A generalisation of Henstock-Kurzweil integral to compact metric spaces
Abbas Edalat · 2025
Read A generalisation of Henstock-Kurzweil integral to compact metric spaces by Abbas Edalat (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-03793.txt.
Introduction
Abbas Edalat generalizes a flexible integral to curved, compact worlds. His 2025 paper extends the Henstock-Kurzweil integral, a gauge-based method that integrates some functions the Lebesgue integral cannot, from the unit interval to compact metric spaces carrying a Borel measure.
Gauges and tagged partitions are rebuilt from intersections of open and closed sets, and shown to arise naturally from simple approximations sitting below the measure. For unbounded functions he defines a new integral as the limit of a net of these approximations, ordered by refinement. The result obeys the basic rules of an integral, agrees with Lebesgue's whenever that exists, and strictly extends it: on the Cantor set he exhibits functions his integral handles that Lebesgue's cannot.
Catalog record
| CREATOR | Abbas Edalat |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 66 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-03793
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A generalisation of Henstock-Kurzweil integral to compact metric spaces Abbas Edalat Abstract We introduce the notion of a gauge and of a tagged partition (subordinate to a given gauge) by intersections of open and closed sets of a compact metric space extending the corresponding notions in Henstock-Kurzweil integration of real-valued functions with respect to the Lebesgue measure on the unit interval. We show that, for the integration of bounded functions with respect to a normalised Borel measure μ \mu on a compact metric space, the notion of a gauge and an associated tagged partition, arise naturally from a normalised simple valuation way-below the Borel measure. Then we consider the integration of unbounded functions with respect to a normalised Borel measure on a compact metric space, for which the Lebesgue integral may fail to exist. A pair of a tagged partition and a gauge defines a simple valuation and we introduce a partial order on these pairs, emulating the partial order of simple valuations in the probabilistic power domain. We define the D μ D_{\mu} -integral of a real-valued function with respect to a Borel measure using the limit of the net of the integrals of the simple valuations induced by pairs of tagged partitions and gauges for the function. The D μ D_{\mu} -integral of functions on a compact metric space with respect to a normalised Borel measure satisfies the basic properties of an integral and generalises the Henstock-Kurzweil integral. We show that when the Lebesgue integral of the function exists then the D μ D_{\mu} -integral also exists and they have the same value. We provide a family of real-valued functions on the Cantor space that are D μ D_{\mu} -integrable but not Lebesgue integrable. A Preprint Key words: Partition-Gauge Pairs; D_{\mu} -Integral; Lebesgue Integral; Henstock-Kurzweil Integral 1 Introduction In this the paper, we develop a generalisation of Henstock-Kurzweil integral for real-valued functions on compact metric spaces with respect to any bounded Borel measure. We first recall a few historical notes in this matter. In 1902, Henri Lebesgue introduced in his PhD thesis what is now called Lebesgue integration. It was clear from the outset that a simple function such as, x\mapsto\frac{1}{x}\sin\frac{1}{x^{3}}:[0,1]\to{\mathbb{R}}_{\bot}, which is intuitively integrable since it has an improper Riemann integral as x\to 0^{+} , is not Lebesgue integrable as it is not absolutely integrable. To redress this situation, Arnaud Denjoy in 1912 and Oskar Perron in 1914 introduced a generalisation of the Lebesgue theory in which unbounded functions can have integrals without being absolutely integrable. While these two theories turned out to be equivalent, they were too complicated to enter into mainstream mathematics. In the 1950’s Ralph Henstock and Jaroslav Kurzweil, working independently, developed a much simpler integration theory, now called Henstock-Kurzweil integration, which is equivalent to the integration theory of Denjoy and Perron. We now recall the definition of the Henstock-Kurzweil integral of a function [ 2 ] . A gauge on [0,1] is a function \gamma:[0,1]\to{\mathbb{R}}^{+} , which