[gmx-users] Normal Mode Analysis in Double-Precision

[Inon Sharony]

Hi all!

I'm trying to perform a normal mode (NM) analysis, but having trouble with the
recommendation of first performing a double precision energy  
minimization before writing
the Hessian. The calculation of the Hessian matrix is done using  
mdrun, and its
diagonalization is done using g_nmeig. I have managed this procedure  
using single-precision, but if the recommended maximal force (the  
minimization convergence criterion) is 1E-5 kJ mole^-1 nm^-1 , I was  
only able so far to get to 1E-2 before getting a message that the  
minimization algorithm has reached machine precision. I have tried  
Conjugate Gradient and L-BFGS.

I also compiled GROMACS in double-precision, but for some reason when  
I try to minimize the energy of the same molecules, I get a  
segmentation fault. I'm not sure where this is coming from (already  
checked the user lists).

Can someone with experience in Normal Mode analysis in GROMACS (or  
just general
experience in using GROMACS with double-precision) help me with my difficulty?

(Currently I'm running pentane or para-dimethylbenzene)

Thanks for your help,
-- 
Inon   Sharony
×™× ×•×Ÿ     ×©×¨×•× ×™
+972(3)6407634
atto.TAU.ac.IL/~inonshar
Please consider your environmental responsibility before printing
this e-mail.