Absoluteness of the Riemann integral
Carlos M. Parra-Londoño, Andrés F. Uribe-Zapata
TL;DR
The article develops a generalized integration framework on Boolean algebras with finitely additive measures and proves its absoluteness across transitive models of $\mathrm{ZFC}$. It then specializes to the Riemann integral on rectangles in $\mathbb{R}^{n}$ by relating it to $\lambda^{n}|_{[a,b]}$, establishing that a bounded function $f$ defined on $[a,b]$ in an inner model $M$ has an absolute extension $g$ in any outer model $N$ with equal integral values, provided such an extension exists. The main contribution is the absoluteness of the Riemann integral between models, including a uniqueness statement up to Lebesgue measure zero, and the construction of an integral on Boolean algebras that underpins this result. The work also clarifies the limitations of absoluteness for the Lebesgue integral due to the non-absoluteness of the real numbers, emphasizing the interplay between model theory, forcing techniques, and classical analysis.
Abstract
This article explores the concept of absoluteness in the context of mathematical analysis, focusing specifically on the Riemann integral on $\mathbb{R}^{n}$. In mathematical logic, "absoluteness" refers to the invariance of the truth value of certain statements in different mathematical universes. Leveraging this idea, we investigate the conditions under which the Riemann integral on $\mathbb{R}^{n}$ remains absolute between transitive models of ZFC, the standard axiomatic system in which current mathematics is usually formalized. To this end, we develop a framework for integration on Boolean algebras with respect to finitely additive measures and show that the classical Riemann integral is a particular case of this generalized approach. Our main result establishes that the Riemann integral over rectangles in $\mathbb{R}^{n}$ is absolute in the following sense: if $M \subseteq N$ are transitive models of ZFC, $a, b \in \mathbb{R}^{n} \cap M$, and $f \colon [a, b] \to \mathbb{R}$ is a bounded function in $M$, then $f$ is Riemann integrable in $M$ if, and only if, in $N$ there exists some Riemann integrable function $g \colon [a, b] \to \mathbb{R}$ extending $f$. In this case, the values of the integrals computed in each model are the same. Furthermore, the function $g$ is unique except for a measure zero set.
