Math 521 Analysis I, Spring 2017

Homework assignments

Homework 1

Due Monday January 30, in class. All students should hand in solutions to Problems 3, 4, 5acd, and 6. Students who registered as Honors students should hand in solutions to all problems.
  1. Let $F=\{0, 1\}$, i.e. $F$ is a set with only two elements, called “$0$” and “$1$”. Define addition and multiplication in $F$ by \[ 0+0=0,\quad 0+1=1+0=1,\quad 1+1=0, \] and \[ 0\cdot0 = 0\cdot 1 = 1\cdot 0 = 0, \quad 1\cdot 1=1. \] Is $F$ a field? Is $F$ an ordered field?
  2. If $A\subset\R$ is a nonempty subset that is bounded from below, then consider $B=\{-x \mid x\in A\}$, and show that $\inf A + \sup B = 0$.
    This is problem 5 from chapter 1 in Rudin. See Roger Cooke’s solutions
  3. If $A$ and $B$ are bounded and non empty subsets of $\R$, then show that $A\subset B \implies \sup A \leq \sup B$.
  4. Let $A, B\subset \R$ be bounded and nonempty sets of real numbers. Let $C = \{a+b \mid a\in A, b\in B\}$ and $D = \{ab \mid a\in A, b\in B\}$.
    1. Prove or disprove: $\sup C = \sup A + \sup B$.
      True.

      First show that $\sup A + \sup B$ is an upper bound for $C$. This is true because every number in $C$ is of the form $a+b$ with $a\in A$ and $b\in B$. Since $\sup A$ is an upper bound for $A$ we have $a\leq \sup A$; similarly, we have $b\leq \sup B$. Therefore $a+b\leq \sup A + \sup B$, and we have shown that $\sup A+\sup B$ is an upper bound for $C$.

      Next show that there is no smaller upper bound, i.e. suppose $m\lt \sup A+\sup B$ and show that $m$ is not an upper bound. To do this we define $r=(\sup A + \sup B)-m$. Since $m\lt \sup A + \sup B$, we know that $r\gt0$. The set $A$ must contain a number $a$ with $a\gt \sup A - \frac12 r$ because $\sup A$ is the least upper bound of $A$. Similarly, there is a $b\in B$ with $b\gt \sup B - \frac12 r$. Then $a+b\in C$ and $a+b\gt (\sup A + \sup B) - r = m$, which implies that $m$ is not an upper bound for $C$. The conclusion is that $\sup A+ \sup B$ is the lowest upper bound for $C$.

    2. Prove or disprove: $\sup D = (\sup A)(\sup B)$.
      Not true. If $A=[-2, -1]$ and $B=[-2,-1]$ then $D=[1,4]$ so that $\sup A = \sup B = -1$, while $\sup D = 4$, which is not the same as $(\sup A)(\sup B)$.
  5. (Variations on the Archimedean property)
    1. Show that the set $A=\{2^n: n\in \N\}$ is unbounded, i.e. show that for every real number $x$ there is an integer $n$ with $2^n\gt x$.
      Suppose $A$ were bounded. Then $m=\sup A$ exists, by the least upper bound axiom. The set $A$ must contain a number in the interval $(\frac12 m, m]$, for otherwise $\frac12m$ would also be an upper bound for $A$ and $m$ would not be the least upper bound. Suppose then that $2^n\in A$ is the number in the interval $(\frac12 m, m]$. Then $2^{n+1}$ also belongs to $A$, while $2^{n+1}=2\cdot 2^n \gt 2\cdot \frac12m = m$. This contradicts the fact that $m$ is an upper bound for $A$.
    2. Show that the set $A=\{n^2 \mid n\in\N\}$ is unbounded.
      For every $n\in\N$ we have $n^2\geq n$, which follows by multiplying both sides of the inequality $n\geq 1$ with $n$ and concluding $n^2\geq n\geq 1$. Given any $x\in \R$ choose a natural number $n\gt x$. Then $n^2\in A$ and $n^2\geq n\gt x$. Thus $A$ is unbounded.
    3. Show that the set $A=\{n! \mid n\in\N\}$ is unbounded.
      Same argument. First show that $n!\geq n$.
    4. Show that the set $A=\{\sqrt{n} \mid n\in\N\}$ is unbounded.
      The set $A$ contains $\N$ as a subset: $A\subset \N$. Let $x\in\R$ be any number. Then choose a natural number $n\in\N$ with $x\lt n$. Since $n=\sqrt{n^2}$ we have $n\in A$, and therefore $x$ cannot be an upper bound for $A$. Thus $A$ is unbounded.
    5. Let $A=\{a_1, a_2, a_3, \ldots\}$ be a set of real numbers where $a_{n+1}\ge a_n+1$ holds for all $n\in\N$. Show that $A$ is unbounded.
  6. Use the Archimedean property to show that for any pair of real numbers $x\lt y$ there is a rational number $r\in \Q$ with $x\lt r\lt y$.
  7. Let \[ a_n = 0.\overbrace{11\dots11}^{n\;{\sf digits}} = \frac{1}{10} + \frac{1}{10^2} + \cdots + \frac{1}{10^n}. \] and let $A= \{a_1, a_2, a_3, \dots\}$. Show that $x=\sup A$ exists, and show that $9x=1$. (Rudin: see section 1.22 on page 11.)
  8. Let $x_0$, $x_1$, $x_2$,…be a sequence of positive real numbers that satisfy $x_{n}\leq \frac23 x_{n-1}$ for all $n\geq 1$. Show that there is an integer $N\in\N$ such that $x_N\lt 1$.
    The idea is that $x_n \leq (\frac 23)^n x_0$, and that $(\frac23)^n$ goes to zero as $n\to\infty$. Here we will prove this without using the words “limit” and “converge” because we have not defined them yet.

