Jordan Decomposition

This post is inspired by Professor Tom Fischer’s writeup.

Recall the Hahn decomposition theorem:

Theorem 1. If \mu is a signed measure on (\Omega, \mathcal F), then there exist disjoint measurable subsets P, N \in \mathcal F such that

\mu (\cdot \cap P) \in [0, \infty),\quad \mu (\cdot \cap N) \in (-\infty, 0] ,\quad P \sqcup N = \Omega.

We call P a positive set, denoted P \geq 0, and N a negative set, denoted N \leq 0, and we call the pair (P, Q) a Hahn decomposition for \mu.

Call a set M is \munull if it is positive and negative: \mu( \cdot \cap M) \in \{0\}.

Define the symmetric difference by

K\triangle L := (K \backslash L) \sqcup (L \backslash K).

We leave it as an exercise to verify that

K \cup L = (K \cap L) \sqcup (K\triangle L).

Problem 1. Let \mu be a signed measure on (\Omega, \mathcal F) and (P_1,N_1), (P_2,N_2) be two Hahn decompositions for \mu. Show that P_1 \triangle P_2 and N_1 \triangle N_2 are \mu-null.

(Click for Solution)

Solution. Fix K \in \mathcal F. Then

\begin{aligned} \mu(K \cap (P_1 \triangle P_2)) &= \mu(K \cap P_1 \backslash P_2) + \mu(K \cap P_2 \backslash P_1) \\ &= \mu(K \cap P_1 \cap N_2) + \mu(K \cap P_2 \cap N_2) \\ &= \mu(K \cap P_1 \cap N_2) + \mu(\emptyset) \\ &= \mu(K \cap P_1 \cap N_2). \end{aligned}

Using Theorem 1,

\begin{aligned} 0 &\leq \mu((K \cap N_2) \cap P_1) \\ &= \mu(K \cap P_1 \cap N_2) \\ &= \mu((K \cap P_1) \cap N_2) \leq 0, \end{aligned}

so \mu(K \cap (P_1 \triangle P_2)) = \mu(K \cap P_1 \cap N_2)  = 0, as required.

We now state the Jordan decomposition theorem.

Problem 2. Let \mu be a finite signed measure on (\Omega, \mathcal F). Construct unique finite measures \mu^+, \mu^- on (\Omega , \mathcal F) such that:

  • \mu = \mu^+ - \mu^-, and
  • for any Hahn decomposition (P, N) of (\Omega, \mathcal F), \mu^+(K) = 0 for K \subseteq N and \mu^-(K) = 0 for K \subseteq P.

We call (\mu^+, \mu^-) the (unique) Jordan decomposition of \mu.

(Click for Solution)

Solution. Use Theorem 1 to construct a Hahn decomposition (P_1, N_1). Define

\begin{aligned} \mu^+ &:= \mu(\cdot \cap P_1) \in [0, \infty),\\ \mu^- &:= -\mu(\cdot \cap N_1) \in [0, \infty). \end{aligned}

Then for any K \in \mathcal F,

\begin{aligned} \mu(K) &= \mu(K \cap P_1) + \mu(K \cap N_1) \\ &= \mu^+(K) - \mu^-(K) \\ &= (\mu^+ - \mu^-)(K). \end{aligned}

Hence, \mu = \mu^+ - \mu^-.

Now fix any Hahn decomposition (P_2, N_2). Then

\begin{aligned} P_1 &= P_1 \cap (P_1 \cup P_2) \\ &= P_1 \cap ((P_1 \cap P_2) \sqcup (P_1 \triangle P_2)) \\ &= (P_1 \cap (P_1 \cap P_2)) \sqcup (P_1 \cap (P_1 \triangle P_2)) \\ &= (P_1 \cap P_2) \sqcup (P_1 \cap (P_1 \triangle P_2)).\end{aligned}

Fix K \subseteq N_2. Then

\begin{aligned} \mu^+(K) &= \mu(K \cap P_1) \\ &= \mu(K \cap P_1 \cap P_2) + \mu(K \cap P_1 \cap (P_1 \triangle P_2)) \\ &= \mu(K \cap \emptyset) + \mu((K \cap P_1) \cap (P_1 \triangle P_2)) \\ &= 0 + 0 = 0. \end{aligned}

Similarly, K \subseteq P_2 implies \mu^-(K) = 0.

Finally, we establish the uniqueness of the measures. Suppose

\mu = \mu_1^+ - \mu_1^- = \mu_2^+ - \mu_2^-.

We need to check that \mu_1^+ = \mu_2^+. To that end, fix any Hahn decomposition (P, N). For any K \subseteq P,

\mu_1^+(K \cap N) = 0 = \mu_2^+(K \cap N)

and

\mu_1^-(K) = 0 = \mu_2^-(K)

so that

\begin{aligned} \mu_1^+(K) &= \mu(P) - \mu_1^-(K) \\ &= \mu(P) - \mu_2^-(K) = \mu_2^+(K). \end{aligned}

Hence, for any K = (K \cap P) \sqcup (K \cap N),

\begin{aligned} \mu_1^+(K) &= \mu_1^+(K \cap P) + \mu_1^+(K \cap N) \\ &= \mu_2^+(K \cap P) + \mu_2^+(K \cap N) = \mu_2^+(K). \end{aligned}

—Joel Kindiak, 7 Jan 26, 1212H

,

Published by


Leave a comment