Continuous maps and homeomorphisms
On the line, continuity was defined with distances. A topology has only open sets. What should it mean for a map between two spaces to be continuous, and when are two spaces really the same space?
Where you are 6 in Topological spaces
Why this exists
A function f : \mathbb{R} \to \mathbb{R} is continuous at a when for every \varepsilon > 0 some \delta > 0 gives |f(x) - f(a)| < \varepsilon whenever |x - a| < \delta (Continuity, The epsilon delta definition of a limit). In words: every open interval about f(a) contains the image of some open interval about a. That sentence needs only open sets, so it makes sense between any two topological spaces.
The second question is when two spaces are the same. For graphs the answer is a Graph isomorphism: when two drawings are the same graph, a bijection matching edges with edges. For spaces it is a bijection matching open sets with open sets, a homeomorphism; through it every property stated with open sets alone passes from one space to the other.
Definition
Throughout, (X, \mathcal{T}) and (Y, \mathcal{T}') are topological spaces and f : X \to Y is a function (One-to-one, onto and bijective functions).
Preimage. For V \subseteq Y, the preimage of V under f is f^{-1}(V) = \{x \in X : f(x) \in V\}. It exists for every function, inverse or not; when f is a bijection it equals the image of V under the inverse function (Composing functions between sets, and when an inverse exists).
Continuous map. f is continuous if f^{-1}(V) \in \mathcal{T} for every V \in \mathcal{T}': the preimage of every open set is open. It is continuous at x \in X if every neighborhood V of f(x) has a neighborhood U of x with f(U) \subseteq V.
Open and closed maps. f is open if f(U) is open for every open U \subseteq X, and closed if f(C) is closed for every closed C \subseteq X.
Homeomorphism. f is a homeomorphism if it is a bijection and both f and its inverse f^{-1} : Y \to X are continuous. Then X and Y are homeomorphic, written X \approx Y.
Topological property. A property of spaces is topological if whenever X has it and X \approx Y, Y has it too.
Intuition
Continuity is tested by pulling back, not by pushing forward. The function h of Figure 1 jumps at x = 1, and the preimage of the open band V keeps a closed end at the jump.
A homeomorphism may stretch and bend without limit (Figure 2 stretches (-1, 1) onto \mathbb{R}), but it may not tear or glue: the map of Figure 3 is a continuous bijection whose inverse tears the circle open at (1, 0).
Results
Theorem 1 (preimages). For any function f : X \to Y, any nonempty family of sets V_i \subseteq Y, i \in I, and any V \subseteq Y,
For g : Y \to Z, (g \circ f)^{-1}(W) = f^{-1}(g^{-1}(W)). Moreover f(f^{-1}(V)) \subseteq V, with equality if f is surjective, and f^{-1}(f(U)) = U for every U \subseteq X if f is injective.
Theorem 2 (tests for continuity). For f : X \to Y the following are equivalent:
- (a) f is continuous;
- (b) f^{-1}(C) is closed for every closed C \subseteq Y;
- (c) f(\overline{A}) \subseteq \overline{f(A)} for every A \subseteq X;
- (d) f is continuous at every point of X;
- (e) f^{-1}(S) is open for every S in a subbasis \mathcal{S} generating \mathcal{T}', for instance a basis for \mathcal{T}'.
Theorem 3 (building continuous maps). (a) Constant maps and \mathrm{id}_X : (X, \mathcal{T}) \to (X, \mathcal{T}) are continuous. (b) Every map out of a discrete space and every map into an indiscrete space is continuous. (c) \mathrm{id}_X : (X, \mathcal{T}_1) \to (X, \mathcal{T}_2) is continuous if and only if \mathcal{T}_1 is finer than \mathcal{T}_2. (d) If f : X \to Y and g : Y \to Z are continuous, so is g \circ f.
Theorem 4 (metric spaces). Let (X, d) and (Y, d') be metric spaces with their metric topologies (Metric spaces, a first look). Then f is continuous if and only if for every x \in X and \varepsilon > 0 there is \delta > 0 with d'(f(x), f(t)) < \varepsilon whenever d(x, t) < \delta. For f : \mathbb{R} \to \mathbb{R} this is continuity as in Continuity.
