Skip to content

Commit

Permalink
fixed small errors, typos, and siplified/organized proof the induced …
Browse files Browse the repository at this point in the history
…uF is Radon
  • Loading branch information
MareoRaft committed Jan 5, 2018
1 parent 6606a32 commit 686e638
Showing 1 changed file with 56 additions and 62 deletions.
118 changes: 56 additions & 62 deletions Real Analysis/real-analysis-notes.tex
Original file line number Diff line number Diff line change
Expand Up @@ -1522,7 +1522,7 @@ \subsection{Regularity}
Similarly, $\mu$ is \de{inner regular on $E$} if and only if \[
\mu(E) = \sup \{\mu(K) \st K \text{ is compact}, K \subset E\}.
\]
Finally, $\mu$ is called \de{regular} is and only if it is inner and
Finally, $\mu$ is \de{regular} if and only if it is inner and
outer regular on every Borel set.
\end{defn}
\begin{defn}
Expand Down Expand Up @@ -1555,7 +1555,7 @@ \subsection{Regularity}
Let $\epsilon > 0$. Then there exists a sequence of $A_n \in \A$ such that
$\sum_{n=1}^\infty \mu(A_n) \leq \mu(E) + \epsilon$ and \(E
\subset \Union A_n\). So,
decompoising $A_n$ into elements of $\Ep$ and renumbering, we
decomposing $A_n$ into elements of $\Ep$ and renumbering, we
can assume without loss of generality that $A_n \in \Ep$, since
if $A_n \not \in \Ep$, we can just take $A_n = A_n \intersect
(-m,m)$ using the assumption $E \subset (-m,m)$. So, now, take $A_n = (\alpha_n,
Expand All @@ -1578,86 +1578,80 @@ \subsection{Regularity}
and $g(2k+1) = -k$. Then, we have \[
\mu(E) = \sum_{n \geq 1} \mu(E \intersect (g(n),g(n)+1)).
\]
So, for all $n$, let $a_{n,i} < b_{n,i}$. Then, \[
\sum_{i=1}^\infty \mu((a_{n,i},b_{n,i})) \leq \sum_{n \geq 1}
Each $E \intersect (g(n),g(n)+1)$ is bounded, so by the previous case that we just proved, there exist $a_{n,i} < b_{n,i}$ such that \[
\sum_{i=1}^\infty \mu((a_{n,i},b_{n,i})) \leq
\mu(E \intersect (g(n),g(n)+1)) + \frac{\epsilon}{2^n}
\]
by the previous bounded base. Finally, this gives us $E \subset
for each $n$. Finally, this gives us $E \subset
\Union_{(n,i) \in \N^2} (a_{n,i},b_{n,i})$ and so \[
\sum_{(n,i) \in \N^2} \mu((a_{n,i},b_{n,i})) \leq \sum_{n \geq
1} \mu(E
\intersect (g(n),g(n)+1)) + \epsilon
1} \mu(E \intersect (g(n),g(n)+1)) + \epsilon = \mu(E) + \epsilon
\]
thereby completing the proof.
thereby proving the lemma.
\end{itemize}
\end{proof}
\subsection*{(2/28/2017) Lecture 9}
With this lemma in hand, we now seek to prove the theorem (\ref{radon-thm}).
\begin{proof}[Proof of Theorem]
We have to show $\mu_F$ is finite on compact sets, outer
We have to show $\mu = \mu_F$ is finite on bounded sets, outer
regular for all sets in $\B_\R$, and inner regular for all sets in
$\B_\R$.
\begin{itemize}
\item ($\mu_F$ is finite on compact sets).
$\mu_F$ is finite on compact sets by definition.
\item ($\mu$ is finite on bounded sets).
If $E$ is bounded, then there exists an open bounded interval $(a,b) \supset E$. Then $\mu(E) \leq \mu((a,b)) = F(b) - F(a) < \oo$.
\item (Outer regularity). We have the $\leq$ direction from
monotonicity and thus equality if $\mu(E) = \infty$. Now, assume
that $\mu(E) < \infty$. Our lemma tells us that \[
\mu(E) = \inf \left\{ \sum \mu((a_i,b_i)) \st a_i < b_i
\text{ finite}, E \subset \Union_{i=1}^\infty (a_i,b_i) \right\}
monotonicity, and we want the $\geq$ direction. Our lemma tells us that \[
\mu(E) = \inf \left\{ \sum \mu((a_i,b_i)) \st E \subset \Union_{i=1}^\infty (a_i,b_i) \right\}.
\]
Next, take $\epsilon > 0$. Then, there exists $a_i,b_i$ such
that $\sum \mu((a_i,b_i)) \leq \mu(E) +\epsilon$ by
subadditivity. Now, take $V = \Union_{i=1}^\infty (a_i,b_i)$
(open) to get $\mu(V) \leq \mu(E) + \epsilon$. But $\epsilon$
was arbitrary, so $\mu(E) = \inf\{ \cdots \}$ as desired.
Now every $\union_{i=1}^\infty (a_i,b_i)$ is open. Therefore
\[
\inf \left\{ \sum \mu(\Theta) \st \Theta \:\text{open} \right\}
\leq \inf \left\{ \sum \mu((a_i,b_i)) \st E \subset \Union_{i=1}^\infty (a_i,b_i) \right\}
\]
where the RHS is $\mu(E)$ as desired.
\item (Inner regularity). The $\geq$ direction is trivial, so we
only need to show $\leq$. We can also reduce to $E$ being
bounded by using continuity from below and intersecting $E
\intersect [-m,m]$ to get $\mu(E) = \lim_{m \to \infty} \mu(E
\intersect [-m,m])$. So, if $\alpha < \mu(E)$, there exists an
$m$ such that $\mu(E \intersect [-m,m]) > \alpha$. Now, if $E$
is bounded, there exists a
compact $K \subset E \intersect [-m,m] \subset E$ with $\mu(K) >
\alpha$. Thus, for all $\alpha$, we have $\sup \{\mu(K)
\st \text{ compact }K \subset E\} \geq \mu(E)$. Now assume that
$E$ is bounded. Then, there exists an $m \in \N$ such that $E
\subset [-m,m]$. Let $\epsilon > 0, F = [-m,m] \setminus E$, and
$\mu(F) < \infty$. Our previous case tells us that there exists
an open $V \supset F$ with $\mu(V) \leq \mu(F) + \epsilon$. So,
we take $K = [-m,m] \intersect V^c$ which is closed and bounded,
and therefore compact.
only need to show $\leq$. We break into two cases:
\begin{itemize}
\item ($E$ unbounded). We can reduce to $E$ being
bounded by using continuity from below and intersecting $E
\intersect [-m,m]$ to get $\mu(E) = \lim_{m \to \infty} \mu(E
\intersect [-m,m])$. So, if $\alpha < \mu(E)$, there exists an
$m$ such that $\mu(E \intersect [-m,m]) > \alpha$. There exists a
compact $K \subset E \intersect [-m,m] \subset E$ with $\mu(K) >
\alpha$. Thus, for all $\alpha$, we have $\sup \{\mu(K)
\st \text{ compact }K \subset E\} \geq \mu(E)$.

