nLab
Sweedler coring

Context

Cohomology

cohomology

Special and general types

Special notions

Variants

Extra structure

Operations

Theorems

Algebra

(,1)-Topos Theory

(∞,1)-topos theory

Background

Definitions

Characterization

Morphisms

Extra stuff, structure and property

Models

Constructions

structures in a cohesive (∞,1)-topos

Contents

Idea

A Sweedler coring is an algebraic structure that is roughly the formal dual of the Čech nerve of a cover: it is used to describe descent in algebraic contexts.

See also monadic descent.

Definition

Let f:RS be the extension of associative unital k-algebras (where k is a commutative unital ring).

The corresponding canonical coring or Sweedler coring is the S-coring

C=S RSC = S\otimes_R S

with coproduct

Δ:CC SCS RS RS\Delta : C\to C\otimes_S C \cong S\otimes_R S\otimes_R S

given by

Δ:s 1s 2s 11s 2\Delta: s_1\otimes s_2 \mapsto s_1\otimes 1 \otimes s_2

and counit

ϵ:CS\epsilon : C\to S

given by

ϵ:s 1s 2s 1s 2.\epsilon: s_1 \otimes s_2 \mapsto s_1 s_2 \,.

The element 11 is a grouplike element in the Sweedler’s coring.

Geometric interpretation

We give a dual geometric interpretation of the Sweedler coring.

Suppose a context of spaces and function algebras on spaces that satisfies the basic axioms of geometric function theory, in that the algebra of functions C(Y 1× XY 2) on a fiber product

Y 1× XY 2 Y 2 Y 1 X\array{ Y_1 \times_X Y_2 &\to& Y_2 \\ \downarrow && \downarrow \\ Y_1 &\to& X }

is the tensor product of the functions on the factors:

C(Y 1× XY 2)=C(Y 1) C(X)C(Y 2).C(Y_1 \times_X Y_2) = C(Y_1) \otimes_{C(X)} C(Y_2) \,.

Then let π:YX be a morphism of spaces and set

R:=C(X)R := C(X)

and

S=C(Y)S = C(Y)

and

(RS):=π *:C(X)C(Y).(R \hookrightarrow S) := \pi^* : C(X) \to C(Y) \,.

The morphism π induces its augmented Čech nerve

(Y× XY× XYY× XYYπX).\left( \cdots \stackrel{\to}{\stackrel{\to}{\stackrel{\to}{\to}}} Y \times_X Y \times_X Y \stackrel{\to}{\stackrel{\to}{\to}} Y \times_X Y \stackrel{\to}{\to} Y \stackrel{\pi}{\to} X \right) \,.

Taking function algebras of this yields, by the above,

(S RS RSS RSSπ *R).\left( \cdots \stackrel{\leftarrow}{\stackrel{\leftarrow}{\stackrel{\leftarrow}{\leftarrow}}} S \otimes_R S \otimes_R S \stackrel{\leftarrow}{\stackrel{\leftarrow}{\leftarrow}} S \otimes_R S \stackrel{\leftarrow}{\leftarrow} S \stackrel{\pi^*}{\leftarrow} R \right) \,.

Writing again C=S RS for the Sweedler coring, this is

(C SCCSπ *R).\left( \cdots \stackrel{\leftarrow}{\stackrel{\leftarrow}{\stackrel{\leftarrow}{\leftarrow}}} C \otimes_S C \stackrel{\leftarrow}{\stackrel{\leftarrow}{\leftarrow}} C \stackrel{\leftarrow}{\leftarrow} S \stackrel{\pi^*}{\leftarrow} R \right) \,.

Descent in terms of coring comodules

Given a morphism f:RS with corresponding Sweedler coring (C=S RS,Δ,ϵ) as above, the category of descent data Desc(S/R) for the categories of right modules along k-algebra extension RS is precisely the category of right C-comodules.

In other words, the objects of Desc(S/R) are the pairs (N,α) where N is a right S-module, and α:NN RS is a right S-module morphism and if we write α(m)= im is i then

  • iα(m i)s i= im i1s i,

  • im is i=m.

In terms of (co)monadic descent

This coring-formulation of descent may be understood as special case of comonadic descent (see also the discussion at Bénabou–Roubaud theorem). We spell this out in detail:

The bifibration in question is

p: Mod CRing

that sends an object in the category Mod of modules to the ring that it is a module over.

A descent datum for a morphism f:RS with respect to this bifibration is a (co)algebra object over the comonad f *f * induced by this morphism. We have that

  • the morphism f * sends an R-module N to the S-module N RS;

  • the morphism f * sends an S-module P to the R-module P SS R, where S R is S regarded as a left S- and a right R-module. So P SS R is just the S-module P with only the right R-action remembered.

Accordingly, the comonad with underlying functor f *f * sends an S-module P to the S-module P SS RS=P SC.

A (co)algebra object for this comad is hence a co-action morphism

PP SCP \to P \otimes_S C

compatible with the monad action. This is precisely a comodule over the Sweedler coring, as defined above.