    Let $m=\inf A$ where $A=\{x_1, x_2, x_3, \dots\}$. Since $0$ is a lower bound for $A$ the greatest lower bound $m$ for $A$ must exist. We will show that $m=0$. If $m$ were positive then, since $m$ is the greatest lower bound of $A$, the set $A$ must contain a number $x_n \in [m, \frac32m)$. But then $A$ also contains $x_{m+1}\leq \frac 23x_n \lt \frac23\cdot\frac32m = m$. Since $A$ apparently contains the number $x_{n+1}$ which is less than $m$, $m$ cannot be a lower bound for $A$, which contradicts the assumption $m=\inf A$. Hence we have shown that $\inf A=0$.

    Since $0$ is the greatest lower bound for $A$, the number $1$ is not a lower bound, and therefore there is an $x_n$ with $x_n\lt 1$.

  9. Let $A = \{x_0, x_1, x_2, \dots\}$ where $x_0$, $x_1$, $x_2$, … are positive real numbers that satisfy $x_{n+1} = \frac{x_n}{x_n+1}$ for all $n\geq 0$. Show that $\inf A=0$.

Homework 2

Due Friday, February 10.
  1. Chapter 2: Problems 2, 3, 4 about countable and uncountable sets.
  2. Chapter 2: Problem 11 about “metric spaces”
  3. Consider the metric $d_5$ on $\R$ from problem 11, chapter 2. Find $B_{1/2}(1)$, i.e. find the “ball with radius $1/2$, centered at $x=1$” for this metric.
  4. Is $d(x, y) = |\arctan(x) - \arctan(y)|$ a metric on the set $\R$?
  5. Prove that the set $\{(x, y)\in\R^2 : 2x+3y \gt 0\}$ is an open subset of $\R^2$.

Homework 3

Due Monday February 27.
  1. Consider the sequence of real numbers given by $a_n=n^{-1}$.
    1. Show that $a_n\to 0$, directly from the definition.
      Given $\varepsilon\gt0$ choose an integer $N$ such that $N\gt 1/\varepsilon$. Then for all $n\ge N$ one has $0\lt a_n \le 1/N \lt \varepsilon$.
    2. Show that if a sequence $x_n$ in a metric space $X$ converges, then any subsequence of $x_n$ also converges and has the same limit.
      Let $y_k = x_{n_k}$ be a subsequence of $x_n$, and let $L$ be the limit of $x_n$. In particular this implies that $n_k\ge k$ for all $k$. Given $\varepsilon\gt0$ we choose $N\in\N$ such that $|x_n-L|\lt \varepsilon$ for all $n\ge N$. Then, if $k\ge N$ we have $n_k\ge k\ge N$ and thus $|y_k-L| = |x_{n_k}-L| \lt \varepsilon$.
    3. Show that $b_n=n^{-2}\to 0$ without using the definition of convergence.
      $b_n = a_{n^2}$, so $b_n$ is a subsequence of the sequence $a_n=1/n$. Therefore it converges, and it has the same limit: $\lim_{n\to\infty} 1/n^2 = 0$.
  2. Let $x_n$ be a convergent sequence of points in a metric space $(X,d)$. Show that the sequence is bounded (i.e. there is a point $a\in X$ and there is a number $R\gt0$ such that $x_n\in B_R(a)$ for all $n\in\N$).
    Let $x_n\to p$, i.e. let $p$ be the limit of the sequence $x_n$. Choose $\varepsilon=1$. There is an $N\in\N$ such that for all $n\ge N$ on has $d(x_n, p)\lt \varepsilon=1$. Thus, if \[ R=\max\left\{1, d(x_1, p), \dots, d(x_{N-1}, p)\right\}, \] then all $x_k$ satisfy $d(x_k, p)\le R$.
  3. Let $x_n$ be a sequence of points in a metric space $(X,d)$. Show that $x_n\to x$ holds if and only if $d(x_n, x)\to0$.
    Consider the sequence of real numbers $a_n = d(x_n, p)$. We are asked to show that $a_n\to0 \iff x_n\to p$.

    Assume $a_n\to 0$. To show $x_n\to p$, let $\varepsilon\gt 0$ be given. By assumption there is an $N\in\N$ such that $|a_n|\lt \varepsilon$ for all $n\ge N$. This implies $d(x_n, p)\lt \varepsilon$ for all $n\ge N$. Therefore $x_n\to p$.

    Assume $x_n\to p$. To show $a_n\to 0$, let $\varepsilon\gt 0$ be given. By assumption there is an $N\in\N$ such that $d(x_n, p)\lt \varepsilon$ for all $n\ge N$. This implies $|a_n|\lt \varepsilon$ for all $n\ge N$. Therefore $a_n\to 0$.

