Skip to Content
Math4121Introduction to Lebesgue Integration (Lecture 14)

Math4121 Lecture 14

Recap

Hankel developed Riemann’s integrability criterion

Definition: Oscillation

Given an interval I[a,b]I\subset[a,b], f:[a,b]Rf:[a,b]\to\mathbb{R} the oscillation of ff at II is

ω(f,I)=supIfinfIf\omega(f,I) = \sup_I f - \inf_I f

Theorem 2.5: Riemann’s Integrability Criterion

A bounded function ff is Riemann integrable if and only if for every σ,ϵ>0\sigma,\epsilon>0 there exists a partition PP of [a,b][a,b] such that

iPΔxi<ϵ\sum_{i\in \mathcal{P}}\Delta x_i<\epsilon

where P={i:ω(f,Ii)>σ}\mathcal{P}=\{i:\omega(f,I_i)>\sigma\}.

Proof:

To prove Riemann’s Integrability Criterion, we need to show that a bounded function ff is Riemann integrable if and only if for every σ,ϵ>0\sigma, \epsilon > 0, there exists a partition PP of [a,b][a, b] such that the sum of the lengths of the intervals where the oscillation exceeds σ\sigma is less than ϵ\epsilon.

QED

Proposition 2.4

For point c[a,b]c\in[a,b], define the oscillation at cc as

ω(f,c)=infcIω(f,I)\omega(f,c) = \inf_{c\in I}\omega(f,I)

Homework question 6: ff is continuous at cc if and only if ω(f,c)=0\omega(f,c)=0.

So we can restate the previous theorem as:

Given σ>0\sigma>0, define Sσ={c[a,b]:ω(f,c)>σ}S_\sigma=\{c\in[a,b]:\omega(f,c)>\sigma\}.

Restate the theorem as:

fR[a,b]f\in\mathscr{R}[a,b] if and only if for every σ,ϵ>0\sigma,\epsilon>0 there exists intervals I1,I2,,InI_1,I_2,\cdots,I_n such that Sσi=1nIiS_\sigma\subset\bigcup_{i=1}^{n}I_i and i=1n(Ii)<ϵ\sum_{i=1}^{n}\ell(I_i)<\epsilon. where (I)\ell(I) is the length of the interval II.

Definition: Outer content

Given a set SS, a finite cover of SS is a collection of intervals C={I1,I2,,In}C=\{I_1,I_2,\cdots,I_n\} such that Si=1nIiS\subseteq\bigcup_{i=1}^{n}I_i.

The length of the cover CC is (C)=i=1n(Ii)\ell(C)=\sum_{i=1}^{n}\ell(I_i).

The outer content of SS is

ce(S)=infcCs(C)c_e(S) = \inf_{c\in C_s}\ell(C)

where CsC_s is the set of all finite covers of SS.

Example:

S={x1,,xn}S=\{x_1,\ldots,x_n\}, then ce(S)=0c_e(S)=0.

  • Let Ii=(xiϵ2n,xi+ϵ2n)I_i=(x_i-\frac{\epsilon}{2n},x_i+\frac{\epsilon}{2n}), so i=1n(Ii)=ϵ\sum_{i=1}^{n}\ell(I_i)=\epsilon

S={1n}n=1S=\{\frac{1}{n}\}_{n=1}^{\infty}, then ce(S)=0c_e(S)=0.

  • In this case, we can only use finite cover, however, there is only one “accumulation point”, so we can cover it with a single interval, and the remaining points can be covered by finite intervals. (for any ϵ>0\epsilon>0, we can construct a finite cover with length ϵ\epsilon that covers all points.)

S=Q[0,1]S=\mathbb{Q}\cap[0,1], then ce(S)=1c_e(S)=1.

  • In this case, by covering the interval with [0,1][0,1], we can get the length of the cover is at most 1.
  • Suppose there exists a cover CC with IC(I)<1\sum_{I\in C}\ell(I)<1, then there must be a gap in the intervals, however, since the Q\mathbb{Q} is dense in R\mathbb{R}, there must be a point in the gap, which is a contradiction.

Theorem 2.5: Hankel’s criterion for Riemann integrability

A function fR[a,b]f\in\mathscr{R}[a,b] if and only if ce(Sσ)=0c_e(S_\sigma)=0 for all σ>0\sigma>0.

The idea is that if the oscillation of a function can be bounded by a finite set that the total length is small, then the function is Riemann integrable.

Hankel’s idea was to apply this theorem to determining how discontinuous a function could be a Riemann integrable function.

A set SS is dense in XX if every point of XX is a limit point of SS.

Definition: Totally discontinuous

ff is totally discontinuous if the points of continuity of ff are not dense.

For example, f(x)={0xQ1xQf(x)=\begin{cases} 0 & x\in\mathbb{Q}\\ 1 & x\notin\mathbb{Q} \end{cases} is totally discontinuous.

Definition: Pointwise discontinuity

ff is pointwise discontinuous if they are dense in [a,b][a,b].

Hankel’s conjecture: ff is pointwise discontinuous, then ff is integrable.

Last updated on