So I natively use LAMMPS for my atomistic simulations and these are the first simulations I have run using Gromacs. Because of this I used the grompp.mdp file from the hexane tutorial as a template/starting point for the iteration runs for my IBI. I had the CG runs stable yesterday morning and then whilst experimenting with various parameters suddenly they exploded and I couldn't figure out what change had caused it. So I have just copied the grompp.mdp file from the tutorial again and modified it accordingly and now the simulations are stable (2 iterations so far... and RDF of first non-bonded pair looks fairly close to the target after being underestimated in the first iteration run).
I did a diff command on the *mdp files which worked, and didn't work respectively. The only difference in parameters (other than cut-off which shouldn't cause such an explosion?) are table-extension=1.0 (in working simulation), =0.0 (in broken simulation) and i was using a v-rescale thermostat in broken simulation (although I'm pretty sure I only started using this to control the uncontrollable temperature rise, to no avail...) And also in the broken simulation rlistlong=2.0 was set whilst this parameter is not set in the working simulation. I think it is important for me to understand why the system was so badly behaved so that I don't waste more time in the future. I also tried running a simulation with the PMFs on a system of just two of my molecules in a large box with plenty of space, i.e no overlapping beads and the simulation still blew up! So something must have been fundamentally wrong in the *mdp file? Thank you Denis and Sikandar for your input regarding my methodology. >From the paper that Denis posted about structural coarse-graining in liquid slabs, it suggests that I could modify the VOTCA potential update in the appropriate script? I could calculate the density profile of my bilayer from the reference simulation and then see what shape I can get away with fitting to it, ideally heavy-side step functions but perhaps hyperbolic if needs be. Here is an example of a study where an Inverse Monte Carlo method has been used to derive CG potentials for bilayers where the RDFs don't go to 1: http://onlinelibrary.wiley.com/doi/10.1002/jcc.23610/full Another point is that currently I am planning to CG using an implicit solvent representation in order to speed up diffusion and see some vesicle formation. I have water in my reference simulation so if I in fact need to include water in my model then it's not the end of the world. In the study I posted just above, they do the same thing, parametrise the structure including water explicitly but just use the CG potentials with implicit solvent. Do you think that pretending my system is a slab of bilayer with vacuum slabs above and below will be sufficient for forming bilayers/vesicles? Best, Dan. On Thursday, 29 January 2015 15:46:21 UTC, sikandar wrote: > > Hi Dan, > > As Denis correctly mentioned, IBI might not work for slab-like system in > which RDFs do not converge to 1. One other option, in addition to IBI-like > algorithm proposed in Denis's paper, is relative entropy minimization > method. We showed that it works for water confined in slab-like channels > http://link.aip.org/link/?JCPSA6/137/214707/1. > > Best, > Sikandar > > On Thu, Jan 29, 2015 at 6:55 AM, Denis Andrienko <[email protected] > <javascript:>> wrote: > >> Hi Dan, >> >> If your RDF does not converge to 1 at large distances (e.g. if you do not >> have a bulk system) the IBI method might simply not work. Have a look at >> this paper: >> http://www2.mpip-mainz.mpg.de/~andrienk/publications/jochum_2012_a.pdf. >> You will see that for slab systems you can still use IBI-like alrorithm, >> but you will have to evaluate the 3-D RDF out of 2-D RDF. >> >> Best, >> Denis >> >> -- >> You received this message because you are subscribed to the Google Groups >> "votca" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected] <javascript:>. >> To post to this group, send email to [email protected] <javascript:> >> . >> Visit this group at http://groups.google.com/group/votca. >> For more options, visit https://groups.google.com/d/optout. >> > > -- You received this message because you are subscribed to the Google Groups "votca" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/votca. For more options, visit https://groups.google.com/d/optout.
