Dear all,
Recently, I developed a fix for induced point dipole moment polarization model like AMBER FF02. The code runs well and the calculated induced dipole is correct. But I encountered some problem with thermostat and barostat. I use NPT ensemble with fix npt command. But the temperature is always 30K higher than the desired temperature. I think it is related to the pairwise force of induced dipole - dipole and dipole - charge interaction since I just use post_force method to add forces to atoms(should it only be used for external force?). Hope someone could give suggestions. I only find the detail of polarization method but not thermostat in journal papers. Thanks in advance.
Regards,
Han
Dear all,
Recently, I developed a fix for induced point dipole moment polarization
model like AMBER FF02. The code runs well and the calculated induced dipole
is correct. But I encountered some problem with thermostat and barostat. I
use NPT ensemble with fix npt command. But the temperature is always 30K
higher than the desired temperature. I think it is related to the pairwise
force of induced dipole - dipole and dipole - charge interaction since I
just use post_force method to add forces to atoms(should it only be used for
external force?). Hope someone could give suggestions. I only find the
detail of polarization method but not thermostat in journal papers. Thanks
in advance.
well, does your simulation conserve energy *without* a thermostat?
i.e. using only fix nve?
axel.
Thanks Axel. I just run it with fix nve for 0.004ns. Neither pe or ke is conserved. What’s the reason for it?
Thanks Axel. I just run it with fix nve for 0.004ns. Neither pe or ke is
conserved. What's the reason for it?
why should those be conserved by themselves? you need to check etotal.
4ps is a very short time. did you equilibrate your input structure
sufficiently?
i cannot help but thinking that you are what is often referred to as
"shooting from the hip", i.e. that you are pushing forward with your
implementation rather quickly without checking things at intermediate
steps as careful as it is needed, and also that you do not possess the
necessary experience in MD in general (or let nobody advise you that
has) for doing such an implementation and validating it.
as you may have already noticed, what is written in a paper is often
skipping over technical details and practical solutions for technical
or practical problems unrelated to the algorithm itself. so you'll
have to account for that as well.
axel.
I forgot to tell you about etotal. It’s not conserved either. Actually, I just tackle MD problem from this summer and there is nobody around that I could ask. My main research field is CFD.
Anyway, thanks for your advice. I’ll run the nve ensemble for longer time and check if the energy is conserved.
I forgot to tell you about etotal. It's not conserved either. Actually, I
just tackle MD problem from this summer and there is nobody around that I
could ask. My main research field is CFD.
please forgive my bluntness, but i consider it not very smart to start
a complex endeavor such as yours without *any* competent guidance and
oversight. posting questions here in this list is totally inadequate.
it is almost guaranteed that you will make many mistakes without
noticing and will waste a lot of time on solving problems for which
better solutions already exist. you are not the first person in a
similar (self-described) situation and there is quite a history of
pretty significant blunders. i don't want to discourage you to
continue your work, but it is hard to underestimate, how much you will
benefit from competent guidance (in person).
also, i don't believe the "there is nobody around" statement. it is
much more likely that you have not looked in the right places. i've
heard this kind of "excuse" many times before and often from people in
places, where i personally knew that people with the proper expertise
are available in the same institution, sometimes even in the same
department.
axel.