The error occurs, in initial guess for non-bonded potential in ibi process.

[Ran Zhao]

Hello, I want to perform an ibi process for coarse-grained models. There are seven CG beads.

I create distributions for all non-bonded pairs and then use the command: $csg_inverse --options settings.xml. 

When the votca is running for step 000, the initial guess is run.

When the specific pair (B-N.dist.tgt) turns to the initial guess, an error occurs:

In inverse.log file:

This is confusing！

I have attached my settings.xml tiles.

Any advice you have would be appreciate!

[Marvin Bernhardt]

Most likely the distribution B-N.dist.tgt contains some noise at the onset. In other words: The RDF starts with zeros for small r, then there are one or two values larger than 1e-10, then one or more values below 1e-10, and then the numbers go up to the first peak. If I remember correctly, the algorithm in the Perl script goes from small to large distances, taking the first value above 1e-10 as start and the next value below 1e-10 as end of the distribution.
Three things you can do:

1. sample longer to make noise disappear
2. define min in the settings.xml to exclude the noise
3. smooth out the noise manually

We have an upcoming publication in JCTC that has a more systematic approach to this problem.

Cheers, Marvin

[Ran Zhao]

Thank you for this helpful advice. It is true that the distribution of B-N contains small values after the last zero points. And after solving this problem, this error vanishes.