  4. Let $(X,d)$ be a metric space, $C\subset X$ a closed subset, and $x_n\in C$ a sequence of points with $x_n\to x$. Show that $x\in C$.
    In order to reach a contradiction, assume that $x\not\in C$. The complement $C^c$ of $C$ is open in $X$, so there is an $\varepsilon\gt0$ such that $B_\varepsilon(x)\subset C^c$. Since $x_n\to x$, there is an $N_\varepsilon\in\N$ for such that $x_n\in B_\varepsilon(x)$ for all $n\ge N$. But then we would have $x_n\not\in C$ for $n\ge N$, which is a contradiction. We conclude that $x\in C$ after all.
  5. Let $x_n\in\R^k$ and $y_n\in \R^k$ be two sequences of vectors for which $x_n\to x$ and $y_n\to y$. Show that $(x_n, y_n) \to (x,y)$. Here $(x,y)$ is the dot product, or inner product of the two vectors. In Rudin’s notation (page 16): the assignment is to show that $x_n\cdot y_n \to x\cdot y$.
    The sequences $x_n\in\R^k$ and $y_n\in\R^k$ converge, so they are bounded, i.e. there is an $R\gt0$ such that $\|x_n\|\le R$, and $\|y_n\|\le R$ for all $n\in\N$. We also have \begin{align*} \left|x_n\cdot y_n - x\cdot y\right| &=\left| x_n\cdot y_n - x_n\cdot y + x_n\cdot y - x\cdot y\right| \\ &\le \left| x_n\cdot y_n - x_n\cdot y\right| + \left|x_n\cdot y - x\cdot y\right| & \text{triangle }\le\\ &\le \left| x_n\cdot (y_n - y)\right| + \left|(x_n - x)\cdot y\right| \\ &\le \| x_n\|\,\|y_n - y\| + \|x_n - x\|\,\| y\| & \text{Cauchy }\le \\ &\le R\| y_n - y\| + R \|x_n - x\| \end{align*} Since $x_n$ and $y_n$ converge we have \[ \lim_{n\to\infty} R\| y_n - y\| + R \|x_n - x\| = 0 \] and hence \[ \lim_{n\to\infty} \left|x_n\cdot y_n - x\cdot y\right| =0. \]
  6. Let $A\subset\R$, and let $\{x_n\}_{n\in\N}$ be a sequence of upper bounds for $A$, i.e. assume that every $x_n$ is an upper bound for $A$. Show that if $x_n\to x$, then $x$ is also an upper bound for $A$.
    Suppose $x$ is not an upper bound for $A$. Then there is a number $a\in A$ with $a\gt X$. Let $\varepsilon= a-x$. Since $x_n\to X$, there is an $N$ such that for all $n\ge N$ one has $|x_n-x|\lt \varepsilon$. This implies, for $n\ge N$, that $x_n \lt x+\varepsilon = a$. Hence $x_n$ is not an upper bound for $A$ if $n\ge N$, contradicting what was given.
  7. [Honors only] For any two points $x,y\in\R^n$ we define \[ d_1(x, y) = |x_1-y_1| + \cdots + |x_n-y_n|. \]
    1. Show that $d_1$ defines a metric on $\R^n$.
    2. Let $d_2$ be the usual Euclidean metric. Prove that for any two points $x,y\in\R^n$ one has \[ d_2(x, y) \leq d_1(x,y) \leq \sqrt{n}\;d_2(x, y). \]
    3. Prove that a sequence of points $\{x_i\in\R^n\}_{i\in\N}$ converges in the $d_1$ distance if and only if it converges in the Euclidean metric. (Notation: denote the components of $x_i\in\R^n$ by $(x_{i1}, \ldots, x_{in})$.)

Homework 4

Due in class Wednesday March 8.
  1. Let $E = \{x\in\mathbb{Q} : 0\le x\le 1\}$, i.e. $E$ consists of all rational numbers in the interval $[0,1]$. Find a sequence of points in $E$ that has no subsequence that converges to some point in $E$.
  2. For each of the following sets say if it is compact or not. Give a short reason. You may quote theorems from the textbook, or from class (if they are from lecture then you must include the statement). All these sets are subsets of the real line or the euclidean plane $\R^2$. In all cases the intended metric is the usual Euclidean distance $d(x,y) = |x-y|$.
    1. $E = (-\infty, 0]$.
      Not compact, because compact sets are always bounded.
    2. $E = \{1\} \cup \{ \frac{n}{n+2} : n\in\N\}$
      The set is bounded because it is contained in $[0,1]$. It is also closed, because its only limit point is $1$, and $1\in E$. Therefore $E$ is compact.
  3. Consider the parabolic segment $E = \{(x, x^2) \in\R^2 : |x|\le 1\}$. Show directly from the definition of a compact subset (summary here) that $E$ is a compact subset of $\R^2$. (Hint: show that if $t_i\in\R$ is a sequence of real numbers with $t_i\to a$, then the points $(t_i, t_i^2)\in\R^2$ converge to the point $(a, a^2)\in\R^2$.)
    Suppose $(x_n, x_n^2)$ is a sequence of points in $E$. Then the corresponding sequence of $x$–coordinates is bounded, and by Bolzano–Weierstrass, there is a subsequence $x_{n_k}$ that converges. If $x_{n_k}\to a$, then $a\in [-1,1]$ because each $x_{n_k}$ lies in this interval. Moreover, the sequence of $y$–coordinates $x_{n_k}^2$ also converges, with $x_{n_k}^2 \to a^2$. Since $x_{n_k}\to a$ and $x_{n_k}^{2} \to a^2$, the points $(x_{n_k}, x_{n_k}^2)$ converge in $\R^2$ to the point $(a, a^2)$, which lies in $E$.

    We therefore have shown that any sequence of points in $E$ has a subsequence that converges to a point in $E$. This mean by definition that $E$ is compact.

  4. Let $\R\supset I_1\supset I_2 \supset I_3 \supset \cdots$ be a decreasing sequence of intervals, i.e. each $I_k\subset \R$ is an interval and for every $k\in\N$ we have $I_{k+1}\subset I_k$. Suppose each $I_k$ is open and nonempty. Is it true that the intersection $\cap_{k\in\N} I_k$ is not empty?
    The intersection can be empty as the following example shows: let $I_n = (0, \frac1n)$. Then no real number $x$ lies in all of the $I_n$ by the Archimedean property. So $\cap_{n\in\N}I_n$ is the empty set.
  5. (Honors only) There is a metric space $X$ which contains a countable set of points $E = \{x_n : n\in\N\}$ for which the distance between any two points is $d(x_n, x_m) = 1$ (for all $n\neq m$).
    1. Show that for any $p\in X$ the ball $B_{1/2}(p)$ cannot contain more than one point of the set $E$.
    2. Can $X$ be $\R^2$ or $\R^3$? (Hint: try to draw a set with three points in the plane so that the distance between any two of them is exactly 1. Then try to draw such a set with four points. Can you find four points in three dimensional Euclidean space that all have unit distance to each other?)
    3. Show that $E$ is a closed and bounded subset of $X$.
    4. Show that $E$ is not a compact subset of $X$.
    5. Show that $E$ is complete.

