nLab wide pullback (Rev #30, changes)

Showing changes from revision #29 to #30: Added | Removed | Changed

\tableofcontents

Definition

A wide pullback in a category π’ž\mathcal{C} is a product (of arbitrary cardinality) in a slice category π’žβ†“C\mathcal{C} \downarrow C. In terms of π’ž\mathcal{C}, this can be expressed as a limit over a category obtained from a discrete category by adjoining a terminal object.

Yet more explicitly, the wide pullback of a family of morphisms f i:A i→Cf_i\colon A_i \to C (a wide cospan) is an object PP equipped with projection p i:P→A ip_i\colon P\to A_i such that f ip if_i p_i is independent of ii, and which is universal with this property.

Binary wide pullbacks are the same as ordinary pullbacks, a.k.a. fiber products.

Of course, a wide pushout is a wide pullback in the opposite category.

Properties

On the other hand, together with a terminal object, wide pullbacks generate all limits:

Proposition

A category CC with all wide pullbacks and a terminal object 11 is complete. If CC is complete and F:C→DF\colon C \to D preserves wide pullbacks and the terminal object, then it preserves all limits.

Proof

To build up arbitrary products ∏ i∈Ic i\prod_{i \in I} c_i in CC, take the wide pullback of the family c iβ†’1c_i \to 1. Then to build equalizers of diagrams f,g:c⇉df, g\colon c \rightrightarrows d, construct the pullback of the diagram

d ↓δ c β†’βŸ¨f,g⟩ dΓ—d\array{ & & d \\ & & \downarrow \delta \\ c & \underset{\langle f, g \rangle}{\to} & d \times d }

From products and equalizers, we can get arbitrary limits.

In dependent type theory

In dependent type theory, the wide pullback of a family of functions f:∏ i:IA(i)β†’Cf:\prod_{i:I} A(i) \to C is given by the type

Β Wide pullbacks

βˆ‘ y:C∏ i:Iβˆ‘ x:A(i)f(i,x)= Cy\sum_{y:C} \prod_{i:I} \sum_{x:A(i)} f(i, x) =_C y

In dependent type theory, the wide pullback of a family of functions f:∏ i:IA(i)β†’Cf:\prod_{i:I} A(i) \to C is given by the type

βˆ‘ y:C∏ i:Iβˆ‘ x:A(i)f(i,x)= Cy\sum_{y:C} \prod_{i:I} \sum_{x:A(i)} f(i, x) =_C y

Since dependent type theory is the internal logic of locally cartesian closed (infinity,1)-categories, this implies that any locally cartesian closed (infinity,1)-category with a terminal object is complete.

Wide pushouts

Recall that the binary pushout of a family of functions f:∏ x:𝟚Aβ†’C(x)f:\prod_{x:\mathbb{2}} A \to C(x) indexed by the boolean domain 𝟚\mathbb{2} is a higher inductive type ⨆ x:𝟚 A,fC(x)\bigsqcup_{x:\mathbb{2}}^{A, f} C(x) generated by the functions

in:∏ x:𝟚C(x)→⨆ x:𝟚 A,fC(x)\mathrm{in}:\prod_{x:\mathbb{2}} C(x) \to \bigsqcup_{x:\mathbb{2}}^{A, f} C(x)
glue:∏ z:Ain(0,f(0,z))=in(1,f(1,z))\mathrm{glue}:\prod_{z:A} \mathrm{in}(0, f(0, z)) = \mathrm{in}(1, f(1, z))

The glue constructor can be rephrased to the constructor

glue:∏ z:Aβˆ‘ c:⨆ x:𝟚 A,fC(x)∏ x:𝟚in(x,f(x,z))=c\mathrm{glue}:\prod_{z:A} \sum_{c:\bigsqcup_{x:\mathbb{2}}^{A, f} C(x)} \prod_{x:\mathbb{2}} \mathrm{in}(x, f(x, z)) = c

which by the type theoretic axiom of choice is the same as the constructor

glue:βˆ‘ C:A→⨆ x:𝟚 A,fC(x)∏ z:A∏ x:𝟚in(x,f(x,z))=C(z)\mathrm{glue}:\sum_{C:A \to \bigsqcup_{x:\mathbb{2}}^{A, f} C(x)} \prod_{z:A} \prod_{x:\mathbb{2}} \mathrm{in}(x, f(x, z)) = C(z)

or the two constructors

contr:A→⨆ x:𝟚 A,fC(x)\mathrm{contr}:A \to \bigsqcup_{x:\mathbb{2}}^{A, f} C(x)
glue:∏ z:A∏ x:𝟚in(x,f(x,z))=contr(z)\mathrm{glue}:\prod_{z:A} \prod_{x:\mathbb{2}} \mathrm{in}(x, f(x, z)) = \mathrm{contr}(z)

Then generalizing from the boolean domain 𝟚\mathbb{2} to any index type II, the wide pushout of a family of functions f:∏ x:IAβ†’C(x)f:\prod_{x:I} A \to C(x) is a higher inductive type ⨆ x:I A,fC(x)\bigsqcup_{x:I}^{A, f} C(x) generated by the functions

in:∏ x:IC(x)→⨆ x:I A,fC(x)\mathrm{in}:\prod_{x:I} C(x) \to \bigsqcup_{x:I}^{A, f} C(x)
contr:A→⨆ x:I A,fC(x)\mathrm{contr}:A \to \bigsqcup_{x:I}^{A, f} C(x)
glue:∏ z:A∏ x:Iin(x,f(x,z))=contr(z)\mathrm{glue}:\prod_{z:A} \prod_{x:I} \mathrm{in}(x, f(x, z)) = \mathrm{contr}(z)

Wide pushouts are an example of a non-recursive higher inductive type.

Notions of pullback:

References

The terminology wide pullback appears in:

  • Paul Taylor, Quantitative domains, groupoids and linear logic, Category Theory and Computer Science: Manchester, UK, September 5–8, 1989 Proceedings. Springer Berlin Heidelberg, 1989.

Wide pullbacks are considered under the term fibred product in:

  • Robert ParΓ©, Simply connected limits. Can. J. Math., Vol. XLH, No. 4, 1990, pp. 731-746, CMS

Revision on February 10, 2024 at 06:15:36 by John Walsh See the history of this page for a list of all contributions to it.