timeintegration nve_sphere

Submitted by JoshuaP on Mon, 01/26/2015 - 15:56

Hi,

I have a question about the functions initial_integrate and final_integrate from fix_nve_sphere. Where is the difference in the functions? When are these function used?

Thanks
Joshua

Forums:

aaigner's picture

aaigner | Wed, 01/28/2015 - 13:09

Hi Joshua!
LIGGGHTS uses a velocity verlet integration scheme, which is an two step approach. Between the two integration steps the force is updated. Check the function run in the file verlet.cpp (seach for modify->initial_integrate) to get an idea how one timestep works.

Bests
Andreas

JoshuaP | Wed, 02/18/2015 - 13:29

Hi Andreas,
so the initial and final integration are used to implement the velocity verlet with half step. But if I use the shortened velocity verlet with the acceleration from this step and the previous one, it is not necessary to use initial and final integration? So I can just save the old acceleration in a new variable and use it for the next step, update it and dont need the half step integration, am I right?
Is the halfstep verlet faster or why is it used?

Many thanks
Joshua

ckloss's picture

ckloss | Mon, 02/23/2015 - 12:33

Hi Joshua,

if you implement a different integration scheme you may or may not need both functions. If it uses old accelerations, I would store them in a fix property/atom

>>Is the halfstep verlet faster or why is it used?
It is fast, efficient, does not need to store any old values and only adds terms which have the same order

Christoph