Homework 5

Due in class, Friday March 17.
  1. Let $x_n$ be a bounded sequence of real numbers.
    1. Suppose that $x_{n_k}$ is a convergent subsequence. Show that \[ \liminf x_n \leq \lim x_{n_k} \leq \limsup x_n. \]
    2. Show that there is a subsequence $x_{n_k}$ that converges to $\limsup x_n$.
    3. Suppose $A\gt \limsup x_n$. Show that $\{n\in\N \mid x_n\geq A\}$ is finite.
    4. Suppose $B\lt \limsup x_n$. Show that $\{n\in\N \mid x_n\geq B\}$ is infinite.
  2. In this problem assume that square roots of non–negative real numbers have been shown to exist and have been shown to satisfy the properties familiar from algebra. Let $x_n = \sqrt{n}$.
    1. Show that $x_{n+1}-x_n\to0$.
    2. Is $x_n$ a Cauchy sequence?

    a.  $0\leq \sqrt{n+1}-\sqrt{n} = \frac{1}{\sqrt{n+1}+\sqrt{n}} \lt \frac1{\sqrt n}$ so $x_n\to0$.

    b.  Cauchy sequences in $\R$ always converge, so $x_n=\sqrt n$ cannot be a Cauchy sequence. Alternatively, Cauchy sequences in any metric space are bounded, so $x_n=\sqrt n$ cannot be a Cauchy sequence.

    1. Show that for any two bounded sequences of real numbers $x_n$ and $y_n$ one has \[ \limsup \bigl(x_n+y_n\bigr) \leq \limsup x_n + \limsup y_n. \]
      By definition \[ \limsup_{n\to\infty}(x_n+y_n) = \lim_{n\to\infty} \Bigl(\sup\{x_k+y_k : k\ge n\}\Bigr). \] For every $k\ge m$ we have \[ x_k \le \sup\{x_l : l\ge m\}, \qquad y_k \le \sup\{y_l : l\ge m\}. \] Hence, for all $k\ge m$, \[ x_k+y_k \le \sup\{x_l : l\ge m\} + \sup\{y_l : l\ge m\}. \] The quantity on the right does not depend on $k$ and thus it is an upper bound for $\{x_k+y_k : k\ge m\}$. It is therefore not smaller than the least upper bound: \[ \sup\{x_k+y_k : k\ge m\} \le \sup\{x_l : l\ge m\} + \sup\{y_l : l\ge m\}. \] If we now take the limit for $m\to\infty $ on both sides we get \[ \limsup_{k\to\infty} x_k+y_k \le \limsup_{k\to\infty} x_k + \limsup_{k\to\infty} y_k. \]
    2. Find two sequences $x_n$ and $y_n$ with $\limsup x_n = \limsup y_n =1$ and $\limsup (x_n+y_n) =0$.
      There are many choices. One possible choice is $x_n=(-1)^n$, and $y_n=-x_n$. Then $\limsup x_n =\limsup y_n =1$ while $x_n+y_n = 0$ for all $n\in\N$, so that $\limsup (x_n+y_n) = 0$.
    3. If $A, B\subset \R$ are bounded and nonempty sets of real numbers,, and if $C = \{a+b \mid a\in A, b\in B\}$, is it then possible that $\sup C \lt \sup A + \sup B$? (This part of the problem addresses the difference between sets of real numbers and sequences of real numbers.)
      This was a problem in Homework 1.
  3. Let $x_n$ be the sequence defined by $x_1=1$, and $x_{n+1} = \frac{3x_n}{1+x_n}$.
    1. Prove by induction that $x_n\leq 2$ for all $n\in\N$.
    2. Prove by induction that $x_n$ is an increasing sequence.
    3. Compute $\lim x_n$.
    (Hint: consider the function $f(x) = 3x/(1+x) = 3 - 3/(1+x)$, and show that $x\lt x'$ implies $f(x)\lt f(x')$ for all $x, x'\geq0$. Also consider $f(1)$ and $f(2)$. )

