question about distance based clustering

[jude vishnu]

Hai everyone,

I have a  question regarding the distance based clustering in freud. What is the criterion for choosing the cut off distance? Could someone comment on this? If we change the cutoff,  the number of clusters formed can also change. So what would be the best way to determine this cutoff in a system.

Regards,

Jude

[Tommy Waltmann]

Hi Jude,

The distance is something we leave up to the user to decide. Try a few different distances and see what makes sense intuitively to call a "cluster" in your system. If you provide more information about your system, I may be able to help recommend something, but I would still try a few different distances to see what will work best.

Best,

Tommy

[Bradley Dice]

A common value for the cutoff distance is around the minimum after the first peak in the radial distribution function. That way most of the bonds defining the cluster are neighbors in a particle's first coordination shell.

[jude vishnu]

Hai,

Thank you for your suggestion. Could you look at my rdf and tell me what is the minima after the peak here? I am attaching the text file

 The reason I am a bit confused here is because, I want to look into percolation in my system. Visually it looks like there is percolation. The percolation is caused by polymers arranging themselves in a region in my system. So when doing a normal rdf, I will always end up having peak at close to 1 sigma. Because most of bonded beads will lie there. Shouldn't I be looking at rdf which excludes the nearest bonded polymer beads? What do you think? Do you have a suggestion on how to do this?

Regards,

Jude