27.08.2014 Views

Real and Complex Analysis (Rudin)

Real and Complex Analysis (Rudin)

Real and Complex Analysis (Rudin)

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

COMPLEX MEASURES 117<br />

This notation is perhaps not the best, but it is the customary one. Note that<br />

I p.1 (E) ~ I p.(E) I, but that in general I p.1 (E) is not equal to I p.(E) I.<br />

It turns out, as will be proved below, that I p.1 actually is a measure, so that<br />

our problem does have a solution. The discussion which led to (3) shows then<br />

clearly that I p.1 is the minimal solution, in the sense that any other solution A. has<br />

the property A.(E) ~ I p.1 (E) for all E E IDl.<br />

The set function I p.1 is called the total variation of p., or sometimes, to avoid<br />

misunderst<strong>and</strong>ing, the total variation measure. The term" total variation of p." is<br />

also frequently used to denote the number I p.1 (X).<br />

If p. is a positive measure, then of course I p.1 = p..<br />

Besides being a measure, I p.1 has another unexpected property: I p.1 (X) < 00.<br />

Since I p.(E) I :s; I p.1 (E) :s; I p.1 (X), this implies that every complex measure p. on<br />

any cr-algebra is bounded: If the range of p. lies in the complex plane, then it<br />

actually lies in some disc of finite radius. This property (proved in Theorem 6.4) is<br />

sometimes expressed by saying that p. is of bounded variation.<br />

6.2 Theorem The total variation I p.1 of a complex measure p. on IDl is a positive<br />

measure on IDl.<br />

PROOF Let {E;} be a partition of E E IDl. Let ti be real numbers such that<br />

ti < I p.1 (EJ. Then each Ei has a partition {Aij} such that<br />

L I p.(Aij) I > ti<br />

)<br />

(i = 1, 2, 3, ... ).<br />

(1)<br />

Since {Ai)} (i,j = 1, 2, 3, ... ) is a partition of E, it follows that<br />

L ti :s; L I p.(Aij) I :s; I p.1 (E).<br />

i i,)<br />

(2)<br />

Taking the supremum of the left side of (2), over all admissible choices of {t i},<br />

we see that<br />

L 1p.I(Ei):s; Ip. I (E). (3)<br />

i<br />

To prove the opposite inequality, let {A)} be any partition of E. Then for<br />

any fixed j, {A j 11 E i } is a partition of A), <strong>and</strong> for any fixed i, {A) 11 Ei} is a<br />

partition of E;. Hence<br />

~ I p.(A)) I = ~I~ p.(Aj 11 E;)I<br />

:s; L L 1p.(A) 11 EJI<br />

) i<br />

= L L 1p.(A) 11 E;)I:s; L 1p.I(Ei)·<br />

; ) ;<br />

(4)

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!