conserved quantity increases linearly

[Bence Balázs Mészáros]

Dear all,

I would like to run PIMD simulations of water using 32 beads on 298K with a qtip4pf force engine, however, the conserved quantity increases linearly quite significantly (more than 10 eV after 10 ps). When I decrease the timestep from 0.25 to 0.1 fs, the drift almost disappears (~0.2 eV after 10 ps). I would like to know, if there's something wrong with my setup, or the 0.1 fs timestep is necessary for PIMD in this case. I attached all my inputs, some of the main parameters:

thermostat: pile_g

tau: 100 fs

mode: NVT

(The standard pimd example with 4 beads gave similarly increasing conserved quantity: i-pi/examples/features/path_integral_molecular_dynamics/standard_pimd_liquid_water)

Best regards,

Bence Mészáros

[Mariana Rossi]

Hi Bence, thanks for reporting. If this is ok, could you try two things? First, check if the drift is the same with the latest version of i-PI in the repository and i-PI version 2.6 (there is a tag that can be downloaded from the repository). Then also check whether the drift is unchanged if you set the flag <fixcom> to False explicitly. I would be grateful if you can report back on these two tests. Depending on how they turn out we analyze further! Thanks.

[Bence Balázs Mészáros]

Thanks for the quick reply. I've tried both suggestions (and I've also tried False fixcom in i-pi 2.6.1), and still have the same issue, unfortunately.

[Mariana Rossi]

Hi Bence! Ok great, thank you. Then it is not a bug but rather how this is. I think your original inputs did not come in the first message. Could you finally send them?  Then we can look at this more closely.

[Bence Balázs Mészáros]

Dear Mariana,

My apologies, I forgot to attach them.