nLab
generalized universal bundle

Context

Homotopy

Higher category theory

higher category theory

Basic concepts

Basic theorems

Applications

Models

Morphisms

Functors

Universal constructions

Extra properties and structure

1-categorical presentations

Contents

Idea

Universal bundles – or mapping cocylinders – are intermediate steps in the computation of homotopy fibers, dual to the that way mapping cone are intermedite steps in the computation of homotopy cofibers.

It is familiar from topology that one can form the path fibration PXX of a topological space. This can be understood as an example of a general construction where one computes homotopy pullbacks of the point – or, if things are not groupoidal, comma objects.

Since universal bundles are examples of this construction, we here speak of generalized universal bundles. Another appropriate term might be generalized path fibrations.

One generalizaton of “generalized universal bundles” is that the objects in question need not be groupoidal, i.e. they behave like directed spaces. In this case the homotopy pullbacks familiar from topology are replaced by comma object constructions. This is useful in various applications. For instance the constructions category of elements and Grothendieck construction can be understood as such directed homotopy pullbacks of the point.

See also

and in particular

Definition

Let C be a closed monoidal category with interval object I. Then for any pointed object ptpt BB in C the generalized universal B-bundle is (if it exists) the morphism

p:E ptBBp : \mathbf{E}_{pt} B \to B

which is the total composite vertical morphism of the pullback diagram

E ptB pt pt B [I,B] d 1 B d 0 B.\array{ \mathbf{E}_{\mathrm{pt}}B &\to& pt \\ \downarrow && \downarrow^{pt_B} \\ [I,B] &\stackrel{d_1}{\to}& B \\ \downarrow^{d_0} \\ B } \,.

So the object E pt:=[I,B]× Bpt is defined to be the pullback of the diagram [I,B]d 1Bpt Bpt and the morphism E ptBB is the composite of the left vertical morphism in the above diagram which comes from the definition of pullback and d 0.

Then a (generalized) ”B-bundle” on some object X is a morphism PX which is the pullback of the generalized universal B-bundle E pt along a “classifying morphism” g:XB

P E pt X g B\array{ P &\to& \mathbf{E}_{pt} \\ \downarrow && \downarrow \\ X &\stackrel{g}{\to}& B }

This can be understood as a “(directed) homotopy pullback” of the point:

If one defines, as one does, a (possiby directed) homotopy between two morphisms f,g:AB to be a morphism η:A[I,B] such that d 0 *η=f and d 1 *η=g, then P is the “lax pullback” (really comma object) of the point along g

P * pt B X g B.\array{ P &\to& * \\ \downarrow &\Downarrow& \downarrow^{pt_B} \\ X &\stackrel{g}{\to}& B } \,.

The fiber of the generalized universal bundle is the loop monoid Ω ptB:

Ω ptB P * pt B * X g B.\array{ \Omega_{pt} B &\to& P &\to& * \\ \downarrow &\Downarrow& \downarrow &\Downarrow& \downarrow^{pt_B} \\ * &\to& X &\stackrel{g}{\to}& B } \,.

the sequence

Ω ptBiE ptBpB\Omega_{pt}B \stackrel{i}{\to} \mathbf{E}_{pt} B \stackrel{p}{\to} B

is exact in that i is the kernel of p in the sense of kernels of morphisms of pointed objects (see there).

Examples

Groupoid incarnations of universal principal bundles

In (higher) categorical contexts, take the interval object to the the interval category I:={ab}. Then

Ordinary G-principal bundles

For C= Cat, B:=BG a one-object groupoid corresponding to a group G with the unique point, E ptBG=EG=G//G is the action groupoid of G acting on itself. The sequence of groupoids is

GG//GBG.G \to G//G \to \mathbf{B}G \,.

This is the universal G-bundle in its groupoid incarnation. It is a theorem by Segal from the 1960s that indeed this maps, under geometric realization to the familiar universal G-bundle in Top. Moreover, it can be seen that every G-principal bundle PX in the ordinary sense is the pullback of EG in the following sense:

the G-bundle PX is classified by a nonabelian G-valued 1-cocycle (the transition function of any of its local trivializations), which is an anafunctor

X^ g BG π X.\array{ \hat X &\stackrel{g}{\to}& \mathbf{B}G \\ \downarrow^\pi \\ X } \,.

