1. Category and Functor
Category
In category theory, we will encounter many presentations in terms of diagrams. Roughly speaking, a diagram is a collection of ‘objects’ denoted by and ‘arrows’ between them denoted by as in the examplesWe will always have an operation to compose arrows. The diagram is called commutative if all the composite paths between two objects ultimately compose to give the same arrow. For the above examples, they are commutative if
Definition 1.1. A category consists of
1. | A class of objects: (a category is called small if its objects form a set). We will write both and for an object in . |
2. | A set of morphisms: for each . An element will be called a morphism from to , and denoted by When is clear from the context, we will simply write for . |
3. | A composition operation between morphismswhich will be denoted in terms of a diagram by |
These are subject to the following axioms:
1. | Associativity: holds, and will be denoted by without ambiguity. This property can be expressed in terms of the following commutative diagram |
2. | Identity: for each , there exists called the identity element such thati.e. we have the following commutative diagrams |
Definition 1.2. A subcategory of (denoted by ) is a category such that
1. | |
2. | |
3. | compositions in coincide with that in under the above inclusion. |
is called a full subcategory of if .
Definition 1.3. A morphism is called an isomorphism (or invertible) if there exists such that and , i.e. we have the following commutative diagram Two objects are called isomorphic if there exists an isomorphism .
Example 1.4. We will frequently use the following categories.
1. | , the category of sets: |
2. | , the category of vector spaces over a field : is a subcategory of , but not a full subcategory. |
3. | , the category of groups:It has a full subcategory |
4. | , the category of rings: is a subcategory of , but not a full subcategory. has a full subcategory |
The main object of our interest is
• | whose objects are topological spaces and |
• | whose morphisms are continuous maps. |
Example 1.5. Let and be two categories. We can construct a new category , called the product of and , as follows.
• | An object of is a pair of objects and . |
• | A morphism is a pair of . |
• | Compositions are componentwise. |
Quotient category and homotopy
Definition 1.6. Let be a category. Let be an equivalence relation defined on each and compatible with the composition in the following sense The compatibility can be represented by the following diagram We say defines an equivalence relation on . The quotient category is defined by
• | |
• | . |
Exercise 1.7. Check the definition above is well-defined.
One of the most important equivalence relations in algebraic topology is the homotopy relation.
Let . Let denote the topological product of .
Definition 1.8. Two morphisms in are said to be homotopic, denoted by , ifWe will also write or to specify the homotopy . This can be illustrated as
Let be a morphism in . We define its homotopy classWe denote
Theorem 1.9. Homotopy defines an equivalence relation on .
Proof. We first check that defines an equivalence relation on morphisms.
• | Reflexivity: Take such that for any . |
• | Symmetry: Assume we have a homotopy . Then reversing asi.e. taking , gives as required. |
• | Transitivity: Assume we have two homotopies and , then putting them together gives as |
We next check is compatible with compositions.
Let and . Assume and . Then
We denote the quotient category of under homotopy relation bywith morphisms .
Definition 1.10. Two topological spaces are said to have the same homotopy type (or homotopy equivalent) if they are isomorphic in .
Example 1.11. and are homotopy equivalent, but not homeomorphic. In other words, they are isomorphic in , but not isomorphic in . As we will see, and are not homotopy equivalent.
There is also a relative version of homotopy as follows.
Definition 1.12. Let and such that . We say is homotopic to relative to , denoted byif there exists such that
We will also write or to specify the homotopy .
Functor
Definition 1.13. Let be two categories. A covariant functor (resp. contravariant functor) consists of
1. | , |
2. | . We denote by(resp. , denoted by |
satisfying
1. | (resp. ) for any composable morphisms (resp. reversing all arrows in the diagram on the right). |
2. | . |
is called faithful (or full) if is injective (or surjective) . is called fully faithful if is both full and faithful.
Example 1.14. The identity functor maps for any object and morphism .
Example 1.15. ,defines a covariant functor anddefines a contravariant functor.
Functors of these two types are called representable (by ).
Example 1.16. The forgetful functor (mapping a group to its set of group elements) is representable by the free group with one generator, i.e. . We have a set isomorphism
Example 1.17. Let be an abelian group. Given , we will study its n-th cohomology with coefficients in . It defines a contravariant functorWe will see that this functor is representable by the Eilenberg-Maclane space if we work with the subcategory of CW-complexes.
