AMBER Archive (2009)

Subject: Re: Re: [AMBER] momentum in MD simulation

From: dhacademic (dhacademic_at_gmail.com)
Date: Wed Sep 02 2009 - 12:37:50 CDT


Dear Prof. Case,

Thanks for the reply. I am sorry for my vague description about the simulations in the last email. Langevin dynamics (gamma_ln is set to 2) and explicit solvent (TIP3P) are used in the simulations (NPT and PBC, room temperature) of polymer conformation. Based on your message, I think the problem that the non-zero momentum of the system during simulations does comes from the using of thermostat. I have done the following tests to estimate the momentum of the system (dipeptide in cubic water box). The order of magnitude of momentum (mass * velocity in atomic unit) of the whole system is shown after the size of box:
  
NVE
10Å: ~10-6
30Å: ~10-7

NVT (ntt=2, Andersen)
10Å: ~10-4
30Å: ~10-7

NVT (ntt=3, Langevin)
10Å: ~10-2
30Å: ~10-3

It seems that Langevin dynamics always have some drift. The "nscm" and "Tautp" parameters are found to have little influence on momentum, and the change of 'Gamma_ln' will induce the change of momentum. However, it's order of magnitude will not change. Does this mean that Andersen temperature coupling scheme is better in controlling velocity?

Best,

Hao

2009-09-02



dhacademic



发件人： case
发送时间： 2009-09-01 07:50:00
收件人： AMBER Mailing List
抄送：
主题： Re: [AMBER] momentum in MD simulation
 
On Mon, Aug 31, 2009, dhacademic wrote:
>
> I found that the center of mass of my system (PBC) will change after MD
> simulations by using PMEMD. So I check the momentum of the system. The
> total momentum of the whole system should be zero. But in my case, the
> momentum in x, y and z directions are all nonzero. What's wrong?
The argument that "the total momentum of the whole system should be zero"
needs to be treated carefully. If the initial velocities don't have zero
net momentum, then Newton's equations will preserve that, and you will see
a drift. If you are using a thermostat, there can be fluctuations in the
momentum due to that. In all circumstances, round-off errors can also
lead to non-zero momentum.
Your post doesn't give any details about what sort of simulation you are
running, or what the starting conditions were. Nor is it clear how far from
zero your momenta are. It's probably useful to run a small periodic system
where you can control these things, and to run experiments to identify the
important parameters. (Check the nscm parameter as well.)
....good luck...dac
_______________________________________________
AMBER mailing list
AMBER_at_ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber

_______________________________________________
AMBER mailing list
AMBER_at_ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber