You are currently browsing the category archive for the ‘CT.category-theory’ category.

This post is a continuation of the previous post – A mild introduction to Categories. Before I jump into the technical details of Categorical stuff like Functors, I would like to introduce the notion of a universal mapping property (henceforth referred to as UMP).

I won’t bog you down with the details of jargon that one has to absorb before getting the definition. Those interested in the definition of a UMP may refer the Wikipedia article here. Loosely speaking, a UMP of an object in a category is a property that characterizes it completely. In this post, I shall only present the universal properties of some familiar objects. We start with the property of a kernel of a group homomorphism.

**Kernel of a group homomorphism: **Let **Grp **be the category of groups. (One may similarly define the UMP of the kernel of a ring, module, field homomorphism). Let and be groups and be a group homomorphism. To familiarize ourselves with the categorical notation, we continue this example in categorical language- Let (**Grp**) and . There exists an object (**Grp**) and a morphism such that the pair satisfies the following properties:

- as a morphism (group homomorphism) from to .
- Whenever $latex (K’, \iota’ \in \text{Mor}(K’,
~~A))$ is such that , then there is a unique morphism such that .~~

In group-theoretic parlance, kernel (of a homomorphism) is a group which has an embedding into the preimage and whose image under the homomorphism is trivial. The group is unique in that any other group that has these properties must factor uniquely via the kernel. This property of the kernel, in fact, characterizes it.

We begin with examples that create new rings from old ones:

**Direct sum of -Modules:** Given -modules , the direct product of and is an -module along with (inclusion) maps such that given any -module and any -module homomorphisms , there is a unique homomorphism such that .

Note: This can be generalized to an arbitrary family of submodules of an -module. The existence follows from by taking all but finitely many nonzero elements of each submodule.

**Polynomial ring in one variable over a ring: **Given rings , a ring-homomorphism and an element , there exists a unique ring-homomorphism extending such that .

**Quotient Ring: **Given a ring and its ideal (possibly trivial or the whole ring), the quotient ring is a ring along with ring homomorphism such that

- ,
- For every ring and homomorphism with , there is a unique ring homomorphism such that .

The UMP of a free module (or free group) is in some sense, more fundamental.

**Free -module: **Let be a set and be a ring. A free -module over is an -module along with an (injective) set map such that given any -module and a set map , there is a unique -module homomorphism such that .

The UMP of a ring of fractions of a ring is often useful.

**Ring of fractions of a ring: **Given a domain and a multiplicatively closed subset of , (that contains 1 but no zerodivisors of ), the ring of fractions of is a ring (denoted by) along with a map such that given a ring and ring homomorphism such that consists of units (of ), then there is a unique homomorphism such that .

Note: If we take to be a domain and then we get the field of fractions of .

We give two examples to show that UMPs exist for non-*algebraic *objects too. (whatever that means!)

**Quotient of a set:** Let be a set with an equivalence relation on it. The quotient of with respect to is a set with a map such that given any set and a map with whenever , then there is a unique map such that .

**Product of sets: **Let be an arbitrary family of sets. We define the product set of ‘sto be a set alongwith (projection) maps such that for any __ (complete this!)

The above two properties – quotient and product of sets generalize immediately and can be used to define the UMP for Quotient topology and Box topology in the category **Top** of topological spaces. To state the UMP of the Product topology, one uses an analog of Direct Sum of Modules stated above.

**Exercise:** Deduce the UMP for cokernel of a module homomorphism.

**Uniqueness of a UMP: **The universal property of an object need not be unique. A trivial example to see this is as follows: The identity element of a group (denote it by ) satisfies the two following properties:

- Given any group , there is a unique homomorphism .
- Given any group , there is a unique homomorphism .

**Future post:**

- A small application of UMP.

Galois, Abel, Cauchy, Cayley and other stalwarts of their era developed group theory. Kummer, Dedekind, Noether were among those who enriched our understanding of rings and ideals, especially over and over . This marked the beginning of what is coined as “Abstract Algebra” in many undergraduate texts and curricula. But as Herstein puts it, *abstractness*, being a relative term, this algebra of rings, fields, modules and vector-spaces might not be* so* *abstract*, as far as today’s development in Algebra is concerned. Perhaps I might call today’s research in Algebraic Geometry, the pinnacle of *abstraction. *Indeed, Grothendieck took the subject to dizzying heights of abstraction. There was another *abstractness* introduced by founders of homological algebra including McLane and Eilenberg. “Category Theory” is sometimes notoriously humoured as “Abstract Nonsense”.

