Calculation of metadynamics forces on-the-fly from COLVARS?

[Praveen-Prabhakar]

Hi all,

I am running a well-tempered metadynamics simulation in which I am biasing the dihedral angle (phi) of a peptide and I wish to calculate the forces due to the bias potential on-the-fly. On checking the COLVARS manual, I see the option, "outputAppliedForce" - which can give the applied force trajectory for this colvar and this would be the -d(Vbias)/d(phi)? Or is the keyword "outputAppliedForce" specific to the ABF method?

Apologies if this has been asked before.

Best,
Praveen

[giacomofiorin]

Hi @Praveen-Prabhakar the quantity reported by outputAppliedForce is the force from metadynamics, plus forces from any other biases.

[Praveen-Prabhakar]

@giacomofiorin thank you for the quick response. I understand. Just a small follow-up question,

I have attached the image of the colvars.traj file. Here, in the well-tempered metadynamics simulation, I am adding bias every 500 steps, hence I can understand at the 500th step there is an applied force due to the bias. The applied forces corresponding to the steps after 500th step are non-zero, but we are not adding bias at these time-steps and I was wondering if it is the propagation of the previously applied forces added during the bias step (500th) ? .. I apologize if I am missing something obvious about well-tempered metadynamics here.

Best,
Praveen

[giacomofiorin]

Hi Praveen, metadynamics is an incrementally added biasing potential with a finite range (in the form of Gaussian functions). There will be a force as long as you are within that range. Additionally, for computational efficiency this force is discretized onto a grid, so it will not change in the steps after 500 as long the CV remains within the same bin (or until a new Gaussian is added)

[Praveen-Prabhakar]

@giacomofiorin Thank you so much for the explanation. Appreciate it!

Best,