Hi,
I have very general and basic questions considering NAMD. In the input you define timesteps, but you also define steps per cycle.
If I understand correctly, atoms will be moved to another place after x timesteps (defined in the cycle). Also, electrostatics and Van der Waals will be calculated once every y cycles, but that is a bit confusing for me. Why define the "steps per cycle" value, when you could just use the timesteps? Would it be bad if I used 1 step per cycle?
Best,
Tomek.
Well, everything moves - it's a dynamics simulation. The point is that atoms don't move especially far over very short time periods (e.g. 10-20 fs), so there is no reason to update the neighbor list every step, because it's wasted time - the neighbors don't change. Over longer times, molecules move in and out of the neighbor list range. So the settings in NAMD basically say "how many steps can elapse before the neighbor list is re-populated." So within one cycle (i.e. before the neighbor list is repopulated), the lists of atoms experiencing short-range interactions don't change. That doesn't mean the forces don't change, of course, because atoms move around.
Setting stepspercycle to 1 would result in big-time performance loss at no benefit to the stability of the simulation. Put simply, this setting indicates the neighbor searching interval. See, for instance:
http://www.ks.uiuc.edu/Research/namd/wiki/?NamdPerformanceTuning
and
http://www.ks.uiuc.edu/Research/namd/2.6/olddocs/ug/node24.html#section:electdesc
The first link provides good advice; leave the "typical" values of stepspercycle = 20 and parlistspercycle = 2, you get a neighbor list update every 10 steps, which is pretty normal.
Hmm,
Thank You for your reply! but if it's possible could you give me a primitive anwser, what is the difference between the timesteps in one cycle? if the system doesn't move than everything should more or less still be the same, until the next cycle, or am i much wrong?
Well, everything moves - it's a dynamics simulation. The point is that atoms don't move especially far over very short time periods (e.g. 10-20 fs), so there is no reason to update the neighbor list every step, because it's wasted time - the neighbors don't change. Over longer times, molecules move in and out of the neighbor list range. So the settings in NAMD basically say "how many steps can elapse before the neighbor list is re-populated." So within one cycle (i.e. before the neighbor list is repopulated), the lists of atoms experiencing short-range interactions don't change. That doesn't mean the forces don't change, of course, because atoms move around.
- Badges
- Science topic