Theorem 5 (homeomorphisms). (a) \mathrm{id}_X is a homeomorphism, and inverses and composites of homeomorphisms are homeomorphisms; so \approx is reflexive, symmetric and transitive. (b) A continuous bijection is a homeomorphism if and only if it is open, if and only if it is closed. (c) If f is a homeomorphism, then for all U, A \subseteq X and x \in X: U is open (closed) if and only if f(U) is open (closed); f(\overline{A}) = \overline{f(A)} and f(\operatorname{int} A) = \operatorname{int} f(A); and x is a limit point of A if and only if f(x) is a limit point of f(A).
Why it is true
Theorem 1. Proof. x \in f^{-1}\big(\bigcup V_i\big) if and only if f(x) lies in some V_i, that is, x lies in some f^{-1}(V_i); for intersections read “every” for “some”. Next, x \in f^{-1}(Y \setminus V) if and only if f(x) \notin V, and x \in (g \circ f)^{-1}(W) if and only if f(x) \in g^{-1}(W). Every point of f(f^{-1}(V)) is some f(x) \in V; if f is surjective, each v \in V is some f(x) with x \in f^{-1}(V). Always U \subseteq f^{-1}(f(U)); if f is injective and f(x) = f(u) with u \in U, then x = u \in U. \square
Theorem 2. Proof. (a) \Leftrightarrow (b): by Theorem 1, f^{-1}(Y \setminus C) = X \setminus f^{-1}(C), and the closed sets are the complements of the open sets.
(b) \Rightarrow (c): f^{-1}\big(\overline{f(A)}\big) is closed and contains A, so it contains \overline{A} (Theorem 2(a) of Closed sets, closure and limit points). Applying f and Theorem 1, f(\overline{A}) \subseteq \overline{f(A)}.
(c) \Rightarrow (b): let C be closed and A = f^{-1}(C). Since f(A) \subseteq C, f(\overline{A}) \subseteq \overline{f(A)} \subseteq \overline{C} = C, so \overline{A} \subseteq f^{-1}(C) = A, and A is closed.
(a) \Rightarrow (d): take U = f^{-1}(V). (d) \Rightarrow (a): let V be open. Each x \in f^{-1}(V) has a neighborhood U_x with f(U_x) \subseteq V, so U_x \subseteq f^{-1}(V), and f^{-1}(V) = \bigcup_x U_x is open.
(a) \Rightarrow (e) holds because subbasis members are open. (e) \Rightarrow (a): a basis element S_1 \cap \cdots \cap S_n has preimage f^{-1}(S_1) \cap \cdots \cap f^{-1}(S_n), open; an open V is a union of basis elements, so f^{-1}(V) is a union of open sets. A basis for \mathcal{T}' covers Y, and its finite intersections are open and include its members, so as a subbasis it generates \mathcal{T}'. \square
Theorem 3. Proof. (a) A constant map has preimages X or \varnothing, and \mathrm{id}_X^{-1}(V) = V. (b) In a discrete space every set is open; an indiscrete Y has open sets \varnothing, Y, with preimages \varnothing, X. (c) \mathrm{id}_X^{-1}(V) = V, so continuity says exactly \mathcal{T}_2 \subseteq \mathcal{T}_1. (d) (g \circ f)^{-1}(W) = f^{-1}(g^{-1}(W)), open when W is. \square
Theorem 4. Proof. Suppose the condition holds, V is open and x \in f^{-1}(V). Some ball B(f(x), \varepsilon) \subseteq V, and the condition gives \delta with f(B(x, \delta)) \subseteq V, so B(x, \delta) \subseteq f^{-1}(V): the preimage is open. Conversely, if f is continuous, the ball B(f(x), \varepsilon) is open, so its preimage is an open set containing x and contains some B(x, \delta), which is the condition. \square
Theorem 5. Proof. (a) \mathrm{id}_X is its own inverse, continuous by Theorem 3(a). If f is a homeomorphism with inverse g, then g is a bijection with inverse f, both continuous. If f : X \to Y and g : Y \to Z are homeomorphisms, g \circ f is a bijection with inverse f^{-1} \circ g^{-1} (Composing functions between sets, and when an inverse exists), and both are continuous by Theorem 3(d).
(b) Let g = f^{-1}. For U \subseteq X, g^{-1}(U) = \{y : g(y) \in U\} = f(U). So g is continuous if and only if f(U) is open for every open U, that is, f is open; using Theorem 2(b) instead, if and only if f is closed.
