AI summary by TwinMind
Lecture 6: The Double Dual and the Outer Measure of a Subset of Real Numbers
MIT OpenCourseWare
1h 20m 58s
Hahn-Banach Theorem Recap and Applications
- Theorem statement: If V is normed space, M is subspace of V, u is bounded linear functional on M (satisfying |u(t)| ≤ constant × ||t|| for all t ∈ M), then exists bounded linear functional U on whole space V extending u with same bound
- Proof method: Used Zorn's lemma (minor error in previous lecture notes, corrected in online notes)
- Key application: For any non-zero v ∈ V, exists bounded linear functional f ∈ V' such that ||f|| = 1 and f(v) = ||v||
- Proof: Define u on span(v) by u(λv) = λ||v||, apply Hahn-Banach to extend
Double Dual and Reflexivity
- Double dual definition: V'' = (V')' = space of bounded linear functionals on V'
- Canonical embedding: For fixed v ∈ V, define T_v: V' → ℂ by T_v(v') = v'(v)
- T_v is linear and bounded: |T_v(v')| ≤ ||v'|| · ||v||
- Isometry theorem: Map T: V → V'' defined by T(v) = T_v is isometric (||T_v|| = ||v||)
- Proof uses previous application of Hahn-Banach theorem
- Reflexivity: Normed space V is reflexive if canonical map V → V'' is onto
- Examples: l^p spaces reflexive for 1 < p < ∞
- Non-reflexive: l^1 (dual is l^∞, but (l^∞)' ≠ l^1), c_0 (dual is l^1, but (l^1)' = l^∞ ≠ c_0)
Motivation for Lebesgue Measure and Integration
- Problems with Riemann integration:
- Limited convergence theorems (only uniform convergence)
- L^1_R (Riemann integrable functions on [0,1]) with norm ||f||_1 = ∫|f|dx not complete (not Banach space)
- Need complete spaces for functional analysis applications
- Solution: Develop Lebesgue measure and integration theory
Lebesgue Measure Construction Plan
- Desired properties for measure m:
- Defined on all subsets of ℝ (impossible)
- m(interval) = length of interval
- Countable additivity: m(⋃E_n) = Σm(E_n) for disjoint sets
- Translation invariance: m(x + E) = m(E)
- Impossibility: No function satisfying all four properties exists
- Compromise: Drop requirement 1, define m on large class of "measurable" sets
Outer Measure Construction
- Definition: For A ⊆ ℝ, outer measure m*(A) = inf{Σl(I_n) : I_n open intervals, A ⊆ ⋃I_n}
- Basic properties:
- m*({single point}) = 0
- m*(countable set) = 0 (proved in detail)
- Monotonicity: A ⊆ B ⟹ m(A) ≤ m(B)
- Subadditivity: m(⋃A_n) ≤ Σm(A_n) (proved in detail using ε-argument)
- Translation invariance: Will be homework exercise
- Next step: Show m*(interval) = length of interval, then restrict to measurable sets
