Altanis

2.6Closed Sets and Limit Points

Updated 29 Jun 2026Chapter (PDF)

[2.6.1]Definition(Closed Set)#

A subset AA of a topological subspace XX is said to be closed if its complement XAX \setminus A is open.

[2.6.2]Example(Examples of Closed Sets)#

Consider a closed interval [a,b]R[a, b] \subseteq \bR equipped with the standard topology. Then note [a,b]C=(,a)(b,)[a, b]^C = (-\infty, a) \cup (b, \infty) is an open set, and so [a,b][a, b] is closed.

In general, for some ([x1,y1]××[xn,yn])Rn([x_1, y_1] \times \cdots \times [x_n, y_n]) \subseteq \bR^n, we observe the complement is open, so closed boxes in Rn\bR^n are closed (as per the name).

In the topological space endowed with the discrete topology, every subset is open, and so every subset is closed.

[2.6.3]Theorem#

Let XX be a topological space. Then:

  1. \emptyset and XX are closed.

  2. The finite union of closed sets is closed.

  3. The arbitrary intersection of closed sets is closed.

Proof.
  1. Note C=X\emptyset^C = X and XC=X^C = \emptyset are both open, so they're both closed.

  2. Let A1,,AnA_1, \dots, A_n be finitely many closed sets. Then

    (k=1nAk)C=X(k=1nAk)=k=1n(XAk)\left( \bigcup_{k = 1}^n A_k \right)^C = X \setminus \left( \bigcup_{k = 1}^n A_k \right) = \bigcap_{k = 1}^n (X \setminus A_k)

    is the finite intersection of open sets, which is open. Thus the finite union is closed.

  3. Let {Aλ}λΛ\{A_\lambda\}_{\lambda \in \Lambda} be a collection of closed sets. Then

    (λΛAk)C=X(λΛAk)=λΛ(XAk)\left( \bigcap_{\lambda \in \Lambda} A_k \right)^C = X \setminus \left( \bigcap_{\lambda \in \Lambda} A_k \right) = \bigcup_{\lambda \in \Lambda} (X \setminus A_k)

    is the arbitrary union of open sets, which is open. Thus the arbitrary intersection is closed.

Now we define a closed set with respect to a topology. If YY is a subspace of XX, then a set AYA \subseteq Y is closed in YY if AA is closed in the subspace topology of YY (equivalently, if YAY \setminus A is open in YY).

[2.6.4]Theorem(Criterion for Closed Set in Subspace)#

Let XX be a topological space, and let YXY \subseteq X be a subspace. Then a subset AYA \subseteq Y is closed if and only if AA is the intersection of a closed subset of XX with YY.

Proof.

\Longrightarrow: Suppose AA is closed in YY. By definition, its complement YAY \setminus A is open in YY. By the definition of the subspace topology, there exists an open set UU in XX such that YA=UYY \setminus A = U \cap Y. Then,

A=Y(YA)=Y(UY)=YU=Y(XU).A = Y \setminus (Y \setminus A) = Y \setminus (U \cap Y) = Y \setminus U = Y \cap (X \setminus U).

Since UU is open in XX, its complement XUX \setminus U is closed in XX. Letting C=XUC = X \setminus U, we have A=CYA = C \cap Y, where CC is a closed set in XX.

\Longleftarrow: Suppose A=CYA = C \cap Y, where CC is a closed set in XX. We wish to show that AA is closed in YY. Consider the complement of AA in YY:

YA=Y(CY)=YC=Y(XC)open set in X.Y \setminus A = Y \setminus (C \cap Y) = Y \setminus C = Y \cap \underbrace{(X \setminus C)}_{\text{open set in } X}.

Since CC is closed in XX, its complement XCX \setminus C is open in XX. By the definition of the subspace topology, the intersection of an open set in XX with YY is open in YY. Therefore, YAY \setminus A is open in YY, which implies that AA is closed in YY, completing the backward direction.

Note this is analogous to the fact that a set open in a subspace topology is simply an open set in the parent topology intersected with the subspace.

[2.6.5]Theorem#

Let XX be a topological space, and let YXY \subseteq X be a subspace. If a subset AYA \subseteq Y is closed in YY, and YY is closed in XX, then AA is closed in XX.

Proof.

AA is the intersection of some closed subset of XX with YY, which is also a closed subset of XX—thus AA is closed in XX.

