Forum Navigation
You need to log in to create posts and topics.

selecting an appropriate cutoff distance for cluster analysis


I am attempting to look at vacancy clusters during an impact simulation. I have used modifiers (defect analysis, expression selection, delete selection etc) such that I only have the vacancies for my cluster analysis. However, I am unsure what value to use for my cutoff distance. This is obviously very important in deciding how the clusters are chosen. I have FCC silver, so would it simply be the nearest neighbor distance (2.88)?




Hi Liam,

you're right, the cutoff in the cluster analysis modifier controls if particles in the direct and indirect reach of one or more other particles are considered to belong to the same cluster. Any two particles from the same cluster are connected by a continuous path consisting of links that are shorter than your specified cutoff so to say.

I'm wondering if what you suggested above might be too short? I uploaded a sketch of an fcc unit cell where blue spheres are regular sites and unfilled spheres are supposed to represent vacancies. Would you, for example, consider case (2) still a vacancy cluster? If yes, then the next-nearest neighbor distance would be too short as cutoff, it would merely group the two vacancies in case (1) into the same cluster. In case (3) however the three vacancies would belong to the same cluster again if you use the next-nearest neighbor cutoff criterion, since they are connected by a continuous path.


Uploaded files:
  • fcc.png

Perhaps I could use somewhere between 1 and 2x the nearest neighbor distance. I have consulted some literature but am not seeing a standard answer.

New for our users in China: OVITO on WeChat 

Official OVITO WeChat channel operated by Foshan Diesi Technology Co., Ltd.