Continuity, defined and proved
Calculus called a function continuous when its limit equals its value, which says nothing at the end of a domain or at an isolated point, and it took the continuity of roots, sines and compositions on trust. What exactly is continuity at a point of any domain, and how are those facts proved?
Where you are 5 in Continuity
Why this exists
Continuity at a point defined continuity at a by \lim_{x \to a} f(x) = f(a) for f defined around a, and took the continuity of roots, sines and compositions on trust. Analysis needs functions on [0, \infty) or \mathbb{Q}, a definition that proofs such as those of Extreme and intermediate value theorems, proved can use, and functions beyond any picture (Example 6). This lesson defines continuity by \varepsilon and \delta, ties it to the limits of Limits of functions, epsilon and delta and to sequences, proves the algebra of continuous functions, and recasts continuity through the open sets of Open, closed and compact sets in R.
Definition
Throughout, S \subseteq \mathbb{R} and f : S \to \mathbb{R}.
Continuity at a point. Let c \in S. Then f is continuous at c if for every \varepsilon > 0 there is \delta > 0 such that \lvert f(x) - f(c) \rvert < \varepsilon for every x \in S with \lvert x - c \rvert < \delta. Otherwise f is discontinuous at c.
Continuous on a set. For A \subseteq S, f is continuous on A if it is continuous at every c \in A. It is a continuous function if it is continuous on S.
Unlike a limit, c must lie in S, the target is f(c), and x = c is tested, at no cost. Only points of S are tested, so at an end of an interval domain the definition is one-sided by itself. The \delta may depend on both \varepsilon and c.
Preimage. For U \subseteq \mathbb{R}, the preimage of U under f is f^{-1}(U) = \{x \in S : f(x) \in U\}. No inverse function is assumed.
Intuition
Draw the band f(c) - \varepsilon < y < f(c) + \varepsilon. Continuity at c says that however thin the band, some strip c - \delta < x < c + \delta is narrow enough that the graph over it, at points of S, stays inside the band (Figure 1).
A sequence in S converging to c eventually enters every strip, so its values eventually enter every band (Theorem 3).
Results
Throughout, f, g : S \to \mathbb{R} and c \in S.
Theorem 1 (isolated points). If c is not a limit point of S, then f is continuous at c.
Theorem 2 (limit form). If c is a limit point of S, then f is continuous at c if and only if \lim_{x \to c} f(x) = f(c).
Theorem 3 (sequential criterion). f is continuous at c if and only if f(x_n) \to f(c) for every sequence \{x_n\} of points of S with x_n \to c. So if one such sequence has \{f(x_n)\} divergent, or convergent to a number other than f(c), then f is discontinuous at c.
Theorem 4 (algebra). If f and g are continuous at c and k \in \mathbb{R}, then f + g, f - g, kf and fg are continuous at c. If also g(c) \ne 0, then \displaystyle \frac{f}{g}, defined on S_0 = \{x \in S : g(x) \ne 0\}, is continuous at c.
Theorem 5 (composition). Let g : A \to \mathbb{R} and f : B \to \mathbb{R} with g(A) \subseteq B. If g is continuous at c \in A and f is continuous at g(c), then f \circ g : A \to \mathbb{R} is continuous at c.
Theorem 6 (standard functions). (a) Every polynomial is continuous on \mathbb{R}, and a rational function \displaystyle \frac{p}{q} is continuous on its domain \{x : q(x) \ne 0\}. (b) For all real x and c, \lvert \sin x - \sin c \rvert \le \lvert x - c \rvert; so \sin and \cos are continuous on \mathbb{R}. (c) \lvert x \rvert is continuous on \mathbb{R}.
Theorem 7 (open and closed sets). Let f : \mathbb{R} \to \mathbb{R}. (a) f is continuous if and only if f^{-1}(U) is open for every open set U. (b) If f is continuous, then f^{-1}(F) is closed for every closed set F.
The continuity of \displaystyle \sqrt{x} on [0, \infty) is Example 2.
Why it is true
Theorem 1. Proof. Some (c - \delta, c + \delta) contains no point of S but c. Whatever \varepsilon is, this \delta tests only x = c, where \lvert f(c) - f(c) \rvert = 0 < \varepsilon. \square
Theorem 2. Proof. The limit asks for \lvert f(x) - f(c) \rvert < \varepsilon at the points of S with 0 < \lvert x - c \rvert < \delta; continuity asks it there and at x = c, where it always holds. So a \delta serves one exactly when it serves the other. \square
Theorem 3. Proof. Let f be continuous at c, let x_n \in S with x_n \to c, and let \varepsilon > 0. Choose \delta serving \varepsilon, then N with \lvert x_n - c \rvert < \delta for n > N. For n > N, \lvert f(x_n) - f(c) \rvert < \varepsilon.
Conversely, suppose f is discontinuous at c. Negating the definition (Negating a quantified statement): there is \varepsilon > 0 such that for every \delta > 0 some x \in S has \lvert x - c \rvert < \delta and \lvert f(x) - f(c) \rvert \ge \varepsilon. For \displaystyle \delta = \frac{1}{n} call this point x_n. Then x_n \in S and x_n \to c, but \lvert f(x_n) - f(c) \rvert \ge \varepsilon for every n, so f(x_n) \not\to f(c). The last sentence of the theorem is the contrapositive of the first half. \square
Unlike in Limits of functions, epsilon and delta, these sequences may take the value c.
Theorem 4. Proof. Let x_n \in S with x_n \to c. By Theorem 3, f(x_n) \to f(c) and g(x_n) \to g(c). The sum law of Convergence of sequences and the product law of Limits of sequences, epsilon and N (with constant sequences, and f - g = f + (-1)g) give the limits of f + g, f - g, kf and fg along \{x_n\}, and Theorem 3 converts back. For the quotient take x_n \in S_0; every g(x_n) \ne 0 and g(c) \ne 0, so the quotient law of Limits of sequences, epsilon and N applies. \square
Theorem 5. Proof. Let \varepsilon > 0. As f is continuous at g(c), there is \eta > 0 with \lvert f(y) - f(g(c)) \rvert < \varepsilon for every y \in B with \lvert y - g(c) \rvert < \eta. As g is continuous at c, there is \delta > 0 with \lvert g(x) - g(c) \rvert < \eta for every x \in A with \lvert x - c \rvert < \delta. For such x, y = g(x) lies in B within \eta of g(c), so \lvert f(g(x)) - f(g(c)) \rvert < \varepsilon. \square
The limit version in Limits of functions, epsilon and delta needed f(x) \ne b near c; continuity tests the outer function at g(c) itself.
Theorem 6. Proof. (a) A constant is continuous with any \delta, and x with \delta = \varepsilon. By Proof by induction, x^k = x \cdot x^{k-1} is continuous by Theorem 4. A polynomial is a finite sum of constant multiples of powers and a rational function is a quotient of polynomials, so Theorem 4 finishes.
(b) For \displaystyle 0 < t < \frac{\pi}{2}, \sin t < t (The squeeze theorem and the limit of sine x over x); for \displaystyle t \ge \frac{\pi}{2} > 1, \lvert \sin t \rvert \le 1 < t; and \sin is odd. So \lvert \sin t \rvert \le \lvert t \rvert for every t. By the sum-to-product identity and \lvert \cos \rvert \le 1,
so \delta = \varepsilon works. Then \displaystyle \cos x = \sin\left(x + \frac{\pi}{2}\right) is continuous by Theorem 5.
(c) \big\lvert \lvert x \rvert - \lvert c \rvert \big\rvert \le \lvert x - c \rvert by the reverse triangle inequality (Result 11 of The real numbers as an ordered field), so \delta = \varepsilon works. \square
Theorem 7. Proof. (a) Let f be continuous, U open and c \in f^{-1}(U). Since f(c) \in U, some (f(c) - \varepsilon, f(c) + \varepsilon) \subseteq U. Continuity gives \delta with f(x) in that interval whenever \lvert x - c \rvert < \delta, so (c - \delta, c + \delta) \subseteq f^{-1}(U), and f^{-1}(U) is open. Conversely, let every such preimage be open, and let c \in \mathbb{R} and \varepsilon > 0. The interval U = (f(c) - \varepsilon, f(c) + \varepsilon) is open (Theorem 1 of Open, closed and compact sets in R) and c \in f^{-1}(U), so some (c - \delta, c + \delta) \subseteq f^{-1}(U): that is, \lvert f(x) - f(c) \rvert < \varepsilon whenever \lvert x - c \rvert < \delta.
(b) f(x) \notin F exactly when x \notin f^{-1}(F), so f^{-1}(\mathbb{R} \setminus F) = \mathbb{R} \setminus f^{-1}(F). The left side is open by (a), so f^{-1}(F) is closed (Theorem 4 of Open, closed and compact sets in R). \square
Worked examples
A delta for a quadratic
Let f(x) = 3x^2 - 5 on \mathbb{R}. Prove that f is continuous at every c. For c = 2 and \varepsilon = 0.1, compare the \delta of the proof with the largest \delta that works.
SolutionFactor the difference:
The factor \lvert x - c \rvert is what \delta controls; \lvert x + c \rvert must be bounded. If \lvert x - c \rvert < 1, the triangle inequality gives \lvert x + c \rvert \le \lvert x - c \rvert + 2\lvert c \rvert < 1 + 2\lvert c \rvert. Given \varepsilon > 0, let
chosen before any x is seen. If \lvert x - c \rvert < \delta, then \lvert f(x) - f(c) \rvert < 3\delta(1 + 2\lvert c \rvert) \le \varepsilon.
At c = 2 and \varepsilon = 0.1 the proof gives \displaystyle \delta = \frac{1}{150} \approx 0.00667. Exactly, \lvert 3x^2 - 12 \rvert < 0.1 means \displaystyle \frac{119}{30} < x^2 < \frac{121}{30}, which near 2 is \displaystyle \sqrt{119/30} < x < \sqrt{121/30}. The right end is nearer: \displaystyle \frac{11}{\sqrt{30}} - 2 \approx 0.00832, against \displaystyle 2 - \sqrt{119/30} \approx 0.00835. So the largest \delta is about 0.00832.
The square root
Prove that \displaystyle r(x) = \sqrt{x} is continuous on [0, \infty).
SolutionProof. At c = 0, let \varepsilon > 0 and \delta = \varepsilon^2. The tested points are 0 \le x < \varepsilon^2, and since the square root is increasing, \displaystyle \lvert \sqrt{x} - 0 \rvert < \varepsilon. Points left of 0 are not in the domain.
At c > 0, multiply and divide by \displaystyle \sqrt{x} + \sqrt{c} \ge \sqrt{c} > 0:
With \displaystyle \delta = \varepsilon\sqrt{c}, every x \ge 0 with \lvert x - c \rvert < \delta has \displaystyle \lvert \sqrt{x} - \sqrt{c} \rvert < \varepsilon. \square
At c = 4 with \varepsilon = 0.5 this gives \delta = 1, the window of Figure 1.
Layla’s confectionery
Layla runs a confectionery, and Mateo programs its chocolate melter. Over 20 minutes the temperature in degrees Fahrenheit is to follow
(a) Find k so that T is continuous on [0, 20]. (b) With that k, find from the definition a \delta serving \varepsilon at t = 5, and the largest \delta for \varepsilon = 0.5. (c) Show by a sequence that every other k makes T discontinuous at 5.
Solution(a) Each piece is a polynomial (Theorem 6), and near a point inside a piece’s interval a small \delta tests only that piece. At t = 15 both formulas give 81. At t = 5, T(5) = 115 while the middle formula gives 81 + 100k; by (b) and (c), continuity there holds exactly when these agree, so k = 0.34 (Figure 2).
(b) For t \le 5, \lvert T(t) - 115 \rvert = 9(5 - t). For 5 < t \le 15, (t - 15)^2 - 100 = (t - 5)(t - 25), so
since 25 - t < 20. With \displaystyle \delta = \min\left\{10, \frac{\varepsilon}{9}\right\}, both pieces give \lvert T(t) - 115 \rvert \le 9\lvert t - 5 \rvert < \varepsilon. For \varepsilon = 0.5 the left piece allows exactly \displaystyle 5 - t < \frac{1}{18}; the right allows t - 5 up to the root of 0.34u(20 - u) = 0.5, \displaystyle u = 10 - \sqrt{100 - 25/17} \approx 0.0738. So the largest \delta is \displaystyle \frac{1}{18} \approx 0.0556.
(c) If k \ne 0.34, let \displaystyle t_n = 5 + \frac{1}{n}, so 5 < t_n \le 6. By the sequence laws
and Theorem 3 makes T discontinuous at 5.
Determined by the rationals
Let f, g : \mathbb{R} \to \mathbb{R} be continuous with f(r) = g(r) for every r \in \mathbb{Q}. Prove that f = g. Then find \displaystyle f(\sqrt{2}) for a continuous f with f(r) = r^2 + r for every rational r.