[2.6.6]Definition(Interior, Closure of Set)#

Let AA be a subset of a topological space XX. The interior of AA is said to be the union of all open subsets of AA, denoted Int(A)\Int(A). The closure of AA is said the be the intersection of all closed subsets of XX that contain AA, denoted A\bar{A}.

[2.6.7]Remark(Relationship between Interior and Closure)#

Note the interior of a subset AA of a topological space XX is the union of every single open subset of AA, which is strictly the largest open subset of AA. Analogously, the closure of a subset AA of a topological space XX is the intersection of all closed subsets of XX that contain AA, which is strictly the smallest closed superset of AA. Then

Int(A)AA.\Int(A) \subseteq A \subseteq \bar{A}.

Of course, if AA is open, then Int(A)=A\Int(A) = A, and analogously if AA is closed, then A=AA = \bar{A}.

Let AA be a subset of the topological space XX, but also a subset of a subspace YXY \subseteq X. Of course, the closure of AA as a subset of YY is not necessarily the same as the closure of AA as a subset of XX.

[2.6.8]Theorem(Closure of Subset under Subspace)#

Suppose XX is a topological space, YY is a subspace of XX, and AYXA \subseteq Y \subseteq X. Then the closure of AA under YY, say ClY(A)\Cl_Y(A) equals ClX(A)Y\Cl_X(A) \cap Y.

Proof.

We show ClY(A)=ClX(A)Y\Cl_Y(A) = \Cl_X(A) \cap Y by proving a forward and backward inclusion.

\subseteq: Suppose xClY(A)x \in \Cl_Y(A). Then note xx is in every closed superset of AA contained in YY, and of course ClX(A)Y\Cl_X(A) \cap Y is one example, so xClX(A)Yx \in \Cl_X(A) \cap Y.

\supseteq: Suppose xClX(A)Yx \in \Cl_X(A) \cap Y. Then let CC be any arbitrary closed superset of AA such that ACYA \subseteq C \subseteq Y. If xCx \in C, then it follows that xClY(A)x \in \Cl_Y(A). Since CC is closed in YY, and YY is a subspace of XX (and thus endowed with the subspace topology), there is some closed subset CXXC_X \subseteq X such that C=CXYC = C_X \cap Y. But A(C=CXY)A \subseteq (C = C_X \cap Y), implying ACXA \subseteq C_X. But ClX(A)\Cl_X(A) is the minimal superset of AA that is closed, so AClX(A)CXA \subseteq \Cl_X(A) \subseteq C_X. Thus xCx \in C, completing the proof.

The definition of the closure of a set, while elegant, is impractical for explicit computation. For a topological space XX and subset AA, one would need to identify every closed subset of XX containing AA, then intersect them all—a collection far too large to work with directly. Instead, we may characterize membership in A\bar{A} in a more tractable ways: via the concept of limit points.

[2.6.9]Definition(Neighborhood)#

A set UU is said to be a neighborhood containing xx if UU is an open set containing xx.

[2.6.10]Theorem(Characterization of Closure of Set)#

Suppose AA is a subset of the topological space XX. Then the following statements are true.

  1. xAx \in \bar{A} if and only if every neighborhood UU containing xx intersects AA.

  2. xAx \in \bar{A} if and only if each BBB \in \mathcal{B}—where B\mathcal{B} is a basis for the topology of XX—containing xx intersects AA.

Proof.
  1. We proceed by contrapositive. That is, we show that xAx \notin \bar{A} if and only if there exists some open neighborhood UXU \subseteq X of xx. First, suppose xAx \notin \bar{A}. Note XAX \setminus \bar{A} is an open subset of XX that must contain xx, but does not intersect AA by construction. Conversely, suppose some open neighborhood UXU \subseteq X of xx exists that doesn't intersect AA. Then XUX \setminus U is a closed superset of AA that does not contain xx, and so xAx \notin \bar{A}, completing the proof.

  2. Basis elements that contain xx are open neighborhoods about xx, and open neighborhoods about xx that intersect AA contain basis elements.

[2.6.11]Remark(Neighborhood Characterization of Closure)#

From real analysis, recall the open interval A=(0,1)RA = (0, 1) \subseteq \bR (with R\bR endowed with the standard topology), and note A=[0,1]\bar{A} = [0, 1]. Now consider 0A0 \in \bar{A}. (1)(1) from the previous theorem says that all neighborhoods about 00 must intersect AA. Indeed, any open interval (ε,ε)A(-\epsilon, \epsilon) \cap A \ne \emptyset, and so 00 is in the closure for AA.