(For instance X^ could be the Čech groupoid of a cover of X.)

The universal groupoid bundle EGBG may now be pulled back along this anafunctor to yield the groupoid bundle g *EGX given by the total left vertical morphism in

g *EG EG X^ g BG π X.\array{ g^* \mathbf{E}G &\to& \mathbf{E}G \\ \downarrow && \downarrow \\ \hat X &\stackrel{g}{\to}& \mathbf{B}G \\ \downarrow^\pi \\ X } \,.

This bundle of groupoids is weakly equivalent to the G-principal bundle we started with in that there is a morphism of bundles of groupoids (with P regarded as a bundle of discrete groupoids)

g *EG P X.\array{ g^* \mathbf{E}G &&\stackrel{\simeq}{\to}&& P \\ & \searrow && \swarrow \\ &&X } \,.

In fact that horizontal morphism is an acyclic fibration in the folk model structure, i.e. a k-surjective functor for all k.

This is recalled in the following reference.

G-principal 2-bundles

For C=2Cat, strict 2-categories , B:=BG a strict one-object 2-groupoid corresponding to a strict 2-group G with the unique point, E ptBG=EG was described under the name INN(G) in

  • Urs Schreiber, David Roberts, The inner automorphism 3-group of a strict 2-group, Journal of Homotopy and Related Structures, Vol. 3(2008), No. 1, pp. 193-244 (arXiv)

This was shown to be action bigroupoid of G acting on itself in

  • Igor Bakovič, Bigroupoid 2-torsors PhD thesis, Munich (2008) (pdf).

One can show that every G-principal 2-bundle as described in

  • Toby Bartels, 2-Bundles (arXiv)

  • Christoph Wockel,

  • Igor Bakovič, Bigroupoid 2-torsors PhD thesis, Munich (2008) (pdf).

indeed is recovered as the pullback of EGBG along the corresponding cocycle, along the lines described above.

The way this works is indicated briefly in the last section of Roberts-Schreiber above. A more detailed description for the moment is in the notes

Universal n-category bundles: n-subobject classifiers

One can take B to be something very different from the familiar classifying groupoids. Taking it to be nCat yields the subobject classifiers of higher toposes:

It was David Roberts in the blog comment

who first pointed out that these (higher) subobject classifiers are just generalized universal bundles in the above sense.

These cases for n=0 and n=1 have been considered in the context of universal category bundles in

The discussion there becomes more manifestly one of bundles if one regards all morphisms CSet appearing there as being the right legs of anafunctors.

There is a well-understood version of this for n=(,1), i.e. for (∞,1)-categories. This is described at universal fibration of (∞,1)-categories.

Action groupoids as generalized bundles

A morphism ρ:BF to a pointed object F (needs not be a basepoint preserving morphism!) can be regarded as a representation of B on the point of F. The pullback of the universal F-bundle along this morphism

ρ *E ptFB\rho^* \mathbf{E}_{pt} F \to B

can be addressed as the F-bundle ρ-associated to the universal B-bundle E ptB.

If B is a groupoid, then ρ *E ptF is the action groupoid of B acting on the point of F.

Further pulling this back along a cocycle g:X^B of a B-principal bundle yields the ρ-accociated bundle of that.

For instance for B=BG and F=Vect with ρ:BGVect a representation of the group G on a vector space V, the ρ-associated Vect-bundle on BG is

VV//GBG.V \to V//G \to \mathbf{B}G \,.

Pulling that further back along the cocycle g:X^BG classifying a G-principal bundle PX, one obtains the familiar vector bundle P× GVX which is ρ-associated to P, along the lines described above:

g *ρ *E ptF P× GV X.\array{ g^* \rho^*\mathbf{E}_{pt}F &&\stackrel{\simeq}{\to}&& P\times_G V \\ & \searrow && \swarrow \\ &&X } \,.