% vim: tw=50
% 18/10/2022 11AM

\begin{enumerate}[(1)]
    \setcounter{enumi}{10}
    \item Let $X$ be any set. Define a metric $d$
        on $X$ by
        \begin{flashcard}[discrete-metric]
        \prompt{Discrete metric?}
        \[ \cloze{d(x, y) = \begin{cases}
            0 & x = y \\
            1 & x \neq y
        \end{cases}} \]
        \end{flashcard}
        Easy to check this works. This is called
        the \emph{discrete metric} on $X$.
    \item
        \begin{flashcard}[p-adic-metric]
        Let $X = \ZZ$. Let $p$ be a prime. The
        \emph{$p$-adic metric} on $\ZZ$ is the
        metric $d$ defined by
        \[ \cloze{d(x, y) = \begin{cases}
            0 & x = y \\
            p^{-a} & \text{if $x \neq y$ and $x -
            y = p^am$ with $p \nmid m$}
        \end{cases}} \]
        \end{flashcard}
        ``Two numbers are close if the difference
        is divisible by a large power of $p$''.
        Triangle inequality:
        \begin{itemize}
            \item Easy if any two of $x, y, z$ are
                the same, so assume $x, y, z$ are
                all distinct.
            \item Let $x - y = p^am$ and $y - z =
                p^bn$ where $p \nmid m$, $p \nmid
                n$ and without loss of generality
                $a \le b$. So $d(x, y) = p^{-a}$
                and $d(y, z) = p^{-b}$. Now:
                \begin{align*}
                    x - z
                    &= (x - y) + (y - z) \\
                    &= p^am + p^b n \\
                    p^a(m + p^{b - a} n)
                \end{align*}
                so $p^a \mid x - z$ so $d(x, z)
                \le p^{-a}$. But $d(x, y) + d(y,
                z) \ge d(x, y) = p^{-a}$, so
                triangle inequality does hold.
        \end{itemize}
\end{enumerate}

\begin{definition*}
    Let $(X, d)$ be a metric space. Let $(X_n)$ be
    a sequence in $X$ and let $x \in X$. We say
    $(X_n)$ converges to $x$ and write ``$x_n \to
    x$'' or ``$x_n \to x$ as $n \to \infty$'' if
    \[ \forall \eps > 0 ~ \exists N ~ \forall n
    \ge N \quad d(x_n, x) < \eps .\]
    Equivalently $x_n \to x$ if and only if
    $d(x_n, x) \to 0$ in $\RR$.
\end{definition*}

\begin{hiddenflashcard}
    \begin{definition*}[Convergence in a metric
        space]
        Let $(X, d)$ be a metric space, and
        $(x_n)$ \cloze{be a sequence in $X$. We
        say $x_n \to x$ if and only if $d(x_n, x)
        \to 0$.}
    \end{definition*}
\end{hiddenflashcard}

\begin{proposition}
    Limits are unique. That is, if $(X, d)$ is a
    metric space, $(x_n)$ a sequence in $X$, $x, y
    \in X$ with $x_n \to x$ and $x_n \to y$, then
    $x = y$.
\end{proposition}

\begin{proof}
    For each $n$,
    \begin{align*}
        d(x, y)
        &\le d(x, x_n) + d(x_n, y) \\
        &\le d(x_n, x) + d(x_n, y) \\
        &\to 0 + 0 = 0
    \end{align*}
    So we would need $d(x, y) \to 0$ as $n \to
    \infty$, but $d(x, y)$ is constant, so $d(x,
    y) = 0$. So $x = y$. 
\end{proof}

\begin{remark*}
    This justifies talking about \emph{the} limit
    of a convergent sequence in a metric space,
    and writing $x = \lim_{n \to \infty} x_n$ if
    $x_n \to x$.
\end{remark*}

\subsubsection*{Remarks on the definition}

\begin{enumerate}[(1)]
    \item Constant sequences obviously converge.
        Moreover, eventually constant sequences
        converge.
    \item Suppose $(X, d)$ is a metric space and
        $Y$ is a subspace of $X$. Suppose $(x_n)$
        is a sequence in $Y$ which converges in
        $Y$ to $x$. Then also $(x_n)$ converges in
        $X$ to $x$. However the converse is false.
        For example, in $\RR$ with the usual
        metric then $\frac{1}{n} \to 0$ as $n \to
        \infty$. Consider the subspace $\RR
        \setminus \{0\}$. Then $\left( \frac{1}{n}
        \right)_{n \ge 1}$ is a sequence in $\RR
        \setminus \{0\}$ but it doesn't converge
        in $\RR \setminus \{0\}$. This is because
        by uniqueness of limits, it would have to
        converge to 0, but $0 \not\in \RR
        \setminus \{0\}$.
\end{enumerate}

