Enriched category - Ciallo Wiki

1 Motivation and Intuition

1.1 Limits of ordinary categories

In an ordinary category, between any two objects \(A\) and \(B\) there is a hom-set \(\mathrm{Hom}(A,B)\). The category records what morphisms exist and how they compose, but it treats the collection of morphisms as largely featureless: there is no intrinsic way to say, inside the category, that some morphisms are close, continuous in a parameter, graded by degree, or linear in coordinates—unless such information is encoded indirectly via additional structure outside the hom-sets.

This limitation becomes visible whenever the morphisms naturally carry extra mathematical data. For example, in many settings one can talk about homomorphisms between vector spaces (which themselves form vector spaces), or about continuous maps (which sit inside topological function spaces), or about homology-type structures (which are organized by chain complexes). Ordinary categories can represent these facts only by enlarging the objects themselves or by external bookkeeping.

1.2 Enrichment as “morphisms with structure”

Enrichment remedies this by replacing the hom-sets with hom-objects in a fixed ambient category. Instead of having merely a set of morphisms from \(A\) to \(B\), one has an object \(\mathcal{V}(A,B)\) in some chosen base (often denoted \(\mathcal{V}\) or a monoidal category \(\mathbf{V}\)). Composition is then required to be a morphism in the base category, compatible with the extra structure carried by these hom-objects.

Conceptually, an enriched category packages “morphisms with structure” into the hom-objects. The same basic categorical axioms still apply—identities and associativity—but they are expressed using the operations of the ambient category, such as tensor products, internal homs, limits, or enrichment-specific maps.

1.3 Examples of enriched hom-objects

Common ambient choices lead to familiar enrichments:

Vector spaces: hom-objects are vector spaces; composition is bilinear.
Topological spaces: hom-objects are topological spaces of maps; composition and identities are continuous.
Posets and preorders: hom-objects represent order degrees; composition respects the order.
Metric or generalized metric structures: hom-objects carry distance-like data; composition interacts with triangle inequalities or related constraints.
Chain complexes: hom-objects have differentials; composition respects the differential with sign rules.

These examples share a pattern: the “space of morphisms” itself has internal mathematical structure, and that structure is made functorial in the categorical variables.

1.4 Enriched vs. internal categories

An internal category is a category whose objects, morphisms, source, target, identities, and composition live inside a single ambient category with finite limits (such as sets, topological spaces, or schemes). Enrichment instead fixes an ambient monoidal category that governs how hom-objects compose, but does not necessarily interpret the category as living with object/morphism maps inside that ambient category.

Both approaches generalize categories, but they generalize different aspects. Internal categories generalize the *data of morphisms as an internal object* with categorical operations between them. Enriched categories generalize the *hom-sets as structured objects* whose composition is a morphism in a base category. The two frameworks can coincide in special cases (for example, when the base supports a suitable encoding), but they should not be conflated.

2 Foundations

2.1 Enrichment base: monoidal categories

Enrichment is formulated relative to a monoidal category, which provides the algebraic mechanism for combining hom-objects.

A monoidal category \((\mathbf{V},\otimes,I)\) supplies:

a tensor product \(\otimes\),
a unit object \(I\),
associativity and unit constraints that relate different ways of tensoring.

Different enrichments use different monoidal categories, but the pattern is the same: hom-objects must be composable via \(\otimes\), and identities come from the unit object \(I\).

2.1.1 Tensor product and unit object

The tensor product is the formal way to “combine” hom-objects. Given objects \(A,B,C\), the enriched homs \(\mathbf{V}(A,B)\) and \(\mathbf{V}(B,C)\) are combined using \(\otimes\) to form something that can map into \(\mathbf{V}(A,C)\).

The unit object \(I\) acts as the source of identities. In an enriched category, the identity element is not merely an element of a set; it corresponds to a morphism from \(I\) into the appropriate hom-object.

2.1.2 Associativity and unit constraints

Because tensoring is not strictly associative in most monoidal categories, one uses natural isomorphisms:

associators that compare \((X\otimes Y)\otimes Z\) with \(X\otimes (Y\otimes Z)\),
left and right unit constraints comparing \(I\otimes X\) and \(X\otimes I\) with \(X\).

These constraints affect how the associativity axiom of enriched categories is expressed and verified.