SolutionProof. Let x \in \mathbb{R}. For each n \ge 1 the interval \displaystyle \left(x, x + \frac{1}{n}\right) contains a rational r_n (The Archimedean property and density of the rationals), so r_n \to x. By Theorem 3, f(r_n) \to f(x) and g(r_n) \to g(x). These are one and the same sequence, which has only one limit, so f(x) = g(x). \square
For the second part, x^2 + x is continuous (Theorem 6) and agrees with f on \mathbb{Q}, so \displaystyle f(\sqrt{2}) = 2 + \sqrt{2}.
Sets cut out by a continuous function
Let f(x) = x^3 - 4x. Find \{x : f(x) \ge 0\} and \{x : f(x) < 0\}, and say why the first must be closed and the second open. Then show that a continuous function can map an open interval onto a set that is not open, and \mathbb{R} onto a set that is not closed.
Solutionf(x) = x(x - 2)(x + 2) changes sign only at -2, 0 and 2, and is positive for large x. So
Theorem 7 predicts the types without factoring: these are the preimages of the closed set [0, \infty) and the open set (-\infty, 0) under a continuous function.
Images behave differently. x^2 maps (-1, 1) onto [0, 1), not open, as no neighborhood of 0 lies in it. \displaystyle \frac{1}{1 + x^2} maps \mathbb{R} onto (0, 1], not closed, as its points \displaystyle \frac{1}{1 + n^2} tend to 0.
Continuous exactly at the irrationals
Define P : \mathbb{R} \to \mathbb{R} by \displaystyle P(x) = \frac{1}{q} if \displaystyle x = \frac{p}{q} with p \in \mathbb{Z}, q \ge 1 and no common factor, and P(x) = 0 if x is irrational. Prove that P, Thomae’s function (Figure 3), is discontinuous at every rational and continuous at every irrational.
Proof. Let \displaystyle c = \frac{p}{q} in lowest terms, so \displaystyle P(c) = \frac{1}{q} > 0. The numbers \displaystyle x_n = c + \frac{\sqrt{2}}{n} are irrational (else \displaystyle \sqrt{2} = n(x_n - c) would be rational) and tend to c, and P(x_n) = 0 \to 0 \ne P(c). By Theorem 3, P is discontinuous at c.
Let c be irrational and \varepsilon > 0. Choose an integer K \ge 1 with \displaystyle \frac{1}{K} < \varepsilon (The Archimedean property and density of the rationals). For each q < K, a fraction \displaystyle \frac{p}{q} in (c - 1, c + 1) has p in an open interval of length 2q, so there are at most 2q of them. Hence the set E of rationals in (c - 1, c + 1) with lowest-terms denominator less than K is finite, and c \notin E. So
is a minimum of finitely many positive numbers, hence positive. Let \lvert x - c \rvert < \delta. If x is irrational, P(x) = 0. If x is rational, it lies in (c - 1, c + 1) but not in E, so its denominator q is at least K and \displaystyle \lvert P(x) - 0 \rvert = \frac{1}{q} \le \frac{1}{K} < \varepsilon. \square
Boundaries
- Continuity is pointwise. The \delta depends on c (Example 1); one \delta for every c is Uniform continuity.
- “Continuous on A” is not “f|_A is continuous”. The first tests x in all of S, the second only in A (Exercise 14).
- Preimages, not images. Images of open or closed sets need not be open or closed (Example 5). Theorem 7 needs the domain \mathbb{R}: under \displaystyle \sqrt{x} on [0, \infty) the preimage of (-1, 1) is [0, 1), not open.
- Where continuity can fail. No function on \mathbb{R} is continuous exactly at the rationals; that needs the Baire category theorem of a later analysis course.
Where this goes next
Extreme and intermediate value theorems, proved combines Theorem 3 with Bolzano-Weierstrass to prove the extreme and intermediate value theorems. Differentiability and the derivative rules, proved shows that differentiable functions are continuous. Metric spaces, a first look replaces \lvert x - y \rvert by a distance, and Theorem 7(a) becomes how topology defines continuity.
Exercises
Epsilon and delta
- Prove that f(x) = 7 - 4x is continuous at every c. For \varepsilon = 0.02, find the largest \delta that works.
- Let f(x) = x^2 + 2x. Prove that f is continuous at 1 with \displaystyle \delta = \min\left\{1, \frac{\varepsilon}{5}\right\}, and for \varepsilon = 0.05 compare this \delta with the largest one, to five decimal places.
