[gmx-developers] Fluctuation properties

[David van der Spoel]

On 1/27/10 9:11 PM, Michael Shirts wrote:
> Hi, David-
>
>    
>> we are trying to get accurate fluctuation properties out of mdrun. For
>> this purpose I have built in a calculation of the enthalpy (Etot + pV)
>> and store that in the energy file for maximum accuracy.
>>
>> Which NPT algorithm/integrator combination is recommended that minimally
>> disturbs the fluctuations?
>> Initial tries with Leap Frog/Nose Hoover/Parrinello Rahman give us
>> fluctuations that are quite a bit too large (using GAFF):
>> methanol: cp_exp=81, cp_gaff=141
>> ethanol:cp_exp=122, cp_gaff=211
>>      
> Sounds like a great time to try the velocity verlet code with the
> Martyna-Tuckerman-Tobias-Klein pressure integrator that I just
> integrated into the CVS last month!   Theoretically, it should give
> the correct fluctuations.
>
> It conserves the conserved quantity correctly -- but there might be
> something else subtle going on, so this would be a good test to run.
>
> To use:
> use 'md-vv' for integrator
> and 'MTTK' for pcoupl.
> tcoupl should remain 'Nose-Hoover'.
>
> (Yes, I need to catch up the documentation -- I've been working on it,
> but it isn't checked in yet).
>    

Trying to run as you advertise it (512 TIP4P/PME/LJ-Cut):
Statistics over 7101 steps [ 5.0000 thru 19.2000 ps ], 2 data sets
All averages are over 7101 frames

Energy                      Average       RMSD     Fluct.  Tot-Drift
-------------------------------------------------------------------------------
Total Energy               -17316.8    169.118    154.362   -239.337  
(kJ/mol)
Conserved En.              -17304.4    7.48269    1.90526    25.0665  
(kJ/mol)

Is this as expected?

And another point: is this supposed to run in parallel? It runs fine on 
1 cpu but gives me a SEGV on 4 cores...

> Note also that 'md-vv-avek' is another integrator that uses the
> average of the half step kinetic energies to determine the
> temperature, instead of the full step velocities.  I'm not 100%
> certain about the validity of this way of doing things this way when
> combined with full step pressure control, so run 'md-vv' also if you
> do 'md-vv-avek', just for comparison's sake.
>
> Note that pV should be the volume times the applied pressure, not the
> instantaneous pressure -- this was wrong a few months ago,but I posted
> a correction 2-3 months ago.  Don't know what version you are using
> for these calculations, so that may not be an issue anyway.
>    


-- 
David.
________________________________________________________________________
David van der Spoel, PhD, Professor of Biology
Dept. of Cell and Molecular Biology, Uppsala University.
Husargatan 3, Box 596,  	75124 Uppsala, Sweden
phone:	46 18 471 4205		fax: 46 18 511 755
spoel at xray.bmc.uu.se	spoel at gromacs.org   http://xray.bmc.uu.se/~spoel
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++