Skip to the Main Content

Note:These pages make extensive use of the latest XHTML and CSS Standards. They ought to look great in any standards-compliant modern browser. Unfortunately, they will probably look horrible in older browsers, like Netscape 4.x and IE 4.x. Moreover, many posts use MathML, which is, currently only supported in Mozilla. My best suggestion (and you will thank me when surfing an ever-increasing number of sites on the web which have been crafted to use the new standards) is to upgrade to the latest version of your browser. If that's not possible, consider moving to the Standards-compliant and open-source Mozilla browser.

June 26, 2008

Category Algebras

Posted by Urs Schreiber

Currently Masoud Khalkali over from Noncommutative Geometry Blog is giving an introductory lecture series on – right – noncommutative geometry, here at the Hausdorff institute.

He is following his notes

Masoud Khalkali
Very basic noncommutative geometry
arXiv:math/0408416 .

Today he started talking about noncommutative algebras arising as groupoid algebras (or “groupoid convolution algebras”, special cases of category algebras, see p. 58) motivating them by their ubiquitousness in noncommutative geometry:

The good news is that most of the noncommutative spaces which are currently in use in noncommutative geometry are constructed by this method.
p. 53

Here I want to use this opportunity to point out how category algebra works from the point of view of groupoidification along the lines of the discussion in An Exercise in Groupoidification: The Path Integral.

For VV a (finite dimensional) vector space, I write FinSet VFinSet_V for the category of sets over VV. An object is a set SS with a map to VV, SV S\to V and a morphism is a commuting triangle S S V. \array{ S &\to& S' \\ & \searrow \swarrow \\ & V } \,. The VV-cardinality operation is a monoidal map ||:(FinSet V,)(V,+) |\cdot| : (FinSet_V, \oplus) \to (V, +) sending a finite set labeled by VV to the sum of the labels of its elements.

Let BEnd(V)\mathbf{B} End(V) be the category whose single object is the vector space VV and whose space of morphisms is End(V)End(V), with composition of morphisms being the composition of these endomorphisms.

Take CC to be any finite category whose category algebra we want to consider. Form the cartesian product category C×BEnd(V) C \times \mathbf{B} End(V) and consider finite sets over the collection of morphisms: S Mor(C×BEnd(V)). \array{ S \\ \downarrow \\ Mor(C \times \mathbf{B} End(V)) } \,. Notice that for each fixed morphisms of CC this is an element in FinSet End(V)FinSet_{End(V)}. Under the End(V)End(V)-cardinality it is therefore an End(V)End(V)-valued function on Mor(C)Mor(C). If V=V = \mathbb{C} (and assuming we are talking about complex vector spaces), then this is a complex function on Mor(C)Mor(C).

Given two such sets over Mor(C×BEnd(V))Mor(C \times \mathbf{B}End(V)) S S Mor(C×BEnd(V)) Mor(C×BEnd(V)) \array{ S &&&& S' \\ \downarrow &&&& \downarrow \\ Mor(C \times \mathbf{B} End(V)) &&&& Mor(C \times \mathbf{B} End(V)) } representing two End(V)End(V)-valued functions on Mor(C)Mor(C) we take the correspondence space given by the composition operation in C×BEnd(V)C \times \mathbf{B}End(V)

S Mor(C×BEnd(V)) t× sMor(C×BEnd(V)) S d 0 d 1 Mor(C×BEnd(V)) Mor(C×BEnd(V)) Mor(C×BEnd(V)), \array{ S && Mor(C \times \mathbf{B} End(V)) {}_t \times_s Mor(C \times \mathbf{B} End(V)) && S' \\ \downarrow & {}^{d_0}\swarrow &\downarrow^\circ& \searrow^{d_1} & \downarrow \\ Mor(C \times \mathbf{B} End(V)) && Mor(C \times \mathbf{B} End(V)) && Mor(C \times \mathbf{B} End(V)) } \,, pull back both sets SS and SS' to the total space and take fiberwise cartesian product there to get the new set d 0 *Sd 1 *Sd_0^* S \otimes d_1^* S' d 0 *Sd 1 *S Mor(C×BEnd(V)) t× sMor(C×BEnd(V)) d 0 d 1 Mor(C×BEnd(V)) Mor(C×BEnd(V)) Mor(C×BEnd(V)) \array{ && d_0^* S \otimes d_1^* S' \\ && \downarrow \\ && Mor(C \times \mathbf{B} End(V)) {}_t \times_s Mor(C \times \mathbf{B} End(V)) && \\ & {}^{d_0}\swarrow &\downarrow^\circ& \searrow^{d_1} & \\ Mor(C \times \mathbf{B} End(V)) && Mor(C \times \mathbf{B} End(V)) && Mor(C \times \mathbf{B} End(V)) } and then push that down along the composition map \circ Mor(C×BEnd(V)) t× sMor(C×BEnd(V)) d 0 d 1 Mor(C×BEnd(V)) Mor(C×BEnd(V)) Mor(C×BEnd(V)) d 0 *Sd 1 *S. \array{ && Mor(C \times \mathbf{B} End(V)) {}_t \times_s Mor(C \times \mathbf{B} End(V)) \\ & {}^{d_0}\swarrow &\downarrow^\circ& \searrow^{d_1} & \\ Mor(C \times \mathbf{B} End(V)) && Mor(C \times \mathbf{B} End(V)) && Mor(C \times \mathbf{B} End(V)) \\ && \uparrow \\ && \int_\circ d_0^* S \otimes d_1^* S' } \,. Under End(V)End(V)-cardinality, the resulting finite set d 0 *Sd 1 *S\int_\circ d_0^* S \otimes d_1^* S' is the category algebra product of the two original End(V)End(V)-valued functions on Mor(C)Mor(C).

