AI summary by TwinMind
Lecture 5: Zorn’s Lemma and the Hahn-Banach Theorem
MIT OpenCourseWare
1h 10m 24s
Introduction and Theorem Statement
- Proving Hahn-Banach theorem: extends bounded linear functionals from subspace to entire normed space
- Answers whether dual space (all bounded linear functionals) is non-trivial for every norm space
- Key tool: Zorn's Lemma from set theory
Zorn's Lemma
- Statement: If every chain in non-empty partially ordered set E has upper bound, then E has maximal element
- Partially ordered set: set with relation like ≤ satisfying three properties
- Chain: subset where any two elements can be compared (one ≤ other)
- Maximal element: element not ≤ anything other than itself (not necessarily upper bound)
Warm-up: Hamel Basis Theorem
- Hamel basis of vector space V: linearly independent set H where every element of V is finite linear combination of elements from H
- Example: {(1,0), (0,1)} is Hamel basis for R²
- Theorem: Every vector space has Hamel basis
Proof Strategy Using Zorn's Lemma
- Define E = set of all linearly independent subsets of V
- Partial order on E by inclusion: e ≤ e' iff e ⊆ e'
- For chain C in E, define c = ⋃(all e ∈ C)
- Show c is linearly independent using chain property
- Apply Zorn's Lemma to get maximal element H
- Prove H spans V by contradiction: if not, can extend H to larger linearly independent set
Hahn-Banach Theorem Statement
- Given: V normed space, M subspace of V, u: M → ℂ linear with |u(t)| ≤ c‖t‖ for all t ∈ M
- Conclusion: ∃ continuous extension U: V → ℂ with same bound, U|_M = u
Key Lemma for Hahn-Banach
- Given: Same assumptions plus x ∉ M
- Conclusion: Can extend u to M' = M + span{x} with same bound constant
- M' consists of elements t + ax where t ∈ M, a ∈ ℂ
Proof Strategy for Hahn-Banach
- Apply Zorn's Lemma to all continuous extensions of u
- Get maximal element U on subspace N
- Use Lemma to show N = V (if not, can extend further, contradicting maximality)
Hahn-Banach Proof Details
- Define E = {(V,N) : N subspace, M ⊆ N, V continuous extension of u to N}
- Partial order: (V₁,N₁) ≤ (V₂,N₂) if N₁ ⊆ N₂ and V₂|_{N₁} = V₁
- For chain C, define N = ⋃(all Nᵢ), V on N by V(t) = Vᵢ(t) if t ∈ Nᵢ
- Show V well-defined using chain property
- Apply Zorn's Lemma, get maximal (U,N)
- If N ≠ V, take x ∉ N, extend using Lemma, contradicting maximality
Lemma Proof Setup
- Need to show every element in M' = M + span{x} has unique representation t + ax
- Uniqueness: if t + ax = t̃ + ãx, then (a-ã)x = t̃-t ∈ M
- If a ≠ ã, then x ∈ M (contradiction), so a = ã, thus t = t̃
Extension Construction
- Define U'(t + ax) = u(t) + aλ for chosen λ ∈ ℂ
- WLOG assume bound constant c = 1
- Need |u(t) + aλ| ≤ ‖t + ax‖ for all t ∈ M, a ∈ ℂ
- Reduces to: |u(t) - λ| ≤ ‖t - x‖ for all t ∈ M
Choosing λ Parameter
- First choose real part α of λ
- Define w(t) = Re(u(t)) (real part of u(t))
- Note: w(t) ≤ |u(t)| ≤ ‖t‖
- Use triangle inequality and supremum arguments
- Show: sup_{t∈M}[w(t) - ‖t-x‖] ≤ inf_{t∈M}[w(t) + ‖t-x‖]
- Choose α between these bounds
- Similar process for imaginary part completes proof
