Natural Transformations — Mathematics Knowledge System

Overview

A natural transformation is a morphism between functors. If functors are 'translations' between mathematical universes, natural transformations are 'consistent' ways of converting one translation into another — with the consistency condition expressed by a commutative square for every morphism in the source category. Natural transformations were historically the motivation for inventing categories and functors in the first place: Eilenberg and Mac Lane introduced the entire categorical apparatus just to make precise what it means for two constructions to be 'naturally' (rather than arbitrarily) isomorphic.

Intuition

Imagine two functors $F, G: \mathcal{C} \to \mathcal{D}$ as two different 'lenses' for viewing the same source category. A natural transformation $\eta: F \Rightarrow G$ is a family of morphisms — one for each object — that coherently converts F's view to G's view. The naturality condition says: it doesn't matter whether you first apply F to an arrow then convert, or first convert then apply G to the arrow; you get the same result. This is the categorical way of saying 'the conversion doesn't depend on any arbitrary choices'.

Formal Definition

Definition

Let $F, G: \mathcal{C} \to \mathcal{D}$ be functors. A natural transformation $\eta: F \Rightarrow G$ is a family of morphisms $\eta_A: FA \to GA$ in $\mathcal{D}$, indexed by objects $A \in \mathcal{C}$, such that for every morphism $f: A \to B$ in $\mathcal{C}$, the following naturality square commutes:

\begin{array}{ccc} FA & \xrightarrow{Ff} & FB \\ \downarrow_{\eta_A} & & \downarrow_{\eta_B} \\ GA & \xrightarrow{Gf} & GB \end{array}

The naturality square for a morphism f: A → B

naturality-square

\eta_B \circ Ff = Gf \circ \eta_A \quad \text{for all } f: A \to B \text{ in } \mathcal{C}

Algebraic form of the naturality condition

naturality-condition

\text{If each } \eta_A \text{ is an isomorphism, } \eta \text{ is called a natural isomorphism: } F \cong G

natural-isomorphism

Notation

Notation	Meaning
$\eta: F \Rightarrow G$	η is a natural transformation from functor F to functor G
$\eta_A$	The component of η at object A, a morphism FA → GA
$F \cong G$	F and G are naturally isomorphic (η is a natural isomorphism)
$[\mathcal{C}, \mathcal{D}]$	Functor category: objects are functors C→D, morphisms are natural transformations
$\mathcal{D}^{\mathcal{C}}$	Alternate notation for the functor category [C,D]

Properties

Vertical composition

Natural transformations compose vertically: if \eta: F \Rightarrow G and \mu: G \Rightarrow H, then (\mu \circ \eta)_A = \mu_A \circ \eta_A defines a natural transformation \mu \circ \eta: F \Rightarrow H.

Example: Functors and their natural transformations form a category $[\mathcal{C}, \mathcal{D}]$.

Horizontal composition

If \eta: F \Rightarrow G (between \mathcal{C} \to \mathcal{D}) and \mu: H \Rightarrow K (between \mathcal{D} \to \mathcal{E}), their horizontal composite \mu * \eta: H \circ F \Rightarrow K \circ G has component (\mu * \eta)_A = \mu_{GA} \circ H\eta_A = K\eta_A \circ \mu_{FA}.

Condition: Both expressions agree by naturality of μ.

Functor category

For small \mathcal{C}, the functor category [\mathcal{C}, \mathcal{D}] has functors as objects and natural transformations as morphisms. Identity natural transformation has (\text{id}_F)_A = \text{id}_{FA}.

