System size effect
- saleheen
-
Topic Author
- Offline
- New Member
-
Less
More
- Posts: 7
- Thank you received: 0
2 years 2 months ago #957
by saleheen
Hi Dr. Maginn!
I was wondering if the calculated properties for a solvent is system size dependent in a gcmc simulation in Cassandra. In our lab group, I ran a gcmc simulation of water at -44.90 kJ/mol chemical potential and at 298 K with an orthorhombic box of 44*48*40 angstrom with 12 angstrom cutoff for both VdW and electrostatic interactions. One of my lab mates ran the same calculation with the same setup but with a 32*32*40 angstrom orthorhombic box. While in my simulation, I could see at -44.90 kJ/mole chemical potential, I am still observing liquid water, in my lab mate's calculation, the box quickly becomes empty indicating it's in the vapor phase. I have attached the input file that we both are using, the only differences being the box size and the number of water molecules in the initial configuration (I used make_config 3000 while my lab mate used make_config 1500, his box being smaller sized). We also use exactly the same force field (TIP4P/2005). At the end of 10 million steps, I can see my water density increasing (~295.67 kg/m3), while in his calculation the box becomes empty after 60 thousand steps. Can you please have a look at the input file and share with us what can be the potential issue here?
Thank you!
Mohammad
I was wondering if the calculated properties for a solvent is system size dependent in a gcmc simulation in Cassandra. In our lab group, I ran a gcmc simulation of water at -44.90 kJ/mol chemical potential and at 298 K with an orthorhombic box of 44*48*40 angstrom with 12 angstrom cutoff for both VdW and electrostatic interactions. One of my lab mates ran the same calculation with the same setup but with a 32*32*40 angstrom orthorhombic box. While in my simulation, I could see at -44.90 kJ/mole chemical potential, I am still observing liquid water, in my lab mate's calculation, the box quickly becomes empty indicating it's in the vapor phase. I have attached the input file that we both are using, the only differences being the box size and the number of water molecules in the initial configuration (I used make_config 3000 while my lab mate used make_config 1500, his box being smaller sized). We also use exactly the same force field (TIP4P/2005). At the end of 10 million steps, I can see my water density increasing (~295.67 kg/m3), while in his calculation the box becomes empty after 60 thousand steps. Can you please have a look at the input file and share with us what can be the potential issue here?
Thank you!
Mohammad
Please Log in to join the conversation.
- ryangmullen
-
- Offline
- Administrator
-
Less
More
- Posts: 124
- Karma: 4
- Thank you received: 24
2 years 2 months ago #958
by ryangmullen
Hi Mohammad,
The thermodynamically stable phase, and therefore density, in a GCMC simulation is independent of the box size. However, the simulation time required to reach that phase will depend on your initial configuration. If you run at liquid water conditions, but start from an equilibrated gas phase configuration, you will need to nucleate a liquid water droplet and wait for that droplet to grow and fill your box. If you start from an equilibrated liquid water configuration (though perhaps equilibrated at a different pressure or temperature), then you will get to the “right” density much sooner.
If you start from a high energy configuration, like what make_config will provide, the box may empty to metastable gas phase density even if running at liquid water muVT. I would therefore not rely on make_config to create a good initial configuration for a GCMC simulation.
Ryan
The thermodynamically stable phase, and therefore density, in a GCMC simulation is independent of the box size. However, the simulation time required to reach that phase will depend on your initial configuration. If you run at liquid water conditions, but start from an equilibrated gas phase configuration, you will need to nucleate a liquid water droplet and wait for that droplet to grow and fill your box. If you start from an equilibrated liquid water configuration (though perhaps equilibrated at a different pressure or temperature), then you will get to the “right” density much sooner.
If you start from a high energy configuration, like what make_config will provide, the box may empty to metastable gas phase density even if running at liquid water muVT. I would therefore not rely on make_config to create a good initial configuration for a GCMC simulation.
Ryan
The following user(s) said Thank You: ejmaginn, saleheen
Please Log in to join the conversation.
- ejmaginn
-
- Offline
- Administrator
-
Less
More
- Posts: 37
- Thank you received: 10
2 years 2 months ago #959
by ejmaginn
Ryan is correct. Let me add that we have found some issues with the orthorhombic box and recommend you use a cubic box if you can. For simulating a fluid phase, there should be no problems with a cubic box.
You may also want to look at mosdef_cassandra ( github.com/MaginnGroup/mosdef_cassandra ). You can easily run NPT simulations to equilibrate a box and then switch to GCMC. This will give you better equilbration as Ryan says.
-Ed
You may also want to look at mosdef_cassandra ( github.com/MaginnGroup/mosdef_cassandra ). You can easily run NPT simulations to equilibrate a box and then switch to GCMC. This will give you better equilbration as Ryan says.
-Ed
The following user(s) said Thank You: saleheen
Please Log in to join the conversation.
- saleheen
-
Topic Author
- Offline
- New Member
-
Less
More
- Posts: 7
- Thank you received: 0
- saleheen
-
Topic Author
- Offline
- New Member
-
Less
More
- Posts: 7
- Thank you received: 0
2 years 2 months ago #961
by saleheen
Thank you Dr. Maginn! Just to clarify, does that mean a cubic box with 'make_config' is good enough to start a GCMC simulation or should I always try to start with an initial configuration of a cubic box filled with solvent molecules? I generally use these fluid phase simulations to identify correct chemical potentials for saturated liquids, and usually I calculate the chemical potential of the fluid in a canonical ensemble at the process conditions and scan the surrounding range of chemical potentials in Cassandra to check what value gives me a steep drop off in density vs chemical potential plot. Do you think it's a good way to identify the correct chemical potential for saturated liquids ?
Please Log in to join the conversation.
Time to create page: 0.122 seconds