(c) If U is open, f(U) = g^{-1}(U) is open; if f(U) is open, U = f^{-1}(f(U)) is open (Theorem 1). Closed sets likewise. Theorem 2(c) gives f(\overline{A}) \subseteq \overline{f(A)}, and applied to g and f(A) it gives g\big(\overline{f(A)}\big) \subseteq \overline{A}; apply f to get the reverse inclusion. f(\operatorname{int} A) is open and lies in f(A), so it lies in \operatorname{int} f(A); the same with g gives g(\operatorname{int} f(A)) \subseteq \operatorname{int} A, the reverse inclusion. Finally, the neighborhoods of f(x) are exactly the sets f(U) with U a neighborhood of x, and since f is injective, f(U) \cap \big(f(A) \setminus \{f(x)\}\big) = f\big(U \cap (A \setminus \{x\})\big), empty exactly when U \cap (A \setminus \{x\}) is. \square
Worked examples
Alert levels in an exhibition hall
Hiro runs an exhibition hall with five galleries a, b, c, d, e. Three motion sensors cover S_1 = \{a, b\}, S_2 = \{b, c, d\} and S_3 = \{d, e\}. With the sensors as a subbasis, X = \{a, b, c, d, e\} is a space whose open sets are the regions the sensors can confirm a visitor is in. The alert panel shows a level in Y = \{0, 1, 2\}, with open sets \varnothing, \{2\}, \{1, 2\}, Y (the level is at least 2, 1, 0). (a) Is f continuous, where f(b) = 2, f(a) = f(c) = 1, f(d) = f(e) = 0? Is g, where g(b) = g(d) = 2, g(c) = g(e) = 1, g(a) = 0? (b) Samir mirrors the floor plan by \rho, which swaps a with e and b with d and fixes c. Is \rho : X \to X a homeomorphism? Is the swap \sigma of a and c?
SolutionThe preimages of \varnothing and Y are \varnothing and X, so only \{2\} and \{1, 2\} need testing. Only S_2 covers c, so every open set containing c contains S_2 = \{b, c, d\}.
(a) f^{-1}(\{2\}) = \{b\} = S_1 \cap S_2 is open, but f^{-1}(\{1, 2\}) = \{a, b, c\} contains c and not d, so it is not open, and f is not continuous. For g, g^{-1}(\{2\}) = \{b, d\} = (S_1 \cap S_2) \cup (S_2 \cap S_3) and g^{-1}(\{1, 2\}) = \{b, c, d, e\} = S_2 \cup S_3 are open, so g is continuous: the panel can be wired from the sensors under g, not under f.
(b) \rho is its own inverse, so \rho^{-1}(S) = \rho(S): \rho(S_1) = \{e, d\} = S_3, \rho(S_2) = S_2, \rho(S_3) = S_1. By Theorem 2(e) \rho is continuous, and so is \rho^{-1} = \rho: \rho is a homeomorphism. The bijection \sigma is not even continuous: \sigma^{-1}(S_1) = \{b, c\} contains c and not d.
A jump seen by two topologies
Let h(x) = 3 - x for x < 1 and \displaystyle h(x) = \frac{x}{2} for x \ge 1 (Figure 1). Show that h : \mathbb{R} \to \mathbb{R} is not continuous, but h : \mathbb{R}_{\ell} \to \mathbb{R} is, with the standard topology on the target.
SolutionTake V = (0, 1). For x < 1, h(x) = 3 - x > 2, never in V; for x \ge 1, \displaystyle \frac{x}{2} \in V exactly when x < 2. So h^{-1}(V) = [1, 2), which is not open in \mathbb{R}: every (1 - r, 1 + r) contains \displaystyle 1 - \frac{r}{2} \notin [1, 2). So h : \mathbb{R} \to \mathbb{R} is not continuous.
In \mathbb{R}_{\ell} use Theorem 2(d). Let V be a neighborhood of h(x), containing some (h(x) - \varepsilon, h(x) + \varepsilon). If x < 1, let \delta = \min\{\varepsilon, 1 - x\} and U = [x, x + \delta): for t \in U, t < 1 and |h(t) - h(x)| = t - x < \varepsilon. If x \ge 1, let U = [x, x + 2\varepsilon): for t \in U, \displaystyle |h(t) - h(x)| = \frac{t - x}{2} < \varepsilon. Each U is a neighborhood of x in \mathbb{R}_{\ell} with h(U) \subseteq V, so h : \mathbb{R}_{\ell} \to \mathbb{R} is continuous.
