, single-bead water and then use the triple-w.py or did you solvate with PW directly? In the latter case you often don't get enough water in the box (you can somewhat counter this by setting the -radius flag lower than the for martini recommended 0.21(because PW contains dummy particles ~0.1 works but it is more error-prone to use this, I recommend the triple-w.py script from the website).
I hope it's either of these, but maybe others have better ideas!
Please Log in or Create an account to join the conversation.
- visvaldas
-
Topic Author
- Visitor
6 years 11 months ago - 6 years 11 months ago #7244
by visvaldas
Replied by visvaldas on topic Inhomogeneity (lumping) of very large CG simulation
sorry I just found out the density of PW is too low! The box has 70 thousand PWs, while the box of similar size has 240 thousand CG W(aters)! I only have to find why I get this when I try to solvate my system. I got the "polarize-water.gro" off martini website for the solvation... I can't guess what's wrong but I will surely report.
Sorry, the temperature is 372, I'll fix the original post.
Sorry, the temperature is 372, I'll fix the original post.
Last edit: 6 years 11 months ago by visvaldas. Reason: specify water is CG.
Please Log in or Create an account to join the conversation.
- Pim
-
- Offline
- Expert Boarder
Less
More
- Posts: 105
6 years 11 months ago - 6 years 11 months ago #7245
by Pim
Replied by Pim on topic Inhomogeneity (lumping) of very large CG simulation
OK, as I mentioned I recommend solvating with the normal Martini water, not the polarize-water.gro and then running the triple-w.py script. Alternatively you can use a different radius on gmx solvate as mentioned, but I recommend the first strategy.
Also check directly after solvating if your resulting gro file has a homogeneous water density (easiest to see in orthographic view in VMD), sometimes a gmx solvate bug plays up and it puts too many waters in.
Also check directly after solvating if your resulting gro file has a homogeneous water density (easiest to see in orthographic view in VMD), sometimes a gmx solvate bug plays up and it puts too many waters in.
Last edit: 6 years 11 months ago by Pim. Reason: edited after your edit.
Please Log in or Create an account to join the conversation.
- visvaldas
-
Topic Author
- Visitor
6 years 11 months ago #7249
by visvaldas
Replied by visvaldas on topic Inhomogeneity (lumping) of very large CG simulation
Thank you Pim for for all your input and for the tip about triple-w.py (however previously polarize-water.gro worked out okay). Yes, I found out I used -radius 0.22 for the solvation command which was wrong and lead to very few waters.
Have a very nice day!
Have a very nice day!
Please Log in or Create an account to join the conversation.
Time to create page: 0.096 seconds