---
title: "Polynomial functors and polynomial monads"
authors: ["Nicola Gambino", "Joachim Kock"]
year: 2013
venue: "Math. Proc. Cambridge Philos. Soc."
arxiv: "0906.4931"
doi: "10.1017/S0305004112000394"
source: "https://arxiv.org/abs/0906.4931"
topics: [curiosity]
seed_rank: 1126
seed_batch: "curiosity-2026-08-21"
reviewed: "2026-08-21"
pool: "maths-foundations"
relevance_score: 9
lineage: polynomial-functors
cites:
  - title: "Framed bicategories and monoidal fibrations"
    url: "https://arxiv.org/abs/0706.1286"
    year: 2008
    arxiv: "0706.1286"
    doi: null
  - title: "Categories of containers"
    url: "https://doi.org/10.1007/3-540-36576-1_3"
    year: 2003
    arxiv: null
    doi: "10.1007/3-540-36576-1_3"
  - title: "Locally cartesian closed categories and type theory"
    url: "https://doi.org/10.1017/S0305004100054335"
    year: 1984
    arxiv: null
    doi: "10.1017/S0305004100054335"
  - title: "Functorial Semantics of Algebraic Theories"
    url: "https://doi.org/10.1073/pnas.50.5.869"
    year: 1963
    arxiv: null
    doi: "10.1073/pnas.50.5.869"
see:
  - "154-locally-cartesian-closed-categories-and-type-theory"
  - "048-functorial-semantics-of-algebraic-theories"
  - "029-categories-for-the-working-mathematician-2nd-ed"
  - "1057-profunctor-optics-modular-data-accessors"
---

# Polynomial functors and polynomial monads

## One-sentence takeaway

A polynomial functor is the composite \(\Pi\)-\(\Sigma\)-\(\Delta\) of a span \(I\leftarrow B\to A\to J\) in a locally cartesian closed category; they form a framed bicategory, and the free monad on a polynomial endofunctor is again polynomial.

## Why it matters here

Why you might love this: the same diagram that in type theory is a dependent sum/product (a container, a species, an operad, a W-type) is here a 1-cell in a framed bicategory, so base change is a fibration not a coherence slog. The free-monad theorem says "the inductive type generated by a polynomial signature is still a polynomial" — trees with typed arities are the free algebra, internally. Bridge: Anoptic's ECS is a polynomial functor (entities = positions, components = directions); gameplay rules that generate trees of events are free polynomial monads; optics/lenses (1057) sit in the same framed-bicategory neighbourhood.

## Key ideas

- In an LCCC, a polynomial from \(I\) to \(J\) is a diagram \(I \xleftarrow{s} B \xrightarrow{p} A \xrightarrow{t} J\); the associated functor is \(t_! \circ p_* \circ s^*\).
- Polynomial functors assemble into a double category / framed bicategory (Shulman): horizontal 1-cells are polynomials, vertical 1-cells are base change, framing handles reindexing.
- Theorem: the free monad on a polynomial endofunctor is polynomial. Operads, multicategories, and containers (Abbott–Altenkirch–Ghani) fall out as slices.
- Two biequivalent presentations: a strict framed 2-category of functors, and a bicategory of representing diagrams.

## Caveats

- General LCCCs, not a Set-only combinatorics paper; finitary/Set specialisations are cited (Kock, Abbott).
- Not a cubical/QTT/optics remint: those already sit at 1054/1055/1057. This is the algebra that makes containers and W-types the same picture.
- Strength / monoidal issues need extra structure; the paper isolates what is purely polynomial.

## Links

- arXiv: [0906.4931](https://arxiv.org/abs/0906.4931)
- Journal: [Math. Proc. Cambridge Philos. Soc. 154 (2013) 153–192](https://doi.org/10.1017/S0305004112000394)