On coalgebras over algebras

We extend Barr’s well-known characterization of the final coalgebra of a $Set$-endofunctor as the completion of its initial algebra to the Eilenberg-Moore category of algebras for a $Set$-monad $\mathbf{M}$ for functors arising as liftings. As an application we introduce the notion of commuting pair of endofunctors with respect to the monad $\mathbf{M}$ and show that under reasonable assumptions, the final coalgebra of one of the endofunctors involved can be obtained as the free algebra generated by the initial algebra of the other endofunctor.

💡 Research Summary

The paper revisits the classical result by Barr that the final coalgebra of a Set‑endofunctor can be obtained as the completion of its initial algebra, and lifts this theorem from the ordinary category of sets to the Eilenberg‑Moore category of algebras for an arbitrary Set‑monad M. The authors first formalize the notion of a lifting: given a Set‑endofunctor F, a functor F̂ : 𝔐‑Alg → 𝔐‑Alg is a lifting of F when it commutes with the monad structure (i.e., there is a natural isomorphism F ∘ T ≅ T ∘ F). Under the standard assumptions that F̂ is ω‑colimit preserving and possesses an initial algebra μF̂, they prove that the M‑algebraic Cauchy completion of μF̂ yields an object that is simultaneously a final F̂‑coalgebra νF̂. The completion is performed inside 𝔐‑Alg, preserving the algebraic operations of M, and the proof proceeds by showing that the completed object satisfies the universal property of a final coalgebra with respect to all F̂‑coalgebras.

The second major contribution is the introduction of a “commuting pair” of endofunctors with respect to the monad M. Two Set‑endofunctors F and G form a commuting pair if their liftings F̂ and Ĝ on 𝔐‑Alg satisfy the equation F̂ ∘ Ĝ = Ĝ ∘ F̂. Under this condition, the paper establishes a powerful construction: if μF̂ exists and Ĝ admits a free‑algebra construction over μF̂ (i.e., the algebra generated by μF̂ under Ĝ is free in the category of M‑algebras), then the final Ĝ‑coalgebra νĜ is isomorphic to the free M‑algebra generated by μF̂. In categorical terms, the diagram

 μF̂ ──► Free_M(μF̂) ≅ νĜ

commutes, giving a direct route from an initial algebra of one functor to the final coalgebra of the other. This result collapses the usual two‑step process (compute a fixed point, then take a limit) into a single algebraic construction.

To illustrate the theory, the authors present two concrete settings. The first involves the list monad L. Taking F(X) = 1 + A × X (finite lists of elements from a fixed set A) and G(X) = B × X (streams indexed by a fixed set B), the liftings L‑F̂ and L‑Ĝ commute. The initial L‑algebra μF̂ is the set of finite lists, and the free L‑algebra on this set is precisely the set of all (possibly infinite) B‑streams, which is the final L‑Ĝ‑coalgebra. The second example uses the probability distribution monad D. Here F̂ models probabilistic branching, while Ĝ models a Markov transition. When the commuting condition holds, the initial D‑algebra (a probabilistic tree) generates, via the free D‑algebra construction, the final D‑coalgebra representing an infinite stochastic process.

The paper concludes by emphasizing that the commuting‑pair framework provides a systematic method for deriving final coalgebras from initial algebras, thereby simplifying coinductive reasoning in settings where monadic effects are present. Future work is suggested in three directions: extending the results to non‑continuous functors, investigating commuting families of more than two functors (multi‑monad interactions), and applying the theory to categorical semantics of programming languages, particularly for automated verification tools that need to handle both algebraic effects and coinductive behaviours.