Equilibration of membrane water system (pressure in NVT)

From: Rene Hamburger UNI KL (hamburge_at_physik.uni-kl.de)
Date: Fri Oct 26 2018 - 02:32:38 CDT

Dear all,

I'm having trouble getting my system equilibrated.

For better overview:
1. My overall goal
2. my system
3. what i did so far
4. my questions

1. My goals: Get an equlibrated system to analyse diffusion of 3 small
molecules in water in the presence of a membrane + use ABF scheme to
calculated interaction landscape between polymer and molecules.

2.My current system consists of
  2 polymer membranes (both single chain in amorphous phase of 4k
monomers = 24k atoms) plane normal vector along z axis. size: x,yA , zA
   with water+ions (45600 water molecules) and 3 small molecules in
between. Total atomnumber ~ 185k
system view along z axis (used pbc): | water(10A) | membrane(30A) |
water ----- water(80A) | membrane(30A) | water(10A) |

3. Throughout i use the langevin thermostat with settings
  langevin on
  langevinDamping 10
  langevinHydrogen no
  langevinTemp $temperature

and the langevin piston Nose-Hoover method as barostat with:

useGroupPressure no
langevinPistonTarget 1.01325 ;# pressure in bar ->1 atm
langevinPistonPeriod 200. ;# oscillation period around 200 fs
langevinPistonDecay 100. ;# oscillation decay time of 100 fs
langevinPistonTemp 300 ;# coupled to heat bath

exclude scaled1-4
1-4scaling 1.0
switching on
switchdist 10
cutoff 12
pairlistdist 14
# integrator
timestep 1 # in fs
nonbondedFreq 1
fullElectFrequency 2
stepspercycle 20
outputPressure 100
xstFreq 100

I made the system ready so far that the temperature is properly
controlled and density is only fluctuating around constant value, I had
pressure components between -10 and 10 bar.
I tried to reach equilibrium in pressure/volume size in NPT with p->1bar.
After figuring out the pressure methods coupling in NAMD (like
useConstantArea, useConstantRatio and flexibleCell) in an anisotropic
NPT simulation I managed to get an average value of 1 bar in each
diagonal component of the pressure tensor (averaging over PAVG, 10ns sim

As I like to do the above mentioned productions runs (diffusion, ABF) in
NVT I now want to change to constant volume, which is where I have trouble.

The boxvector sizes vary with an sigma ~ 0.2A for x,y and 0.45A for z.
As I've read in literature one should use the average boxvectors from
the NPT then but using those or values close to them doesn't seem to get
me there because of large flutuations coming from the incompressibility
of water and/or to short simtime.

4. my questions:

  4.1. Is it reasonable to use anisotropic pressure coupling for a
membrane system?
       For biological membranes I've seen in literature that it's most
common to use semi-isotropic coupling (ConstantShape or even ConstantArea)

  4.2. How large or small are the boxvector fluctuations for similar
heterogeneous system in your experience?
       Though I may just have to do several dozen ns longer such that
the box size statistics get better?
  4.3. Can I even expect to obtain the desired pressure in NVT?
  4.4 Is it necessary to check/evaluate the diagonal components each or
should I mainly care about the trace?
4.5 Is there another common way to go from NPT to NVT?

Thanks in advance for any comments and hints.

Kind regards,
René Hamburger
PhD student

This archive was generated by hypermail 2.1.6 : Mon Dec 31 2018 - 23:21:28 CST