In classical arithmetic, divisibility is a binary state: a number either divides another, or it does not. However, across fields as diverse as probability theory, algebraic geometry, and description logic, researchers are finding that "divisibility"—the ability to decompose an object into parts—is better understood as a structured spectrum. The unifying mathematical tool for this spectrum is the semigroup, an algebraic structure that captures how parts combine to form wholes.
The Spectrum of Divisibility
In probability theory, we often distinguish between "indivisible" distributions and "infinitely divisible" ones (like the Gaussian distribution, which can be decomposed into an arbitrary number of independent components). As S. Albeverio and Jiang-Lun Wu argue in Partly Divisible Probability Distributions, this binary classification misses a vast middle ground.
They propose that probability distributions exist on a continuum. A distribution might be "partly divisible," meaning it can be split into a specific set of independent, identically distributed components, but not just any number of them. The key insight is that the set of valid "division factors" for a given distribution isn't random; it forms a closed semigroup. This means if you can divide a distribution into $n$ parts and also into $m$ parts, the algebraic structure of the semigroup dictates the other possible ways it can be decomposed.
Approximate Divisors and Numerical Semigroups
This idea of structured decomposition extends to the very definition of a "multiple." In The family of $a$-floor quotient partial orders, Harry Richman and J. C. Lagarias 0001 introduce "approximate divisor" relations. By using a scaling parameter within a floor function, they bridge the gap between standard divisibility and simple addition.
When we look at the set of "multiples" generated by these approximate relations, they form a numerical semigroup. This is a subset of non-negative integers closed under addition. The power of this framing is that it allows researchers to use combinatorial tools like the Frobenius number—the largest integer that cannot be expressed as a sum of elements in the semigroup. This number effectively quantifies the "gaps" in the divisibility structure, showing exactly where the "approximate multiples" become a continuous stream of integers.
Achieving Closure in Logic and Geometry
The realization that "achievable values" form semigroups is also transforming how we query complex data. In description logic, when we ask for the number of individuals satisfying a certain property, there isn't always a single answer. Instead, there is a "spectrum" of possible counts across different valid models.
Quentin Manière and Marcin Przybylko demonstrate in Spectra of Cardinality Queries over Description Logic Knowledge Bases that for certain types of queries, this spectrum is a subsemigroup under addition. The intuition is elegant: if a knowledge base allows for a model with $A$ individuals and another model with $B$ individuals, the underlying logic often allows these models to be "added" or combined to produce a model with $A+B$ individuals.
A similar logic appears in algebraic geometry. Amey Kulkarni shows in On the semigroup of an ordinary multiple point that the geometric properties of functions near a curve's singularity can be mapped to a "branch semigroup." Here, the "order of vanishing" along different branches of a curve behaves like a vector of integers that must satisfy specific additive conditions to be valid.
Why the Semigroup Perspective Works
The reason this perspective is so productive across different fields is that it moves the focus from individual values to compositional rules.
Whether we are looking at the infinite divisibility of Generalized Gaussian distributions—as explored by Alex R. Dytso, Ronit Bustin, and Shlomo Shamai in Analytical properties of generalized Gaussian distributions—or the possible answers to a database query, the core question is the same: If I have these parts, what wholes can I build?
By identifying these sets as semigroups, researchers gain access to a rigorous language for describing "gaps" (Frobenius numbers), "limits" (infinite divisibility), and "structure" (closed subsemigroups), turning what looks like a scattered set of values into a predictable mathematical landscape.