now, now. there is no reason to get upset about being corrected.
Its not the issue of changing parameters without understanding. So far, there are very little options to play with the parameters. While using the chain tool, I changed the distance between monomers as 4.03 which I got from several papers for my polymer and the reduced unit (no distance less than....) as 4.08. Then rest of the work has been done by the chain tool!!!
Then for the micelle example file, the soft running portion is pretty much straight forward. I changed the unit real because in my main simulation I am using all the units in real unit (Can I not use the real unit for soft running?) and the bond harmonic potential I got the parameters from literature. you can say that the potentials are wrong but then i have nothing to say because i didn't build the potentials myself!! But based on these potentials that literature has represented further good analysis.
this is exactly what i am talking about. what is your purpose here? to
take a system that is far away from equilibrium and has overlapping
particles to something that is (much) closer to what you need. does it
matter whether parameters are perfect to <1%?
not at all. you just want to get close, so that you can use the proper
parameters and the proper model without the simulation barfing all
over you. so you have to think what is going to happen. if particles
overlap, they will repel each other, quite strongly if you have a
normal interaction. that is supposed to been taken care of by
"growing" the soft potential. but you also want the chains to be at
the proper bond lengths. so what does that mean for your parameters?
- the repulsion of your atom cores should be rather weaker than stronger
- the bond strength should be rather stronger than weaker.
- lengths and sizes matter, energies can be off.
now with *this* kind of thinking you have to look at your parameters again.
- bond length is fine. anything from 4.0 to 4.1 would probably do
- bond strength. way to wimpy. typical bond strengths in classical
models are easily two orders of magnitude larger and i would probably
go even larger just for safety
- on the other hand, the LJ epsilon is often 1-2 orders of magnitude
smaller in real units than in reduced units
have you actually visualized the trajectory of this step? do your
bonds stretch too far or not?
now considering that you may (still) have fast traveling atoms and
probably longer chains than the micelle example you have to anticipate
that it may be more difficult to unoverlap the chains, so that has to
be taken into account with the neighbor list parameters. also the
differences in units, too. with this in mind, it seems like a good
idea to have at least 1/2 sigma as neighbor list skin and update the
neighbor list every step to be save. how much is your average sigma?
how does it compare to 0.3? why the need to use delay 5 since you are
just running short simulation to unoverlap you initial configuration
and performance is really not a big concern.
BUT, you *did* have warnings about very long bonds and subsequent
problems, yet you chose to ignore the warnings and not investigate if
things are behaving as they should. how can you expect people not to
tell you that you act unreasonably?
criticism is part of the business and it will get much more harsh and
unfair as you move on. don't see it as a personal attack (what would
be the benefit of that?), but as a challenge (which actually has large
benefits in the long run). you won't be motivated to do a better
effort the next time by giving you a pat on the back, even when things
don't look right.
The portion you are saying strange is obtained from the example file which is not very hard to understand. Is the example file strange then!!!
no. it works for what it does and for the system at hand. but there is
a difference between just blindly copying something and adapting
something and that difference has apparently not yet fully settled in
your mind.
axel.