Skip to content

Generating σ-fields

For a sample space Ω\Omega, let A\cA be a set of subsets of Ω\Omega. Suppose we want to generate a σ\sigma-field using A\cA. Note that there may be many different σ\sigma-fields F\cF such that AF\cA\subseteq\cF. Hence, to make this generation unique, we will focus on the smallest such F\cF.

  1. For A={}\cA=\{\emptyset\} or A={Ω}\cA=\{\Omega\}, σ(A)={,Ω}\sigma(\cA)=\{\emptyset, \Omega\}, the trivial σ\sigma-field.
  2. For A={A}\cA=\{A\} for some AΩA\subset\Omega, σ(A)={,A,Ac,Ω}\sigma(\cA)=\{\emptyset, A, A^c, \Omega\}.
  3. For A={{ω}:ωΩ}\cA=\{\{\omega\}: \omega\in\Omega\}, σ(A)={A:A\sigma(\cA)=\{A: A is countable or AcA^c is countable }\}.

Denoted by B\cB, the Borel σ\sigma-field is defined on the sample space Ω=(0,1]\Omega=(0,1]. It is the σ\sigma-field generated by B0={finite unions of disjoint subintervals of Ω}\cB_0 = \{\text{finite unions of disjoint subintervals of }\Omega\} (recall that we saw this set earlier as a counterexample for σ\sigma-fields). Note that one can define the Borel σ\sigma-field on R\R in the same way as we did for Ω\Omega.

It is well known that B\cB can also be generated by open, closed, and half-open sets. Formally, let

  • Q={(a,b]Ω}\cQ=\{(a, b]\subseteq\Omega\}
  • R={(a,b)Ω}\cR=\{(a, b)\subseteq\Omega\}
  • S={[a,b]Ω}\cS=\{[a, b]\subseteq\Omega\}
  • T={[a,b)Ω}\cT=\{[a, b)\subseteq\Omega\}

The notation above may feel a little weird, but essentially it says that Q\cQ is the set of all intervals of Ω\Omega that are open on the left and closed on the right (similarly for the other three sets).