Homepage Solution manuals Terence Tao An Introduction to Measure Theory Exercise 1.3.6 (Measurability of a function and the area under its graph)

Exercise 1.3.6 (Measurability of a function and the area under its graph)

Let f : d [0,+) be an unsigned Lebesgue measurable function. Show that the region under its graph is a Lebesgue measurable subset of d+1.

Answers

Notice that we can represent the area under the graph as the union of the two following sets:

{(x,t) d × : 0 t f(x)} = {(x,t) d × : 0 t < f(x)} {(x,t) d × : t = f(x)}

We argue that the former is Lebesgue measurable, while the latter is a null set.

  • {(x,t) d × : 0 t < f(x)}
    Since f is an unsigned Lebesgue measurable function, by Lemma 1.3.9 (iv) we can find an increasing sequence (fn)n of unsigned simple functions fn fn+1 < f such that f(x) = sup {fn(x) : n } for all x d. Consider the areas under the curves of these simple functions:

    {(x,t) d × : 0 t f n(x)}.

    This set is Lebesgue measurable for any n since it can be represented as a union of measurable sets

    {(x,t) d × : 0 t f n(x)} = i=1m {(x,t) E i × : 0 t ci} = i=1mE i×[0,ci]

    where c1,,cm are the finite number of values that fn takes on the corresponding Lebesgue measurable sets E1,,Em. In other words, the theorem is at least true for simple functions. The trick is to represent the area under f as the countable union of the Lebesgue measurable areas under the simple functions:

    n {(x,t) d × : 0 t f n(x)} = {(x,t) d × : 0 t < f(x)}.

    It is easy to verify that this is indeed the case:

    • From fn < f we trivially have

      n : {(x,t) d × : 0 t f n(x)} {(x,t) d × : 0 t f(x)}

    • Pick (x,t) from the right-hand side, i.e., t f(x). Set 𝜖 := f(x) t, and find an n such that f(x) fn(x) 𝜖. Then, (x,t) {(x,t) d × : 0 t fn(x)} since fn(x) f(x) 𝜖 = t.

    Thus, the strict area under the graph is Lebesgue measurable by Lemma 1.2.13 (vi).

  • {(x,f(x)) : x d}
    Assume that f : d is measurable. We prove that {(x,f(x)) : x d} is a Lebesgue null set.It is enough to prove the claim when f : B is measurable and B d is a box. Then, since d is a countable union of boxes, the claim will follow. Let G := {(x,f(x)) : x B}, and for n , let Fn := G (B × [n,n + 1]). It is enough to prove that Fn is a null set for every n , so without loss of generality, we show it for the unit interval product F := F0.Fix some k , and let I1,,Ik1 be a sequence of intervals [0, 1 k],[1 k, 2 k],,[k1 k ,1]. Then, for any j = 0,,k 1 we have

    m(G (B × I j)) m(f1(I j) × Ij) m(f1(I j)) ×1 k.

    which implies, by Exercise 1.2.22 and some basic set theory

    m(F) = m ( j=0k1G (B × I j)) = j=0k1m(G (B × I j)) j=0k1m(f1(I j) × Ij)

    At this point we use the assumption that f is Lebesgue measurable, and apply Lemma 1.3.9 (ix) which states that the inverse must remain Lebesgue measurable

    j=0k1m(f1(I j)) ×1 k = 1 km ( j=0k1f1(I j)) = 1 k × m(f1([0,1])) 1 k × m(B)

    Since k was arbitrary, we can take k and we are done.

User profile picture
2020-08-30 00:00
Comments
  • This is overcomplicating, simply using (4) of Lemma 7 is sufficient.
    isn2025-01-28