For some set AA, our theorem says that the closure A\bar{A} contains AA, as well as additional points that, when taking an open neighborhood around, always intersect AA. We formalize this idea and relate them back to the closure.

[2.6.12]Definition(Limit Point)#

Suppose AA is a subset of a topological space XX. We say xXx \in X is a limit point if every open neighborhood UXU \subseteq X about xx intersects AA at some point other than xx itself. That is, for every open set UXU \subseteq X with xUx \in U, it follows that U(A{x})U \cap (A \setminus \{x\}) \ne \emptyset. Said differently, xx is a limit point if it belongs to the closure of A{x}A \setminus \{x\}.

[2.6.13]Theorem(Closure and Limit Points)#

Let AA be a subset of the topological space XX. Let AA' be the set of all limit points of AA. Then

A=AA.\bar{A} = A \cup A'.
Proof.

We prove equality by forward and backward inclusion.

\subseteq: Suppose xAx \in \bar{A}. If xAx \in A, then xAAx \in A \cup A'. Otherwise, if xAx \notin A, then A=A{x}A = A \setminus \{x\}. But note that, since xAx \in \bar{A}, every open neighborhood UXU \subseteq X of xx intersects AA nontrivially, so U(A{x})U \cap (A \setminus \{x\}) \ne \emptyset. Thus xAAAx \in A' \subseteq A \cup A'.

\supseteq: Suppose xAAx \in A \cup A'. If xAx \in A, xAx \in \bar{A}, obviously. Now suppose xAx \notin A', and so xAx \in A'. Analogous to before, this means A{x}=AA \setminus \{x\} = A. Since xAx \in A', for every open neighborhood UXU \subseteq X of xx, U(A{x})=UA0U \cap (A \setminus \{x\}) = U \cap A \ne 0. Thus xAx \in \bar{A}, completing the proof.

[2.6.14]Corollary#

A subset of a topological space is closed if and only if it contains its limit points.

Proof.

Let AA be a subset of topological space XX, and let AA' be the set of limit points. First, suppose AA is closed. Then A=A=AAA = \bar{A} = A \cup A', meaning AAA' \subseteq A. Now suppose AA contains all its limit points, and so AAA' \subseteq A. Then AAAA' \cup A \subseteq A and AAAA \subseteq A' \cup A, meaning A=AA=AA = A \cup A' = \bar{A}, completing the proof.

[2.6.15]Remark(Prelude to Hausdorffness)#

Recall the standard topology on R\bR. A few nice properties are evident in its study.

  1. For each x0Rx_0 \in \bR, the singleton {x0}\{x_0\} is closed. For any x:xx0x: x \ne x_0, we may construct an open neighborhood U=(xε,x+ε)U = (x - \epsilon, x + \epsilon) specifically to enforce x0Ux_0 \notin U by choosing ε>0\epsilon > 0 to be small enough.

  2. For any convergent real sequence (xn)R(x_n) \subseteq \bR, we say xnxx_n \to x if, for every open neighborhood UU of x, there is some NZ+N \in \bZ_+ such that xnUx_n \in U for every nNn \ge N. In R\bR, this limit is unique—that is, if xnx1x_n \to x_1 and xnx2x_n \to x_2, then x1=x2x_1 = x_2.

These don't transfer to general topological spaces though. Consider the triplet S={a,b,c}S = \{a, b, c\} endowed with topology {,{b},{a,b},{b,c},S}\{\emptyset, \{b\}, \{a, b\}, \{b, c\}, S\}.

  1. We see that {b}\{b\} is not closed, since S{b}={a,c}S \setminus \{b\} = \{a, c\} is not in the topology (i.e., not open).

  2. Moreover, with R\bR's definition of convergence, let's define xn=bx_n = b and see what it converges to. For every open neighborhood of bb ({a,b},{b},{b,c},{a,b,c}\{a, b\}, \{b\}, \{b, c\}, \{a, b, c\}), we see xn=bx_n = b is in each neighborhood for every nZ+n \in \bZ_+. The open neighborhoods of aa ({a,b},{a,b,c}\{a, b\}, \{a, b, c\}) and cc ({b,c},{a,b,c}\{b, c\}, \{a, b, c\}) also contain bb. Thus the conclusion is that lim(xn){a,b,c}\lim(x_n) \in \{a, b, c\}, which does not agree with R\bR's topological properties.

