xyz file and simulation time

[Maia Courtiel]

Dear I-PI users and developers,

I am using i-PI to perform PIMD simulations on water clusters, such as [H2O]20, employing a Deep Neural Network Potential built with the deepMD-kit.

My i-PI input file (input.xml) reads the geometry from the "init.xyz" file, which typically looks like this:

60
#CELL(abcABC):   1000.0   1000.0   1000.0    90.00000    90.00000    90.00000  Step:     0  Bead:       0 x_centroid{angstrom}  cell{angstrom} 
 O     -5.421070    0.418670   -1.840697
 H     -4.506376    0.491649   -2.164162
(...)

However, the other day, I noticed that one of the simulations was six times faster (150,000 steps instead of 25,000 in 5 hours). The only difference I found was an error in the init.xyz file:

60
#CELL(abcABC):   1000.0   1000.0   1000.0    90.00000    90.00000    90.00000  Step:     0  Bead:       0 x_centroid{angstrom}  cell{angstrom  O     -0.950641    2.368749    1.129194
 O     -5.421070    0.418670   -1.840697
 H     -4.506376    0.491649   -2.164162

(...)

Upon further testing, I confirmed that this was the only difference. I am perplexed as to why this error would significantly speed up my simulation. Could someone please explain what might be happening here?

Best regards,

Maïa COURTIEL

[Maia Courtiel]

Hello everyone,

I'd like to follow up on my previous message as I still haven't found an answer, even after searching on my own. I'm wondering if my message might have been overlooked or lost among others. 

If someone could take a moment to look into it, it would be greatly appreciated and helpful.

Thank you very much. 

Kind regards.

Maïa COURTIEL

[Michele Ceriotti]

Have you looked at the outputs for the simulation with an error in the input? 

It seems that when there is a malformed comment line it assumes that the cell is in atomic units, which would 

make the initial configuration crazily compressed, and if e.g. you are doing NPT it will explode in a gas of atoms, 

that I suppose can be computed very quickly because they don't have neighbors. 

Without having more details on what you're doing and what you observe besides the speedup, there isn't much more we can do to help.

Michele