Measure Theory

non-measurable set

A non-measurable set is a subset of the line so badly behaved that no consistent length can be assigned to it. Any attempt to give it a Lebesgue measure that respects translation invariance and additivity leads to a contradiction. Such sets show why one cannot simply measure every subset of R, and why a sigma-algebra is unavoidable.

Precisely, a non-measurable set is a subset of R that is not a member of the Lebesgue sigma-algebra: Lebesgue measure is undefined on it. The existence of such sets is a theorem, but their construction relies essentially on the axiom of choice; no non-measurable set can be exhibited by an explicit formula or a constructive procedure. The classic witness is the Vitali set.

The deeper lesson is a no-go result: there is no translation-invariant measure defined on all subsets of R that assigns to each interval its length. The Vitali argument turns translation invariance and countable additivity against each other to produce a contradiction. The honest caveat is set-theoretic: in models of mathematics without the axiom of choice (Solovay’s model), it is consistent that every subset of R is Lebesgue measurable, so non-measurability is a consequence of choice, not a brute fact.

In R^3 the Banach–Tarski paradox pushes this even further: a solid ball can be cut into finitely many non-measurable pieces and reassembled into two balls of the original size. Non-measurable pieces have no volume, so no contradiction with conservation of volume arises.