Error bars from ABF calculations

Dear Lammpsusers,

I have performed free energy calculations using ABF method using LAMMPS. I want to know how to calculate error bars for obtained free energies.

Looking forward to the reply.


Linking here my reply on the same topic on another list:

Best regards to you and Bala,

Hi Giacomo,

I have gone through the link you mentioned in the previous mail. I found that I have to calculate the correlation length and the global variance of the forces.

By using the formula A2 in the paper titled as “Assessing the efficiency of free energy calculation methods”, I am able to calculate global variance, but When it is coming to calculate the correlation length of the series I am using the rerun command in LAMMPS input file and have added the following lines in colvar block.

corrFunc on
corrFuncLength 150
corrFuncType coordinate
corrFuncOutputFile correaltion.dat

Actually, my previous dump file has 1000 frames (thermo 5000) but when I am rerunning simulation using previous dump file, LAMMPS was able to read only 200 frames.

My Questions Are:

1). How to tell LAMMPS to read all the frames from dump file? Here I am using rerun statement as (…/dump.npt first 0 last 50000000 every 0 dump x y z vx vy vz scaled no)
2). The correaltion.dat file contains the correlation at each time. I think I have to fit the function to exp(-t/T).
3). How do I merge the correlation lengths for each window to get final correlation length?

Here I am attaching my colvar file for rerun and LAMMPS input as well.

Awaiting to your reply,


in.PCP.hexane (743 Bytes) (1.13 KB)

The corrFunc option currently does not support computing the ACF of the force. You can just use outputTotalForce yes in the variable’s definition, and analyze the resulting colvars.traj.