Eric,
Actually, in the current implementation in pmemd, there is a fairly simple solution for semiisotropic pressure coupling:
ntp=3
csurften=0.0
Why does this work? Starting with (in LaTex):
\gamma = \frac{h_z}{2} (P_{zz}-\overline{P_t})
where gamma is the surface tension, h_z is the instantaneous box dimension in the direction of surface tension, P_{zz} is the pressure in the direction of surface tension, P_t is the average tangential pressure.
reduces to the semiisotropic coupling when there is no surface tension. If my math is correct, then:
\gamma = 0
means
P_{zz} = \overline{P_t} = \frac{P_{xx} + P_{yy}}{2}
Where P_{xx} and P_{yy} are the tangential components of pressure. In pmemd, this constraint works by setting P_{xx} = P_{yy} and solving for P_{xx}.
It is an additional constraint that removes one degree of freedom from the simulation. Lipid14 simulations were run with *anisotropic* pressure coupling.
It depends on the lipid bilayer system that you're trying to simulate. It's possible on very long timescales that bilayer system may not be stable and will collapse in one dimension (i.e. the box dimension decreases significantly and the simulation stops). If that is a major problem with your simulations, you may want additional semi-isotropic coupling. It's definitely something to take note of in these bilayer simulations.
All the best,
Ben Madej
UCSD Chemistry and Biochemistry
SDSC
________________________________________
From: Jason Swails [jason.swails.gmail.com]
Sent: Tuesday, May 20, 2014 10:11 AM
To: amber.ambermd.org
Subject: Re: [AMBER] cant run ntp=3 without csurften
On Tue, 2014-05-20 at 10:48 -0600, Eric Hill wrote:
> Hi Amber devs,
> Now that lipid14 is published, I was wondering if you had a suggestion
> to work around the following issue: Amber requires that csurften be set
> for ntp=3 (semiisotropic coupling). This would be nice to use in my
> membrane simulations. If there is a workaround possible, it would be
> great to know.
I think the appropriate workaround is to just use anisotropic pressure
scaling. The semi-isotropic barostat seems like a hack to me to correct
for deficiencies in the force field (as is the constant surface tension
term). Since Lipid14 no longer needs constant surface tension, it
likely does not need semi-isotropic coupling, either.
If your system is truly isotropic, an anisotropic barostat will result
in isotropic scaling on average. Likewise, if your system is truly
semi-isotropic (like you would hope for an ideal layer), the anisotropic
barostat will result in semiisotropic scaling on average.
Good luck,
Jason
--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue May 20 2014 - 11:30:02 PDT