Next, we show that $K
\subset E$. If $x \in K$ and $x \not \in E$, we get $x \in
[-m,m] \setminus E = F \subset V$, yielding that $x \in V^c$ and
so $x \in E$, a contradiction. Similarly, we also have that $V
\supset F$, and so $\mu(V) \leq \mu(F) + \epsilon$ giving us
$\mu(V \setminus F) \leq \epsilon$.
\item ($E$ bounded). There exists an $m \in \N$ such that $E
\subset [-m,m]$. Let $\epsilon > 0, F = [-m,m] \setminus E$, and
$\mu(F) < \infty$. Our previous case tells us that there exists
an open $V \supset F$ with $\mu(V) \leq \mu(F) + \epsilon$ and thus $\mu(V \setminus F) \leq \epsilon$. So,
we take $K = [-m,m] \intersect V^c$ which is closed and bounded,
and therefore compact. By design, $K \subset E$.\\

Finally, we show $E \setminus K \subset V \setminus F$. We note
that
\begin{align*}
E \setminus F & = E \intersect K^c \\
& = E \intersect ([-m,m] \intersect V^c)^c \\
& = E \intersect ([-m,m]^c \union V) \\
& = (E \intersect [-m,m]^c) \union (E \intersect
V)\\
& = E \intersect V
\end{align*}
since $E \intersect [-m,m]^c = \emptyset$. Furthermore,
\begin{align*}
V \setminus F & = V \intersect F^c \\
& = V \intersect ([-m,m] \intersect F^c)^c \\
& = V \intersect ([-m,m]^c \union F) \\
& = (V \intersect F) \union (V \intersect [-m,m]^c)
\end{align*}
Thus, we get that $\mu(E \setminus K) \leq \mu(V \setminus F)
\leq \epsilon$ for all $\epsilon$.
Finally, we show $E \setminus K \subset V \setminus F$. We note
that
\begin{align*}
E \setminus F & = E \intersect K^c \\
& = E \intersect ([-m,m] \intersect V^c)^c \\
& = E \intersect ([-m,m]^c \union V) \\
& = (E \intersect [-m,m]^c) \union (E \intersect
V)\\
& = E \intersect V
\end{align*}
since $E \intersect [-m,m]^c = \emptyset$. Furthermore,
\begin{align*}
V \setminus F & = V \intersect F^c \\
& = V \intersect ([-m,m] \intersect F^c)^c \\
& = V \intersect ([-m,m]^c \union F) \\
& = (V \intersect F) \union (V \intersect [-m,m]^c)
\end{align*}
Thus, we get that $\mu(E \setminus K) \leq \mu(V \setminus F)
\leq \epsilon$ for all $\epsilon$.
\end{itemize}
\end{itemize}
\end{proof}
\begin{defn}
We say $\mu$ is \de{tight} if, for all $\epsilon > 0$, there
We say that a probability measure $\mu$ is \de{tight} if, for all $\epsilon > 0$, there
exists a compact $K \subset \R$ such that $\mu(K) \geq
1-\epsilon$.
\end{defn}
Expand All @@ -1674,7 +1668,7 @@ \subsection{Regularity}
\begin{defn}
(Aside). We say a sequence of Borel probability measures on $\R$,
$(\mu_n)_{n \geq 1}$, \de{converges weakly} to a Borel probability
measure such that, for all $f \from \R \to \R$ bounded an
measure $\mu$ if, for all $f \from \R \to \R$ bounded an
continuous, \[
\int f d \mu_n \to \int f \d\mu \text{ as } n \to \infty
\]
Expand Down

0 comments on commit 686e638

Please sign in to comment.