Should config.nrg should no longer follow the initial configuration if the system changes during MD?

[M. X. L (M. X.)]

Dear MBX users and developers,

I'm using MB-pol with i-Pi software for PIMD simulations.

My systems are some water clusters, and I followed the example in plugins/i-pi/examples/molecular_dynamics/gas_phase/3h2o/100K/1-nvt.

Simulating a system about 5ps with a timestep of 0.5fs, a proton transfer led to the appearance of OH- and H3O+ ions, and then the system immediately collapsed.

I think the problem should be in config.nrg.

config.nrg should no longer follow the initial configuration after the system changes.

I would appreciate it if someone could give me some suggestions.

 

These are my mbx.json input.

{

   "Note" : "This is a cofiguration file",

   "MBX" : {

       "box" : [],

       "twobody_cutoff"   : 100.0,

       "threebody_cutoff" : 4.5,

       "dipole_tolerance" : 1E-8,

       "dipole_max_it"    : 100,

       "dipole_method"     : "cg",

       "alpha_ewald_elec" : 0.0,

       "grid_density_elec" : 2.5,

       "spline_order_elec" : 6,

       "alpha_ewald_disp" : 0.0,

       "grid_density_disp" : 2.5,

       "spline_order_disp" : 6,

       "ignore_2b_poly" : [],

       "ignore_3b_poly" : [],

       "ttm_pairs" : []

   } ,  

   "i-pi" : {

       "port" :      ,

       "localhost" : "     "

   }

}

Thank you very much for your time.

[M. X. L (M. X.)]

I changed the timestep to 0.25fs and 0.2fs. 

Δt=0.25fs: The system did not collapse

Δt=0.2fs: The system did collapse.
These are my inputs outputs

I would appreciate it if someone could give me some suggestions.

I am very grateful if someone can provide assistance.

[M. X. L (M. X.)]

I conducted a series of tests, and I'm here to update my test results.

Initially, I thought it was the configuration change that caused the system to crash. First, I tested different time steps: 0.5fs, 0.25fs, 0.2fs.

The System         0.5fs                                    0.25fs                                       0.2 fs

(H2O)7            behaved normally                  collapsed                                behaved normally

(H2O)8            collapsed                               behaved normally                   collapsed

It's strange that the structure don't collapse with large timestep, but do collapse with smaller timestep.

And the system collapses completely in one step, with the atomic coordinates expanding many times over.  

21
# CELL(abcABC): 1889.72610  1889.72610  1889.72610    90.00000    90.00000    90.00000  Step:       29320  Bead:       0 x_centroid{angstrom}  cell{atomic_unit}
       O  5.84501e+00  9.10788e+00  6.17925e+00
       H  6.66861e+00  8.61139e+00  5.88785e+00
       H  5.52557e+00  9.57844e+00  5.40719e+00
       O  7.75549e+00  7.45337e+00  5.40449e+00
       H  7.65488e+00  7.06963e+00  4.55653e+00
       H  7.67473e+00  6.68880e+00  5.95303e+00
       O  7.37846e+00  5.00178e+00  7.12204e+00
       H  6.75640e+00  5.27242e+00  7.84038e+00
       H  8.25463e+00  4.99230e+00  7.49784e+00
       O  5.50858e+00  5.90805e+00  9.01360e+00
       H  4.84291e+00  6.45178e+00  8.55564e+00
       H  5.49463e+00  6.09110e+00  9.93531e+00
       O  4.85327e+00  5.28441e+00  5.31851e+00
       H  4.24903e+00  5.86684e+00  5.76304e+00
       H  5.31979e+00  4.87686e+00  6.01058e+00
       O  6.51446e+00  6.26900e+00  3.01123e+00
       H  5.96553e+00  5.88616e+00  3.67785e+00
       H  6.53901e+00  5.67632e+00  2.32604e+00
       O  4.05752e+00  7.39902e+00  7.13330e+00
       H  3.26617e+00  7.77545e+00  7.49575e+00
       H  4.61247e+00  8.21566e+00  6.67493e+00
21
# CELL(abcABC): 1889.72610  1889.72610  1889.72610    90.00000    90.00000    90.00000  Step:       29324  Bead:       0 x_centroid{angstrom}  cell{atomic_unit}
       O  2.80997e+06 -5.52843e+07  5.17669e+06
       H  6.23580e+07 -1.43449e+08 -9.84222e+07
       H  4.20895e+07  1.15454e+08 -5.15092e+07
       O -3.75897e+07  2.68224e+07  1.99361e+07
       H  5.31736e+06 -1.11306e+07 -2.78032e+07
       H -3.37919e+07 -2.04906e+06  1.42899e+07
       O -9.28690e+06  4.51907e+06  2.07168e+06
       H -2.49895e+07  2.22688e+07  3.29610e+07
       H  4.79434e+07 -1.74286e+07 -2.30965e+05
       O  4.68413e+06 -2.96253e+04 -1.50668e+07
       H  5.57763e+06  1.35277e+08 -8.38480e+07
       H  2.70235e+07  1.38334e+07  1.29502e+08
       O  3.96554e+06  7.95051e+06 -1.14218e+06
       H  4.34647e+07  1.92682e+08  5.69615e+07
       H  7.20002e+06 -4.16376e+07 -3.00407e+07
       O -1.43221e+06 -3.00475e+05  2.87594e+06
       H -1.49765e+07 -9.04759e+06  9.37323e+06
       H  4.10175e+06 -1.16443e+06 -1.26320e+07
       O  3.84612e+07  1.05959e+05 -1.41559e+07
       H -9.86449e+08 -5.78006e+08  3.56129e+08
       H  7.89545e+08  5.81810e+08 -2.89896e+08

Later, I thought there was some issues with inputs, so I modified the dipole_tolerance and dipole_method, but the system still collapsed.

Then, I thought that my system might have fallen into some bugs, so I modified the initial configuration, made slight adjustments to it, and tested the isomer of (H2O)8, but the system still collapsed.

I also considered whether this was a matter of randomness, so I changed the random seed, but the system still collapsed.

I hope this can help users who encounter similar problems.

[Richa Rashmi]

Dear M. X. L.,

We recommend using a smaller timestep with PIGLET simulations, as PIGLET noise may contain very high frequency components. We found that a timestep of 0.05 fs works well for your system. Additionally, we tested PIMD simulations for your system with both nbead=32 and nbead=64, and they were stable. Given that you are working with water clusters, PIMD simulations should indeed be feasible.

Hope this helps!

Best regards, 

MBX team

[M. X. L (M. X.)]

I didn't expect you to run some tests for me!

Thank you so much for your selfless help!!!!