Likewise \mathrm{id} : \mathbb{R}_{\ell} \to \mathbb{R} is continuous (Theorem 3(c)), but its inverse \mathrm{id} : \mathbb{R} \to \mathbb{R}_{\ell} is not, as the preimage of [0, 1) is [0, 1). A continuous bijection need not be a homeomorphism.
An interval stretched onto the line
Give (-1, 1) and \mathbb{R} the metric |s - t|. Show that \displaystyle f(x) = \frac{x}{1 - x^2} is a homeomorphism from (-1, 1) onto \mathbb{R}, and deduce that every open interval (a, b) is homeomorphic to \mathbb{R}.
f is a quotient of polynomials whose denominator is positive on (-1, 1), so it is continuous in the \varepsilon-\delta sense of Continuity, hence continuous by Theorem 4.
Fix y. For x \in (-1, 1), f(x) = y means yx^2 + x - y = 0. If y = 0, then x = 0. If y \ne 0, the quadratic has two real roots with product \displaystyle \frac{-y}{y} = -1, so at most one lies in (-1, 1), and the root
does, because \displaystyle \sqrt{1 + 4y^2} > 2|y|. The last form also gives 0 at y = 0. So f is a bijection with inverse \displaystyle g(y) = \frac{2y}{1 + \sqrt{1 + 4y^2}}, continuous on \mathbb{R} since its denominator is at least 2. By Theorem 4, f is a homeomorphism.
For a < b, \displaystyle x \mapsto \frac{2x - a - b}{b - a} maps (a, b) onto (-1, 1) with continuous inverse \displaystyle u \mapsto \frac{(b - a)u + a + b}{2}; composing with f (Theorem 5(a)) gives (a, b) \approx \mathbb{R}. Boundedness is not a topological property.
Wrapping an interval around the circle
Let S^1 = \{(x, y) \in \mathbb{R}^2 : x^2 + y^2 = 1\} with the metric \lVert p - q \rVert (Dot product, length and orthogonality in n dimensions), [0, 1) with |s - t|, and e(s) = (\cos 2\pi s, \sin 2\pi s). Show that e : [0, 1) \to S^1 is a continuous bijection but not a homeomorphism.
By \cos \alpha \cos \beta + \sin \alpha \sin \beta = \cos(\alpha - \beta),
Since |\sin u| \le |u|, \lVert e(s) - e(t) \rVert \le 2\pi |s - t|, and \displaystyle \delta = \frac{\varepsilon}{2\pi} works in Theorem 4: e is continuous. Each point of S^1 is (\cos \theta, \sin \theta) for exactly one \theta \in [0, 2\pi) (The unit circle), so e is a bijection.
The set \displaystyle U = [0, \frac{1}{2}) is the ball \displaystyle B(0, \frac{1}{2}) of [0, 1), hence open. Its image is the upper half circle, the points (\cos \theta, \sin \theta) with 0 \le \theta < \pi, all with y \ge 0, and it contains (1, 0). Given r > 0, let \displaystyle t = \frac{1}{7}\min\{r, 1\}. The point e(1 - t) = (\cos 2\pi t, -\sin 2\pi t) has y < 0, as 0 < 2\pi t < \pi, so it is not in e(U); yet its distance from (1, 0) = e(0) is 2\sin \pi t \le 2\pi t < r. No ball about (1, 0) lies in e(U), so e is not open, and by Theorem 5(b) it is not a homeomorphism.
Telling spaces apart
With the metric |s - t|, show that (a) \mathbb{Z} and \mathbb{Q} are not homeomorphic; (b) the standard line and (\mathbb{R}, \mathcal{T}_f), the finite-complement topology, are not homeomorphic; (c) \mathbb{N} \approx \mathbb{Z}.
SolutionIn (a) and (b) bijections exist (Countable sets; \mathrm{id}_{\mathbb{R}}), so the obstruction must be topological.
