Choice of SMASS for AIMD simulations

[rmz4ed]

Hi,

In the VASP wiki description of the SMASS parameter in the INCAR file(wiki/index.php/SMASS), it specifies that for the Nose thermostat, SMASS >=0 should be used. I am wondering if there is any method for choosing what specific value to use for the SMASS parameter when using the Nose thermostat?

Thank you!

[ferenc_karsai]

In principle the Nose-Hoover (SMASS) M_s needs to be such that the frequency of the fictitious thermostat position variable s with N particles, omega=(1/PI)sqrt(6(N-1)k_BT/M_s), couples to the characteristic frequency of the real system. For further reading please see: Hunenberg, Adv.Polym. Sci. 173, 105 (2005)

What are characteristic frequencies: For solids these are the phonon frequencies (for example from a simple phonon calculation).
In the paper "NoseHoover chains: The canonical ensemble via continuous dynamics" by Martyna, Tuckerman and Klein (1992) a good choice for the Nose-Hoover mass is given as M_s=Nk_BT/omega^2.
For liquids something like 'phonon'-frequencies might be obtained from the spectrum of the velocity auto-correlation function.

It should be emphasized that any value for SMASS is ok as long as Nose-Hoover thermostat and ionic movement don't decouple leading to a non canonical ensemble. This might happen only if frequency of the fictitious thermostat and the characteristic frequencies of your system are orders of magnitude different.
How to check for correctness? On requires that the overall temperature is correct but also that the "temperature" of individual degrees of freedom averages to the predefined simulation T (so that the equipartition principle is not violated - one doesn't want to have cold and hot modes even if they possibly average to the right T). This can be checked from the velocities in the vasprun.xml.

[rmz4ed]

Thank you very much for the detailed response!

[jun_yin2]

In principle the Nose-Hoover (SMASS) M_s needs to be such that the frequency of the fictitious thermostat position variable s with N particles, omega=(1/PI)sqrt(6(N-1)k_BT/M_s), couples to the characteristic frequency of the real system. For further reading please see: Hunenberg, Adv.Polym. Sci. 173, 105 (2005)

What are characteristic frequencies: For solids these are the phonon frequencies (for example from a simple phonon calculation).
In the paper "NoseHoover chains: The canonical ensemble via continuous dynamics" by Martyna, Tuckerman and Klein (1992) a good choice for the Nose-Hoover mass is given as M_s=Nk_BT/omega^2.
For liquids something like 'phonon'-frequencies might be obtained from the spectrum of the velocity auto-correlation function.

It should be emphasized that any value for SMASS is ok as long as Nose-Hoover thermostat and ionic movement don't decouple leading to a non canonical ensemble. This might happen only if frequency of the fictitious thermostat and the characteristic frequencies of your system are orders of magnitude different.
How to check for correctness? On requires that the overall temperature is correct but also that the "temperature" of individual degrees of freedom averages to the predefined simulation T (so that the equipartition principle is not violated - one doesn't want to have cold and hot modes even if they possibly average to the right T). This can be checked from the velocities in the vasprun.xml.

Hi,

I have learned a lot from your previous answer. But I have one more question about the velocities in vasprun.xml. I do not quite understand how to check the temperature correctness from velocities in vasprun.xml? I have seen that only two velocities block in vasprun.xml.