PLUMED metadynamics with i-pi

[Sriram Vignesh AK]

Hello everyone,

I am trying to estimate quantum free energy barrier for an isomerization process using i-pi and plumed. The barrier height for the reaction is around 25 kcal/mol(~105 kJ/mol). So i am using the following plumed input.
I am applying a bias on the required dihederal(RC). But the isomerization happens only once from reactant to products and doesn't come back - hence it samples the product region way too much giving me wrong energy profiles(the minima at phi=0 keeps going up in energy whereas its supposed to be isoenergetic with phi = 3.14)

Is there something wrong with my input or what's the correct way to do it?

# default units are LENGTH=nm ENERGY=kJ/mol TIME=ps
#MOLINFO STRUCTURE=rctnt_E.xyz
q: TORSION ATOMS=4,9,7,8 #H5-N3-C1-N2
mtd: METAD ARG=q PACE=20 SIGMA=0.35 HEIGHT=80.00 FILE=plumed/HILLS BIASFACTOR=10 TEMP=300 GRID_MIN=-pi GRID_MAX=pi
PRINT ARG=q  STRIDE=1 FILE=plumed/COLVAR
FLUSH STRIDE=1

Thanks in advance

-Sriram

[Michele Ceriotti]

I was going to suggest you had a look at https://atomistic-cookbook.org/examples/pi-metad/pi-metad.html but the input looks reasonable. 

It's hard to tell based on your input, but the simulation looks pretty short so this is likely just non-converged metadynamics. In fact, I suspect this is more a metadynamics issue that i-PI or quantum nuclei being problematic. Hills that are 80 kJ/mol tall, with a deposition rate of 20 8fs are a very much on the insane side. Also, given you have a large barrier, I'd probably use a larger biasfactor. Try to run with something slightly more reasonable, e.g.

mtd: METAD ARG=q PACE=500 SIGMA=0.2 HEIGHT=10.00 FILE=plumed/HILLS BIASFACTOR=40 TEMP=300 GRID_MIN=-pi GRID_MAX=pi 

Best,

Michele

[Sriram Vignesh AK]

Thank you Prof. Michele for your quick response. I understand your concerns and will try to implement the suggestions you had given. I had given strong biases so that it samples both wells faster; but yeah that did not work well. (I had ran simulation till 45 ps, still i could not get the isomerization happen again). I would use a larger biasfactor and see.
Thank you once again for the suggestions.

With regards,

Sriram

[Michele Ceriotti]

Also, run all the tests with classical MD, I don't expect you'll get so dramatic changes that a setup that works well classically won't work with PIMD.