ProofComplete

Differentiation of Measures - Key Proof

ProofProof Sketch of the Fundamental Theorem of Calculus (Part 2)

We prove that if F:[a,b]RF: [a, b] \to \mathbb{R} is absolutely continuous, then F(x)F(a)=axF(t)dtF(x) - F(a) = \int_a^x F'(t) \, dt.

Step 1 (F' exists a.e.): Since FF is absolutely continuous, it has bounded variation. By Lebesgue's theorem on monotone functions, FF is differentiable almost everywhere.

Step 2 (F' is integrable): We show FL1[a,b]F' \in L^1[a, b]. For any partition a=x0<x1<<xn=ba = x_0 < x_1 < \cdots < x_n = b: i=1nF(xi)F(xi1)Vab(F)<\sum_{i=1}^{n} |F(x_i) - F(x_{i-1})| \leq V_a^b(F) < \infty

where Vab(F)V_a^b(F) is the total variation. As the partition refines, this sum approaches abF(t)dt\int_a^b |F'(t)| \, dt (by definition of Lebesgue integral for the derivative). Thus FF' is integrable.

Step 3 (Main equality): Define G(x)=F(a)+axF(t)dtG(x) = F(a) + \int_a^x F'(t) \, dt. We must show G(x)=F(x)G(x) = F(x) for all xx.

By the first part of FTC, GG is absolutely continuous and G(x)=F(x)G'(x) = F'(x) a.e.

Define ϕ(x)=F(x)G(x)\phi(x) = F(x) - G(x). Then:

  • ϕ(a)=F(a)G(a)=0\phi(a) = F(a) - G(a) = 0
  • ϕ(x)=F(x)G(x)=0\phi'(x) = F'(x) - G'(x) = 0 a.e.

Step 4 (Zero derivative implies constant): We must show that if ϕ\phi is absolutely continuous on [a,b][a, b] with ϕ(a)=0\phi(a) = 0 and ϕ(x)=0\phi'(x) = 0 a.e., then ϕ(x)=0\phi(x) = 0 for all xx.

For any ϵ>0\epsilon > 0 and x[a,b]x \in [a, b], by absolute continuity there exists δ>0\delta > 0 such that for any finite collection of disjoint intervals {(ai,bi)}\{(a_i, b_i)\} with biai<δ\sum |b_i - a_i| < \delta: ϕ(bi)ϕ(ai)<ϵ\sum |\phi(b_i) - \phi(a_i)| < \epsilon

Step 5: Cover [a,x][a, x] by intervals so small that ϕ<ϵ|\phi'| < \epsilon on most of each interval (possible since ϕ=0\phi' = 0 a.e.). Using absolute continuity and the fact that ϕ=0\phi' = 0 a.e., we deduce: ϕ(x)ϕ(a)=ϕ(x)=axϕ(t)dt=0|\phi(x) - \phi(a)| = |\phi(x)| = \left|\int_a^x \phi'(t) \, dt\right| = 0

Thus ϕ(x)=0\phi(x) = 0 for all xx, so F(x)=G(x)=F(a)+axF(t)dtF(x) = G(x) = F(a) + \int_a^x F'(t) \, dt.

Remark

Key insights in the proof:

  1. Absolute continuity is essential: The proof crucially uses both parts of absolute continuity: differentiability a.e. and the ability to recover the function from its derivative.

  2. Measure zero sets: The fact that F=0F' = 0 outside a measure-zero set is sufficient because integration ignores measure-zero sets.

  3. Comparison to Riemann integral: For Riemann integration, the FTC requires continuous derivatives. Lebesgue integration allows FF' to be merely integrable, greatly expanding applicability.

  4. Monotone case first: The proof for monotone functions (which is simpler) uses the Vitali covering lemma. The general case for absolutely continuous functions builds on this.

  5. Connection to measures: The theorem can be rephrased as: the Radon-Nikodym derivative of the measure dμF(x)=dF(x)d\mu_F(x) = dF(x) with respect to Lebesgue measure is F(x)F'(x) almost everywhere.

This theorem completes the circle: differentiation and integration are inverse operations in the Lebesgue setting, with absolutely continuous functions playing the role of "primitives."