On 08/11/2015 09:31 PM, Smith Lee wrote:
> Dear All,
>
> For crystal structure refine, will you please tell me what is radius of convergence in plain language?
>
> Best regards.
>
> Smith
How does this sound?:
Convergence is reached when further refinement does not significantly change the structure.
The structure is in some kind of minimum, not necessarily a global minimum but at least a local minimum. The radius of convergence is how far you can move the structure (in parameter space) away from that minimum,and still have it refine back to the same minimum. If you move further you have gone over a "ridge" and refinement will take it down the other side of the ridge into a new minimum. You have moved it within radius of convergence of that new minimum. The visual idea is the radius of sphere in parameter space, such that if you stay within that sphere it will refine back to the minimum in the center. But of course it would not be a sphere unless the parameters are normalized to their roughness, maybe not even then.
But it is just a concept - no one actually reports a radius of convergence in terms of Angstroms and A**2. It is used to think about strategy.
For example: you are rebuilding your structure and you think this Phe is skewed in the density- would fit better if rotated 10* around chi2. But you don't change it because you realize such a small change would be within the radius of convergence. If that really were better, it would have already gone there. If you move it there, further refinement will just bring it back. But - if you think it should be in rotamer 2 and now it is in rotamer 3, occupying some density that looks more like a water, then you need to change it. The new position is better but it will never get there by itself because that is outside the radius of convergence. When you do change it, there is no use taking a lot of trouble fit as well as possible, because if you just get it within radius of convergence of the best position, the refinement program will take it to the same place regardless exactly how you put it.
eab
|