Skip to content
Part IA Michaelmas, Lent Term

Sections, Retractions, and Idempotents

Section-Retraction Pairs

Definition

A section-retraction pair is a pair of functions s:BAs : B \to A and r:ABr : A \to B such that:

rs=idBr \circ s = \text{id}_B

  • ss is called the section (or left inverse)
  • rr is called the retraction (or right inverse)

Intuition

  • ss embeds BB into AA
  • rr projects back from AA to BB
  • Composing rr after ss returns to where we started in BB

Section-retraction pair diagram

Examples

  • BABB \hookrightarrow A \twoheadrightarrow B where ss is inclusion and rr is identity on BB
  • s:NZs : \mathbb{N} \to \mathbb{Z}, s(n)=ns(n) = n; r:ZNr : \mathbb{Z} \to \mathbb{N}, r(n)=nr(n) = |n| fails since r(s(1))r(s(-1)) is undefined
  • Correct: s:NZs : \mathbb{N} \to \mathbb{Z}, s(n)=ns(n) = n; r:ZNr : \mathbb{Z} \to \mathbb{N}, r(n)=max(n,0)r(n) = \max(n, 0)

Properties

Section is Injective

If ss is a section of rr, then ss is injective.

Proof: If s(b1)=s(b2)s(b_1) = s(b_2), then b1=r(s(b1))=r(s(b2))=b2b_1 = r(s(b_1)) = r(s(b_2)) = b_2.

Retraction is Surjective

If rr is a retraction of ss, then rr is surjective.

Proof: For any bBb \in B, b=r(s(b))b = r(s(b)), so bb is in the image of rr.

Converse Requires Choice

If s:BAs : B \to A is injective, there exists a retraction r:ABr : A \to B (on image of ss) iff we can “pick” values for elements outside the image.

This requires the Axiom of Choice in general.

Idempotents

Definition

A function e:AAe : A \to A is idempotent if ee=ee \circ e = e.

Intuition

Applying ee once or twice gives the same result. ee “freezes” elements to a fixed form.

Examples

  • Projection (x,y)(x,0)(x, y) \mapsto (x, 0) is idempotent
  • Identity is idempotent
  • Constant function xcx \mapsto c is idempotent iff c=cc = c (always true)

Examinable Proof: Idempotents and Section-Retraction Pairs

Theorem: There is a bijection between:

  • Section-retraction pairs (s,r)(s, r) with s:BAs : B \to A, r:ABr : A \to B, rs=idBr \circ s = \text{id}_B
  • Idempotents e:AAe : A \to A

From Section-Retraction to Idempotent

Given (s,r)(s, r) with rs=idBr \circ s = \text{id}_B, define e=sr:AAe = s \circ r : A \to A.

Idempotence:

ee=srsr=sidBr=sr=ee \circ e = s \circ r \circ s \circ r = s \circ \text{id}_B \circ r = s \circ r = e

From Idempotent to Section-Retraction

Given idempotent e:AAe : A \to A, define:

  • B=Im(e)={e(a):aA}B = \text{Im}(e) = \{e(a) : a \in A\}
  • r:ABr : A \to B by r(a)=e(a)r(a) = e(a)
  • s:BAs : B \hookrightarrow A the inclusion

Verification: r(s(b))=r(b)=e(b)r(s(b)) = r(b) = e(b).

If bBb \in B, then b=e(a)b = e(a) for some aa. So e(b)=e(e(a))=e(a)=be(b) = e(e(a)) = e(a) = b by idempotence.

Thus rs=idBr \circ s = \text{id}_B.

These are Inverses

Starting with (s,r)(s, r): the associated idempotent srs \circ r has image isomorphic to BB (via ss).

Starting with ee: the section-retraction pair recovers ee as srs \circ r.

Splitting Idempotents

The process of going from ee to (s,r)(s, r) is called splitting the idempotent.

The set B=Im(e)B = \text{Im}(e) is the “fixed points” of ee, and ee is the identity on BB.


Summary

  • Section-retraction: rs=idBr \circ s = \text{id}_B
  • Sections are injective; retractions are surjective
  • Idempotent: ee=ee \circ e = e
  • Bijection: (s,r)sr(s, r) \mapsto s \circ r and e(inclusion,eIm(e))e \mapsto (\text{inclusion}, e|_{\text{Im}(e)})