(a) In \mathbb{Z}, \displaystyle B(n, \frac{1}{2}) = \{n\}, so every one-point set is open. In \mathbb{Q} none is: B(q, r) contains the rational \displaystyle q + \frac{1}{m} once \displaystyle m > \frac{1}{r}. A homeomorphism f : \mathbb{Z} \to \mathbb{Q} would make f(\{0\}) = \{f(0)\} open, by Theorem 5(c).
(b) In the standard line, (0, 1) and (2, 3) are disjoint nonempty open sets. In \mathcal{T}_f, nonempty open sets U, V have finite complements, so the complement of U \cap V, the union of those two, is finite, and U \cap V \ne \varnothing. A homeomorphism would carry (0, 1) and (2, 3) to disjoint nonempty open sets of \mathcal{T}_f (Theorem 5(c), with f injective), and there are none.
(c) Both spaces are discrete, as in (a). By Theorem 3(b) every bijection between them is continuous both ways, for instance \displaystyle \varphi(n) = \frac{n}{2} for even n and \displaystyle \varphi(n) = -\frac{n + 1}{2} for odd n, which lists \mathbb{Z} as 0, -1, 1, -2, 2, \ldots
Continuity in the finite-complement topology
Let \mathcal{T}_f be the finite-complement topology on \mathbb{R}. (a) Prove that f : (\mathbb{R}, \mathcal{T}_f) \to (\mathbb{R}, \mathcal{T}_f) is continuous if and only if f is constant or every fiber f^{-1}(\{y\}) is finite. (b) Decide this for polynomials of degree at least 1 and for \cos.
Solution(a) Proof. The closed sets of \mathcal{T}_f are \mathbb{R} and the finite sets. By Theorem 2(b), f is continuous if and only if f^{-1}(K) is finite or \mathbb{R} for every finite K. As f^{-1}(K) = \bigcup_{y \in K} f^{-1}(\{y\}) (Theorem 1), finite fibers make every such preimage finite, and a constant f gives \varnothing or \mathbb{R}. Conversely, if f is continuous, each fiber is finite or \mathbb{R}, and a fiber equal to \mathbb{R} makes f constant. \square
(b) For a polynomial p of degree n \ge 1, p^{-1}(\{y\}) is the set of roots of p - y, at most n points, so p is continuous. \cos is not constant and \cos^{-1}(\{1\}) contains 2\pi k for every k \in \mathbb{Z}, so it is not continuous, though it is continuous on the standard line.
Boundaries
- A continuous bijection need not be a homeomorphism (Examples 2 and 4).
- Continuous maps need not be open or closed (Exercise 15), and Theorem 2(c) is only an inclusion.
- Distances are not topological: boundedness is lost (Example 3).
- Continuity depends on both topologies (Examples 2 and 6): finer domains and coarser targets admit more continuous maps.
Where this goes next
Subspace and product topologies restricts continuous maps to subsets and builds maps into products. Quotient spaces, or gluing glues points, as Example 4 nearly does. Connectedness and Compactness and open covers are topological properties kept by continuous images, the tools that prove [0, 1] \not\approx (0, 1). In Homotopy continuous maps are themselves deformed.
Exercises
Finite spaces
- Let X = \{1, 2, 3, 4\} have open sets \varnothing, \{1\}, \{2\}, \{1, 2\}, \{1, 2, 3\}, X, and Y = \{u, v, w\} have open sets \varnothing, \{u\}, \{u, v\}, Y. Which maps X \to Y are continuous? (a) 1, 2 \mapsto u, 3 \mapsto v, 4 \mapsto w; (b) 1, 3 \mapsto u, 2 \mapsto v, 4 \mapsto w; (c) 2 \mapsto u, 1, 3, 4 \mapsto v.
- With Y as in Exercise 1, how many of the 27 maps Y \to Y are continuous, and which are homeomorphisms?
- Samir sets up a second hall with rooms p, q, r, s and two sensors covering \{p, q, r\} and \{r, s\}, a subbasis. (a) List the open sets. (b) Find every homeomorphism from this space to itself.
- Let \Sigma = \{0, 1\} have open sets \varnothing, \{1\}, \Sigma. (a) Prove that f : X \to \Sigma is continuous if and only if f^{-1}(\{1\}) is open, so continuous maps X \to \Sigma correspond one to one with open sets of X. (b) How many of the 16 maps from the space X of Exercise 1 to \Sigma are continuous?
