In order to obtain molecular dynamics (MD) simulations trajectory for computing phonon dispersion relation at finite temperature, I performed NVT MD simulation using VASP for my target system.

Here is the temperature & total free energy profile I obtained (temperature in the unit of Kelvin, and total free energy in the unit of eV, the simulation timestep is 1 fs).

t/e profile

Clearly, there is something wrong with my simulation, the temperature keeps fluctuating over a wide range, while the total free energy is constantly decreasing. I have optimized my structure using the standard conjugate gradient method before MD simulation (under the same accuracy level).

What does this temperature & total free energy profile mean? Should I let it run for a bit longer until the total free energy converges?

  • $\begingroup$ +1 Nice first question! Welcome to our new community, and thank you for contributing your question here! We hope to see much more of you in the future :) $\endgroup$ Commented Apr 29, 2021 at 2:53
  • 2
    $\begingroup$ It might be necessary to post your INCAR, POSCAR, and KPOINTS files to get a good answer to your question. $\endgroup$
    – mt.huebsch
    Commented Apr 29, 2021 at 4:36

1 Answer 1


In principle, temperature fluctuations are perfectly normal at the nanoscale. In fact, fluctuations are intrinsic to any ensemble. The magnitude (variance) of the fluctuations of the kinetic energy in the NVT ensemble is inversely correlated with the number of degrees of freedom, and if you have a small system, these will necessarily be large, even at equilibrium. You can, however, monitor the average temperature to check if it corresponds to the desired value.

The total and potential energy should also fluctuate around some average value. If they do not do this, the system is not yet in equilibrium. I cannot directly see what is happening in your system, but a decreasing energy is not strange per se. Even if you minimize a structure, this only guarantees you start from a local minimum. There's no reason the MD run finds no new, deeper minima.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .