[gmx-users] Re: Potential energy from a previous configuration mismatch

[Justin Lemkul]

On 11/30/12 11:00 AM, Mark Abraham wrote:
> On Fri, Nov 30, 2012 at 2:35 PM, Justin Lemkul <jalemkul at vt.edu> wrote:
>
>>
>>
>> On 11/30/12 8:21 AM, escajarro wrote:
>>
>>> I do have constraints, all the bond lengths are fixed and they are
>>> specified
>>> in the topology file. Am I doing something wrong? I copy here the
>>> definition
>>> of one of the models I use. I also tried using in the input the line:
>>>
>>> constraints              = all-bonds
>>>
>>> instead of
>>>
>>> constraints              = none
>>>
>>> with the same result
>>>
>>> <http://gromacs.5086.n6.**nabble.com/file/n5003394/**Clipboard.jpg<http://gromacs.5086.n6.nabble.com/file/n5003394/Clipboard.jpg>
>>>>
>>>
>>>
>>>
>> That all looks fine, and the results should be the same under those two
>> circumstances.  The problem is likely due to the fact that L-BFGS cannot
>> actually make use of constraints (a bug in mdrun), so that is the source of
>> the difference.  The EM is being done without constraints, and then the MD
>> starts with constraints.  Use a different EM algorithm and see if the
>> outcome improves.
>
>
> If L-BFGS is silently not using the constraints, then I think that is a
> real bug and it would be good to report it so we might sometime fix it. Or
> maybe it's an undocumented "feature" we need to document? Which is it,
> please?
>

 From the recent discussion, it seems that L-BFGS is silently ignoring the 
constraints, as Berk observed there is no check for this combination, which 
explained David's finding that bond lengths seem to vary randomly:

http://lists.gromacs.org/pipermail/gmx-developers/2012-November/006433.html

-Justin

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

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

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