Homework 6

This homework will not be collected, but is part of the practice problems for the upcoming midterm.
  1. Suppose $X$ is a metric space, and that $x_n\in X$ is a sequence in $X$ for which $\lim_{n\to\infty} d(x_n, x_{n+1}) = 0$. Is such a sequence always a Cauchy sequence?
  2. Which of the following series converge? In this problem assume that the standard functions from calculus (sine, cosine, natural logarithm, exponential function) have been defined and that their familiar properties have been proved.
    1. $\sum_{n=1}^\infty \frac{n}{n^3+1}$
    2. $\sum_{n=1}^\infty \left(\frac{n}{2n^2-1}\right)^2$
    3. $\sum_{n=1}^\infty \frac{1}{n (\ln n)^2}$
    4. $\sum_{n=1}^\infty \frac{\ln n}{n^2}$
    5. $\sum_{n=1}^\infty \frac{(-1)^{n+1}}{\ln (n+1)}$
    6. $\sum_{n=1}^\infty \frac{\sin(an)}{\ln (n+1)}$ where $a\in\R$ is some constant.
  3. For which values of $x\in\R$ does the series \[ f(x) \stackrel {\sf def}= \sum_{n=1}^\infty \frac{x^n}{1+x^{2n}} \] converge?
  4. Let $\sum a_n$ and $\sum b_n$ be two convergent series.
    1. Show that $\sum (a_n+b_n)$ also converges.
      We have to show that the partial sums of $\sum (a_n+b_n) $ converge. The partial sums are
      \begin{multline}s_n = \sum_{k=1}^n (a_n+b_n) = (a_1+b_1) +(a_2+b_2) + \cdots +(a_n+b_n)\\ = (a_1+a_2+\cdots +a_n) + (b_1+b_2+\cdots +b_n). \end{multline}
      Let $n\to\infty$ in this equation, and we get
      $\displaystyle \sum_{k=1}^\infty (a_n+b_n) = \sum_1^\infty a_k + \sum_1^\infty b_k $
    2. If the two series $\sum a_n$ and $\sum b_n$ converge absolutely, does it follow that $\sum (a_n+b_n)$ also converges absolutely?
      If $\sum a_n$ and $\sum b_n$ converge absolutely, then $\sum|a_n|$ and $\sum|b_n|$ both converge, and therefore $\sum c_n$ with $c_n=|a_n|+|b_n|$ also converges by part (a) of this problem. The triangle inequality implies $|a_n+b_n|\leq |a_n|+|b_n| = c_n$ and since $\sum c_n$ converges, the comparison test then implies that $\sum|a_n+b_n|$ also converges. This means, by definition, that the series $\sum(a_n+b_n)$ converges absolutely.
    3. Suppose the series $\sum a_n$ converges. Does it follow that $\sum a_{2n}$ converges?
      No. Consider the series $a_n = (-1)^{n+1}/n$. Then
      $\sum a_n = 1-\frac{1} {2} + \frac13 - \frac14 + \frac15 - \cdots$
      which converges (alternating series test), while
      $\sum a_{2n} = -\frac12 - \frac14 - \frac16 - \frac18 - \cdots$
      which diverges (e.g. condensation test).
    4. Suppose the series $\sum a_n$ converges absolutely. Does it follow that $\sum a_{2n}$ converges absolutely?
      Yes. We have to show that $\sum|a_{2n}|$ converges. Since the terms in this series are positive, we only have to check that the partial sums are bounded. This follows from the fact that $\sum|a_n| \lt \infty$:
      $\sum_{k=1}^n |a_{2k}| = |a_2| + |a_4| + \cdots + |a_{2n}| \leq |a_1|+|a_2| + |a_3| + \cdots + |a_{2n}| \leq \sum_{k=1}^\infty |a_k|$.

Homework 7

Due in class, Monday April 10.

In the following problems we let $X$ and $Y$ be a metric spaces, and write $d$ for the metric on both spaces.

  1. Suppose $f:X\to Y$ is a function for which there is a number $C\gt0$ such that

    $ d(f(p), f(q)) \leq C d(p, q)$ holds for all $p,q\in X$. $(*)$

    Show that $f$ is continuous. (The condition $(*)$ is called a Lipschitz condition.)
    We prove continuity directly from the $\epsilon$-$\delta$ definition. To prove that $f$ is continuous at $p\in X$, let $\epsilon\gt0$ be given. Choose $\delta \lt \epsilon/C$. Then for all $x\in B_\delta(p)$ one has \[ d(f(x), f(p)) \leq C d(x, p) = C \delta \lt \epsilon, \] so that $f(x) \in B_\epsilon(f(p))$. This implies $f$ is continuous at $p$.
  2. Let $p\in X$ be some given point, and consider the function $f(x) = d(p,x)$. Show that $f$ is continuous by proving that $f$ satisfies a Lipschitz condition (as in the previous problem).
    We first show that $f(x)\leq f(y) + d(x, y)$ holds for all $x,y\in X$ by using the triangle inequality:
    $f(x) = d(x,p) \leq d(x,y) + d(y,p) = d(x,y) + f(y)$.
    Thus
    $f(x)-f(y)\leq d(x, y)$
    for all $x,y\in X$. Switching $x$ and $y$ around we find that
    $f(y) - f(x) \leq d(y,x) = d(x,y)$.
    Together, these two inequalities imply $|f(x) - f(y)|\leq d(x,y)$.
  3. Let $E\subset X$ be a subset and consider
    $f(x) = \inf \{ d(x,p)\mid p\in E\} $
    Prove that $f$ satisfies a Lipschitz condition.
    We again prove $f(x)-f(y)\leq d(x,y)$ for all $x,y\in X$. Let $x$ and $y$ be given. By definition of $f$ we have $f(x) \leq d(x, p)$ for every $p\in E$. The triangle inequality implies
    $f(x) = d(x, p) \leq d(x, y) + d(y, p)$ i.e. $f(x) - d(x, y) \leq d(y,p)$.
    This holds for all $p\in E$, solutions
    $f(x) - d(x, y) \leq \inf \{d(y,p) \mid p\in E\} = f(y)$.
    Therefore $f(x) - f(y) \leq d(x, y)$.

    Since this holds for all $x,y$ we also have $f(y)-f(x) \leq d(y,x) = d(x,y)$, and thus

    $|f(x) - f(y)| \leq d(x,y)$.