There is some property of R\bR's standard topology that makes it nicer than SS's topology, in that singletons are closed and convergent sequences have a unique limit.

[2.6.16]Definition(Hausdorffness)#

A topological space XX is said to be Hausdorff if, for any pair of distinct points x1,x2Xx_1, x_2 \in X, there exists two disjoint neighborhoods U1,U2XU_1, U_2 \subseteq X about x1x_1 and x2x_2 respectively.

[2.6.17]Theorem(Singletons are Closed in Hausdorff Space)#

Every singleton subset of a topological space XX is closed.

Proof.

Suppose {x0}X\{x_0\} \subseteq X. Let xXx \in X be such that xx0x \ne x_0. By Hausdorffness of XX, there are disjoint open neighborhoods U1,U2XU_1, U_2 \subseteq X about xx and x0x_0. Thus xAx \notin \bar{A}. This implies A={x0}\bar{A} = \{x_0\}, meaning the singleton is closed.

The property that singleton subsets of a topological space are closed is weaker than the space being Hausdorff. We extract this condition as the T1T_1 axiom.

[2.6.18]Definition(T1T_1 Space)#

A topological space XX is said to be T1T_1 if every singleton subset of XX is closed.

From a geometric perspective, a space is T1T_1 if for any two distinct points x,yx, y, there exists an open neighborhood of xx that does not contain yy. The space is Hausdorff if, for any two distinct points xx and yy, there exist disjoint open neighborhoods of xx and yy. Thus Hausdorffness is a stronger condition than T1T_1. We will see shortly that although T1T_1 spaces preserve our intuition about limit points, T1T_1 alone is not strong enough to guarantee the uniqueness of sequence limits, whereas Hausdorff spaces do.

[2.6.19]Theorem#

Let AA be a subset of the topological space XX satisfying the T1T_1 axiom. Then xx is a limit point of AA if and only if every neighborhood of XX contains infinitely many points of AA.

Proof.

()(\Longrightarrow): Suppose xx is a limit point of AA. Then there is some open neighborhood UXU \subseteq X of xx such that (U(A{x})=S)(U \cap (A \setminus \{x\}) = S) \ne \emptyset. Suppose SS is finite, so S={x1,,xn}S = \{x_1, \dots, x_n\}. Note that SS is a finite union of singleton subsets of XX, and since singletons are closed in T1T_1 spaces, SS is closed. Thus XSX \setminus S is open. Let U=U(XS)U' = U \cap (X \setminus S). Then UU' is open, xUx \in U, but U(A{x})=U \cap (A \setminus \{x\}) = \emptyset. We have constructed an open neighborhood about xx that does not intersect A{x}A \setminus \{x\}, meaning xx is not a limit point, forcing a contradiction. Thus U(A{x})U \cap (A \setminus \{x\}) is infinite.

()(\Longleftarrow): Every open neighborhood about xx intersects AA infinitely amount of times, so every open neighborhood about xx intersects A{x}A \setminus \{x\} infinitely many times, meaning xx is a limit point.

[2.6.20]Theorem(Uniqueness of Limits in Hausdorff Space)#

Suppose XX is a Hausdorff topological space. Then a sequence (xn)n=1X(x_n)_{n = 1}^\infty \subseteq X converges to, at most, one limit point.

Proof.

Suppose xnx_n converges, so xnxx_n \to x. Let x0xx_0 \ne x: we show xn↛x0x_n \not \to x_0, confirming xx is the only limit for the sequence. Since xx0Xx \ne x_0 \in X, a Hausdorff space, we can construct disjoint open neighborhoods UU and U0U_0 about xx and x0x_0 respectively such that UU0=U \cap U_0 = \emptyset. Recall that since xnxx_n \to x, there exists some NZ+N \in \bZ_+ such that xnUx_n \in U for every nNn \ge N. But since U0U_0 is disjoint from UU, it follows that the only terms of the sequence U0U_0 can possibly contain are x1,,xN1x_1, \dots, x_{N - 1}. In other words, the tail for (xn)(x_n) is not contained in U0U_0, so xn↛x0x_n \not \to x_0, completing the proof.

