Product Measures

Core Idea

Given measure spaces (X, ℱ, μ) and (Y, 𝒢, ν), the product measure space (X × Y, ℱ ⊗ 𝒢, μ ⊗ ν) is defined on the product σ-algebra generated by measurable rectangles A × B, with (μ ⊗ ν)(A × B) = μ(A)ν(B).

Explainer

From your study of measure spaces, you know that a measure μ on (X, ℱ) assigns sizes to measurable sets in X, with the key property of countable additivity. The product measure construction asks: given two separate measure spaces, how do you build a coherent measure on their Cartesian product X × Y? The answer starts with the most elementary two-dimensional sets: measurable rectangles A × B, where A ∈ ℱ and B ∈ 𝒢. For these, the right measure is obvious — area should equal width times height: (μ ⊗ ν)(A × B) = μ(A) · ν(B).

The analogy to ordinary area helps build intuition. On ℝ² with Lebesgue measure, the area of a rectangle [a,b] × [c,d] is (b−a)(d−c) = length × width. The product measure construction is the rigorous generalization of this: you start by declaring what measure means on rectangles, then extend to all measurable sets using the Carathéodory extension theorem. The product σ-algebra ℱ ⊗ 𝒢 is the σ-algebra generated by all measurable rectangles — it contains all the sets you can build from rectangles by taking complements, countable unions, and intersections.

The subtlety is in that extension step. Measurable rectangles don't cover all of X × Y — you can have complicated sets, like diagonal strips or fractal-shaped regions, that are not rectangles and not even unions of finitely many rectangles. The Carathéodory extension theorem guarantees that the assignment on rectangles extends uniquely to a full measure on ℱ ⊗ 𝒢, as long as the rectangle assignment is consistent (which it is, since μ and ν are already measures). This uniqueness is essential: there's exactly one product measure consistent with assigning measure μ(A)ν(B) to rectangles.

The payoff comes with Fubini's theorem, which is the immediate downstream application of product measures. Fubini says that integrating a function over X × Y with respect to μ ⊗ ν is the same as iterated integration — first over X, then over Y (or vice versa). In other words, ∬ f d(μ ⊗ ν) = ∫[∫ f dμ] dν. This is the rigorous foundation for switching the order of integration in multivariable calculus, and product measures are the machinery that makes it work in full generality.