ensemble
/ on-SOM-bul /
Suppose you want the typical outcome of one coin toss, but you only have one coin. You could imagine instead a huge room full of identical coins all tossed at once, and just look at the fractions. An ensemble is exactly that imaginary crowd: a vast collection of identical copies of your system, each prepared the same way but free to be in any of its allowed microscopic states.
More precisely, an ensemble is a conceptual set of all the microstates compatible with a fixed set of macroscopic conditions, together with the probability of each. By averaging a property over this whole imagined collection — an ensemble average — you predict what a single real system does over time. The most common types are named by what they hold fixed: the microcanonical fixes energy, the canonical fixes temperature, and the grand canonical fixes temperature and chemical potential.
Why it matters: the ensemble is the formal stage on which statistical mechanics is performed. It lets you replace the impossible task of following one system through time with the tractable task of averaging over many copies. The honest caveat is that this swap relies on a deep assumption (called ergodicity) that the two kinds of average really do agree — true for ordinary systems but not guaranteed for every one.
To predict the pressure of gas in a sealed flask at fixed temperature, you do not film one flask for hours. Instead you imagine countless identical flasks, each frozen in a different molecular arrangement, and average the wall-pushing force across them all — a canonical ensemble at work.
Averaging over many imagined copies stands in for following one system over time.
The ensemble is a thinking tool, not a real pile of flasks. The whole point is that for a large system the ensemble average and the long-time average of a single system come out the same.