2.6.1Problems#

[2.6.21]Problem#

Let XX be an ordered set endowed with the order topology. Show that (a,b)[a,b]\bar{(a, b)} \subseteq [a, b]. Under what conditions does equality hold?

Solution.

Note (a,b)\bar{(a, b)}, by definition, is the smallest closed superset of (a,b)(a, b). Thus, if [a,b][a, b] is a closed superset of (a,b)(a, b), then (a,b)[a,b]\bar{(a, b)} \subseteq [a, b] follows immediately. Note that [a,b]c=(,a)(b,)[a, b]^c = (\leftarrow, a) \cup (b, \rightarrow) is a union of two open sets (so it's open), making [a,b][a, b] closed, completing the proof.

Equality is achieved when [a,b](a,b)[a, b] \subseteq \bar{(a, b)}, since we have forward and backward inclusion. Since [a,b]=(a,b){a}{b}[a, b] = (a, b) \cup \{a\} \cup \{b\}, simply knowing that a,ba, b are limit points of (a,b)(a, b) forces equality. We show aa (resp. bb) is a limit point of (a,b)(a, b) if aa (resp. bb) has no immediate successor (resp. immediate predecessor).

Suppose aa has no immediate successor. Let UXU \subseteq X be an open neighborhood about aa. Then note that we can locally refine UU by choosing some basis element of the order topology, VV, such that aVUa \in V \subseteq U. Note a basis element is of the form V=(a,c)V = (a, c), where c:a<cbc: a < c \le b. In any case, since aa has no immediate successor, there is some x<cx < c for which xVUx \in V \subseteq U. This means xU((a,b){a})=U(a,b)x \in U \cap ((a, b) \setminus \{a\}) = U \cap (a, b), meaning aa is a limit point for (a,b)(a, b).

[2.6.22]Problem#

Let A,BA, B be subsets of a topological space XX. Prove the following.

  1. If ABA \subseteq B, then AB\bar{A} \subseteq \bar{B}.

Solution.
  1. Suppose ABA \subseteq B. Let xAx \in \bar{A}. If xAx \in A, then xBBx \in B \subseteq \bar{B} trivially, so suppose xAx \notin A. Then xx is a limit point, so every open neighborhood UXU \subseteq X about xx is such that U(A{x})U \cap (A \setminus \{x\}) \ne \emptyset. But ABA \subseteq B, so this implies U(B{x})U \cap (B \setminus \{x\}) \ne \emptyset, meaning xx is a limit point of BB too, so xBx \in \bar{B}.

[2.6.23]Problem#

Show that every order topology is Hausdorff.

Solution.

Suppose an ordered set XX is endowed with the order topology. Let xyXx \ne y \in X, and without loss of generality, suppose x<yx < y. Suppose yy is xx's immediate successor: then there exist open neighborhoods U=(,y)xU = (-\infty, y) \ni x and V=(x,)yV = (x, \infty) \ni y where UV=U \cap V = \emptyset. If yy is not xx's immediate successor, there exists t:x<t<yt: x < t < y. Then U=(,t)xU = (-\infty, t) \ni x and V=(t,)yV = (t, \infty) \ni y such that UV=U \cap V = \emptyset, completing the proof.

[2.6.24]Problem#

Show the product of two Hausdorff spaces is Hausdorff.

Solution.

Let (x1,y1)(x2,y2)X×Y(x_1, y_1) \ne (x_2, y_2) \in X \times Y, where XX and YY are Hausdorff topological spaces. Without loss of generality, suppose x1x2x_1 \ne x_2. By Hausdorffness of XX, there exists open neighborhoods UXx1U_X \ni x_1 and VXx2V_X \ni x_2 such that UXVX=U_X \cap V_X = \emptyset. There also exists some open neighborhood Vy1,y2V \ni y_1, y_2. Then (UX×V)(x1,y1)(U_X \times V) \ni (x_1, y_1) and (VX×V)(x2,y2)(V_X \times V) \ni (x_2, y_2) are disjoint open neighborhoods of X×YX \times Y, completing the proof.

[2.6.25]Problem#

Show the subspace of a Hausdorff space is Hausdorff.

Solution.