2.2 Hom-objects and enriched hom notation

Let \(\mathcal{C}\) be an enriched category. For each pair of objects \(A,B\in \mathrm{Ob}(\mathcal{C})\), there is a hom-object \[ \mathcal{C}(A,B)\in \mathbf{V}. \] This hom-object generalizes \(\mathrm{Hom}(A,B)\) from sets to structured objects. The notation \(\mathcal{C}(A,B)\) is standard even though its internal nature depends on \(\mathbf{V}\).

2.3 Identity morphisms in the enriched setting

In an ordinary category, for each object \(A\) there is an identity morphism \(\mathrm{id}_A\in \mathrm{Hom}(A,A)\). In a \(\mathbf{V}\)-enriched category, the identity is specified by a morphism in \(\mathbf{V}\): \[ I \longrightarrow \mathcal{C}(A,A). \] This morphism selects the “identity behavior” inside the structured hom-object. When the base is a category like vector spaces, this is analogous to choosing a vector corresponding to the identity linear map.

2.4 Composition as a morphism in the base category

The enriched version of composition is also a morphism in \(\mathbf{V}\). For objects \(A,B,C\), there is a map in the base category: \[ \mathcal{C}(B,C)\otimes \mathcal{C}(A,B)\longrightarrow \mathcal{C}(A,C). \] Thus, composing morphisms is not an external operation on elements; it is encoded as a structural morphism whose existence and compatibility come from the ambient tensor product.

2.5 Coherence conditions

Enriched categories require axioms analogous to unit laws and associativity, but now stated using the associativity and unit constraints of the monoidal base.

Associativity: composing three hom-objects in either parenthesization yields the same result after applying the associator constraints.
Unit laws: using the identity maps and tensor unit constraints yields the appropriate left and right neutrality for composition.

These coherence conditions ensure that repeated composition is well-defined independently of how tensor products are parenthesized.

3 Enriched Categories in Specific Bases

3.1 Categories enriched over vector spaces

When the base \(\mathbf{V}\) is the monoidal category of vector spaces over a field (with tensor product), enriched categories become linear categories.

3.1.1 Linear categories and bilinear composition

A \(\mathbf{Vect}\)-enriched category has hom-objects that are vector spaces and whose composition map is bilinear: \[ \mathcal{C}(B,C)\otimes \mathcal{C}(A,B)\to \mathcal{C}(A,C). \] Bilinearity ensures that composing maps is linear in each argument separately, matching the behavior expected in representation theory and module categories.

3.1.2 Enriched functors and natural transformations

An enriched functor between linear enriched categories sends objects to objects and acts linearly on each hom-space in a way compatible with identities and composition. An enriched natural transformation between such functors respects the linear structure as well: its components are morphisms in the base that satisfy an enriched version of naturality.

3.2 Categories enriched over topological spaces

Let the base be topological spaces with cartesian product (or another standard monoidal structure). Then enriched hom-objects are topological spaces.

3.2.1 Continuous composition and identity maps

The hom-object \(\mathcal{C}(A,B)\) is a space of morphisms. The composition map \[ \mathcal{C}(B,C)\times \mathcal{C}(A,B)\to \mathcal{C}(A,C) \] must be continuous, and the identity assignment is likewise a continuous morphism from the unit object (often a one-point space) into \(\mathcal{C}(A,A)\). This captures “continuous families of morphisms” and ensures that categorical operations vary continuously.

3.2.2 Enriched homotopical perspectives

When topological enrichment is combined with homotopical methods, enriched structures can be used to model higher coherence and invariants that depend continuously on parameters. In such contexts, the hom-spaces and their topology provide more information than bare hom-sets.

3.3 Categories enriched over posets and preorders

A monoidal category can be built from poset-like data where the tensor product encodes composition of order relationships.

3.3.1 Order-enriched homs

If the base captures order (e.g., with hom-objects being truth values or elements of a lattice), then an enriched category can encode comparisons between morphisms using order rather than membership in a set. Composition becomes an order-preserving operation, and identities act as “maximal” or “reflexive” elements within hom-objects.

3.3.2 Galois-like interpretations

In order-enriched contexts, enriched relations often correspond to adjunction-like phenomena between order structures. While the details depend on the base chosen, the overarching idea is that enrichment can transform order-theoretic properties into compositional laws.

When the base is a category of metric-like objects (or more general “distance” structures), enriched homs record quantitative data.

3.4.1 Enriched distance-like homs

Instead of hom-sets, one has objects that measure how different morphisms are, or how “far” one morphism is from another. Composition typically must satisfy an inequality-type constraint reflecting the geometry of distances—most famously the triangle inequality in metric settings.