**What is a Category?**

A category, roughly speaking, consists of objects and relations between objects. To be more formal, I would say a category is a collection of objects (**Obj**) and for any two objects **Obj**, a (possibly empty) set (**Mor**()) called as morphisms from to (which can be thought of as the set of all maps from to ); and for any three objects **Obj**, a law of composition (i.e. a map)

**Mor**() **Mor**() **Mor**()

satisfying a few properties that we may want them to satisfy. Before I start-off with some examples, let me tell a fundamental category that can stand as a guiding example in better understanding properties of categories. **Set **is a category whose objects are sets (not classes of sets; for I don’t want to start off paradoxes again :P) and morphisms are functions (or plain set-maps).

The properties a category (**Obj**, **Mor**) is expected to satisfy are obvious —

- Two sets
**Mor**() and**Mor**() are disjoint unless and , in which case they are equal. - For each object of
**Obj**, there is a morphism**Mor**() which acts as left and right identity for the elements of**Mor**() and**Mor**() respectively, for all objects**Obj**(). - The law of composition is associative whenever defined (whatever that means!)

(In (2) above, note that **Mor**() may contain more than one object; there are so many bijections from a set onto itself.)

Notice that these properties are satisfied by the aforementioned category **Set** quite trivially. Categories are abundant in Mathematics. Here are a few to begin with:

(PS: Fat rigorous books in Maths make comments like: *Henceforth, by abuse of notation, we will denote a category by its objects. *I refrain from making such obvious remarks.)

(PPS: Ignore those examples below which you don’t follow.)

**PSet**, (meaning, pointed sets) with nonempty sets with a fixed element (special point) of every set as objects, and morphisms as set maps with the property that special points are mapped to special points,**Grp**, or the category with groups as objects and group-homomorphisms as morphisms,- R-
**Mod**, or the category of R-modules with R-module-homomorphisms as morphisms, **FD**––**Vec**, or finite-dimensional vector spaces over a field with -linear homomorphisms as morphisms,**PPTop**, or Pointed path-connected topological spaces with a special point and with continuous functions (respecting special-points) serving as morphisms,**Top**– with objects as open sets of a topological space and inclusion maps as morphisms. Note that here,**Mor**() is empty if and consists of one object, the inclusion map from to otherwise.**Hol**, where objects are regions (open & path-connected subsets) of and morphisms are holomorphic functions on them,- –
**Var**Let be an algebraically closed field. (Affine or Projective) algebraic varieties constitute objects in this category and (variety) morphisms are morphisms here. (PS: If you know this category, then you know most of the categories above :P)

We also have something known as Subcategories; a subcategory is a category whose objects and morphisms are the objects and morphisms of another (bigger) category. A few examples to illustrate this —

**PSet**is a subcategory of**Set**.**Ab**, the category of abelian groups is a subcategory of**Grp**,**Rng**= Category of rings (possibly without identity),**Ring**= Category of rings with identity and**CRing**= Category of commutative rings,**Field**= Category of fields. Then**Field****CRing****Ring****Rng**(PS: This fancy notation is by Jacobson.)

**Why study Category Theory?**

Categories were introduced to make rigorous the meaning of the word *natural* used in many contexts. As an example, consider this. Given a -vector space and a basis of , we have an explicit map from to its dual, , but there is no such explicit map without a basis. However, we have a *natural* map from to its bidual, , namely, ; .

A category, say of groups, has group-homomorphisms that take a group to another retaining the *juice* or* information* of the original group. We can then study a* tough* group by looking at its image in a relatively *easier* group and can deduce information about the tough group. In a somewhat similar fashion, we have maps (called as *functors*) that take objects and morphisms of one category to another. Thus, study of a tougher category can be simplified by looking at corresponding objects (and morphisms) in the other category.

For instance, a geometric assertion on affine varieties can be transferred to a statement in commutative algebra with the use of a functor that associates to each affine variety, its coordinate ring. Similarly, one observes that homeomorphism of path-connected topological spaces implies group-isomorphism of their fundamental groups. Thus, if the fundamental groups are different, then the two spaces cannot be homeomorphic.

More on Functors in a future post!

Future posts:

- Functors
- Products & coproducts
- Limits & colimits
- Yoneda Lemma

## Recent Comments