Are Statistical Ensembles Just Mathematical Tools or Realistic Concepts?

[parton]

Recently I was asked whether the concept of a statistical ensemble is actually realistic and I recognized that I can not really answer this question myself. I think that it is more a mathematical "tool", because of the probability nature rather than a realistic "thing", but still it helps us to describe statistical systems.

What do you think?

 

[Mapes]

I agree. McQuarrie (in Statistical Mechanics, which I have handy) describes an ensemble as a "(mental or virtual) collection of a very large number of systems, each constructed to be a replica on a thermodynamic (macroscopic) level of a particular thermodynamic system of interest."