nLab double category of algebras

Idea

For a 2-monad TT, there are several naturally defined 2-categories of TT-algebras, depending on whether we take the morphisms to be lax, colax (oplax), strong/pseudo, or strict. Of these, the lax and colax cases TAlg lT Alg_l and TAlg cT Alg_c are the most general, since a strict morphism is also strong, and a strong morphism can be regarded as either lax or colax (using the structure morphisms or their inverses).

There is no natural 2-category that contains both lax and colax morphisms, since a lax morphism cannot be composed with a colax morphism and obtain any sort of TT-morphism. However, lax and colax TT-morphisms can be combined as the horizontal and vertical morphisms in a double category. This double category allows a characterization of strong morphisms as companion pairs, of doctrinal adjunctions as conjoint pairs, and its 2-cells are relevant for defining oplax/lax comma objects.

Definition

Let TT be a 2-monad on a 2-category đť’¦\mathcal{K}. The objects of the double category TAlgT \mathbf{Alg} are the TT-algebras, the horizontal morphisms are the lax TT-morphisms, and the vertical morphisms are the colax TT-morphisms. The 2-cells are 2-cells in đť’¦\mathcal{K} between composites of underlying morphisms, such that a certain cube of structure 2-cells commutes: consider a square

where horizontal arrows are lax TT-morphism, and vertical ones are colax. A 2-cell α:gf⇒kh\alpha : g f \Rightarrow k h fills the square and it is such that the equality of pasting diagrams

holds. This means that a certain diagram of 2-cells, that can be obtained translating the above equality into a commutative hexagon, is commutative.

For a concrete example in the case of monoidal categories, take lax monoidal functors F:A→BF:A\to B and K:C→DK:C\to D and colax monoidal functors G:B→DG:B\to D and H:A→CH:A\to C, this cube becomes two commutative hexagons for the composition and unit constraints:

G(Fx⊗Fy) → GF(x⊗y) → KH(x⊗y) ↓ ↓ GFx⊗GFy → KHx⊗KHy → K(Hx⊗Hy).\array{ G(F x \otimes F y) & \to & G F (x\otimes y) & \to & K H (x\otimes y) \\ \downarrow &&&& \downarrow \\ G F x \otimes G F y & \to & K H x \otimes K H y & \to & K (H x \otimes H y). }
G(I B) → GF(I A) → KH(I A) ↓ ↓ I D = I D → K(I C)\array{ G(I_B) &\to & G F(I_A) & \to & K H(I_A)\\ \downarrow &&&& \downarrow \\ I_D & = & I_D & \to & K(I_C) }

If 𝒦\mathcal{K} is a strict 2-category, then TAlgT \mathbf{Alg} is a strict double category; otherwise it is “pseudo in both directions” in some sense (such double categories can be defined, but are tricky; see double category). As the objects we can use either strict TT-algebras (if TT is a strict 2-monad) or pseudo algebras (if TT is either a strict or a pseudo 2-monad).

Properties

  • The horizontal 2-category of TAlgT \mathbf{Alg} is the 2-category TAlg lT Alg_l of lax TT-morphisms, and its vertical 2-category is TAlg cT Alg_c.

  • There is a forgetful double functor from TAlgT \mathbf{Alg} to the double category Q(đť’¦)\mathbf{Q}(\mathcal{K}) of quintets in đť’¦\mathcal{K}. In fact, TAlgT \mathbf{Alg} can be enhanced to a triple category whose “transversal morphisms” are strict TT-morphisms (or pseudo ones, if TT is only pseudo, in which case we have to deal with “triply pseudo triple categories”), and the induced forgetful functor from this triple category to the triple category of “quintets and commutative cubes” in đť’¦\mathcal{K} is monadic in an appropriate sense.

  • A companion pair in TAlgT \mathbf{Alg} consists of two isomorphic morphisms in đť’¦\mathcal{K} between a pair of TT-algebras along with a pseudo TT-morphism structure on one (hence both). The category of companion pairs between two TT-algebras is thus equivalent (though not isomorphic) to the category of pseudo TT-morphisms.

  • A conjoint pair in TAlgT \mathbf{Alg} is precisely a “colax/lax” or doctrinal adjunction, and the theorems about doctrinal adjunction can naturally be expressed as lifting properties of the forgetful functor from TAlgT \mathbf{Alg} to Q(đť’¦)\mathbf{Q}(\mathcal{K}).

  • The universal 2-cell of an oplax/lax comma object has the structure of a square in TAlgT \mathbf{Alg}. Its universal property ought to be related to TAlgT \mathbf{Alg} as well, but this is still unclear.

References

Last revised on August 7, 2024 at 10:01:00. See the history of this page for a list of all contributions to it.