[gmx-users] Position restraints and double precision

Miguel Caro miguel.caro at aalto.fi
Mon Apr 14 15:14:45 CEST 2014


Thank you for your reply. I will look into freezegrps, it might be just 
what I need.

Question 1) was more related to whether needing such large numbers might 
be indicative of a problem with my calculation, given that they are 5 
orders of magnitude larger than the default. Of course I would expect 
oscillations from the harmonic potential but I was not expecting to be 
able to "see" them (as in see atoms oscillate during VMD visualization).

Thanks,
Miguel

On 2014-04-14 15:55, Justin Lemkul wrote:
>
>
> On 4/14/14, 8:45 AM, Miguel Caro wrote:
>> Hi all,
>>
>> I am trying to do an MD simulation of a small hexagonal box where I 
>> need some of
>> the atoms to stay fixed. This is because I am mixing ab initio and 
>> classical MD
>> and some of the atoms in my system should remain within the 
>> configuration
>> predicted by DFT.
>>
>> As far as I know, GROMACS only allows energy penalties, rather than true
>> position constraints, and so I have to resort to extremely high 
>> energy penalties
>> for some of the atoms:
>>
>>     [ position_restraints ]
>>     ;  i funct       fcx        fcy        fcz
>>         1    1       100000000       100000000       100000000
>>         2    1       100000000       100000000       100000000
>>     etc...
>>
>> The above restraints still lead to a small oscillation of the "fixed" 
>> atoms.
>> These numbers being so large means that the number of significant 
>> figures
>> allowed for the total energy during the energy minimization is 
>> limited to very
>> few decimal places using single precision. As a matter of fact, GROMACS
>> complains that convergence to machine precision is achieved before 
>> the required
>> precision.
>>
>> I guess I have two questions:
>>
>> 1) Why do I need such huge energy penalties to fix my atoms?
>>
>
> Because, as you say above, position restraints don't fix the position 
> of atoms.  You can apply a bias to disfavor movement, but it will 
> still move. Fortunately, you can actually fix atoms using freezegrps.
>
>> 2) Will the precision issue also affect an MD run or only the EM run? 
>> Should I
>> reinstall GROMACS in double precision?
>>
>
> The precision issue that you're referring to affects EM algorithms; I 
> doubt there would be any practical problem during MD.  Most of the 
> truly sensitive calculations during MD are actually done in double 
> precision, regardless of the chosen output precision.
>
> -Justin
>

-- 
*Dr. Miguel Caro*
/Postdoctoral researcher/
Department of Electrical Engineering and Automation,
and COMP Centre of Excellence in Computational Nanoscience
Aalto University, Finland
Personal email: *mcaroba at gmail.com*
Work: *miguel.caro at aalto.fi*
Website: http://mcaroba.dyndns.org


More information about the gromacs.org_gmx-users mailing list