Absolute integrability is the central safeguard: the integral of the function's absolute value must be finite on the product space. Under this condition, the relevant iterated integrals exist almost everywhere and agree with the multiple integral. This requirement prevents the order of integration from producing ambiguous or incompatible results.
“Almost everywhere” allows an exceptional set of parameter values whose measure is zero. Thus, the theorem does not require every individual slice of the product space to behave perfectly. Instead, it guarantees the needed existence and equality for all parameter values outside that negligible set, which is the natural standard in measure-theoretic integration.
The exchange is justified by absolute integrability, which controls the total size of the function across the product space. That control ensures that integrating first with respect to one variable and then the other yields the same value as reversing the order. The result turns a potentially complicated multidimensional calculation into equivalent successive procedures.
Tonelli’s theorem addresses nonnegative functions and provides related conclusions without requiring the same absolute-integrability formulation used by Fubini’s theorem. This distinction is important because nonnegative integrands avoid cancellation between positive and negative contributions. In practice, Tonelli’s result often supplies the appropriate justification when an integrand is known to be nonnegative.
First, identify the product space and verify the relevant integrability condition, or verify nonnegativity when using Tonelli’s theorem. Next, choose one variable for the inner integral and the remaining variable for the outer integral. If useful, reverse that order and compare the resulting iterated calculation, since the theorem guarantees agreement under its hypotheses.
In probability, the theorem permits multidimensional integrals associated with joint distributions and expectations to be handled as successive integrations. In mathematical physics, it supports calculations over multidimensional domains by reducing them to ordered one-variable integrations. These applications rely on checking the theorem’s hypotheses before simplifying the original multiple integral.