  4. Suppose $K\subset X$ is compact, and let $p\in X$ be some point that does not belong to $K$. Show that there is a point $q\in K$ that is nearest to $p$, i.e. a point $q\in K$ such that $d(p, q')\geq d(p,q)$ holds for all $q'\in K$.
  5. If $X=Y=[0, \infty)$, does $f(x) =\sqrt{x}$ satisfy a Lipschitz condition?
    No. If $f$ did satisfy a Lipschitz condition then there would be a constant $C$ such that
    $\frac{|f(x) - f(y)|} {|x-y|} \leq C$
    would hold for all $x, y$ with $x\neq y$. In particular this would hold if $y=0$. Thus
    $\frac{|f(x) - f(y)|} {|x-y|} = \frac{\sqrt x - 0} {x-0} =\frac{1} {\sqrt x} \leq C$
    for all $x\gt0$. This is not true.
  6. (Honors only)
    1. Suppose $x_n\in\R$ is a sequence that converges to $\lim x_n=L$ and consider the sequence $a_n = (x_1+\cdots+x_n)/n$ (i.e. $a_n$ is the average of the first $n$ numbers in the sequence $x_n$). Show that $a_n\to L$.
    2. If $x_n\in\R$ is any sequence and $a_n = (x_1+\cdots+x_n)/n$ converges, does it then follow that the sequence $x_n$ also converges?
    3. Suppose $a_n$ is a bounded sequence of real numbers, and define for any $r\in (0, 1)$ the quantity \[ A(r) = (1-r) \sum_{n=1}^\infty a_n r^n. \] Show that the series defining $A(r)$ converges if $0\leq r\lt 1$. Then show that if the sequence $a_n$ converges, then \[ \lim_{r\nearrow1} A(r) = \lim_{n\to\infty} a_n \]
    4. Compute $A(r)$ and $\lim_{r\nearrow1}A(r)$ in the case that $a_n=1$ for odd $n$ and $a_n=0$ for even $n$.

Homework 8

Due in class Wednesday April 19.
    1. If $X$ is a metric space, and $f:X\to \R$ is a continuous function then show that the zeroset $\{x\in X \mid f(x)=0\}$ is a closed subset of $X$ (hint: use this Theorem and read the corresponding example.)
      This the corollary to theorem 4.8 in Rudin (page 87.)
    2. If $f:\R\to\R$ is any function, then the graph of $f$ is by definition the set \[ G_f = \bigl\{(x,y) \in\R^2 : y=f(x)\bigr\}. \] Use part a. of this problem to prove that if $f$ is continuous, then $G_f$ is a closed subset of $\R^2$.
    The function $F:\R^2\to\R$ defined by $F(x, y) = y-f(x)$ is continuous. Its zeroset is the graph $G_f$, so the graph must be a closed set.

    To prove that $F(x,y) = y-f(x)$ is continuous one can argue as follows: the two functions $(x,y)\to x $ and $ (x,y)\to y $ are the coordinate functions, so we know they are continuous (we proved this here).

    The function $f(x)$, seen as a function on $\R^2$, is the composition of the coordinate function $ (x, y)\to x $ and the given function $x\to f(x)$, which is ggiven to be continuous. Therefore $ (x, y)\in\R^2 \to f(x) \in\R $ is continuous.

    Finally, $y-f(x)$ is the difference of continuous functions, which is again continuous.

  1. Use the Mean Value Theorem to prove the following statements about continuous functions $f, g:[a,b]\to\R$ that are differentiable on $(a, b)$:
    1. If $f'(x)=0$ for all $x\in (a, b)$ then $f$ is constant.
    2. If $f'(x)\lt g'(x)$ for all $x\in (a, b)$, and if $f(a) = g(a)$ then $f(x)\lt g(x)$ for all $x\in (a, b]$.
      An approach that doesn’t work: We apply the MVT to the function $f$ and also to the function $g$ on the interval $[a,x]$. We find that there are $c_1, c_2\in (a, x)$ such that
      $\frac{f(x)-f(a)} {x-a} = f'(c_1)$ and $\frac{g(x)-g(a)}{x-a} = g'(c_2)$.
      and therefore
      $f(x) = f(a) + f'(c_1) (x-a)$ and $g(x) = g(a) + g'(c_2)(x-a)$.
      We could now hope to use $f(a) = g(a)$ and $f'(c)\lt g'(c)$ to compare the expressions for $f(x)$ and $g(x)$, but this runs into a problem because the derivatives $f'$ and $g'$ are eveluated at different points: the points $c_1$ and $c_2$ that we got by applying the MVT might be different. The fix is to change the argument so that we only apply the MVT once, and so that all derivatives are evaluated at the same “mystery point” $c\in(a, x)$.

      An approach that does work.  Consider the function $h(x) = g(x)-f(x)$. We know that $h(a)=0$ and that $h'(x) = f'(x) - g'(x) \gt0$ for all $x\in(a, b)$. Apply the MVT to $h$ on the interval $[a,x]$: there is a $c\in(a, b)$ such that

      $\frac{h(x)-h(a)} {x-a} = h'(c) \gt 0$
      Hence $h(x)-h(a)\gt0$, and, since $h(a)=0$, $h(x)\gt0$. By definition of $h$ this implies $f(x) \gt g(x)$.

    3. Assume that there is a number $M\in\R$ such that $|f'(x)| \leq M$ for all $x\in (a, b)$. Show that $f$ is uniformly continuous by proving that $f$ satisfies a Lipschitz condition.
  2. Let $f:\R\to\R$ be the function defined by $f(x) = 0$ for all $x\neq 0$, and $f(0) = 1$. Show that $f$ is Riemann integrable on the interval $[-1, 1]$ and prove from the definition that $\int_{-1}^1 f(x) dx = 0$. (hint: use this criterion for Riemann integrability and look for step functions $s\leq f\leq t$ whose partition points are $0, -r, r, 1$ and choose $r$ appropriately.)
  3. [Honors only] Let $f:\R\to\R$ be a uniformly continuous function. Show that there is a constant $C\gt 0$ such that $f(x) \le Cx$ for all $x\ge 1$. (Hint: what does the definition of uniform continuity say when $\epsilon=1$?)
  4. [Honors only]  Consider the following variation on the Weierstrass function: \[ f(x) = \sum_{n=-\infty}^\infty 2^{-n} \sin \bigl(3^n x\bigr)\,. \] Note that the difference with the original Weierstrass function is that here the sum is taken over all $n\in\Z$, including all negative numbers.

