Homework 1

Table of Contents

Problem 1

Prove the remaining implications in Proposition 1.1.18:

Let (x(k))k\p{x^{\p{k}}}_k be a sequence in Rn\R^n.

  1. If (x(k))k\p{x^{\p{k}}}_k converges with respect to d1d_{\ell^1}, then (x(k))k\p{x^{\p{k}}}_k converges with respect to d2d_{\ell^2}.
  2. If (x(k))k\p{x^{\p{k}}}_k converges with respect to dd_{\ell^\infty}, then (x(k))k\p{x^{\p{k}}}_k converges with respect to d1d_{\ell^1}.
  3. If (x(k))k\p{x^{\p{k}}}_k converges with respect to d1d_{\ell^1}, then (x(k))k\p{x^{\p{k}}}_k converges with respect to dd_{\ell^\infty}.
Solution.
  1. Given x,yRnx, y \in \R^n,

    (d1(x,y))2=(i=1nxiyi)2=i=1nxiyi2+ijxiyixjyj0i=1nxiyi2=(d2(x,y))2.\begin{aligned} \p{d_{\ell^1}\p{x, y}}^2 = \p{\sum_{i=1}^n \abs{x_i - y_i}}^2 &= \sum_{i=1}^n \abs{x_i - y_i}^2 + \underbrace{\sum_{i \neq j} \abs{x_i - y_i} \abs{x_j - y_j}}_{\geq 0} \\ &\geq \sum_{i=1}^n \abs{x_i - y_i}^2 \\ &= \p{d_{\ell^2}\p{x, y}}^2. \end{aligned}

    Thus, because metrics are non-negative,

    d2(x,y)d1(x,y).d_{\ell^2}\p{x, y} \leq d_{\ell^1}\p{x, y}.

    In particular, if limkd1(x(k),x(0))=0\lim_{k\to\infty} d_{\ell^1}\p{x^{\p{k}}, x^{\p{0}}} = 0, then by the squeeze theorem,

    d2(x(k),x(0))d1(x(k),x(0))k0.d_{\ell^2}\p{x^{\p{k}}, x^{\p{0}}} \leq d_{\ell^1}\p{x^{\p{k}}, x^{\p{0}}} \xrightarrow{k\to\infty} 0.
  2. Given x,yRnx, y \in \R^n,

    d1(x,y)=i=1nxiyii=1nmaxj=1,,nxjyj=nd(x,y).\begin{aligned} d_{\ell^1}\p{x, y} = \sum_{i=1}^n \abs{x_i - y_i} \leq \sum_{i=1}^n \max_{j=1,\ldots,n} \abs{x_j - y_j} &= n d_{\ell^\infty}\p{x, y}. \end{aligned}

    The claim follows from the squeeze theorem like in part (1).

  3. Given x,yRnx, y \in \R^n, for any 1in1 \leq i \leq n, we have

    xiyij=1nxjyj=d1(x,y).\abs{x_i - y_i} \leq \sum_{j=1}^n \abs{x_j - y_j} = d_{\ell^1}\p{x, y}.

    Thus, because this inequality holds uniformly for all 1in1 \leq i \leq n,

    d(x,y)=maxi=1,,nxiyid1(x,y).d_{\ell^\infty}\p{x, y} = \max_{i=1,\ldots,n} \abs{x_i - y_i} \leq d_{\ell^1}\p{x, y}.

    The claim follows from the squeeze theorem like in part (1).

Problem 3

Prove Proposition 1.2.10.

Proposition (1.2.10)

Let (X,d)\p{X, d} be a metric space, let EE be a subset of XX, and let x0x_0 be a point in XX. Then the following statements are logically equivalent.

  1. x0x_0 is an adherent point of EE.
  2. x0x_0 is either an interior point or a boundary point of EE.
  3. There exists a sequence (xn)n=1\p{x_n}_{n=1}^\infty in EE which converges to x0x_0 with respect to the metric dd.
Solution.

We will show (i)     \iff (ii), and then (i)     \implies (iii)     \implies (i).

(i)     \iff (ii):

Recall that X=Int(E)EExt(E)X = \Int\p{E} \cup \partial E \cup \Ext\p{E} and that this is a disjoint union. Thus, it suffices to show that x0Ext(E)x_0 \notin \Ext\p{E}. Notice that

(r>0:B(x0,r)E)    ¬(r>0:B(x0,r)E=).\p{\forall r > 0 : B\p{x_0, r} \cap E \neq \emptyset} \iff \neg \p{\exists r > 0 : B\p{x_0, r} \cap E = \emptyset}.

In other words,

x0 is an adherent point    xExt(E).x_0\text{ is an adherent point} \iff x \notin \Ext\p{E}.

(i)     \implies (iii):

By definition,

r>0:B(x0,r)E.\forall r > 0 : B\p{x_0, r} \cap E \neq \emptyset.

Thus, for each nNn \in \N, we have B(x0,1n)EB\p{x_0, \frac{1}{n}} \cap E \neq \emptyset. Let xnB(x0,1n)Ex_n \in B\p{x_0, \frac{1}{n}} \cap E (this is where we use (countable) choice, but this detail isn't that important to us in this class.) Thus, we get a sequence (xn)nE\p{x_n}_n \subseteq E such that

d(xn,x0)1nn0,d\p{x_n, x_0} \leq \frac{1}{n} \xrightarrow{n\to\infty} 0,

so (xn)n\p{x_n}_n converges to x0x_0 by the squeeze theorem.

