I am applying NEMD (SLLOD) to obtain shear viscosity of water at 293K (20°C) for water models TIP3P, SPC/E & TIP4P/2005.
The setup has worked fine for TIP3P and SPC/E: after applying unit conversions, viscosity values are in good range of the experimental value/literature (slightly more accurate than reported simulations, which are mostly at 300K).
However, for TIP4P/2005 I just can’t seem to make it work. The system setups are the same, the only differences are parameters specific to the water models.
‘’ERROR: Out of range atoms – cannot compute PPPM’’
comes up every time after PPPM Initialization; the simulation never actually gets to carrying out the equilibration run. The error refers to the kspace_style command line. I used pair_style ‘lj/cut/tip4p/long’ with kspace_style ‘pppm/tip4p’.
However, when running on a single processor (instead of 16), the simulation does proceed, with sensible output and viscosity values (accuracy > TIP3P & SPC/E). Dangerous builds: 1 in the equilibration run (10k fs) and 0 for the data gathering run (1ns).
In various attempts to get through this error message I used slightly different potentials with ‘pppm/disp/tip4p’ which gave the error of ‘KSpace style does not yet support triclinic geometries’. Could this also be the case for the ‘pppm/tip4p’ style?
Out of curiosity, I also tried removing the kspace_style line, changing pair_style to ‘lj/cut/tip4p/cut’ - the simulation does proceed, but viscosities come out ~ 80x larger than expected (just to mention in case this helps).
I have run simulations (orthogonal box) for all three water models with no problems before attempting viscosity calculations. I’d ideally like to use NEMD, rather than shear wall or rNEMD, if possible. So far I haven’t come across a direct literature comparison in terms of whether this method with TIP4P/2005 can work or not.
I’m thinking this is some sort of parallel computing problem, however I’m not sure what sort of things I could change/investigate to try and resolve it.
I would be really grateful for some help with this, or any wisdom/opinions/pointers on what might be happening.
Olivera