21.4 Absolute Continuity

Definition 21.4.1 (Absolutely Continuous).label Let $(X, \cm)$ be a measurable space, $\mu$ be a positive measure on $(X, \cm)$, and $\nu$ be a signed measure or vector veasure on $(X, \cm)$, then $\nu$ is absolutely continuous with respect to $\mu$, denoted $\nu \ll \mu$, if for every $E \in \cm$ with $\mu(E) = 0$, $\nu(E) = 0$ as well.

Lemma 21.4.2.label Let $(X, \cm)$ be a measurable space, $\mu$ be a positive measure on $(X, \cm)$, and $\nu$ be a signed measure or vector measure on $(X, \cm)$, then $\nu \ll \mu$ if and only if $|\nu| \ll \mu$.

Proposition 21.4.3.label Let $(X, \cm)$ be a measurable space, $\mu$ be a positive measure on $(X, \cm)$, $E$ be a normed space over $K \in \RC$, and $\nu: \cm \to E$ be a vector measure, then the following are equivalent:

  1. (1)

    $\nu$ is absolutely continuous with respect to $\mu$.

  2. (2)

    For each $\eps > 0$, there exists $\delta > 0$ such that for any $A \in \cm$ with $\mu(A) < \delta$, $\norm{\nu(A)}_{E} < \eps$.

Proof, [Theorem 3.15, Fol99]. $(\neg 2) \Rightarrow (\neg 1)$: Using Lemma 21.4.2, assume without loss of generality that $\nu$ is a positive finite measure.

By assumption, there exists $\eps > 0$ and $\seq{A_n}\subset \cm$ such that for each $n \in \natp$, $\mu(A_{n}) < 2^{-n}$ and $\norm{\nu(A_n)}_{E} \ge \eps$. For each $N \in \natp$, let $B_{N} = \bigcup_{n \ge N}A_{n}$, then $\mu(B_{N}) < 2^{-N+1}$. By Proposition 20.1.5,

\[\mu\paren{\bigcap_{N \in \natp}B_N}= \limv{N}\mu(B_{N}) = 0\]

but by Proposition 21.2.3,

\[\norm{\nu\paren{\bigcap_{N \in \natp}B_N}}_{E} = \limv{N}\norm{\nu(B_N)}_{E} \ge \eps\]

so (1) does not hold.$\square$

Definition 21.4.4 (Uniformly Absolutely Continuous).label Let $(X, \cm)$ be a measurable space, $\mu$ be a positive measure on $(X, \cm)$, $E$ be a normed space over $K \in \RC$, and $\mathcal{U}\subset M(X, \cm; E)$ be a family of $E$-valued vector measures on $(X, \cm)$, then $\mathcal{U}$ is uniformly absolutely continuous with respect to $\mu$ if for every $\eps > 0$, there exists $\delta > 0$ such that for all $A \in \cm$ with $\mu(A) < \delta$, $\norm{\nu(A)}_{E} < \eps$ for all $\nu \in \mathcal{U}$.

Theorem 21.4.5 (Vitali-Hahn-Saks).label Let $(X, \cm, \mu)$ be a measurable space, $\mu$ be a positive measure on $(X, \cm)$, $E$ be a normed vector space over $K \in \RC$, and $\seq{\nu_n}\subset M(X, \cm; E)$ be $E$-valued vector measures on $(X, \cm)$ such that:

  1. (a)

    For each $n \in \natp$, $\nu$ is absolutely continuous with respect to $\mu$.

  2. (b)

    For each $A \in M(X, \cm; E)$, $\nu(A) = \limv{n}\nu_{n}(A)$ exists.

then

  1. (1)

    $\seq{\nu_n}$ is uniformly absolutely continuous with respect to $\mu$.

  2. (2)

    The mapping $\nu$ is a $E$-valued vector measure absolutely continuous with respect to $\mu$.

Proof, [Theorem II.2.1, Yos12]. Let $\cm_{0}$ be the euivalence classes of essentially equal sets in $\cm$, equipped with the Fréchet-Nikodym metric with respect to $\mu$. For each $n \in \natp$, $\nu_{n} \ll \mu$ by assumption (a), and $\nu_{n} \in UC(\cm_{0}; E)$.

(1): Let $\eps > 0$. For each $N \in \natp$, let

\[A_{N} = \bracs{A \in \cm_0 \bigg | \sup_{n \ge N}\norm{\nu_n(A) - \nu_N(A)}_E \le \eps}\]

then since $\seq{\nu_n}\subset UC(\cm_{0}; E)$, $A_{N} \subset \cm_{0}$ is closed. By assumption (b), $\cm_{0} = \bigcup_{N \in \natp}A_{N}$. By the Baire Category Theorem, there exists $N \in \natp$, $A \in \cm_{0}$, and $\delta > 0$ such that for every $m, n \ge N$ and $B \in \cm_{0}$ with $\mu(A \Delta B) \le \delta$, $\norm{\nu_m(B) - \nu_n(B)}_{E} \le \eps$.

Let $B \in \cm$ with $\mu(B) = \mu(\emptyset \Delta B) \le \delta$ and write $B = (A \cup B) \setminus (A \setminus B)$, then $\mu((A \cup B) \Delta B) \le \delta$ and $\mu((A \setminus B) \Delta B) \le \delta$. Thus

\begin{align*}\norm{\nu_m(B) - \nu_n(B)}_{E}&= \norm{(\nu_m - \nu_n)(A \cup B) - (\nu_m - \nu_n)(A \setminus B)}_{E} \\&\le 2\eps\end{align*}

and $\seq{\mu_n}$ is uniformly absolutely continuous with respect to $\mu$.

(2): Let $\seq{A_n}\subset \cm$ and $A \in \cm$ with $A = \bigsqcup_{n \in \natp}A_{n}$, then for each $N \in \natp$,

\begin{align*}\norm{\nu(A) - \sum_{n = 1}^N \nu(A_n)}_{E}&= \norm{\nu\paren{A \setminus \bigsqcup_{n = 1}^N A_n}}_{E} \\&= \limv{k}\norm{\nu_k\paren{A \setminus \bigsqcup_{n = 1}^N A_n}}_{E}\end{align*}

Since $\seq{\nu_k}$ is uniformly absolutely continuous,

\[\lim_{N \to \infty}\sup_{k \in \natp}\norm{\nu_k\paren{A \setminus \bigsqcup_{n = 1}^N A_n}}_{E} = 0\]

so $\nu$ is a vector measure.

Absolute continuity is equivalent to uniform continuity as a mapping $\cm_{n} \to E$, and uniform absolute continuity is equivalent to uniform equicontinuity as mappings $\cm \to E$. By the Arzelà-Ascoli Theorem, $\nu \in UC(\cm_{0}; E)$, so $\nu \ll \mu$.$\square$

Post a Comment

Name:Email:
Please enter the tag of the current page (Z9) to post the comment.
Tag: