Category theory MOC

Adjoint functor

An adjunction of functors is an adjunction in β„­π”žπ”±. #m/def/cat Let 𝖣, 𝖒 be categories. A pair of functors 𝐹 :𝖣 ⇆𝖒 :π‘ˆ form an adjunction, written

https://q.uiver.app/#q=WzAsMixbMCwwLCJcXG1hdGhzZiBEIl0sWzIsMCwiXFxtYXRoc2YgQyJdLFswLDEsIkYiLDIseyJjdXJ2ZSI6MX1dLFsxLDAsIlUiLDIseyJjdXJ2ZSI6MX1dLFszLDIsIiIsMix7ImxldmVsIjoxLCJzdHlsZSI6eyJuYW1lIjoiYWRqdW5jdGlvbiJ9fV1d

or compactly 𝐹 βŠ£π‘ˆ :𝖣 →𝖒, iff there is a natural isomorphism in 𝖲𝖾𝗍(𝖒𝐨𝐩×𝖣) of hom-sets1

πœ‘:𝖣(𝐹×1𝖣)≅𝖒(1π–’Γ—π‘ˆ):πœ‘βˆ’1πœ‘πΆ,𝐷:𝖣(𝐹𝐢,𝐷)≅𝖒(𝐢,π‘ˆπ·):πœ‘βˆ’1𝐢,𝐷.

When adjoints exist they are unique up to natural isomorphism, hence we call 𝐹 the left adjoint of π‘ˆ, and π‘ˆ the right adjoint of 𝐹.

Proof of uniqueness

By duality, it suffices to prove right adjoints are unique up to natural isomorphism. Suppose 𝐹 βŠ£π‘ˆ,𝑉 :𝖣 ⇆𝖒. Then by adjunction

𝖒(1Γ—π‘ˆ)≅𝖣(𝐹×1)≅𝖒(1×𝑉)

hence for any object 𝐷 βˆˆπ–£ we have

γ‚ˆ(π‘ˆπ·)β‰…γ‚ˆ(𝑉𝐷)

naturally and thus

π‘ˆπ·β‰…π‘ˆπ‘‰

naturally (see Yoneda embedding).

The name comes from an analogy to the Adjoint operator. In the archetypal examples, we think of π‘ˆ as forgetful and 𝐹 as free β€” See Free-forgetful adjunction.

Unit and coΓΌnit

We can equivalently rephrase the condition for an adjunction in terms of a unit or coΓΌnit, so named since they form the corresponding data for a monad or comonad induced by the adjunction respectively.

To see that either of these are necessary and sufficient, note2

πœ‚πΆ=πœ‘πΆ,𝐹𝐢(1𝐹𝐢)πœ‘βˆ’1(𝑓)=π‘“β™―πœ–π·=πœ‘βˆ’1π‘ˆπ·,𝐷(1π‘ˆπ·)πœ‘(𝑔)=𝑓♭

This gives us another perspective on adjunctions: They are a weakening of Equivalence of categories.

Properties


#state/tidy | #lang/en | #SemBr

Footnotes

  1. 2010. Category theory, Β§9 ↩

  2. 2020. From categories to homotopy theory, p. 40 ↩