Worked Examples

For each vector space $V$, define $\iota_V: V \to V^{**}$ by $\iota_V(v)(\varphi) = \varphi(v)$ for $v \in V$ and $\varphi \in V^*$. This is a linear map.
$\iota_V(v)(\varphi) = \varphi(v), \quad v \in V,\; \varphi \in V^*$
We must verify: for every linear map $T: V \to W$, the following commutes:
$\begin{array}{ccc} V & \xrightarrow{T} & W \\ \downarrow_{\iota_V} & & \downarrow_{\iota_W} \\ V^{**} & \xrightarrow{T^{**}} & W^{**} \end{array}$
Compute $\iota_W \circ T$: for $v \in V$ and $\psi \in W^*$:
$(\iota_W(T(v)))(\psi) = \psi(T(v))$
Compute $T^{**} \circ \iota_V$: recall $T^{**}(f)(\psi) = f(T^*(\psi)) = f(\psi \circ T)$ for $f \in V^{**}$. So:
$(T^{**}(\iota_V(v)))(\psi) = \iota_V(v)(T^*(\psi)) = \iota_V(v)(\psi \circ T) = (\psi \circ T)(v) = \psi(T(v))$
Both sides equal $\psi(T(v))$, so the square commutes. Hence $\iota$ is a natural transformation. For finite-dimensional spaces, each $\iota_V$ is an isomorphism, making $\iota$ a natural isomorphism.

Answer: The double dual map $\iota_V(v)(\varphi) = \varphi(v)$ is a natural transformation from $\text{Id}$ to $(-)^{**}$. For finite-dimensional spaces it is a natural isomorphism, witnessing $V \cong V^{**}$ naturally.

Practice Problems

Difficulty 7/10

Let $F, G: \mathcal{C} \to \mathcal{D}$ be functors and $\eta: F \Rightarrow G$ a natural transformation. Prove that $\eta$ is a natural isomorphism if and only if each component $\eta_A$ is an isomorphism in $\mathcal{D}$.

Difficulty 6/10

For the identity functor $\text{Id}: \text{Ab} \to \text{Ab}$ and the functor $A \mapsto A/2A$, describe what a natural transformation $\eta: \text{Id} \Rightarrow A/2A$ looks like and verify naturality.

Difficulty 8/10

Prove that the functor categories $[\mathcal{C}, \mathcal{D}]$ form a category by verifying that vertical composition of natural transformations is associative and has identities.

Common Mistakes

Common Mistake

If FA ≅ GA for every object A, then F and G are naturally isomorphic.

Having pointwise isomorphisms is not enough. The isomorphisms must be coherent — the naturality squares must commute. A non-natural isomorphism is just a lucky coincidence at each object, not a systematic relationship.

Every finite-dimensional vector space V is isomorphic to its dual V*, but the isomorphism requires a choice of basis and is not natural (there is no natural isomorphism Id ≅ (-)*).

Common Mistake

Natural transformations are only between functors of the same variance.

Natural transformations can be defined between any two functors with the same source and target category, regardless of whether they arose from covariant or contravariant constructions (after converting to covariant via opposite categories).

Common Mistake

Vertical and horizontal composition of natural transformations are the same operation.

Vertical composition composes transformations between the same pair of functors. Horizontal composition composes transformations across different functor compositions. They satisfy the interchange law: $(\nu \cdot \mu) * (\beta \cdot \alpha) = (\nu * \beta) \cdot (\mu * \alpha)$.

Quiz

Summary

A natural transformation $\eta: F \Rightarrow G$ is a family of morphisms $\{\eta_A: FA \to GA\}$ making all naturality squares commute.
Natural transformations are the morphisms of functor categories $[\mathcal{C}, \mathcal{D}]$, turning the collection of functors into a category itself.
A natural isomorphism has each component $\eta_A$ an isomorphism; this is the correct notion of 'the same' for functors.
The double dual $V \hookrightarrow V^{**}$ is a canonical example: it is natural (in contrast to $V \cong V^*$, which requires choosing a basis).
Vertical and horizontal composition of natural transformations satisfy the interchange law, giving $\text{Cat}$ a 2-categorical structure.

References

BookMac Lane, S. Categories for the Working Mathematician, 2nd ed. Springer, 1998. Ch. II.
BookRiehl, E. Category Theory in Context. Dover, 2016. Ch. 1.
WebsiteWikipedia — Natural transformation
WebsitenLab — natural transformation