    Show that the series converges for all $x\in \R$. (Hint: to deal with the terms where $n\lt 0$ use that $|\sin\theta|\leq|\theta|$ holds for all $\theta\in\R$)

    Also show that the graph of $f$ is self similar, in the sense that $f(3x) = 2f(x)$ holds for all $x\in\R$.

Homework 9

Due in class Wednesday, May 3.
  1. Consider the sequence of functions $f_n:\R\to\R$ defined by \[ f_n(x) = \frac{1}{1+nx^2}. \]
    1. Show that the sequence of functions converges pointwise and find the limit function $f$.
      We compute $\lim_{n\to\infty}f_n(x)$ for each $x\in\R$.

      If $x\neq0$ then $1+nx^2\to\infty$ as $n\to\infty$, so $1/(1+nx^2)\to0$.

      If $x=0$ then $f_n(x) = f_n(0) = 1$ for every $n\in\N$, so $f_n(0)\to 1$ as $n\to\infty$.

      The limit function is therefore

      $f(x) = \lim_{n\to\infty} f_n(x) = \begin{cases} 1 & (x=0) \\ 0 (x\neq0) \end{cases}$
      We have shown that the functions $f_n(x)$ converge pointwise to $f(x)$.

    2. Does $f_n$ converge uniformly to $f$?
      No, the sequence does not converge uniformly.

      One way to show this is to compute $\du(f_n, f) = \sup_{x\in\R}|f_n(x) - f(x)|$. It turns out that $\du(f_n, f)=1$ for all $n\in\N$, and therefore $\du(f_n, f)\not\to0$ as $n\to\infty$, which implies that $f_n$ does not converge uniformly.

      In this problem there is an easier, indirect argument: all of the functions $f_n$ are continuous. If the convergence of $f_n$ to $f$ were uniform then the limit function $f$ would also have to be continuous. But $f$ is not continuous at $x=0$. Therefore $f_n$ does not converge uniformly to $f_n$ on $\R$.

    3. Does the sequence converge uniformly on the interval $[1,2]$?
      Yes, it does. In this case the uniform distance between $f_n$ and $f$ is
      $\displaystyle \du(f_n, f) \stackrel{\sf def}= \sup_{1\leq x\leq 2}|f_n(x) - f(x)|$
      Note that the difference with part (b) of this problem lies in the range of $x$es over which the supremum is taken.

      In this case $f(x)=0$ for all $x\in[1,2]$, and the difference $|f_n(x)-f(x)|$ takes its maximum at $x=1$. Therefore $\du(f_n, f) = 1/(1+n)$, and $\du(f_n, f) \to 0$ as $n\to\infty$. This implies that $f_n(x)\to f(x)=0$ uniformly on $[1,2]$.

  2. Consider the series \[ f(x) = \sum_{n=1}^\infty \frac{x^{n}}{n^2}. \]
    1. Show that the series converges uniformly for $|x|\leq a$ for any $a\lt1$.
    2. Does the series converge uniformly for $|x|\lt1$?
    Let $M_n = \frac1{n^2}$. Then $|f_n(x)|\leq M_n$ for all $x\in [-1,1]$. Also, $\sum_{n=1}^\infty M_n = \sum n^{-2} \lt\infty$. Therefore the Weierstrass M–test applies, and we conclude that the series converges uniformly for $|x|\leq 1$.

    The same arguments, with the same choice of $M_n$ apply to the first question, i.e. the series converges absolutely for $|x|\leq a$ if $0\lt a\lt1$.

  3. Consider the series \[ f(x) = \sum_{n=0}^\infty x^{n}. \]
    1. use the Weierstrass M-test to show that the series converges uniformly for $|x|\leq a$ for any $a\lt1$.
      Let $M_n = a^n$. Then the $n$th term $f_n(x) = x^n$ satisfies $|f_n(x)| \leq M_n$ for all $x\in [-a, a]$. When $a\lt 1$ the series $\sum M_n = \sum _{n=0}^\infty a^n = 1/(1-a)$ converges, and thus Weierstrass’M–test implies that $\sum_{n=0}^\infty x^n$ converges uniformly on $[-a, a]$.
    2. Does the series converge uniformly for $|x|\lt 1$?
      If the series converges uniformly, the the terms must go to zero uniformly, meaning that
      $\displaystyle\lim_{n\to\infty} \Bigl(\sup_{|x|\lt1} |f_n(x)|\Bigr) =0$.
      We compute for our problem that
      $\displaystyle\sup_{|x|\lt 1} |f_n(x)| = \sup_{|x|\lt 1}|x|^n = 1$
      Therefore
      $\displaystyle\lim_{n\to\infty} \Bigl(\sup_{|x|\lt1} |f_n(x)|\Bigr) =1 \neq 0$
      which implies that the series does not converge uniformly.
  4. Consider the series \[ f(x) = \sum_{n=0}^\infty x^{n^2}. \]
    1. Show that the series converges uniformly for $|x|\leq a$ for any $a\lt1$.
      If $|x|\lt 1$ then, since $n^2\ge n$ for all integers $n\geq0$ one has $|x^{n^2}| = |x|^{n^2} \leq |x|^n$. Since the geometric series $\sum |x|^n$ converges if $|x|\lt1$, the comparison test implies that $\sum_0^\infty x^{n^2} $ converges for every $x\in (-1, 1)$. This proves that the series converges pointwise on the interval $(-1,1)$.

      To prove uniform convergence on the interval $[-a, a]$, where $0\lt a\lt1$, we use the Weierstrass M–test. Choose $M_n = a^{n}$. Then we have for all $x$ with $|x|\leq a$ that $|x^{n^2}| \leq a^{n^2} \leq a^n$. We also have $\sum_{n=0}^\infty M_n =\sum a^n \lt\infty$. Weierstrass’ conditions are therefore met, and we can conclude that the series converges uniformly on $(-1, 1)$.

