We have seen that we can transfer commutativity to strong submonads. Can we transfer other good properties, such as being an affine or relevant monad as well?
Another Commutativity Property
As we are now interested in affine and relevant monads, this post will assume we are working in a category with finite products. We have seen that strong natural transformations commute with double strength. This property will be crucial again today, but we will also need another simple equation.
For a natural transformation
the following equation holds
The proof is a straightforward combination of properties of products and naturality.
Affine Monads
Assume that is a affine monad. That is, it is a commutative monad such that the following equation holds
If
a strong monad morphism, then
If is component-wise a monomorphism, then so is
, and so
Combining this with the results of the previous post, is then an affine monad. In categories with pullbacks, we can strengthen this to
Strong submonads of affine monads are affine.
When the base category is set we can go further, to the slogan
Submonads of affine monads are affine.
Relevant Monads
The argument for relevant monads is even easier. Assume is a relevant monad, so the following equation holds:
If
is a strong monad morphism, then:
and so if is component-wise a monomorphism
and from the results of the previous post, is relevant. For categories with pullbacks, we have the slogan
Strong submonads of relevant monads are relevant.
and in the case of set monads we get the punchier
Submonads of relevant monads are relevant.
Cartesian Monads
As Cartesian monads are simply monads which are both affine and relevant, we can combine the previous two results to deduce that:
Strong submonads of Cartesian monads are Cartesian.
Algebraic Intuitions
For set monads, being affine or relevant requires that certain equations hold between terms. If we think of a submonad as dropping some of the algebraic structure whilst retaining all applicable equations, intuitively we would expect being affine or relevant to still hold.
Summary
Commutative, affine, relevant and Cartesian monads were important when we discussed both well-known and more recent sufficient conditions for the existence of distributive laws. These results allow us to extract new such monads from those we already understand.