A PRO, an abbreviation for “product category”, is similar to a PROP but more general, being merely a monoidal category but not necessarily symmetric monoidal. They may be used to describe theories for finitary algebraic and coalgebraic structures which make sense in any monoidal category, for example monoids and comonoids.


A PRO is a strict monoidal category TT in which every object is of the form

x n=xxntimes(productofncopiesofx)x^{\otimes n} = \underset{n \: times}{\underbrace{x \otimes \ldots \otimes x}} \qquad (product \: of \: n \: copies \: of \: x)

for some n0n \geq 0. Somewhat more precisely, a PRO is a strict monoidal category for which the unique strict monoidal functor

(,+,0)(T,,I)(\mathbb{N}, +, 0) \to (T, \otimes, I)

is an isomorphism on objects.

Adam: Is there a variant where instead of a single object xx we start with some set of objects (or the objects of some other category) and require that any object of the new category is isomorphic to the tensor of zero or more objects from that set? I suppose this would be a “multi-sorted PRO” or something like that, although you’d have to specify the base set. Maybe “PRO generated by XYZ” or “PRO over XYZ” or some such thing. If there’s an established term for this, or something close to an established term, I’m keen on hearing about it.

RD : The term “coloured PROP” has been used. See arXiv:1207.2773

Morphisms x mx nx^{\otimes m} \to x^{\otimes n} in a PRO are often thought of as operations which accept mm inputs and produce nn outputs, hence PROs are like nonpermutative operads but for the multiple outputs, which make them more general.

Let CC be a monoidal category. A TT-model in CC (or a CC-representation of TT) is a strong monoidal functor F:TCF: T \to C; the underlying object of the TT-model is the value F(x)F(x). A TT-model is thus an object cc of CC equipped with operations

θ c:c mc n,\theta_c: c^{\otimes m} \to c^{\otimes n},

one for each morphism x mx nx^{\otimes m} \to x^{\otimes n}, reflecting the monoidal category structure of TT. A homomorphism FGF \to G of TT-models F,G:TCF, G: T \to C is a monoidal transformation between FF and GG; it may be equivalently expressed as a morphism ϕ:F(x)G(x)\phi: F(x) \to G(x) in CC which respects the modelings of the TT-operations θ\theta.


The simplex category Δ\Delta is a PRO whose models are monoids in monoidal categories. Similarly, Δ op\Delta^{op} is a PRO whose models are comonoids.

The cube category is a PRO whose models are objects XX equipped with “elements” x 0,x 1:IXx_0, x_1: I \to X and a projection p:XIp: X \to I which is a retraction of both x 0x_0 and x 1x_1, where II denotes the monoidal unit.

The monoidal category of planar thickened 1d tangles (is this the right vocabulary?) is a PRO whose models are noncommutative Frobenius objects in monoidal categories.


Just as the notion of PROP is a symmetric monoidal analogue of PRO, there is a braided monoidal analogue called a PROB. The cartesian monoidal analogue is known as a Lawvere theory.

More generally, given any doctrine DD, let I(D)I(D) be an initial algebra of the doctrine. A PRO-D can then be defined as an algebra TT such that the algebra map I(D)TI(D) \to T induces an isomorphism on objects. This is clearly an evil notion, and yet a useful and general one in practice.

Revised on October 16, 2016 10:18:35 by Anonymous Coward (