    2. Does the series converge uniformly for $|x|\lt1$?
      No, it does not. The reason is the same as for the previous problem. If the series did converge uniformly on $(-1, 1)$, then the terms would have to go to zero uniformly, i.e.
      $\lim_{n\to\infty} \bigl(\sup_{|x|\lt 1} |x^{n^2}|\bigr) = 0$.
      But $\sup_{|x|\lt 1} |x^{n^2}| = 1$, so
      $\lim_{n\to\infty} \bigl(\sup_{|x|\lt 1} |x^{n^2}|\bigr) = 1 \neq 0$,
      and therefore the series does not converge uniformly on $(-1,1)$.
  5. The Bessel function of order zero is defined by \[ J_0(x) = \sum_{n=0}^\infty \bigl(-\tfrac14\bigr)^{n}\frac{x^{2n}}{(n!)^2} \]
    1. Show that the series converges pointwise, i.e. for all $x\in\R$.
    2. Let $L\gt0$ be any number, and show that the series converges uniformly on the interval $[-L,L]$.
    3. Show that $J_0$ is a continuous function.
    4. Show that $J_0$ is a differentiable function.
    5. Show that $J_0'$ is a differentiable function.
  6. Show that Weierstrass’ function \[ f(x) = \sum_{n=0}^\infty 2^{-n} \sin (3^nx) \] is continuous by first proving that the series converges uniformly on $\R$.
    Let $M_n = 2^{-n}$. Then the $n$th term $f_n(x) = 2^{-n}\sin(3^nx)$ satisfies $|f_n(x)|\leq M_n$ for every $x\in\R$, and for every $n\geq0$. Since $\sum_0^\infty M_n = \sum_0^\infty 2^{-n} = 2$ converges, the Weierstrass M-test applies, and we can conclude that the series converges uniformly on $\R$.

    Each term is continuous (assuming that continuity of the Sine function as known), so therefore the sum of the series is also a continuous function.

  7. Let $\cC$ be the set of all continuous functions $f:[0, 2\pi]\to \R$. Let
    $\du(f, g) = \sup_{0\leq x\leq 2\pi} |f(x)-g(x)|$
    be the distance function on $\cC$.
    1. Consider the functions $f_n(x) = \sin (2^n x)$ where $n\geq0$ is an integer. Show that $\du(f_n, f_m) \geq 1$ whenever $n\neq m$.
      Assume that $n\lt m$. For $a=2^{-n-1}\pi$ one has $f_n(a) = \sin \frac\pi2 = 1$ and $f_m(a) = \sin 2^{m-n-1}\pi =0$. Therefore
      $\du(f_n, f_m) = \sup_{0\leq x\leq 2\pi} |\sin 2^nx - \sin 2^mx| \geq |\sin 2^na - \sin 2^ma| = 1-0 =1.$
    2. Is the set $A = \{f_n : n=0,1,2,3,\dots\} = \{\sin x, \sin 2x, \sin 4x, \sin 8x, \dots\}$ a closed subset of $\cC$?
      Yes. To prove this using Rudin’s definition of “closed” we must show that every limit point of the set $A$ belongs to $A$. This particular set $A$ turns out not to have any limit points, so that it is closed by default.

      To see that $A$ has no limit points, suppose $g\in\cC$ is a limit point of $A$, and consider the neighborhood $B_{1/2}(g)$. This neighborhood cannot contain more than one point from $A$, for if $f_n\in B_{1/2}(g)$ and $f_m\in B_{1/2}(g)$ then

      $\du(f_n, f_m) \leq \du(f_n, g) + \du(g, f_m) \lt \frac12 + \frac12 = 1$,
      which can only happen if $n=m$.

      If $B_{1/2}(g)$ contains one point $f_n\neq g$ from $A$, then let $r=\du(g, f_n)$; otherwise let $r=\frac12$. In either case $B_r(g)$ contains no points from $A$, except possibly $g$.

    3. Is the set $A$ a bounded subset of $\cC$?
      Yes. The distance between any two functions is at most 2:
      $\du(f_n, f_m) = \sup_{x\in[0, 2\pi]}|\sin 2^nx - \sin 2^mx| \leq 1+1 = 2.$
      Therefore every $f_n$ belongs to the ball with radius 3 centered at $f_0$, i.e. $A\subset B_3(f_0)$. This implies that $A$ is bounded.
    4. Is the set $A$ compact?
      No, $A$ is not compact. The sequence $(f_n)_{n\in\N}$ lies in $A$, but it has no convergent subsequence. If $f_{n_k}$ were a convergent subsequence, then it would have to be a Cauchy sequence, which implies that there is an $N\in\N$ such that for all $k,\ell\geq N$ one has $\du(f_{n_k}, f_{n_\ell}) \le 1$ (definition of Cauchy sequence with $;\epsilon=1$). But if we choose $k\neq \ell$ then $\du(f_{n_k}, f_{n_\ell}) \geq 1$. The contradiction shows that $f_{n_k}$ cannot be a convergent subsequence, and thus the set $A$ is not compact.
  8. [Honors only] Let $X$ be a metric space and $f_n:X\to\R$ a sequence of functions. Let $A, B \subset X$ be subsets of $X$.
    1. If $f_n$ converges uniformly on $X$, then show that $f_n$ converges uniformly on $A$.
    2. Suppose that $f_n$ converges uniformly on both subsets $A$ and $B$. Show that $f_n$ converges uniformly on $A\cup B$.
  9. [Honors only] Show that the series \[ f(x) = \sum_{n=1}^\infty (-1)^{n+1} \frac{x^n}{\sqrt n} \] converges uniformly for $x\in[0,1]$.