Initial random velocities via kinetic energy

[RPMchemMSU]

Dear authors,

To set initial velocities, I use veloc random $2, where $2 is, according to manual, kinetic energy per atom in eV.
I recently understood that I don't completely understand how these energies are interpreted. It is stated later that velocities are set as unit vectors of random direction with norm which corresponds to classic energy sqrt(2E/m_i).

To set that energy I was calculating vibrational energy of the molecule at certain T, substract ZPE, and then divide it by the number of nuclei. I am now concerned, is what I am doing the correct way? Or should I also include in this quantity ZPE of the system? Do I need to account for energy of translational and rotational motion as well?

In short, what is the most correct way of determining which value to set if I want initial conditions to be at certain temperature T?

Best regards,
Paul

[maisebastian]

Dear Paul,
the most standard way of preparing initial conditions at a certain temperature depends on the system. For small, harmonically rigid molecules without flexible degrees of freedom like torsions or anharmonic bonds, I suggest using wigner.py with the -t option to sample from a non-zero temperature harmonic oscillator.

For flexible molecules or assemblies, one should do dynamics in the ground state (e.g., with Amber or SHARC itself). Ideally, one uses a quantum thermostat for this, but SHARC currently does not have any option to do this conveniently.

I recommend that you check the tutorial of SHARC 4.

Best,
Sebastian

[RPMchemMSU]

Dear Sebastian,

I am trying to generate initial conditions by running dynamics in the ground state using sharc, but for this simulation to run I must specify initial conditions for it too. Would it be correct if I generate initial velocities for this ground state run using Wigner distribution?

Best regards,
Paul

[maisebastian]

Dear Paul,
yes ,that sounds like a reasonable approach to generate a Wigner distribution (at a desired temperature) and then run SHARC MD with it. I suggest you check whether you want to have constant energy or constant temperature and use the Langevin thermostat if the latter. If you would need all initial conditions to have the same energy (I never needed that before), then the only way would be to start from a single geometry with a well-defined random Ekin.

Best,
Sebastian