Zorn's Lemma

If every chain has an upper bound, a maximal element exists — light up the chains one by one in the divisibility poset of 12's divisors. Equivalent to the axiom of choice, the master key of algebra.

every chain has an upper bound⇒∃ maximal m\text{every chain has an upper bound} \Rightarrow \exists\, \text{maximal } m
1∣2∣4∣12,1∣3∣6∣121 \mid 2 \mid 4 \mid 12,\quad 1 \mid 3 \mid 6 \mid 12

Examples & Exercises

Prove that every nonempty finite poset has a maximal element (the finite version of Zorn's lemma — no choice axiom needed).

P≠∅ finite poset ⇒ ∃ m∈P maximalP \ne \varnothing\ \text{finite poset} \ \Rightarrow\ \exists\, m \in P\ \text{maximal}
The divisor poset with the chain C0={1,2,4,12}C_0 = \{1, 2, 4, 12\} highlighted: in this finite poset the proof's "keep climbing" stops at 12.
Show proof
  1. Let PP be a nonempty finite poset. Pick any x0∈Px_0 \in P. If x0x_0 is maximal, we are done. Otherwise some x1∈Px_1 \in P with x0<x1x_0 < x_1 exists (that is exactly what "x0x_0 is not maximal" means).
  2. Repeat: if xkx_k is not maximal, pick xk+1x_{k+1} with xk<xk+1x_k < x_{k+1}. The terms of x0<x1<x2<⋯x_0 < x_1 < x_2 < \cdots are all distinct (antisymmetry and transitivity give xi<xjx_i < x_j for i<ji < j). But PP is finite, so infinitely many distinct elements cannot be drawn.
  3. Hence the process must stop — some xkx_k is maximal. Finiteness making "climbing forever" impossible is the entire proof; for infinite posets this argument breaks, and Zorn's lemma (with the axiom of choice) is needed.

In P={1,2,3,4,6,12}P = \{1, 2, 3, 4, 6, 12\} ordered by divisibility, verify directly that every chain has an upper bound in PP, and find all maximal elements.

C chain ⇒ max⁡C exists and bounds C,maximal(P)={12}C\ \text{chain} \ \Rightarrow\ \max C\ \text{exists and bounds}\ C, \qquad \text{maximal}(P) = \{12\}
The chain C1={1,2,6,12}C_1 = \{1, 2, 6, 12\} is highlighted: its only upper bound is {12}\{12\}, and 12 is the unique maximal element of the whole poset.
Show proof
  1. Let C⊆PC \subseteq P be any chain. Since PP has only 6 elements, CC is finite, and a finite chain has a greatest element max⁡C\max C (a finite totally ordered subset). Every c∈Cc \in C satisfies c∣max⁡Cc \mid \max C, so max⁡C∈P\max C \in P is an upper bound of CC. Every chain has an upper bound in PP.
  2. Concretely, for the three maximal chains: C0={1,2,4,12}C_0 = \{1,2,4,12\} has upper bounds {12}\{12\}, C1={1,2,6,12}C_1 = \{1,2,6,12\} has {12}\{12\}, C2={1,3,6,12}C_2 = \{1,3,6,12\} has {12}\{12\} (an upper bound of a chain containing 12 must be a multiple of 12, and the only multiple of 12 in PP is 12). Shorter chains work the same way: e.g. {1,2}\{1, 2\} has upper bounds {2,4,6,12}\{2, 4, 6, 12\} and {1,3}\{1, 3\} has {3,6,12}\{3, 6, 12\}.
  3. Maximal elements: x∈Px \in P is maximal if no element of PP is strictly above it (in divisibility). Since 1∣21 \mid 2, 2∣42 \mid 4, 3∣63 \mid 6, 4∣124 \mid 12, 6∣126 \mid 12, none of 1,2,3,4,61, 2, 3, 4, 6 is maximal. The only multiple of 12 in PP is 12 itself, so 12 is maximal — and unique. Exactly as Zorn's lemma promises.

Prove Zorn's lemma implies the axiom of choice: for a family (Xi)i∈I(X_i)_{i \in I} of nonempty sets, apply Zorn's lemma to the poset of partial choice functions. (The converse AC ⇒\Rightarrow Zorn is not proved here — it needs transfinite recursion, an idea-level argument.)

P={f:dom⁡f⊆I, f(i)∈Xi},f≤g  ⟺  g extends fP = \{f : \operatorname{dom} f \subseteq I,\ f(i) \in X_i\}, \quad f \le g \iff g\ \text{extends}\ f
The chain C2={1,3,6,12}C_2 = \{1, 3, 6, 12\} is highlighted: Zorn's lemma says "if every chain has an upper bound, a maximal element exists", and the proof below applies exactly this to the problem of finding a choice function.
Show proof
  1. Fix a family (Xi)i∈I(X_i)_{i \in I} of nonempty sets. Order the set P={f:dom⁡f⊆I, f(i)∈Xi}P = \{f : \operatorname{dom} f \subseteq I,\ f(i) \in X_i\} of partial choice functions by extension: f≤gf \le g means dom⁡f⊆dom⁡g\operatorname{dom} f \subseteq \operatorname{dom} g and gg agrees with ff on dom⁡f\operatorname{dom} f. PP is nonempty (the empty function belongs to it).
  2. For any chain C\mathcal{C} in PP, let u=⋃Cu = \bigcup \mathcal{C}. The chain condition makes uu a function: if f,g∈Cf, g \in \mathcal{C} are both defined at ii, the larger extends the smaller, so f(i)=g(i)f(i) = g(i). With dom⁡u=⋃f∈Cdom⁡f\operatorname{dom} u = \bigcup_{f \in \mathcal{C}} \operatorname{dom} f and u(i)∈Xiu(i) \in X_i, we get u∈Pu \in P and f≤uf \le u for all f∈Cf \in \mathcal{C} — uu is an upper bound of the chain.
  3. By Zorn's lemma, PP has a maximal element mm. We show dom⁡m=I\operatorname{dom} m = I: otherwise some i0∈I∖dom⁡mi_0 \in I \setminus \operatorname{dom} m exists, and since Xi0≠∅X_{i_0} \ne \varnothing, picking x0∈Xi0x_0 \in X_{i_0} extends mm to m′m' with m′(i0)=x0m'(i_0) = x_0. Then m<m′m < m', contradicting maximality. So mm is a choice function defined on all of II — the axiom of choice holds.
  4. Note: the converse (AC ⇒\Rightarrow Zorn) needs transfinite recursion along ordinals to grow a chain into a maximal one; it stays an idea-level statement here — the Zorn ⇒\Rightarrow AC direction above is proved completely.

1. Number Systems