We can also discuss the action of this category algebra on VV-valued functions on Obj(C)Obj(C) in a similar manner:

For AA a monoid write BA\mathbf{B}A for the corresponding one-object category. A representation of AA on VV is a functor ρ:BAVect \rho : \mathbf{B}A \to Vect such that ρ:(a)(Vρ(a)V). \rho : (\bullet \stackrel{a}{\to} \bullet) \mapsto (V \stackrel{\rho(a)}{\to} V) \,. We write V//A V//A for the corresponding action category, being the pullback V//A Vect * Set * BA ρ Vect Set \array{ V//A &\to& Vect_* &\to& Set_* \\ \downarrow && \downarrow && \downarrow \\ \mathbf{B}A &\stackrel{\rho}{\to}& Vect &\to& Set } which has VV as its space of objects and one morphism of the form vaρ(a)(v) v \stackrel{a}{\to} \rho(a)(v) for each vVv \in V and each aAa \in A, with the obvious composition coming from the product in AA.

For the discussion of category algebras, consider again the monoid A=End(V)A = End(V) and hence the action category V//End(V)V//End(V). Notice that every End(V)End(V)-valued set over Mor(C×BEnd(V))Mor(C \times \mathbf{B}End(V)) pulls back along the canonical projection p:V//ABA p : V//A \to \mathbf{B}A to a finite set over Mor(C×V//End(V))Mor(C \times V//End(V)).

So consider a VV-valued set over Obj(C)Obj(C) in the form of a finite set S Obj(C×V//End(V)))=Obj(C)×V \array{ S \\ \downarrow \\ Obj(C \times V//End(V)) ) = Obj(C)\times V } and then the correspondence space of the category CC p *S S Mor(C×V//End(V)) s t Obj(C)×V Obj(C)×V, \array{ && p^* S' \\ && \downarrow \\ S && Mor(C \times V//End(V)) \\ \downarrow & {}^s\swarrow && \searrow^t \\ Obj(C)\times V &&&& Obj(C)\times V } \,,

where p *Sp^* S' is the pullback along the above pp of an End(V)End(V)-valued set over Mor(C)Mor(C) representing, by the previous discussion, an element in the category algebra of CC (with coefficients in VV).

Then pull back SS along the source map to Mor(C×V//End(V))Mor(C \times V//End(V)), tensor there with p *Sp^* S'

s *Sp *S Mor(C×V//End(V)) s t Obj(C)×V Obj(C)×V, \array{ && s^*S \otimes p^* S' \\ && \downarrow \\ && Mor(C \times V//End(V)) \\ & {}^s\swarrow && \searrow^t \\ Obj(C)\times V &&&& Obj(C)\times V } \,,

and then push down along the target map

Mor(C×V//End(V)) ts *Sp *S s t Obj(C)×V Obj(C)×V \array{ && Mor(C \times V//End(V)) && \int_t s^* S \otimes p^* S' \\ & {}^s\swarrow && \searrow^t & \downarrow \\ Obj(C)\times V &&&& Obj(C)\times V }

to get a VV-valued set over Obj(C)Obj(C). Under VV-cardinality, this is the image of the original VV-valued function represented by SS under the action of the category algebra element represented by SS'.

The reasoning needed for seeing this is precisely the one used for the disucssion of the path integral in this context.

Posted at June 26, 2008 8:50 PM UTC

TrackBack URL for this Entry:

2 Comments & 0 Trackbacks

Re: Category Algebras

Urs wrote:
> The V-cardinality operation is a monoidal
> map
> ∣⋅∣:(FinSet V,⊕)→(V,+)
> sending a finite set labeled by V to the
> sum of the labels of its elements.

Do you want to put a weight of this sum of elements? For instance if you have a finite group acting on a vector space V you could take the S sitting over v in V to be all the G translates of v and the map would then be 1/|G| \sum gv, i.e. the projection of v on the invariant subspace.

Posted by: Maarten Bergvelt on July 1, 2008 5:36 PM | Permalink | Reply to this

Re: Category Algebras

Very good point.

Indeed, these weights will come in when we do push-forward of sets over groupoids (instead of just sets over sets, as I have considered here).

You can see these factors appear in definition 5 of John Baez’s HDA VII.

The way I like to think about them is this:

ultimately, as I was just telling Eric # ,these sets over a groupoid will be thought of as functors from the groupoid to Set\mathrm{Set}. Then we can consider the push-forward literally as the colimit of that set-valued functor and Tom Leinster’s formula applies to give us these weights automatically (or almost automatically).

Posted by: Urs Schreiber on July 1, 2008 5:55 PM | Permalink | Reply to this

Post a New Comment