- (a) Sort the four topologies on \{1, 2\} into homeomorphism classes. (b) The set \{1, 2, 3\} carries 29 topologies. Into how many homeomorphism classes do they fall?
Maps of the line
- Let p(x) = x^2 - 2x on the standard line. Find p^{-1}((1, 4)) and p^{-1}([-1, 3]).
- Which maps are continuous? (a) x \mapsto \lfloor x \rfloor from \mathbb{R}_{\ell} to \mathbb{R}; (b) x \mapsto \lceil x \rceil from \mathbb{R}_{\ell} to \mathbb{R}; (c) x \mapsto -x from \mathbb{R}_{\ell} to \mathbb{R}_{\ell}; (d) x \mapsto -x from \mathbb{R}_{\ell} to \mathbb{R}.
- Show that \displaystyle \varphi(x) = \frac{x}{1 + |x|} is a homeomorphism from \mathbb{R} onto (-1, 1), and find its inverse.
- With the metric |s - t| and \displaystyle P = \left\{\frac{1}{n} : n \ge 1\right\}, decide which pairs are homeomorphic: (a) \mathbb{Z} and P; (b) \mathbb{Z} and P \cup \{0\}; (c) [0, 1] and [0, 1] \cup \{2\}.
Proofs
- Prove that f : X \to Y is continuous if and only if f^{-1}(\operatorname{int} B) \subseteq \operatorname{int} f^{-1}(B) for every B \subseteq Y.
- Prove that if every function from X to every space is continuous, X is discrete, and that if every function from every space to Y is continuous, Y is indiscrete.
- Prove that every strictly increasing surjection f : \mathbb{R} \to \mathbb{R} is a homeomorphism.
- Let f : X \to Y be a continuous surjection and A dense in X. Prove that f(A) is dense in Y.
- For x in a finite space let M_x be the smallest open set containing x. Prove that f between finite spaces is continuous if and only if f(M_x) \subseteq M_{f(x)} for every x.
- Refute each claim with a counterexample on the standard line: (a) a continuous map sends open sets to open sets; (b) a continuous map sends closed sets to closed sets.
Answers
- (a) Continuous: the preimages of \{u\} and \{u, v\} are \{1, 2\} and \{1, 2, 3\}. (b) Not continuous: the preimage of \{u\} is \{1, 3\}, not open. (c) Continuous: the preimages are \{2\} and X.
- 10 are continuous, and only \mathrm{id}_Y is a homeomorphism. A map is fixed by A = f^{-1}(\{u\}) \subseteq B = f^{-1}(\{u, v\}), which must be open: it is u on A, v on B \setminus A, w elsewhere. The open sets form the chain \varnothing \subseteq \{u\} \subseteq \{u, v\} \subseteq Y, with 4 + 3 + 2 + 1 = 10 pairs A \subseteq B. A bijection needs one point in A and two in B, so A = \{u\}, B = \{u, v\}: the identity.
- (a) The sensors meet in \{r\}; the open sets are \varnothing, \{r\}, \{r, s\}, \{p, q, r\} and \{p, q, r, s\}. (b) A homeomorphism permutes the open sets, keeping sizes, so it fixes \{r\} and \{r, s\}, hence r and s. Swapping p and q fixes every open set, so there are two: the identity and that swap.
- (a) The preimages of \varnothing and \Sigma are \varnothing and X, always open, so only \{1\} matters. Each open U gives exactly one continuous map, 1 on U and 0 off it. \square (b) 6, the number of open sets of X.
- (a) Indiscrete, discrete, and \{\varnothing, \{1\}, X\}, \{\varnothing, \{2\}, X\}, which the swap 1 \leftrightarrow 2 carries onto each other: 3 classes. (b) 9 classes. With \{a, b, c\} = \{1, 2, 3\} in some order, the open sets other than \varnothing and X are: none (1 topology); \{a\} (3); \{a, b\} (3); \{a\}, \{a, b\} (6); \{a\}, \{b, c\} (3); \{a\}, \{b\}, \{a, b\} (3); \{a\}, \{a, b\}, \{a, c\} (3); \{a\}, \{b\}, \{a, b\}, \{a, c\} (6); all six (1). The counts add to 29. A homeomorphism keeps the sizes of open sets and inclusions between them, which tells the rows apart.
