Questions: Natural Isomorphisms and Universal Constructions

Question 1

For every finite-dimensional vector space V over ℝ, V ≅ V* (the dual space). A student concludes that the identity functor and the dual functor (V ↦ V*) are naturally isomorphic. What is wrong with this reasoning?

• A. V and V* are not isomorphic for any infinite-dimensional spaces, so a global natural isomorphism cannot exist
• B. Pointwise isomorphism is not sufficient for natural isomorphism; the isomorphisms V ≅ V* require choosing a basis and cannot be made to commute with all linear maps simultaneously
• C. The dual functor is contravariant, so no natural transformation from the identity functor to the dual functor can exist at all
• D. Natural isomorphisms require the domain and codomain categories to be the same, which fails here

Answer: (B)

For each finite-dimensional V, you can find an isomorphism φ_V: V → V*, but doing so requires choosing a basis (mapping basis vectors to dual basis vectors). Natural isomorphism additionally requires that for every linear map f: V → W, the square G(f) ∘ φ_V = φ_W ∘ F(f) commutes — but no basis-dependent choice satisfies this for all linear maps simultaneously. This is the canonical example of pointwise isomorphisms that fail to be natural: the individual isomorphisms do not cohere across morphisms. (The canonical map V → V** is natural; V → V* is not.)

Question 2

The universal property of a product A × B is stated as a natural isomorphism Hom(X, A × B) ≅ Hom(X, A) × Hom(X, B), natural in X. What does naturality in X actually guarantee?

• A. That A × B is the largest object admitting morphisms into both A and B
• B. That the bijection between morphisms into the product and pairs of morphisms is preserved under pre-composition with any g: W → X — the correspondence works coherently for all objects simultaneously, not just pointwise
• C. That A × B is unique up to isomorphism in the category
• D. That the projection maps π_A and π_B are themselves isomorphisms

Answer: (B)

Naturality in X means: for any morphism g: W → X, composing a morphism h: X → A × B with g gives a morphism W → A × B that corresponds — via the natural isomorphism — to composing each component with g. This coherence is what makes the universal property robust: it doesn't just say 'morphisms happen to biject at each X' but that the bijection is functorially compatible. Without naturality, you might have pointwise bijections that give inconsistent results when morphisms are composed, destroying the structural content.

Question 3

Two objects that both satisfy the same universal property in a category are related by a unique isomorphism — there is exactly one isomorphism between them compatible with the universal property.

• T. True
• F. False

Answer: True

This is the fundamental consequence of universal properties: uniqueness up to unique isomorphism. If U and U' both satisfy the same universal property, applying U's universal property to the data presented by U' gives a morphism u: U → U', and vice versa gives u': U' → U. The uniqueness clause forces u ∘ u' = id and u' ∘ u = id, so u is an isomorphism, and it is the unique one compatible with the structure. The natural isomorphism of hom-functors picks out this canonical isomorphism, distinguishing it from arbitrary isomorphisms that may exist between the objects.

Question 4

Two functors F and G are naturally isomorphic if and only if F(X) ≅ G(X) for most object X in the domain category.

• T. True
• F. False

Answer: False

Pointwise isomorphism — F(X) ≅ G(X) at each object — is necessary but not sufficient for natural isomorphism. Natural isomorphism additionally requires that the isomorphisms η_X: F(X) → G(X) commute with all morphisms: for every f: X → Y, we need G(f) ∘ η_X = η_Y ∘ F(f). The canonical counterexample is finite-dimensional vector spaces and their duals: V ≅ V* holds at every object, but no natural transformation witnesses this, because satisfying the naturality square requires choosing a basis. Pointwise isomorphisms with no coherence are structurally meaningless.

Question 5

What is the difference between saying 'F(X) and G(X) are isomorphic for every object X' and saying 'F and G are naturally isomorphic'? Why does the stronger condition matter?

Model answer: Pointwise isomorphism says that at each object X there exists some isomorphism between F(X) and G(X), chosen independently with no coherence requirements across different X. Natural isomorphism requires that the isomorphisms η_X: F(X) → G(X) satisfy the naturality condition: for every morphism f: X → Y, the square G(f) ∘ η_X = η_Y ∘ F(f) commutes. This coherence means the identification of F with G is canonical — it does not depend on arbitrary choices and works consistently across the entire category, not just object by object.

The distinction matters because non-natural isomorphisms depend on choices (like bases) that are invisible to the categorical structure and break functoriality. Natural isomorphisms capture the idea that two functors 'do the same thing' in a way that respects morphisms throughout the category. Universal properties expressed as natural isomorphisms of hom-functors inherit this coherence, which is why objects satisfying universal properties have canonical (not merely arbitrary) isomorphisms between them — the 'unique up to unique isomorphism' principle.