# Request for pmf calculation

Hi,
I want to do pmf (potential mean force) calculation for my system , is there any direct command to calculate pmf in lammps. So that i can avoid Umbrella sampling
for pmf calculation. I will be very thankful for yours kind help.

With Regards,
Pooja Sahu

Hi,
I want to do pmf (potential mean force) calculation for my system , is
there any direct command to calculate pmf in lammps. So that i can avoid
Umbrella sampling for pmf calculation. I will be very thankful for yours kind help.

what is the problem with umbrella sampling?

in combination with WHAM it is in many
cases the best way to compute a PMF.

what do you need the PMF for?
there are alternative approaches

there are multiple options to do those in lammps

axel.

PMF is related to the radial distribution function. So, the easiest
way to calculate PMF is calculate the RDF and then calculate PMF from
w(r) = - k_BT * log(g(r)).

This algorithmic relationship between the PMF and the RDF means that a
relatively small change in the free energy (i.e. a small multiple of
k_BT) may correspond to g(r) changing by an order of magnitude from
its most likely value.

Hope this helps

Hi,
A great thanks for the kind help, but i have one more doubt. Actually I want to calculate pmf to get binding energy of ion with crown-ether in presence as well as absence of water. I am new to simulation, till now i have calculated rdf between two types atoms only. Here, should i use center of mass of crown ether to calculate rdf with ion. will it give binding energy?

now, this is a rather different question. binding energy and PMF are
different. To calculate binding energy between crown ether and ion,
you have to consider all Oxygen atoms which are bound to the ion, I
don't think the CG of the ether can be used to calculate BE (refer to
crown ether structure).

If you want just a quick PMF calculation, calculate the RDF between O
atoms and the ion and PMF from the formula.

PMF is related to the radial distribution function. So, the easiest
way to calculate PMF is calculate the RDF and then calculate PMF from
w(r) = - k_BT * log(g(r)).

that is a very special case (and only works well for dilute monoatomic gases).

there are many other PMFs that people calculate.

axel.

now, this is a rather different question. binding energy and PMF are
different. To calculate binding energy between crown ether and ion,

not at all. the binding free energy for a _one dimensional
reaction coordinate_ is the integral over the PMF.

you have to consider all Oxygen atoms which are bound to the ion, I
don't think the CG of the ether can be used to calculate BE (refer to
crown ether structure).

If you want just a quick PMF calculation, calculate the RDF between O
atoms and the ion and PMF from the formula.

sorry, but this is not good advice. see my other response.

actually, umbrella sampling, steered MD or metadynamics
(the latter with a suitably chosen reaction coordinate)
are the best ways to get this.

axel.

Hi,
A great thanks for the kind help, but i have one more doubt. Actually I
want to calculate pmf to get binding energy of ion with crown-ether in
presence as well as absence of water. I am new to simulation, till now i
have calculated rdf between two types atoms only. Here, should i use center
of mass of crown ether to calculate rdf with ion. will it give binding
energy?

it doesn't. as i mentioned earlier, the relation between
g(r) and PMF cannot be applied here.

you still didn't explain, why umbrella
sampling would be a bad choice?

it gives you the best control over accuracy
and convergence and is embarrassingly
parallel and thus can be run very efficiently
on a parallel machine.

the next best choice would be to run a trajectory
of the bound state and store away a set of restarts
and then run a bunch of steered MD trajectories
where you use a center-of-mass distance collective
variable (using the USER-COLVARS package, or
fix smd) and average over them.

or the external PLUMED code) can also be used.

using USER-COLVARS has the advantage that you
can use the same interface for umbrella sampling,
SMD and MTD. however, i just got a bug report for
the code that is in the current LAMMPS release
and sent out a patch with a bug fix to steve. so
you'd have to use either LAMMPS-ICMS for now
or wait until steve has incorporated the patch.

please also note, that any of these accelerated
sampling techniques require that you spend some
effort learning how they work and how you use
them correctly.

axel.