Suppose XX is a Hausdorff space and YXY \subseteq X is a subspace. Let xyYx \ne y \in Y. Note there exists open neighborhoods UXU \subseteq X of xx and VXV \subseteq X of yy such that UV=U \cap V = \emptyset. The open analogue of these neighborhoods in the subspace are sets of the form UYU \cap Y and VYV \cap Y. Of course, xUYx \in U \cap Y and yVYy \in V \cap Y, and note (UY)(VY)=(UV)Y=Y=(U \cap Y) \cap (V \cap Y) = (U \cap V) \cap Y = \emptyset \cap Y = \emptyset, completing the proof.

[2.6.26]Problem#

Show that a topological space XX is Hausdorff if and only if the diagonal Δ={(x,x):xX}\Delta = \{(x, x): x \in X\} is closed in X×XX \times X.

Proof.

()(\Longrightarrow): Suppose XX is Hausdorff. Suppose (x,y)X×X(x, y) \in X \times X is a limit point for Δ\Delta. Then, for every open neighborhood UX×XU \subseteq X \times X of (x,y)(x, y), it follows that U(Δ(x,y))U \cap (\Delta \setminus (x, y)) \ne \emptyset. For the sake of contradiction, suppose xyx \ne y. Then, by Hausdorffness of XX, there exist disjoint open neighborhoods UXU \subseteq X and VXV \subseteq X such that xUx \in U and yVy \in V. Note then that U×VU \times V is open in X×XX \times X. Since UU and VV are disjoint, there are no tuples with the same entry in each slot when taking their Cartesian product—that is, (U×V)(Δ(x,y))=(U \times V) \cap (\Delta \setminus (x, y)) = \emptyset, which is a contradiction. Thus x=yx = y, implying Δ\Delta contains all its limit points and thus is closed.

()(\Longleftarrow): Suppose Δ\Delta is closed, and so Δc\Delta^c is open. Let xyXx \ne y \in X, then observe (x,y)Δc(x, y) \in \Delta^c. Since Δc\Delta^c is an open subset of X×XX \times X, there is some basis element U×VU \times V (where U,VU, V are open subsets of XX) such that (x,y)U×VΔc(x, y) \in U \times V \subseteq \Delta^c. For U×VU \times V to be a subset of Δc\Delta^c, this implies no elements of U×VU \times V may be in Δ\Delta: thus UV=U \cap V = \emptyset. Thus we have constructed two disjoint open neighborhoods that contain xx and yy respectively, completing the proof.

[2.6.27]Problem#

Endow R\bR with the finite complement topology. For what points of R\bR does xn=1/nx_n = 1/n converge to.

Solution.

Note a set ORO \subseteq \bR is open if O=O = \emptyset, or RO\bR \setminus O is finite (i.e., it misses finitely many points of R\bR). Let xRx \in \bR be arbitrary, and let ORO \subseteq \bR be an arbitrary open neighborhood such that xOx \in O. Note OO can only miss finitely many elements of R\bR, so OO must contain infinitely many elements of xnx_n. That is, there is some NZ+N \in \bZ_+ such that xnOx_n \in O for every nNn \ge N. Thus xnxx_n \to x, but xx is arbitrary, so (xn)(x_n) converges to every real number.

[2.6.28]Definition(Boundary)#

Suppose XX is a topological space and AXA \subseteq X. We define the boundary of AA to be the set

A=AXA.\partial A = \bar{A} \cap \bar{X \setminus A}.
[2.6.29]Remark(Intuition Behind Boundary)#

Suppose XX is a topological space and AXA \subseteq X. Recall that if xAx \in \bar{A}, then every open neighborhood UXU \subseteq X about xx is such that UAU \cap A \ne \emptyset. Equivalently, if xXAx \in \bar{X \setminus A}, then every open neighborhood UXU \subseteq X about xx is such that U(XA)U \cap (X \setminus A) \ne \emptyset. Thus, if x(AXA)=Ax \in (\bar{A} \cap \bar{X \setminus A}) = \partial A, every open neighborhood UXU \subseteq X about xx must have nontrivial intersection with AA and XAX \setminus A. Graphically, drawing an open neighborhood about a point leads to it spilling inside and outside the set.

[2.6.30]Problem#

Suppose XX is a topological space and AXA \subseteq X. Show that Int(A)\Int(A) and A\partial A are disjoint.

Solution.

