Calculating bias potential at each timestep in Metadynamics

[Seunghwan_Kwon]

Hello,

I am trying to calculate energy profile of chemical reaction by using Metadynamics.

I utilized the Python code that was posted on VASP Wiki (https://www.vasp.at/wiki/index.php/Nucl ... 3Cl_-_mMD2) to extract bias potential from HILLSPOT.

As a result, I got the energy profile along with CV. (CV vs Energy)

However, I couldn't get the bias potential energy in terms of timesteps. (Timestep vs Energy)

Is there any way that can calculate energy in terms of timesteps?

Also, I am a little bit confused about the contents of HILLSPOT file.

Is it correct that bias potential is not updated until the timestep reaches to NSW/HILLS_BIN?

That is, bias potential remains the same value until the new bias is updated?

Thank you for reading.

[martin.schlipf]

Could you attach the input and output of your calculation? What specifically is not working?

Regarding the things you are unclear about: The HILLSPOT file has the same format as the PENALTYPOT one. Specifically, it defines the centers of the Gaussians that are added to enforce the constraints imposed in the ICONST file. In addition, each line contains the height and width of the Gaussian. I would expect to get one line of (centers, height, width) every HILLS_BIN steps.

[Seunghwan_Kwon]

Hello, thank you for replying.

I am trying to follow up the Metadynamics calculation studied in this paper: https://doi.org/10.1021/acs.jpcc.0c09108.

I attached the input and output files.

It consists of total 15 Metadynamics iteration with each 1000 fs simulation.

I am trying to make the graph like Figure 2(b) in the paper (https://doi.org/10.1021/acs.jpcc.0c09108), which is the "potential energy vs timesteps".

I am having trouble calculating bias potential at each timesteps.

I tried to calculate CV and bias potential using REPORT file and HILLSPOT file, but I am struggling calculating the fluctuation of bias potential at each timesteps.

I would appreciate if you share any ideas or insights on calculating bias potential at each timesteps like in Figure 2(b).

Thank you for reading.

[martin.schlipf]

But in that paper the potential energy is the DFT energy not the bias potential, isn't it? This energy is written to the OUTCAR file in every step. The bias potential would fluctuate on a much longer timescale or not at all depending on the setting of HILLS_BIN.

You would need to look in the supplementary information of the paper where they claim to provide more information about their computational setup if they included the bias potential in the potential energy.

[Seunghwan_Kwon]

Thank you for replying.

I've been studying about this, and I also concluded that the graph in Figure.2(b) is the potential energy written in OUTCAR.

But my concern now is Figure.2(c).

I want to calculate the acceleration factor, but the instantaneous bias potential is required to calculate the acceleration factor.

I am confused about the instantaneous bias potential.

Is it right that the instantaneous bias potential is calculated by the summation of bias at each step from HILLSPOT?

Thank you for reading.

[martin.schlipf]

As far as I understand that is the correct interpretation. Note that the bias potential should not change during a calculation (unless you manually set HILLS_BIN) but the value of the bias potential can change depending on where you are in the phase space. To make it clearer, if you look at the documentation, the centers of the Gaussian are invariant but the geometric parameters ξ may change during a relaxation.