3.4.2 Examples and typical constraints

Typical constraints include:

composition does not increase distance beyond a controlled bound,
identity morphisms behave as zero-distance elements,
enriched functors are non-expansive (or appropriately Lipschitz) with respect to the metric enrichment.

These conditions allow categorical language to express convergence, stability, and approximation behaviors.

3.5 Categories enriched over chain complexes

Let the base be chain complexes with a tensor product and monoidal unit. Enriched categories then have hom-objects graded by degree and equipped with differentials.

3.5.1 Differential graded enrichment

A differential graded (dg) category is a category enriched over chain complexes such that:

each hom-object is a chain complex,
identities correspond to cycle-like elements,
composition is a chain map.

This is especially common in homological algebra and modern representation theory, where morphism spaces naturally carry chain-level structure.

3.5.2 Signs and composition compatibility

Chain complexes bring subtleties: compatibility with differentials requires sign conventions so that the differential interacts with tensor products correctly. Composition must respect the grading, and the “Leibniz rule” manifests as a condition involving differentials and degrees of elements. These sign rules ensure associativity and unitality lift consistently to the chain level.

4 Enriched Functors and Transformations

4.1 Enriched functors

An enriched functor between \(\mathbf{V}\)-enriched categories is determined by how it acts on objects and on hom-objects.

4.1.1 Mapping objects and hom-objects

Given enriched categories \(\mathcal{C}\) and \(\mathcal{D}\), an enriched functor \(F:\mathcal{C}\to\mathcal{D}\) assigns each object \(A\) of \(\mathcal{C}\) to an object \(F(A)\) of \(\mathcal{D}\), and for each pair \(A,B\) gives a morphism in \(\mathbf{V}\): \[ \mathcal{C}(A,B)\to \mathcal{D}(F(A),F(B)). \] This mapping must preserve the enrichment structure rather than merely sending morphisms as elements.

4.1.2 Compatibility with composition and identities

The functor axioms require commutation with identity maps and composition morphisms. Concretely, the enriched hom-maps induced by \(F\) must intertwine with the base-category composition laws so that:

identities map to identities,
the image of a composite equals the composite of images in the enriched sense.

4.2 Enriched natural transformations

Natural transformations are also defined relative to the enrichment.

4.2.1 Naturality in enriched form

If \(F,G:\mathcal{C}\to\mathcal{D}\) are enriched functors, an enriched natural transformation specifies, for each object \(A\), a morphism in the base category linking \(F(A)\) and \(G(A)\). Naturality then becomes a compatibility condition with enriched composition maps, ensuring commutativity of the appropriate enriched diagrams in \(\mathbf{V}\).

4.2.2 Component conditions

The components are not arbitrary elements; they must satisfy coherence relative to the base structure. In linear enrichment, for instance, the components behave linearly with respect to morphism composition; in dg enrichment, the differential constraints and grading must match.

4.3 Enriched isomorphisms and equivalences

An enriched isomorphism is typically an enriched functor whose induced hom-maps are isomorphisms in the base category, and which has an enriched inverse up to those structural isomorphisms. Enriched equivalence generalizes categorical equivalence: it allows essential surjectivity on objects (in an enriched or reflected sense) and homwise equivalence (e.g., quasi-isomorphism in dg contexts, or isomorphism in strictly enriched settings).

5 Universal Constructions and Limits

5.1 Enriched products and coproducts

Products and coproducts in enriched categories depend on how the ambient base supports corresponding universal properties. Informally, an enriched product of objects is an object that represents morphisms into it in a way consistent with the base structure, not just set-theoretically.

In many enrichments, products and coproducts can be constructed by combining the base’s own limits and colimits, while ensuring that the universal maps are morphisms in \(\mathbf{V}\) and satisfy enriched naturality.

5.2 Enriched limits: ends and coends

Enriched limits are often expressed via ends and coends, which generalize universal wedges and quotients in a way that is compatible with enrichment.

5.2.1 Ends as universal enriched wedges

An end is a universal object (or universal hom-object) that captures families of morphisms subject to dinaturality, with universality measured in the base category. In an enriched setting, the universal property accounts for the structure of hom-objects and uses enriched natural transformations.

Ends frequently appear when describing equalizers of enriched morphisms or when encoding invariants constructed from functor categories.

5.2.2 Coends as universal enriched quotients

Coends dualize ends and often arise in ways that correspond to “gluing” along morphisms. In enriched form, coends provide a structured way to aggregate data from a bifunctor while respecting the ambient composition and tensoring.

5.3 Weighted limits and colimits

A powerful enrichment principle is that limits and colimits can be organized via weights. A weight assigns, in the base category, a measure of how a diagram contributes to a limit or colimit.