\subsubsection*{Examples}

\begin{enumerate}[(1)]
    \item Let $d$ be the Euclidean metric on
        $\RR^n$. Exactly as in $\RR^2$, we have
        $x_n \to x$ if and only if the sequence
        converges in each coordinate in the usual
        way in $\RR$. \myskip
        What about other metrics on
        $\RR^n$? For example let $d_\infty$ be the
        uniform metric
        \[ d_\infty (x, y) = \max_i |x_i - y_i| \]
        which sequences converge in $(\RR^n,
        d_\infty)$? Note that
        \[ d(x, y) = \sqrt{\sum_{i = 1}^n (x_i -
        y_i)^2} \le \sqrt{\sum_{i = 1}^n
        d_\infty (x, y)^2} \]
        so $d(x, y) \le \sqrt{n} d_\infty (x, y)$.
        But also $d_\infty(x, y) \le d(x, y)$. Now
        suppose $(x_n)$ is a sequence in $\RR^n$.
        Then
        \[ d(x_n, x) \to 0 \iff d_\infty (x_n, x)
        \to 0 \]
        So exactly the same sequences converge in
        $(\RR^n, d)$ and $(\RR^n, d_\infty)$. What
        about the $l_1$ metric $d_1$?
        \[ d_1(x, y) = \sum_{i = 1}^n |x_i - y_i| \]
        Similarly $d_\infty(x, y) \le d_1(x, y)
        \le nd_\infty (x, y)$. So again, exactly
        the same sequences converge in $(\RR^n,
        d)$.
    \item Let $X = C([0, 1]) = \{f : [0, 1] \to
        \RR \mid \text{$f$ is continuous}\}$. Let
        $d_\infty$ be the uniform metric on $X$:
        \[ d_\infty(f, g) = \sup_{x \in [0, 1]}
        |f(x) - g(x)| \]
        Now note that
        \begin{align*}
            f_n \to f \text{ in $X, d_\infty$}
            &\iff d_\infty(f_n, f) \to 0 \\
            &\iff \sup_{x \in [0, 1]} |f_n(x) -
            f(x)| \to 0 \\
            &\iff f_n \to f \text{ uniformly}
        \end{align*}
        We also had the $L_1$ metric $d_1$ on $X$:
        \[ d_1(f, g) = \int_0^1 |f - g| \]
        Now
        \begin{align*}
            d_1(f, g)
            &= \int_0^1 |f - g| \\
            &\le \int_0^1 d_\infty(f, g) \\
            &= d_\infty(f, g)
        \end{align*}
        So similarly to previous example,
        \[ f_n \to f \text{ in $(X, d_\infty)$}
        \implies f_n \to f \text{ in $(X, d_1)$} \]
        But converse does not hold, i.e. we can
        find a sequence $(f_n)$ in $X$ such that
        $f_n \to 0$ in $d_1$ metric by $f_n$
        doesn't converge in the $d_\infty$ metric.
        So we want $(f_n)$ such that $\int_0^1
        |f_n| \to 0$ as $n \to \infty$, but
        $(f_n)$ does not converge uniformly. We
        can just take functions like this:
        \begin{center}
            \includegraphics[width=0.6\linewidth]
            {images/f18f5b964ed111ed.png}
        \end{center}
        Then clearly $f_n \to f$ in the $d_1$
        metric but not in the $d_\infty$ metric.
    \item Let $(X, d)$ be a discrete metric space;
        \[ d(x, y) = \begin{cases}
            0 & x = y \\
            1 & x \neq y
        \end{cases} \]
        When do we have $x_n \to x$ in $(X, d)$?
        Suppose $x_n \to x$, i.e.
        \[ \forall \eps > 0 ~ \exists N ~ \forall
        n \ge N \quad d(x_n, x) < \eps \]
        Setting $\eps = 1$ in this, can find $N$
        such that
        \[ \forall n \ge N ~ d(x_n, x) < 1 \]
        i.e. $\forall n \ge N$, $d(x_n, x) = 0$,
        i.e. $\forall n \ge N$, $x_n = x$. Thus
        $(x_n)$ is eventually constant. But we
        know that in \emph{any} metric space,
        eventually constant sequences converge. So
        in this space, $(x_n)$ converges if and
        only if $(x_n)$ eventually constant.
\end{enumerate}

\begin{definition*}
    Let $(X, d)$ and $(Y, e)$ be metric spaces and
    let $f : X \to Y$.
    \begin{enumerate}[(i)]
        \item
            \begin{flashcard}[metric-space-convergence]
            Let $a \in X$ and $b \in Y$. We say
            $f(x) \to b$ as $x \to a$ if
            \cloze{
            \[ \forall \eps > 0 ~ \exists \delta >
            0 ~ \forall x \in X \]
            \[ 0 < d(x, a) < \delta \implies
            e(f(x), b) < \eps \]}
            \prompt{\cloze{\fcemph{Note that $d(x, a) >
            0$}! \\ This means that the value of
            $f(a)$ doesn't affect whether $f(x)
            \to b$ as $x \to a$. This is
            different to continuity.}}
            \end{flashcard}
        \item Let $a \in X$. We say $f$ is
            \emph{continuous} at $a$ if $f(x) \to
            f(a)$ as $x \to a$. That is
            \[ \forall \eps > 0 ~ \exists \delta >
            0 ~ \forall x \in X \]
            \[ d(x, a) < \delta \implies e(f(x),
            f(a)) < \eps \]
        \item If $\forall x \in X$, $f$ is
            continuous at $a$, we say \emph{$f$ is
            a continuous function} or simply
            \emph{$f$ is continuous}.
        \item We say \emph{$f$ is uniformly
            continuous} if
            \[ \forall \eps > 0 ~ \exists \delta >
            0 ~ \forall x, y \in X \]
            \[ d(x, y) < \delta \implies e(f(x),
            f(y)) < \eps \]
        \item Suppose $W \subset X$. We say $f$ is
            \emph{continuous on $W$} (similarly
            for \emph{uniformly continuous on
            $W$}) if the function $f \mid_W$ is
            continuous, as a function $W \to Y$
            where now thinking of $W$ as a
            subspace of $X$.
    \end{enumerate}
\end{definition*}