[gmx-users] no output dgdl file

[Justin A. Lemkul]

Moeed wrote:
> 
>         Hi Justin,
> 
> 
> I was looking all the time for a file with dgdl extension throughout the 
> directory. IIRC in previous versions -dgdl would generate such 
> extension. I found now .xvg file which has dhdl data. (if I am right 
> dgdl is the very dhdl in 4.5.3!) 
> 

The extension has always been .xvg, but the root filename has changed.  It is 
now perhaps a bit more accurate (delta Hamiltonian).

>  
> 
>         I have a little problem with FE output file. Below is the
>         settings and also I am including -dgdl in the command I issue
>         but no dgdl (or dhdl) file generates. I dont figure where the
>         problem lies ! (version 4.5.3).
> 
> 
>     How long has the simulation been running?  Output is buffered, and I
>     find that dhdl.xvg is not updated often.
> 
> 
>         free_energy          =   yes
>         init_lambda          =   0 delta_lambda         =   0
>         sc_alpha             =   0.5
>         sc-power             =   1
>         sc_sigma             =   0.3
>         foreign_lambda       =   0.1
>         dhdl_derivatives     =   yes
>         couple-moltype       =   Polymer
>         couple-lambda0       =   vdw-q   couple-lambda1       =   none
> 
> 
>     Note that turning off all interactions simultaneously (especially
>     when using soft-core) can be very unstable.  You should (de)couple
>     vdW and Coulombic interactions separately.
> 
> 
> Yes I read your comment about decoupling separately earlier on but I was 
> not sure if this could depend upon the system so I was thinking of 
> comparing results from two trials:
> 
> trial 1:
> couple-lambda0       =   vdw-q    
> couple-lambda1       =   none 
> 
>  trail 2 A:
> couple-lambda0       =   vdw-q    
> couple-lambda1       =   vdw
>  trail 2 B
> couple-lambda0       =   vdw
> couple-lambda1       =   none
> 
> So:
> 
> 1- Does this decoupling issue depend on the system or it is a must for 
> all FE studies?

No, it pertains to the very stability of all calculations.  Whenever I've 
decoupled charges and LJ simultaneously, either the simulations crash or the 
resulting energies are not reliable.  There are numerous book chapters and 
papers dedicated to these topics.

> 2- I actually went through many of FE posts and never found how separete 
> decoupling can be done. from  trail 2 A and B I get dgdl files but how 
> to get total vdw-q to none from these two? Please shed some light on the 
> details of procedure..
> 

Add them.  DeltaG is a state function.

> 3- With free_energy turned off my runs take only a few hours but as I 
> shift to free_energy          =   yes, runs take nearly one day foe each 
> lambda.. Is this normal for s system having 2500 atoms?
>  

Yes.  The free energy code slows down runs quite a bit.

-Justin

-- 
========================================

Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
MILES-IGERT Trainee
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================