- x^2 - 2x > 1 gives \displaystyle x < 1 - \sqrt{2} or \displaystyle x > 1 + \sqrt{2}, and x^2 - 2x < 4 gives \displaystyle 1 - \sqrt{5} < x < 1 + \sqrt{5}, so p^{-1}((1, 4)) is the open set \displaystyle (1 - \sqrt{5}, 1 - \sqrt{2}) \cup (1 + \sqrt{2}, 1 + \sqrt{5}). Since x^2 - 2x = (x - 1)^2 - 1 \ge -1 always and x^2 - 2x \le 3 means (x - 3)(x + 1) \le 0, p^{-1}([-1, 3]) = [-1, 3], closed.
- (a) Continuous: the preimage of V is the union of the [n, n + 1) with n \in V \cap \mathbb{Z}, open in \mathbb{R}_{\ell}. (b) Not continuous: the preimage of \displaystyle (-\frac{1}{2}, \frac{1}{2}) is (-1, 0], and no [0, \varepsilon) lies inside it. (c) Not continuous: the preimage of [0, 1) is (-1, 0]. (d) Continuous: the preimage -V of a standard-open V is standard-open, hence open in the finer \mathbb{R}_{\ell}.
- \displaystyle |\varphi(x)| = \frac{|x|}{1 + |x|} < 1. Let \displaystyle \psi(y) = \frac{y}{1 - |y|} on (-1, 1). Since \displaystyle 1 - |\varphi(x)| = \frac{1}{1 + |x|}, \psi(\varphi(x)) = x; since \displaystyle 1 + |\psi(y)| = \frac{1}{1 - |y|}, \varphi(\psi(y)) = y. Both are quotients of continuous functions with nonzero denominators, continuous by Theorem 4.
- (a) Yes. Other points of P lie at least \displaystyle \frac{1}{n(n + 1)} from \displaystyle \frac{1}{n}, so P is discrete, and both are countably infinite; for instance \displaystyle n \mapsto \frac{1}{2n + 1} for n \ge 0 and \displaystyle n \mapsto \frac{1}{-2n} for n < 0 (Theorem 3(b)). (b) No: \{0\} is not open in P \cup \{0\}, as every B(0, r) holds \displaystyle \frac{1}{n} for \displaystyle n > \frac{1}{r}, while every one-point set of \mathbb{Z} is open. (c) No: \{2\} is open in [0, 1] \cup \{2\}, but no one-point set is open in [0, 1]: each B(x, r) holds a point of [0, 1] other than x.
- If f is continuous, f^{-1}(\operatorname{int} B) is open and lies in f^{-1}(B), so it lies in the largest open set inside f^{-1}(B). Conversely, for open V, \operatorname{int} V = V gives f^{-1}(V) \subseteq \operatorname{int} f^{-1}(V) \subseteq f^{-1}(V), so f^{-1}(V) is open. \square
- The identity from X to X with the discrete topology is continuous, so each U \subseteq X, the preimage of the open set U, is open in X. The identity from Y with the indiscrete topology to Y is continuous, so each open V of Y is open in the indiscrete topology: V = \varnothing or V = Y. \square
- f is injective, hence bijective, and f^{-1} is strictly increasing too. For a < b, f((a, b)) = (f(a), f(b)): the inclusion \subseteq is monotonicity, and if f(a) < y < f(b), then y = f(x) for some x, and x \le a or x \ge b would put y outside. Open sets are unions of open intervals, so f is open; by the same argument f^{-1} is open, that is, f is continuous. Theorem 5(b) finishes. \square
- By Theorem 2(c), Y = f(X) = f(\overline{A}) \subseteq \overline{f(A)}. \square
- If f is continuous, f^{-1}(M_{f(x)}) is open and contains x, so contains M_x. Conversely, let V be open and x \in f^{-1}(V). Then M_{f(x)} \subseteq V, so f(M_x) \subseteq V and M_x \subseteq f^{-1}(V); so f^{-1}(V) is a union of open sets. \square
- (a) A constant map x \mapsto 5 sends the open (0, 1) to \{5\}, not open. (b) \displaystyle x \mapsto \frac{1}{1 + x^2} sends the closed set \mathbb{R} to (0, 1], not closed, as 0 is a limit point outside it.