← Back to arXiv
arXivLogicarXiv:2607.25460

Initial algebras from constructive ordinals

The mathematics of infinite structures often requires reasoning about processes that go on for more than finitely many steps, which involves so-called transfinite recursion. Classical mathematics handles this using ordinal numbers, but classical proofs often rely on the law of excluded middle, meaning they work by assuming something is false and deriving a contradiction. Constructive mathematics takes a stricter approach, requiring that proofs actually exhibit or build the objects they claim exist, rather than just ruling out their non-existence. This paper works out how a well-known constructive notion of ordinal, one that has been studied for decades, can support a rich theory of transfinite recursion in this stricter setting.

The main payoff is constructive versions of several important theorems in category theory, a branch of mathematics concerned with abstract structural relationships. One key result is a constructive proof of Adamek's theorem, which is a general tool for building so-called initial algebras. An initial algebra is essentially the most basic or canonical solution to a recursive structural equation, and initial algebras are foundational in computer science and logic because they capture things like the natural numbers, lists, and trees in a very general way. Getting a constructive version of this theorem means you can actually compute or extract the objects being described.

The other main application is a constructive version of Quillen's small object argument, a central technique in homotopy theory and abstract algebra. This argument is used to build special factorization systems, which allow you to decompose maps between mathematical objects in a controlled way. Factorization systems of this kind, called algebraic weak factorization systems, are important in modern approaches to foundations of mathematics, including homotopy type theory. By making these constructions fully constructive, the paper opens the door to implementing them in proof assistants and using them in settings where computational content matters.

Read original →