nLab
universe in a topos

Context

Topos Theory

topos theory

Background

Toposes

Internal Logic

Topos morphisms

Extra stuff, structure, properties

Cohomology and homotopy

In higher category theory

Theorems

Universes

Foundations

Contents

Idea

A universe in a topos is a topos-theoretic version of the notion of Grothendieck universe; see that page for general motivation and applications.

To free the notion from membership-based set theory, we must replace sets of sets by families of sets, just as in passing from power sets to power objects we must replace sets of subsets by families of subsets.

Definition

A universe in a topos is a morphism el:EU satisfying the axioms to follow. We think of el:EU as a U-indexed family of objects (sets), and we define a morphism a:AI (regarded as an I-indexed family of objects) to be U-small if there exists a morphism f:IU and a pullback square

A E a el I f U.\array{A & \to & E\\ ^a\downarrow && \downarrow^{el}\\ I& \underset{f}{\to} & U.}

Note that f is not, in general, unique: a universe can contain many isomorphic sets. With this definition, the pullback of a U-small morphism is automatically again U-small. We say that an object X is U-small if X1 is U-small.

The axioms which must be satisfied are:

  1. Every monomorphism is U-small.

  2. The composite of U-small morphisms is U-small.

  3. If f:AI and g:BA are U-small, then so is the dependent product Π fg (where Π f is the right adjoint to f *:/I/A).

  4. The subobject classifier Ω is U-small.

Note that since 0Ω is a monomorphism, (1) and (4) imply that the initial object 0 is U-small. A predicative universe is a morphism el:EU where instead of (4) we assume merely that 0 is U-small; this makes sense in any locally cartesian closed category. In a topos, the generic subobject 1Ω is a predicative universe, and of course a morphism is Ω-small if and only if it is a monomorphism.

If we assume only (1)–(3), then the identity morphism 1 0:00 of the initial object would be a universe, for which it itself is the only U-small morphism. On the other hand, if has a natural numbers object N, we may additionally assume that N is U-small, to ensure that all universes contain “infinite” sets.

Note that any object isomorphic to a U-small object is U-small; thus in the language of Grothendieck universes this notion of smallness corresponds to essential smallness. Roughly, we may say that (1) corresponds to transitivity of a Grothendieck universe, (3) and (4) correspond to closure under power sets, and (2) corresponds to closure under indexed unions.

Example: universes in SET

We spell out in detail some implications of these axioms for the case that the topos in question is the Categeory of Sets according to ETCS, to be denoted SET.

Write * for the terminal object in SET, the singleton set. Notice that for each ordinary element uU, i.e. *uU, there is the set E u over u, defined as the pullback

E u E * u U\array{ E_u &\to& E \\ \downarrow && \downarrow \\ * &\stackrel{u}{\to}& U }

We think of E as being the disjoint union over U of the E u. In the language of indexed categories, this is precisely the case: the object ESET is the indexed coproduct of the U-indexed family (EU)SET/U.

  • By the definition of U-smallness and the notation just introduced, an object S in SET, regarded as a *-indexed family S*, is U-small precisely if it is isomorphic to one of the E u.

  • If S is a U-small set by the above and if S 0S is a monomorphism so that S 0 is a subset of S, it follows from 1) and 2) that the comoposite (S 0S*)=(S 0*) is U-small, hence that S 0 is U-small. So: a subset of a U-small set is U-small.

    • In particular, let be the initial object, which is a subset Ω of Ω=2. So: the empty set is U-small.
  • Let S, T and K be objects of SET, regarded as *-indexed families f:S*, T* and K*. Notice that (SETS)(f *K,f *T)(SETS)(K×S p 2 S,T×S p 2 S) is canonically isomorphic to SET(K×S,T). Since Π f is defined to be the right adjoint to f *:SETSETS it follows that Π ff *TT S is the function set of functions from S to T. By 3), if S, T are U-small then so is the function set T S.

    • Since by 4) Ω=2 is U-small and for every S the function set 2 SP(S) is the power set of S, it follows that the power set of a U-small set is U-small.
  • Let I be a U-small set, in that I* is U-small, and let SI be U-small, to be thought of as an I-indexed family of U-small sets S i, where S i is the pullback S i S * i I, so that S is the disjoint union of the S i: S=sqcup iIS i. By axiom 2) the composite morphism (SI*)=(S*) is U-small, hence S is a U-small set, hence the I-indexed union of U-small sets iIS i is U-small.

By standard constructions in set theory from these properties the following further closure properties of the universe U follow.

  • For I a U-small set and SI an I-indexed family of U-small sets S i, the cartesian product iIS i is U-small, as it is a subset of P(I×S).

In summary

  • the sets ,*,2 are U-small;

  • a subset of a U-small set is U-small;

  • the power set P(S) of any U-small set is U-small;

  • the function set T S for any two U-small sets is U-small;

  • the union of a U-small family of U-small sets is U-small.

  • the product of a U-small family of U-small sets is U-small.

Axioms of universes

Just as ZFC and other material set theories may be augmented with axioms guaranteeing the existence of Grothendieck universes, so may ETCS and other structural set theories be augmented with axioms guaranteeing the existence of universes in the above sense. For example, the counterpart of Grothendieck’s axiom

  • For every set s there exists a universe U containing s, i.e. sU

would be

  • For every morphism a:AI in , there exists a universe el:EU such that a is U-small.

Consequences

One can show, from the above axioms, that the U-small morphisms are closed under finite coproducts and under quotient objects. See the reference below.

In terms of indexed categories

Recall that an -indexed category is a pseudofunctor opCat. The fundamental -indexed category is the self-indexing 𝔼 of , which takes I to the slice category 𝔼 I=/I and x:IJ to the base change functor x *.

An internal full subcategory of is a full sub-indexed category 𝔽 of 𝔼 (that is, a collection of full subcategories 𝔽 I𝔼 I closed under reindexing) such that there exists a generic 𝔽-morphism, i.e. a morphism el:EU in 𝔽 U such that for any a:AI in 𝔽 I, we have af *(el) for some f:IU. In this case (since is locally cartesian closed) there exists an internal category U 1U in such that 𝔽 is equivalent, as an indexed category, to the indexed category represented by U 1U.

An internal full subcategory is an internal full subtopos if each 𝔽 is a logical subtopos of 𝔼 (closed under finite limits, exponentials, and containing the subobject classifier). A universe in , as defined above, can then be identified with an internal full subtopos satisfying the additional axiom that U-small morphisms are closed under composition.

In the internal logic

In a topos with a universe, we can talk about small objects in the internal logic by instead talking about elements of U. We can then rephrase the axioms of a universe in the internal logic to look more like the usual axioms for a Grothendieck universe, with the morphism el:EU interpreted as a “family of objects” (S u) u:U:

  1. for all u in U, if X is a subset of S u (in the sense that there exists an injection XS u), then there is a v in U such that XS v;
  2. for all u in U, there is a v in U such that the power set P(S u)S v;
  3. there is a v in U such that S v is empty;
  4. for all u in U and all functions f:S uU, there is a v in U such that the disjoint union i:S uS f(i)S v.

In a well-pointed topos, such as a model of ETCS, these “internal” axioms are equivalent to their “external” versions that refer to global elements of U.

Mike: I haven’t actually checked anything in this section, but it seems probable.

References

  • Thomas Streicher, Universes in Toposes, In: From sets and types to topology and analysis: towards practicable foundations for constructive mathematics (ps,pdf)

Revised on October 12, 2012 04:30:40 by David Roberts (192.43.227.18)