CHARMM Development Project Forums User Discussion & Questions Energy terms, Constraints, Restraints, and Solvation size of solvation box

recently, I performed MD for alanine dipeptide in water box. The population of the backbone dihedral angles (phi/psi) were quite different in water boxes with different size. we can fine PPII, alpha-R, alpha-L in cubic water box with its sidelength of 28, but in the box with its sidelength of 42, there were no population for alpha-L.
the force field are the same, and the CMAP is considered.
my question is, how did the box size affect the population of the backbone dihedral angles, and how to decide the most probable size of the solvation box?
thank you so much for your time!

It is unlikely that 28A or 42A side length would make a drastic difference for the alanine dipeptide. Are your results converged? There are tons of literature using this model system, you can try to reproduce some of it.
A series of converged simulations with increasing side lengths will tell you what is needed in order not to have a size dependence.

Thank you so much for your time, I'll check the literature and try to reproduce some of it.
ps: I've try 30 A and 42 A side length for ALAD in CHCl3, the dihedral distribution shows no big difference, but the MD of ALAD in 34 and 42A DMSO boxes are quite different.

The fundamental question is whether your simulations are converged; this must be established before your start making comparisons.

Running replicate simulations, esp. starting from different molecular configurations, is one way to try to answer this question.

Dear Prof. Lennart, I am sure the energy, the temperature, the presure and etc. are convered.

Thank you for your advice, I'll try it with other initio structure (the origin ALAD is extended).

Yes, but is the conformational distribution converged?

Dear sir, I am sorry that I don't know how to make sure the conformational distribution was converged, does it have another criterion ?

AS Rick said, run several independent simulations and see if your distributions are the same; if you have a long simulation you can see if the distributions obtained using different parts of the trajectory are the same. Ideally you should have multiple transitions between the accessible conformations.