[gmx-developers] inconsistent logfiles

[Erik Lindahl]

Vibin R wrote:
> Hi
> 	While doing energy minimisation with different initial structures,
> I am getting different starting points (such as sometimes step0,
> sometimes step1, step4 etc) both from energy trajectory and from the
> logfile)
> 1. What could be the reason, why i am not getting step 0,1,2,...etc
> consistently. What should i do for getting it?
> 2. Will it be meaningful if i assume that the step0 energy values which we
> get from the logfile as the energy of the structure prior to energy
> minimisation?
>                                         

Hi,

I'm not 100% sure I got you right, but this *might* explain it :-)

The steepest descent algorithm works by taking a trial step in the 
direction of the gradient. If the new energy is lower we move, and if 
not we modify the stepsize. So, if step2 wasn't successful we overwrite
that with step3 on the stadard output. This won't work in the energy 
file since we can't go back in a file, but if I recall correctly we only
write successful energies to the files...


Cheers,

Erik