[gmx-users] LINCS vs SHAKE

chris.neale at utoronto.ca chris.neale at utoronto.ca
Sun Oct 31 03:38:05 CET 2010

Mark, I am quite sure that a difference in the overall simulation time  
between two replicas is not going to introduce any bias. Still, you  
bring up a good point. If the energy distribution (or even worse, its  
average) at a given temperature is affected by the timestep, then this  
would be a bad idea. Given the difficulties in converging ST A-values  
or (even harder) SREM PDFs, I'm not sure if the answer to this is  
known, but I would bet that the timestep is going to influence this  
now that you point it out.


-- original message --

On 31/10/2010 10:04 AM, chris.neale at utoronto.ca wrote:
> Justin is probably correct, your system is unstable. However, let me  
> mention another possibility. As the temperature increases, the  
> largest stable timestep decreases in length. Is it possible that  
> over some temperature (>1000K?) a 2 fs timestep is no longer viable?  
> In this case, even a well minimized system with a large number for  
> lincs_iter is still going to give you incorrect sampling and  
> possibly crash. You'd need to reduce you highest temperature or  
> reduce your timestep.
> I don't know if the gromacs REMD algorithm allows you to use a  
> smaller timestep at higher temperatures, but it's something to look  
> into.
It doesn't. Off the cuff, I can't say whether it's sound to simulate at
a smaller timestep at higher temperature and exchange with a simulation
that has done the same number of larger timesteps.

> Chris.
> -- original message --
> For Replica exchange, is there any advantage of in using SHAKE over
> LINCS(other than the stepsize)?
>  I am running an REM simulation and the simulation stops after running for
> variable number of steps (100000, 10000000 etc.) because some bond moves
> more than 30 degrees and LINCS gives a warning.

More information about the gromacs.org_gmx-users mailing list