This viewpoint unifies many classical constructions: when the base is sets, weights reduce to familiar indexing mechanisms. In enriched contexts, weights carry additional structure, allowing more nuanced aggregation than ordinary diagram colimits.

5.4 Enriched Kan extensions

Kan extensions generalize extension of functors along other functors. In enriched category theory, Kan extension can be formulated with enriched ends/coends, with the enriched hom-objects playing the role of measuring maps.

5.4.1 Enriched adjunctions via Kan machinery

A typical pattern is that enriched adjunctions can be characterized by the existence of certain enriched Kan extensions. This connects representability, colimits, and universal properties in a single enriched framework.

6 Adjunctions and Monads (Enriched)

6.1 Enriched adjunctions

An adjunction in an enriched setting consists of enriched functors \(F:\mathcal{C}\leftrightarrows \mathcal{D}:G\) with unit and counit that satisfy the triangular identities, now expressed in the ambient base’s morphisms. The hallmark is an enriched form of hom-isomorphism: \[ \mathcal{D}(F(A),B)\cong \mathcal{C}(A,G(B)) \] as objects in the base category, not merely as sets.

6.2 Enriched monads and algebras

A monad in an enriched context is an endofunctor (enriched) with multiplication and unit maps that respect enrichment. Algebras for enriched monads are objects equipped with an action morphism in the base, satisfying associativity and unit axioms in the enriched sense.

This framework allows algebraic structures (modules, representations, chain-level structures) to be encoded in a way that interacts correctly with the enrichment.

6.3 Kleisli and Eilenberg–Moore constructions (enriched flavor)

The Kleisli category and Eilenberg–Moore category can be built in the enriched setting under suitable assumptions on the base and completeness properties.

Kleisli construction: captures free algebra behavior for the monad, with morphisms reflecting the monadic structure.
Eilenberg–Moore construction: organizes algebras and algebra homomorphisms that preserve the monad action.

In enriched cases like dg categories, these constructions have interpretations aligned with “derived” or homological constraints.

6.4 The enriched Beck theorem overview (conceptual)

Beck’s theorem characterizes when a functor is monadic. In enriched settings, the criterion adapts to enriched limits and the behavior of the comparison functor. Conceptually, it states that if the functor reflects the structure of algebras and preserves the right enriched constructions, then the original situation is equivalent to an Eilenberg–Moore category of some monad.

7 Enriched Yoneda Lemma and Representability

7.1 Enriched Yoneda embedding

The Yoneda embedding generalizes to enriched categories by sending an object \(A\) to the enriched hom-functor \(\mathcal{C}(A,-)\). Rather than producing ordinary set-valued presheaves, one obtains presheaves valued in the base category.

This embedding retains the essence of Yoneda: objects are determined by how they map into other objects, with all structure encoded through the enriched hom-objects.

7.2 Yoneda lemma for enriched homs

The enriched Yoneda lemma states that enriched natural transformations from a representable functor to any other functor correspond to the value of the target functor at the representing object, again expressed within the base category.

As a result, representable functors are fully faithful in the enriched sense: mapping data between representables is recovered from the enriched hom-objects themselves.

7.3 Representable and dense enrichment

Representability generalizes to enriched presheaves: an enriched presheaf is representable if it is isomorphic (in the enriched presheaf category) to a representable functor \(\mathcal{C}(A,-)\).

Density refers to when a collection of representable objects suffices to reconstruct the entire enriched category, often measured via colimits or ends weighted appropriately.

7.4 Enriched presheaves

Enriched presheaves are contravariant (or covariant) enriched functors from an enriched category into the base category.

7.4.1 Presheaf categories in enriched form

The category of enriched presheaves carries an enriched structure itself under appropriate conditions on the base. Hom-objects between presheaves can be constructed using ends, making the presheaf category a central setting for enriched limits, colimits, and representability arguments.

7.4.2 Density and reconstruction ideas

Using density and enriched Yoneda, one can reconstruct enriched functors from their action on representables. This underlies many completion and extension techniques, including enriched variants of classical sheaf-like or completion procedures.

8 Enriched Model Structures (High-level)

8.1 Enrichment-compatible categorical frameworks

In homotopical contexts, one often studies enriched categories that interact well with a model structure (or with another framework organizing “weak equivalences”). The key requirement is compatibility: hom-objects and enriched composition should behave coherently with the chosen notion of equivalence.