Suppose xInt(A)x \in \Int(A). Since Int(A)A\Int(A) \subseteq A, it follows that xAx \in A. For the sake of contradiction, suppose xAx \in \partial A, implying xXAx \in \bar{X \setminus A}. Note XA=(XA)(XA)X \setminus A = (X \setminus A) \cap (X \setminus A)' (where SS' refers to the limit points of SS). Since xAx \in A, x(XA)x \notin (X \setminus A), so it is only possible that x(XA)x \in (X \setminus A)'. This implies that every open subset UXU \subseteq X must be such that U(Ac{x})U \cap (A^c \setminus \{x\}) \ne \emptyset. But note xInt(A)Ax \in \Int(A) \subseteq A, meaning every open neighborhood of xx must be fully contained in AA, which is a contradiction. Thus Int(A)A=\Int(A) \cap \partial A = \emptyset.

[2.6.31]Problem#

Suppose XX is a topological space and AXA \subseteq X. Show A=Int(A)A\bar{A} = \Int(A) \cup \partial A. That is, A\bar{A} is the disjoint union of the interior and boundary of AA.

Solution.

We prove this by forward and backward inclusion. First, suppose xAx \in \bar{A}, and suppose xInt(A)x \notin \Int(A). This implies there are no XX-open neighborhoods of xx that are subsets of AA. Indeed, this implies the only XX-open neighborhoods of xx must intersect XAX \setminus A. Thus xXAx \in \bar{X \setminus A}, so xAx \in \partial A. Thus every xAx \in \bar{A} must be either in Int(A)\Int(A) or A\partial A, and so AInt(A)A\bar{A} \subseteq \Int(A) \cup \partial A. Conversely, suppose xInt(A)Ax \in \Int(A) \cup \partial A. If xInt(A)x \in \Int(A), then xAAx \in A \subseteq \bar{A}. If x(A=AXA)x \in (\partial A = \bar{A} \cap \bar{X \setminus A}), then xAx \in \bar{A}, completing the proof.

[2.6.32]Problem#

Suppose XX is a topological space and AXA \subseteq X. Show that A=\partial A = \emptyset if and only if AA is clopen.

Solution.

Suppose AA is XX-clopen. Since AA is closed, A=AA = \bar{A}. Let xA=Ax \in A = \bar{A} be a limit point of AA. Then note every XX-open neighborhood OxO \ni x is such that OAO \subseteq A, so no open neighborhoods intersect XAX \setminus A. Thus xXAx \notin \bar{X \setminus A}, implying A=\partial A = \emptyset. Conversely, suppose A=AXA=\partial A = \bar{A} \cap \bar{X \setminus A} = \emptyset. If xx is a limit point of AA, then xAx \in \bar{A}, then it cannot be in XAX \setminus A and so xAx \in A, meaning AA is closed. The same argument can be made to show XAX \setminus A is also closed, meaning AA is open as well, completing the proof.

[2.6.33]Problem#

Suppose XX is a topological space and AXA \subseteq X. Show that UU is open if and only if U=UU\partial U = \bar{U} \setminus U.

Solution.

()(\Longrightarrow): Suppose UU is open. We prove the forward direction by forward and backward inclusion. To get started, suppose xUUx \in \bar{U} \setminus U. Then note xUx \notin U, and so xXUXUx \in X \setminus U \subseteq \bar{X \setminus U}, and so xUx \in \partial U. Conversely, suppose xUx \in \partial U. Immediately, we know xUx \in \bar{U}. For the sake of contradiction, suppose xUx \in U. Since UU is open, there is some basic XX-open neighborhood OO such that xOUx \in O \subseteq U. But O(XU)=O \cap (X \setminus U) = \emptyset, so xXUx \notin \bar{X \setminus U}, a contradiction. Thus xUx \notin U, meaning xUUx \in \bar{U} \setminus U. Thus U=UXU=UU\partial U = \bar{U} \cap \bar{X \setminus U} = \bar{U} \setminus U.

()(\Longleftarrow): Suppose U=UU\partial U = \bar{U} \setminus U. To show UU is open, we show XUX \setminus U is closed, or that XU=XUX \setminus U = \bar{X \setminus U}. Let xXUx \in \bar{X \setminus U} be a limit point for XUX \setminus U. Note that if xUx \in U, then xUx \in \bar{U} and so xUXUx \in \bar{U} \cap \bar{X \setminus U}, but xUUx \notin U \setminus \bar{U}, which is contradictory—thus xXUx \in X \setminus U, completing the proof.