Geometric interpretation

Descent for Sweedler corings is a special case of comonadic descent. We describe this in detail and relate it by duality to the geometrically more intuitive monadic descent for codomain fibrations.

Assuming again a suitable geometric context as above, we may identify a module over R=C(X) with (the collection of sections of) a vector bundle (or rather a suitable generalization of that: a coherent sheaf) over X. Similarly for Y. So we write

Vec(X):=RModVec(X) := R Mod

and

Vect(Y):=SModVect(Y) := S Mod

for the corresponding categories of modules. The assignment of such categories to spaces

Vec:ZVec(Z)Vec : Z \mapsto Vec(Z)

extends to a contravariant pseudofunctor

Vec:Spaces opCatVec : Spaces^{op} \to Cat

by assigning to a morphism f:YX of spaces the corresponding functor

Vec(X)C(X)Mod fC(Y)C(Y)ModVec(Y).Vec(X) \simeq C(X) Mod \stackrel{- \otimes_{f} C(Y)}{\to} C(Y) Mod \simeq Vec(Y) \,.

This way Vec becomes a prestack of categories on our category of spaces.

If this prestack satisfies descent along suitable covers, it is a stack.

Geometrically this is the case if for each morphism π:YX that is regarded as a cover, the category Desc(Y,Vec) whose objects are tuples consisting of

  • an object aVec(Y)

  • an isomorphism g:π 1 *aπ 2 *a

  • such that

    π 2 *a π 12 *g π 23 *g π 1 *a π 13 * π 3 *\array{ && \pi_2^* a \\ & {}^{\mathllap{\pi_{12}^* g}}\nearrow && \searrow^{\mathrlap{\pi_{23}^* g}} \\ \pi_1^* a &&\stackrel{\pi_{13}^*}{\to}&& \pi_3^* }

    commutes.

Morphism are defined similarly (see stack and descent for details).

To get the geometric pucture that underlies, by duality, the above comodule definition of descent, we need to reformulate this just a little bit more:

every ordinary vector bundle EX (of finite rank) is the associated bundle EP× O(n)V of an O(n)-principal bundle PX, and as such its sections may be identified with O(n)-equivariant functions PV n on the total space of P.

Using this we may think of the C(X)-module of sections of E as a submodule of the C(X)-module of all functions on P

Γ(E)C(P).\Gamma(E) \subset C(P) \,.

We now reformulate the geometric descent for vector bundles in terms of geometric descent for their underlying principal bundles, and then take functions on everything in sight to obtain the comodule definition of descent that we want to describe:

A descent datum (transition function) for a principal bundle QY may be thought of as the the morphism g in the double pullback diagram

Y× XY× YQ Q g Y× XY Y π Q Y π X.\array{ &&Y\times_X Y \times_Y Q &\to& Q \\ &&\downarrow && \downarrow \\ &{}^{\mathllap{g}}\swarrow&Y \times_X Y &\to& Y \\ &&\downarrow && \downarrow^{\mathrlap{\pi}} \\ Q &\to& Y &\stackrel{\pi}{\to}& X } \,.

Because here Y× XY× YQ is the space whose points consist of a point in a double overlap of the cover and a point in the fiber of Q over that with respect to one patch, and the morphism identifies this with a point in the fiber of Q regarded as sitting over the other patch. Analogously, there is a cocycle condition on g on triple overlaps.

Now, blindly applying our functor that takes functions of spaces to the above diagram yields the double pushout diagram

C(Y× XY× YQ) C(Q) g * C(Y× XY) C(Y) π * C(Q) C(Y) π * C(X).\array{ &&C(Y\times_X Y \times_Y Q) &\leftarrow& C(Q) \\ &&\uparrow && \uparrow \\ &{}^{\mathllap{g}^*}\nearrow&C(Y \times_X Y) &\leftarrow& C(Y) \\ &&\uparrow && \uparrow^{\mathrlap{\pi^*}} \\ C(Q) &\leftarrow& C(Y) &\stackrel{\pi^*}{\leftarrow}& C(X) } \,.

We may restrict to N:=Γ(E)C(Q) as just discussed and switch to the notation from above to get

N SC N g * C S π * N S π * R.\array{ &&N \otimes_{S} C &\leftarrow& N \\ &&\uparrow && \uparrow \\ &{}^{\mathllap{g}^*}\nearrow& C &\leftarrow& S \\ &&\uparrow && \uparrow^{\mathrlap{\pi^*}} \\ N &\leftarrow& S &\stackrel{\pi^*}{\leftarrow}& R } \,.

The morphism

α:=g *:NN SC\alpha := g^* : N \to N \otimes_S C

obtained this way is the co-action morphism from the above algebraic definition.

The further cocycle condition on g similarly translates into the condition that α really satisfies the comodule property.

Various properties of canonical coring correspond to adequate properties of the ring extension. For example, coseparable Sweedler corings correspond to split extension?s (the k-algebra extension RS is split if there is an R-bimodule map h:SR with h(1 S)=1 R).

References

Sweedler corings are named after Moss Sweedler.

Revised on September 11, 2012 13:13:07 by Urs Schreiber (131.174.190.72)