Polynomial Functors and Polynomial Monads

We study polynomial functors over locally cartesian closed categories. After setting up the basic theory, we show how polynomial functors assemble into a double category, in fact a framed bicategory. We show that the free monad on a polynomial endofunctor is polynomial. The relationship with operads and other related notions is explored. Introduction Background. Notions of polynomial functor have proved useful in many areas of mathematics, ranging from algebra [41, 34] and topology [10, 50] to mathematical logic [17, 45] and theoretical computer science [24, 2, 20]. The present paper deals with the notion of polynomial functor over locally cartesian closed categories. Before outlining our results, let us briefly motivate this level of abstraction. Among the devices used to organise and manipulate numbers, polynomials are ubiquitous. While formally a polynomial is a sequence of coefficients, it can be viewed also as a function, and the fact that many operations on polynomial functions, including composition, can be performed in terms of the coefficients alone is a crucial feature. The idea of polynomial functor is to lift the machinery of polynomials and polynomial functions to the categorical level. An obvious notion results from letting the category of finite sets take the place of the semiring of natural numbers, and defining polynomial functors to be functors obtained by finite combinations of disjoint union and cartesian product. It is interesting and fruitful to allow infinite sets. One reason is the interplay between inductively defined sets and polynomial functors. For example, the set of natural numbers can be characterised as the least solution to the polynomial equation of sets X ∼= 1 +X , while the set of finite planar trees appears as least solution to the equation X ∼= 1 + ∑