(iii)     \implies (i):

Let r>0r > 0. Since (xn)n\p{x_n}_n converges to x0x_0, there exists nNn \in \N such that

d(xn,x0)<r    xnB(x0,r).d\p{x_n, x_0} < r \implies x_n \in B\p{x_0, r}.

Since xnEx_n \in E by assumption, this means xnB(x0,r)Ex_n \in B\p{x_0, r} \cap E, i.e., this is a non-empty set.

Problem 5

Prove Proposition 1.3.4(b).

Proposition (1.3.4)

Let (X,d)\p{X, d} be a metric space, let YY be a subset of XX, and let EE be a subset of YY.

  1. EE is relatively open with respect to YY if and only if E=VYE = V \cap Y for some set VXV \subseteq X which is open in XX.
  2. EE is relatively closed with respect to YY if and only if E=KYE = K \cap Y for some set KXK \subseteq X which is closed in XX.
Solution.

While Solution 1 is much easier than Solution 2, I highly recommend that you understand Solution 2 as well since it requires more understanding of the metric space definitions--it's very good practice.

Throughout this problem, I'll be using Proposition 1.2.15 (the long one) many times without mentioning it explicitly.

Solution 1 (using part (i))

"    \implies"

Assume EE is relatively closed with respect to YY. Then YEY \setminus E is open in YY, so by part (i), there exists VXV \subseteq X which is open in XX such that YE=VYY \setminus E = V \cap Y. Notice that

E=Y(YE)=Y(X(VY))=Y((XV)(XY))=(Y(XV))(Y(XY))==Y(XV).\begin{aligned} E &= Y \setminus \p{Y \setminus E} \\ &= Y \cap \p{X \setminus \p{V \cap Y}} \\ &= Y \cap \p{\p{X \setminus V} \cup \p{X \setminus Y}} \\ &= \p{Y \cap \p{X \setminus V}} \cup \underbrace{\p{Y \cap \p{X \setminus Y}}}_{=\emptyset} \\ &= Y \cap \p{X \setminus V}. \end{aligned}

Since VV was open in XX, it follows that KXVK \coloneqq X \setminus V is closed in XX.

"    \impliedby"

Assume that E=KYE = K \cap Y for some set KXK \subseteq X which is closed in XX. Then by a similar computation as above,

YE=Y(KY)=Y(XK).Y \setminus E = Y \setminus \p{K \cap Y} = Y \cap \p{X \setminus K}.

Since KK is closed in XX, VXKV \coloneqq X \setminus K is open in XX. Thus, part (i), YEY \setminus E is open in YY, and hence EE is closed in YY.

Solution 2 (from the definitions)

"    \implies"

Assume that EE is closed in YY. We let KEK \coloneqq \cl{E}, where the closure is taken with respect to XX. We immediately have EEYE \subseteq \cl{E} \cap Y, so it remains to check the reverse inclusion.

Let x0EYx_0 \in \cl{E} \cap Y. We will show that x0x_0 is an adherent point of EE in (Y,dY×Y)\p{Y, d_{Y \times Y}}. Let r>0r > 0. Notice that because x0Yx_0 \in Y,

BY(x0,r)=BX(x0,r)Y.B_Y\p{x_0, r} = B_X\p{x_0, r} \cap Y.

Thus,

BY(x0,r)E=BX(x0,r)EY=BX(x0,r)E(EY).(x0E)\begin{aligned} B_Y\p{x_0, r} \cap E &= B_X\p{x_0, r} \cap E \cap Y \\ &= B_X\p{x_0, r} \cap E && \p{E \subseteq Y} \\ &\neq \emptyset. && \p{x_0 \in \cl{E}} \end{aligned}

Since rr was arbitrary, this means x0x_0 is an adherent point of EE in YY. But EE is closed in YY, so x0Ex_0 \in E.

"    \impliedby"

Assume that E=KYE = K \cap Y, where KK is a closed set in XX. We will show that EE contains all of its adherent points in YY.

Let y0Yy_0 \in Y be an adherent point of EE in (Y,dY×Y)\p{Y, d_{Y \times Y}}. Since E=KYE = K \cap Y, we only need to show that y0Ky_0 \in K. Moreover, KK is closed in (X,d)\p{X, d}, so it further suffices to show that y0y_0 is an adherent point of KK in (X,d)\p{X, d}. Let r>0r > 0 and notice that

E=KY    EK,BY(y0,r)=BX(y0,r)Y    BY(y0,r)BX(y0,r).\begin{gathered} E = K \cap Y \implies E \subseteq K, \\ B_Y\p{y_0, r} = B_X\p{y_0, r} \cap Y \implies B_Y\p{y_0, r} \subseteq B_X\p{y_0, r}. \end{gathered}

In particular,

BX(y0,r)KBY(y0,r)E,B_X\p{y_0, r} \cap K \supseteq B_Y\p{y_0, r} \cap E,

which is non-empty since y0y_0 is an adherent point of EE in (Y,dY×Y)\p{Y, d_{Y \times Y}}. Thus, y0y_0 is an adherent point of KK, which completes the proof.