is meant to generalise the norm of a partition of [0,1] to one that depends on each point x\in[0,1] . A tagged partition P of [0,1] is a finite collection (t_{i},I_{i})_{i=1}^{K} where (I_{i})_{i=1}^{K} is a partition of [0,1] by the closed intervals I_{i} for 1\leq i\leq K and t_{k}\in I_{k} for each 1\leq i\leq K . The tagged partition P=(t_{i},I_{i})_{i=1}^{K} is said to be \gamma -fine or subordinate to \gamma if I_{i}\subset(t_{i}-\gamma(t_{i}),t_{i}+\gamma(t_{i})) for all 1\leq i\leq K . Given a function f:[0,1]\to{\mathbb{R}}_{\bot} , the Riemann sum of the tagged partition P for f is given by, S(f,P)=\sum_{i=1}^{K}f(t_{k})\ell(I_{k}), where \ell(I) is the length of the interval I . A function f:[0,1]\to{\mathbb{R}}_{\bot} has Henstock-Kurzweil integral a\in{\mathbb{R}} if for each \epsilon>0 , there exists a gauge \gamma such that |a-S(f,P)|<\epsilon for all \gamma -fine tagged partitions of [0,1] . The Henstock-Kurzweil integral, which can also be defined for real-valued function on {\mathbb{R}}^{n} , is sometimes called the generalised Riemann integral since it follows the definition of the Riemann integral by extending the notion of a the norm of a partition. While there have been attempts to extend the Henstock-Kurzweil integration to more general spaces, these generalisations have lacked the simplicity of the original Henstock-Kurzweil integral for real-valued functions with respect to the Lebesgue measure on [0,1] . We note that Using the notion of a set-valued map on non-degenerate closed intervals, it is shown in [ 8 ] that the Henstock-Kurzweil integral with respect to [0,1] can also be defined using a notion of lower and upper Riemann sums. 2 Partitioning by crescents In this section, we will define a general notion of integral of real-valued functions with respect to a normalised Borel measure \mu on a compact metric space. This so-called D_{\mu} -integral generalises the Lebesgue integral by allowing which only exists if the integrand is absolutely Lebesgue integrable. It thus generalises the Henstock-Kurzweil integral which is only defined for real-valued functions on Euclidean spaces with respect to the Lebesgue measure. The Henstock-Kurzweil integral of a function f:[0,1]\to{\mathbb{R}} , when its exists, can be evaluated over any subinterval [a,b]\subset[0,1] . This situation is like the Riemann integral and is in contrast to the Lebesgue integral of f which can be evaluated with respect to any measurable subset E\subset[0,1] . We will thus seek a family of elementary subsets of X which plays the same role in D_{\mu} -integration as the family of intervals of [0,1] (i.e., open, closed or half-closed intervals) plays in Henstock-Kurzweil integration. We propose that for a general theory of integration on compact metric spaces, the corresponding notion of elementary set is given by a crescent defined as the intersection of an open set and a closed set. Since X itself is both open and closed, it follows that open sets and closed sets are themselves crescents. Next consider the notion of a partition. In the Henstock-Kurzweil theory, integration is performed with respect to the Lebesgue measure on the unit interval in which the boundary of any interval has zero Lebesgue measure. In this case, one works with the notion of a non-overlapping partition of [0,1] , i.e., a finite set of closed intervals, with pairwise disjoint interiors, whose union is [0,1] . For an arbitrary Borel measure \mu , however, the boundary of a crescent, even in the case of intervals of [0,1] , can have non-zero \mu -measure. We therefore define a partition of a crescent C as a disjoint union of C by a finite number of crescents. In order to have an appropriate set of crescents of X as basic sets for a general integration theory on X , we require two properties for our set which hold trivially for intervals of the unit interval: (i) The set of intervals of [0,1] is closed under finite intersections. (ii) For any two intervals I_{1},I_{2}\subset[0,1] , we have the disjoint union I_{1}=(I_{1}\setminus I_{2})\cup(I_{1}\cap I_{2}) . We will require (i) and a generalisation of (ii) for the set of crescents we employ in order to develop a general theory of integration on compact metric spaces. Let {{\cal B}} be a basis of open subsets of X (including the empty set) which is closed under finite (including empty) intersections. The set of crescents generated by