Sorry about the confusing title.
I am talking about the output frequency to write the trajectory file. Normally, the time-step in MD simulation is 1 or 2fs. And I am using 1ps/frame to output the trajectory file. The file size is big. And when I analysis these files. I choose 10ps/frame. I found it's hard to find difference to analyze the file between 1ps/frame and 10ps/frame. So should I change the frequency to write the trajectory file to 10ps/frame to save space?
How often do you guys store the trajectory files?
It depends on the system you are simulating. For small hydrocarbons the sampling times are relatively small, 10's to 100's of fs, as the size of the molecules of interest increases so it does the sampling times, for macromolecules (proteins) sampling frequencies are in the range of 10's of ps to 10's of ns. The sampling is relative to the size and flexibility of the molecules of interest. Perhaps a more detailed description of your system would give you a better answer to your question.
Assuming your simulation is of a biomolecular system, a sampling interval of 100ps-1ns is appropriate - the processes you're likely to be interested in occur at longer timescales.
It depends on simulation time and size of your system.
Please find the parameter file which our group uses for 100ns simulation. Please notice the nstxout and nstvout for saving coordinate and velocity for your traj.
Thank you all. My simulations are about 200-400ns, I always converted all trajectory file to DCD to save space. I thought 1ps/frame is a waste of space, but my advisor insists that. I just curious how other people deal with this problem.
It depends on the system you are simulating. For small hydrocarbons the sampling times are relatively small, 10's to 100's of fs, as the size of the molecules of interest increases so it does the sampling times, for macromolecules (proteins) sampling frequencies are in the range of 10's of ps to 10's of ns. The sampling is relative to the size and flexibility of the molecules of interest. Perhaps a more detailed description of your system would give you a better answer to your question.
Also, the sampling frequency has to be related to the parameter you want to calculate, diffusional constants, mechanical properties, structural properties, etc.
I agree with Aldo Fabrizzio Combariza Montañez.
In fact, if you want to calculate a vibration spectrum from the velocity autocorrelation function, a sampling of ~3 fs is needed.
It may sound strange, but otherwise much of the vibrational modes cannot be recovered due to Nyquist's Theorem.
So, it all depends on what you're interested in.
Thanks again. I think Aldo Fabrizzio Combariza Montañez gave me the best answer. As a second PhD student, the MD simulation I know is limited. I need to learn more about it.
You are most welcome! Don't hesitate to shout if help is needed!
Writing a PSF file would be helpful if you are going to use the VMD GUI and tools for the trajectory analysis. Certain VMD extensions work with PSF only.
Hi,
Not sure if this is implicit in your question but just in case it helps: remember that you can always save the trajectory from your simulation at 1ps frequency like your advisor asks, and then you can filter that trajectory to any frequency you want for analysis. Working with proteins, I usually use 10ps. But I often create a small trajectory file at say 1ns frequency just for testing my analysis scripts. Once everything works, I run it on the denser trajectory file. Saves a lot of time.
- Badges
- Science topic
- Similar topics
- Chemistry
- Theoretical Chemistry
- Computational Chemistry
- Molecular Dynamics