AI summary by TwinMind
Lecture 5: Complete Metric Spaces
MIT OpenCourseWare
1h 10m 48s
Course Recap and Introduction
- Day 1: Metric space definition and examples
- Day 2: General theory (sequences, Cauchy sequences) - condensed from full real analysis coverage
- Previous work enables stating/proving theorems for metric spaces via analogy
- Compact metric spaces exception - not analogous to Euclidean case (closed and bounded)
- Four-part compactness theorem: topologically compact = sequentially compact = totally bounded and Cauchy complete
- Focus today: complete metric spaces applications
Lipschitz Functions
- Definition: function f: X→Y such that d(f(x), f(y)) ≤ K·d(x,y) for some K ∈ ℝ
- K values: if K < 0, positive definiteness implies everything zero; if K > 0, requires more work
- Theorem: Lipschitz implies continuous
- Proof: let ε > 0, choose δ = ε/K
- d(f(x), f(y)) ≤ K·d(x,y) < K·δ = ε
- Stronger result: Lipschitz implies uniform continuity
- Delta doesn't depend on x, only on f and ε
Uniform Continuity
- Definition: f: X→Y uniformly continuous if ∀ε > 0, ∃δ > 0 such that d(x,y) < δ ⟹ d(f(x), f(y)) < ε
- Key difference from continuity: δ independent of point x
- Proposition: If f: X→Y continuous and X compact, then f uniformly continuous
- Proof uses Lebesgue number lemma
- Cover X with balls of radius δc around each point c
- By compactness, ∃δ > 0 such that ∀x, ball of radius δ around x fits in some cover ball
- Triangle inequality completes proof
Integral Operator Application
- Proposition: Let f: [a,b]×[c,d]→ℝ be continuous
- Define G(y) = ∫ₐᵇ f(x,y)dx
- Then G is continuous
- Proof: Use uniform continuity to swap limit and integral
- lim_{n→∞} G(yₙ) = lim_{n→∞} ∫ₐᵇ f(x,yₙ)dx = ∫ₐᵇ f(x,y)dx = G(y)
Contractions and Fixed Points
- Definition: Contraction is Lipschitz function f: X→X with 0 ≤ K < 1
- Fixed Point: x such that f(x) = x
- Geometric intuition: distances between image points shrink by factor K < 1
Banach Fixed Point Theorem (Contraction Mapping Theorem)
- Theorem: Let X be non-empty Cauchy complete metric space, f: X→X contraction
- Then ∃ unique fixed point
- Proof:
- Pick arbitrary x₀ ∈ X
- Define sequence: xₙ₊₁ = f(xₙ)
- Show sequence is Cauchy: d(xᵢ₊₁, xᵢ) ≤ Kⁱ·d(x₁, x₀)
- d(xₙ, xₘ) ≤ Σᵢ₌ₙᵐ⁻¹ Kⁱ·d(x₁, x₀) = Kⁿ/(1-K)·d(x₁, x₀)
- Since K < 1, can make arbitrarily small for large n
- By completeness, ∃x such that xₙ → x
- x = lim f(xₙ₋₁) = f(lim xₙ₋₁) = f(x) (by continuity)
- Uniqueness: if y = f(y), then d(x,y) = d(f(x), f(y)) ≤ K·d(x,y)
- Since K < 1, implies d(x,y) = 0, so x = y
Differential Equations Application
- Example: For λ ∈ ℝ, f,g continuous on [a,b], k continuous on [a,b]×[a,b]
- Consider Tf(x) = g(x) + λ∫ₐᵇ k(x,y)f(y)dy
- Question: for which λ is T a contraction?
- Analysis: d(Tf₁, Tf₂) ≤ |λ|·sup|k(x,y)|·(b-a)·d(f₁, f₂)
- Contraction when |λ| < 1/(C·(b-a)) where C = sup|k(x,y)|
- By Banach theorem: ∃ unique f such that Tf = f
- Leads to existence/uniqueness for differential equations
Completions of Metric Spaces
- Motivation: ℝ is completion of ℚ
- Three construction methods for ℝ from ℚ:
- Dedekind cuts
- Least upper bound property
- Equivalence classes of Cauchy sequences
- Focus on equivalence classes method
Equivalence Classes Construction
- Define equivalence: Cauchy sequences {aₙ} ~ {bₙ} if d(aₙ, bₙ) → 0
- Verify equivalence relation:
- Reflexive: d(aₙ, aₙ) = 0
- Symmetric: d(aₙ, bₙ) = d(bₙ, aₙ)
- Transitive: triangle inequality
- ℝ = set of equivalence classes of Cauchy sequences in ℚ
General Completion Theorem
- Theorem: For any metric space (M,d), ∃ completion (M̄, d̄) such that:
- M ⊆ M̄
- d̄ restricts to d on M
- M̄ is Cauchy complete
- M̄ = closure of M
- Proof strategy: Embed M in C^∞(M) (continuous bounded functions)
- For fixed m' ∈ M, define Gₘ(p) = d(p,m) - d(p,m')
- This gives isometric embedding: d(m₁,m₂) = d_∞(Gₘ₁, Gₘ₂)
- Take closure of embedded M in complete space C^∞(M)
Key Lemma: C^∞(M) is Complete
- C^∞(M) = space of continuous bounded functions on M
- Metric: d_∞(f,g) = sup_{x∈M} |f(x) - g(x)|
- Lemma: C^∞(M) is Cauchy complete
- Lemma: Closed subset of complete metric space is complete
Applications of Completions
- Banach spaces: Completion of normed vector spaces (major topic in 18.102)
- Hilbert spaces: Completion of inner product spaces
- Lebesgue integration: Riemann integrable functions have "holes"
- Example: indicator function on rationals not Riemann integrable
- Complete C₀ᶜ(ℝ) under L¹ metric: d(f,g) = ∫|f(x) - g(x)|dx
- Completion gives L¹ functions with Lebesgue integral
- Generalizes to Lᵖ spaces: complete under Iₚ metric
- Lᵖ spaces are Cauchy complete norm spaces
Conceptual Framework
- Metric spaces allow viewing functions as points
- Functions become objects to manipulate rather than input-output machines
- Equivalence classes of Cauchy sequences: fundamental construction principle
- Applications span analysis, differential equations, functional analysis
