[gmx-users] Problem with NPT equilibration

[Mark Abraham]

Bing Bing wrote:
> Dear all,
> I've tried to simulate a membrane system with POPC and protein following
> Justin's tutorial. The system was minimized with 2 stages. First using
> restraint on everything accept water for 10000 SD steps:-
> Steepest Descents converged to Fmax < 1000 in 1785 steps
> Potential Energy  = -1.69457097061162e+06
> Maximum force     =  9.15408957613777e+02 on atom 24360
> Norm of force     =  7.53074616222725e+03
> I've proceed with 2nd stage minimization without restraint,
> Steepest Descents converged to Fmax < 1000 in 844 steps
> Potential Energy  = -1.87072963657420e+06
> Maximum force     =  9.14155621660386e+02 on atom 24360
> Norm of force     =  7.84646890755063e+03
> The energy curve seems ok. With this, i went on with NVT equilibration (
> restraint on protein) for 100ps. Both energy and temperature plot look fine
> also.
> But it failed while running NPT 1ns(restraint on protein also), it stop with
> Range checking error, whereby the ci value exceeded the ci cutoff. it is
> suggested that the minimzation might not done properly. But the minimization
> were converged in both stages. Is there anything i missed out here?

That all looks fairly reasonable, but it seems you still need to be more 
gentle. If the box density differs markedly from the expected value 
(presumably about 1 g/mL for protein+membrane in water) then the NPT 
transition will be sharp. You may need to choose a smaller integration 
step size during the equilibration, or go back and choose a better box 
size. Obviously, use nstxout=1 and watch where the breakage starts to 
happen. That may clue you in to a problem.

Mark