At a high level, this means that enrichment-compatible mapping objects should preserve or reflect the intended homotopical information, so that derived constructions are stable under categorical operations.

8.2 Enriched homotopy-theoretic intuition (overview)

Enriched hom-objects can encode not just whether morphisms exist, but also how morphisms vary up to homotopy, along parameters, or along higher-dimensional structures. In such settings, the enriched structure plays the role of a “controlled space” of morphisms, where homotopical meaning is accessible through the base category.

8.3 Transporting enrichment through constructions

When passing to categories of functors, localizations, completions, or homotopy categories, one needs to transport the enrichment so that it remains valid after taking derived or localized versions of hom-objects. This typically involves verifying that the enrichment operations are compatible with replacements (e.g., cofibrant/fibrant objects in model category language) so that compositions and identities remain well-behaved.

9 Relationships to Other Frameworks

9.1 Enriched categories vs. ordinary categories

Ordinary categories correspond to enrichment over the monoidal category of sets (with cartesian product or similar monoidal structure). In this sense, ordinary hom-sets are recovered as hom-objects that carry no additional internal structure besides their elements.

Thus, enriched categories can be viewed as “categories with structured hom-spaces,” where the structure is chosen according to the mathematical phenomenon being studied.

9.2 Enriched categories vs. internal categories

Internal categories treat categorical structure as internal data in a base category with finite limits, while enriched categories treat homs as objects in a monoidal base and express composition via base morphisms. Both generalize categories, but their emphases differ:

internal categories: emphasize internalizing objects and morphisms,
enriched categories: emphasize enriching homs and composition.

Understanding the distinction clarifies which type of universality and which kind of coherence are being imposed.

9.3 Enriched categories vs. ∞-categorical enrichment (overview)

Enrichment can be pushed further by moving to higher categories, where hom-objects are themselves higher-dimensional and composition is only associative up to coherent higher homotopies. In such frameworks, enrichment merges with ∞-categorical ideas: the base is no longer merely a 1-category with strict tensoring but a higher structure where associativity and identities are controlled by higher cells.

The overall relationship is conceptual: both enrichments aim to encode additional structure in morphisms, but ∞-categorical enrichment accounts for higher coherences directly.

9.4 Enriched profunctors and spans (conceptual)

Profunctors generalize functors by relating objects via bimodules-like data rather than direct mapping on objects. Enriched profunctors incorporate the base-category structure in their defining hom-objects or action maps, leading to a richer “relation calculus.”

In many treatments, enriched profunctors can be composed using coends weighted by the enrichment base, connecting profunctor composition to the universal constructions discussed earlier.

10 Standard Constructions and Computation Patterns

10.1 Building new enriched categories from old ones

Enriched categories can be formed from existing ones through standard category-theoretic operations adapted to enrichment, such as:

forming functor categories (enriched natural transformations),
taking enriched subcategories when hom-objects are inherited appropriately,
constructing products and coproducts in the enriched sense.

These constructions depend on the base category supporting the relevant limits/colimits and on the enrichment maps being stable under the operations performed.

10.2 Enriched completion and completion-style processes

Completion procedures aim to add limits, colimits, or other universal features so that an enriched category becomes “closed” under a desired construction. Typical strategies use:

presheaf categories (via enriched Yoneda),
sheaf-like localization principles (in settings where a Grothendieck topology exists),
completion by representables or by weighted colimits.

The enriched viewpoint helps preserve the structure in hom-objects during completion rather than collapsing it to sets.

10.3 Practical “diagram schemas” for composition

In computations, composition in enriched categories is often implemented via the base-category composition morphisms and tensor products of hom-objects. Diagrammatic reasoning typically tracks:

which hom-objects are tensored,
how associators and unit constraints re-parenthesize,
how the enriched functor maps intertwine with enriched composition.

These schemas reduce errors by making coherence explicit, especially when dealing with non-strict monoidal bases.

10.4 Common pitfalls and coherence checks

Several recurring sources of difficulty include:

Ignoring associator/unit constraints: non-strict monoidal categories require careful coherence.
Assuming elementwise definitions: enriched axioms are morphism-level constraints in the base category; elementwise arguments can fail when hom-objects are not sets.
Overlooking sign and grading rules: in dg-enrichment, differential compatibility depends on correct degree bookkeeping.
Mismatched naturality: enriched naturality uses enriched hom-objects and enriched composition, so the relevant commutative diagrams are in \(\mathbf{V}\), not in sets.

Checking coherence conditions systematically—often by verifying that the required enriched diagrams commute—is the standard safeguard.