Constraints with Langevin Middle Integrator

[Ethan Meitz]

Why would you not always need to apply that velocity correction though? It seems that the forces used in the position update should always be the same ones used in the velocity update.

And within OpenMM is this correction added when you use constraints automatically or do you have to define a custom integrator?

[Peter Eastman]

I don't understand your question. It is always applied. It's part of the algorithm.

All the standard integrators apply constraints correctly. They use different algorithms in which constraints come up in different ways, but each algorithm handles them in a correct way.

[Ethan Meitz]

Oh, you said that the (x1 - x)/dt correction is part of RATTLE and not the constraint algorithm (SHAKE, CCMA etc.). I was asking if that correction is used by OpenMM when you use a constraint algorithm or do you have to define a custom integrator?

[Peter Eastman]

I already said: all standard integrators apply constraints correctly. That correction is part of RATTLE. Other algorithms account for velocities in other ways, but all of them handle them correctly.

[Ethan Meitz]

Thanks for your patience and help, I learned a lot from this!