Page 1 of 1
Zero-timestep extrapolation
Posted: Tue Jan 14, 2014 12:49 pm
by varelse
Is it possible to estimate the order in which the zero-timestep extrapolation increases accuracy? I want to check if I need to run next DMC calculations, or just use the result of a single run.
Re: Zero-timestep extrapolation
Posted: Tue Jan 14, 2014 1:00 pm
by Dario_Alfe
Extrapolations are always dangerous, as we do not know what is the functional form of the energy as function of time step.
Re: Zero-timestep extrapolation
Posted: Tue Jan 14, 2014 1:08 pm
by varelse
Well, it should be linear if the timestep is small enough.
Re: Zero-timestep extrapolation
Posted: Tue Jan 14, 2014 1:45 pm
by Dario_Alfe
maybe, but you don't know how small is small enough. Often you find that it looks linear in some range, then you keep decreasing the time step and find that it departs from the linearity that you thought you were observing.