Homepage Solution manuals Terence Tao An Introduction to Measure Theory Exercise 1.7.20 (Product of Dirac measures and product of counting measures)

Exercise 1.7.20 (Product of Dirac measures and product of counting measures)

Let (X,X), (Y,Y ) be measurable spaces.

(i)
Show that the product of two Dirac measures on (X,X), (Y,Y ) is a Dirac measure on (X × Y,X ×Y ).
(ii)
If X,Y are at most countable, show that the product of the two counting measures on (X,X), (Y,Y ) is the counting measure on (X × Y,X ×Y ).

Answers

(i)
The product of two Dirac measures is a Dirac measure.
Let δx : BX {0,1} and δy : BY {0,1} be the Dirac measures of the original measure spaces. We argue that the product measure, which we denote as δ, is indeed itself a Dirac measure given by δ : BX ×BY ,δ(A) = δ(x,y)(E) = 1A(x,y).

The trick is to use Exercise 1.7.9(i) and find a set E × F in the generating set B0 = {E × FE BX,F BY } such that δ(E × F A) = 0. Since E × F is in the generating set, we can profit from the fact that the product of measures δx δy and the product measure δ coincide on this domain. In other words, by additivity of δ we have

0 = δ ([E × F] A) = δ(E × F) δ(A) = δx(E) δy(F) δ(A).

Hence,

δx(E) δy(F) = δ(A)

From here on, we differentiate two mutually exclusive cases:

  • either (x,y) A, in which case we must also have (x,y) E × F and so δ(E) = δx(E) δy(E) = 1;
  • or (x,y)A, in which case we could, at first glance, have (x,y) (E × F) A. But then, repeating the procedure from the previous part, but this time applied to (E × F) A, we obtain a cover E× FB0 of (E × F) A such that δ(E× F) = δ([E × F] A) = 1 as (x,y) E× F - a contradiction to δ([E × F] A) = 0 as per construction of E × F.
(ii)
The product of the two counting measures is the counting measure.
Mimicking the strategy from the previous part of the exercise, we find a set E × F in the generating set B0 = {E × FE BX,F BY } such that #(E × F A) = 0 (Exercise 1.7.9 (i)). In other words, by additivity of # we have 0 = # ([E × F] A) = #(E × F) #(A) = |E × F| #(A).

Hence,

|E × F| = #(A)

Now suppose, for the sake of contradiction, that #(A)|A|. We then have one of the following cases:

  • |A| > #(A). In this case |A| > |E × F| although A E × F - a contradiction.
  • |A| < #(A). Then, we at least have some (n,m) (E × F) A. Repeating the procedure, we can find a cover E× FB0 of (E × F) A with # (E× F) = # ([E × F] A). But (n,m) [E × F] A E× F - a contradiction to the fact that |E× F| = 0.
User profile picture
2021-09-14 00:00
Comments