Example 1.18. We define a contravariant functor are continuous real functions on . A classical result of Gelfand-Kolmogoroff says that two compact Hausdorff spaces are homeomorphic (i.e. isomorphic in ) if and only if and are ring isomorphism (i.e. isomorphic in ).
Proposition 1.19. Let be a functor. Suppose is an isomorphism in , then is an isomorphism in .
Natural transformation
Definition 1.20. Let be two categories and be two functors. A natural transformation consists of morphismssuch that the following diagram commutes for any (here if are covariant and if are contravariant) is called a natural isomorphism if is an isomorphism for any and we write .
Example 1.21. We consider the following two functors Given a commutative ring , is the group of invertible matrices with entries in , and is the multiplicative group of invertible elements of . We can identify .
The determinant defines a natural transformation where is the determinant of the matrix. The naturality of is rooted in the fact that the formula for determinant is the same for any coefficient ring. In this way, we can say precisely that taking the determinant of a matrix is a natural operation.
Example 1.22. Let and . We have
• | A natural transformation for (contravariant) representable functors . |
• | A natural transformation for (covariant) representable functors . |
Example 1.23. The above example is a special case of the following construction. Let .
• | Let be a contravariant functor. Then any induces a natural transformation by assigning to . |
• | Let be a covariant functor. Then any induces a natural transformation by assigning to . |
Definition 1.24. Let be functors and be two natural transformations. The composition is a natural transformation from to defined by
Definition 1.25. Two categories are called isomorphic if such that . They are called equivalent if and two natural isomorphisms . In these cases, we say gives an isomorphism/equivalence of categories.
In applications, isomorphism is a too strong condition to impose for most interesting functors. Equivalence is more realistic and equally good essentially. The following proposition is very useful in practice.
Proposition 1.26. Let be an equivalence of categories. Then is fully faithful.
Functor Category
Definition 1.27. Let be a small category, and be a category. We define the functor category
• | Objects: functors from to |
• | Morphisms: natural transformations between two functors (which is indeed a set since is small). |
The following Yoneda Lemma plays a fundamental role in category theory and applications.
Theorem 1.28 (Yoneda Lemma). Let be a category and . Denote the two functors
1. | Contravariant version: Let be a contravariant functor. Then there is an isomorphism of sets This isomorphism is functorial in . |
2. | Covariant version: Let be a covariant functor. Then there is an isomorphism of sets This isomorphism is functorial in . |
The precise meaning of functoriality in is that we have natural isomorphisms of functors The required isomorphisms in the above Yoneda Lemma are those maps described in Example 1.23.
One important consequence is that we have set isomorphisms which are functorial both in and . This gives rise to a fully faithful functor
Duality
Many concepts and statements in category theory have dual descriptions. It is worthwhile to keep eyes on such dualities. Roughly speaking, the dual of a category-theoretical expression is the result of reversing all the arrows for morphisms, changing each reference to a domain to refer to the target (and vice versa), and reversing the order of composition.
For example, let for a category. We can define its opposite category by declaring
• | ; |
• | is a morphism in if and only if is a morphism in ; |
• | the composition of two morphisms in is the same as the composite in . |
A contravariant functor is the same as a covariant functor . With this help, we can work entirely with covariant functors or contravariant functors. For example, the two statements in Yoneda Lemma are actually the same if we consider opposite categories.
An another example, we will often consider the lifting problem by finding a map such that the following diagram is commutativeThe dual problem is the extension problem by finding a map such that the dual diagram is commutative
Adjoint functor
Let be two categories, and let be two (covariant) functors. The rulesdefine two functors We say and are adjoint to one another (more precisely, is the left adjoint, is the right adjoint), if there is a natural isomorphism that is, for each , we have a set isomorphism and this isomorphism is functorial both in and in . We sometimes write adjoint functors as
Example 1.29 (Free vs Forget). Let be a set, and denote the free abelian group generated by . This defines a functor Forgetting the group structure defines a functor (such functor is often called a forgetful functor)These two functors are adjoint to each otherIn fact, many "free constructions" in mathematics are left adjoint to certain forgetful functors.
Proposition 1.30. Letbe adjoint functors. Then there are natural transformations
is called the unit of the adjunction. is called the counit of the adjunction.