nbeads="1" for PLUMED interface

[Barak Hirshberg]

Hello everyone,

I have a question regarding the nbeads keyword in the bias section, when running i-PI with PLUMED: (taken from the zundel example)

<bias>

            <force forcefield="plumed" nbeads="1"></force>
</bias>

Does this control what coordinates are transferred to PLUMED to calculate CVs and the bias?

Does nbeads = 1 mean transferring the coordinates of the centroid?

If I change it to the actual number of beads will PLUMED have access to all coordinates of all beads?

Thanks,

Barak

[Michele Ceriotti]

It does RP contraction so yes nbeads=1 means transferring the centroid.

If you use nbeads=total_nbeads it'll transfer all the beads, but *one at a time*.

There's not obvious way to transfer all the coordinates at once, for the moment. 

It's a question that keeps resurfacing though so we're open to discussing how

one could implement that feature.

Michele

[Chenghan Li]

Hi Michele,

This is Chenghan. Thanks a lot for the information!

I just want to make sure I understand correctly what you were saying:

The current i-Pi only allows one to apply bias forces on the centroid or one bead but cannot calculate and apply bias forces simultaneously on every bead. Am I correct?

If that is the case, do you have any software in your mind that could potentially do such that thing?

Also thank Barak asked the question that I want to ask.

Best Regards,

Chenghan

[Michele Ceriotti]

It allows to compute *the same bias on every bead*, not atm to compute a bias that depends on the relative position of beads. 

Do you want a metadynamics-like, adaptive bias, or a fixed bias?

[Chenghan Li]

Hi Michele,

My bias is position-dependent, so it should be different for each bead in principle. That’s why I want to apply individual bias on each bead. 

If the current I-Pi implementation does not allow me to do that yet, do you have anything coming to your mind that may have that functionality at this moment? Maybe it is possible that i-Pi constructs multiple instances of some external force driver that evokes its own PLUMED?

Best Regards,

Chenghan

[Michele Ceriotti]

Wasn't clear. There is no problem in computing a position dependent bias on each bead. That's the standard mode. 

What you can't do (and often needed for fancy PI metadyn stuff) is to add a bias that depends on the relative position of two beads.

That is, you can compute V(q1), V(q2) [the index being the bead index] but not V(q1-q2)

[Chenghan Li]

Hi Michele,

Thanks for the detailed explanation! Sorry but I am a little confused here.

For example, if I have a harmonic bias potential V(x)=x^2 and three beads located at x=-1,0,1 at some moment. 

Does i-Pi allow me to apply bias force of (-2)*(-1)=2 on the left bead, 0 on the middle one and (-2)*1=-2 on the right bead? If so, could you please give me some instructions on how to do that (eg. nbeads=?)?

My understanding about "the same bias on every bead" is like all 0 on the three beads if bias is applied to the centroid. Is that correct?

My bias implemented in PLUMED is a force field correction so it is better regarded as part of PES. So I think it is more justifiable if applying the bias on individual bead independently.

Correct me if I am wrong and thanks a lot for your time! 

Best Regards,

Chenghan