- Let g(x) = \dfrac{x}{x + 1} on [0, \infty). Prove that g is continuous at 1 with \delta = 2\varepsilon, and find the largest \delta for \varepsilon = 0.1.
Sequences, pieces and discontinuities
- Let f(x) = \dfrac{\lvert x^2 - 9 \rvert}{x - 3} for x \ne 3 and f(3) = 6. Use two sequences to show that f is discontinuous at 3. Could another value of f(3) make f continuous there?
- Find every point at which f(x) = x\lfloor x \rfloor is continuous.
- Mateo’s second program for Layla’s melter is T(t) = 68 + 12t on [0, 4], T(t) = a - 1.4(t - 4)^2 on (4, 9] and T(t) = b + 1.2(t - 9) on (9, 14]. Find a and b making T continuous on [0, 14], and then find T(12).
- For Thomae’s function P of Example 6 and \displaystyle \varepsilon = \frac{1}{4}, find the largest \delta that serves \varepsilon at \displaystyle c = \sqrt{2}.
Building continuous functions; open and closed sets
- Find the largest set on which \displaystyle h(x) = \dfrac{\sqrt{4 - x^2}}{2\cos x - 1} is defined, and show that h is continuous there, naming the theorems used.
- Find \{x : x^4 - 5x^2 + 4 \le 0\} and \{x : x^4 - 5x^2 + 4 > 0\}, and say before computing why one is closed and the other open.
Proofs
- Prove that if f is continuous at c, so is \lvert f \rvert. Show that the converse fails.
- Prove that if f and g are continuous at c, so is \max\{f, g\}, the function x \mapsto \max\{f(x), g(x)\}.
- Let a < b < d, let f be continuous on [a, b] and g continuous on [b, d], with f(b) = g(b). Prove that h, equal to f on [a, b] and to g on (b, d], is continuous on [a, d].
- Prove that \displaystyle \sqrt[3]{x} is continuous on \mathbb{R}. (The difference of cubes factors.)
- Let D(x) = 1 for x \in \mathbb{Q} and D(x) = 0 otherwise. Prove that D|_{\mathbb{Q}} is continuous but D is continuous at no point. What does this say about “D is continuous on \mathbb{Q}”?
- Let f : \mathbb{R} \to \mathbb{R} satisfy f(x + y) = f(x) + f(y) for all x and y, and be continuous at 0. Prove that f is continuous everywhere and that f(x) = f(1)\,x for every x.
- Let f, g : \mathbb{R} \to \mathbb{R} be continuous. Prove that \{x : f(x) < g(x)\} is open and \{x : f(x) = g(x)\} is closed.
Answers
- Largest \delta = 0.005. \lvert f(x) - f(c) \rvert = 4\lvert x - c \rvert, so \displaystyle \delta = \frac{\varepsilon}{4} serves every \varepsilon, and 4\lvert x - c \rvert < 0.02 exactly when \lvert x - c \rvert < 0.005.
- \delta = 0.01 against the largest, \displaystyle \sqrt{4.05} - 2 \approx 0.01246. \lvert x^2 + 2x - 3 \rvert = \lvert x - 1 \rvert \, \lvert x + 3 \rvert, and \lvert x - 1 \rvert < 1 gives \lvert x + 3 \rvert < 5, so the difference is below 5\delta \le \varepsilon. For \varepsilon = 0.05 the condition is 3.95 < (x + 1)^2 < 4.05, near 1 that is \displaystyle \sqrt{3.95} - 1 < x < \sqrt{4.05} - 1; the left end is \displaystyle 2 - \sqrt{3.95} \approx 0.01254 away.
- Largest \displaystyle \delta = \frac{1}{3}. \displaystyle \left\lvert \frac{x}{x + 1} - \frac{1}{2} \right\rvert = \frac{\lvert x - 1 \rvert}{2(x + 1)} \le \frac{\lvert x - 1 \rvert}{2} for x \ge 0, below \displaystyle \frac{\delta}{2} = \varepsilon. For \varepsilon = 0.1, as g is increasing, 0.4 < g(x) < 0.6 exactly when \displaystyle \frac{2}{3} < x < \frac{3}{2}, at distances \displaystyle \frac{1}{3} and \displaystyle \frac{1}{2} from 1.
- For x > 3, f(x) = x + 3; for -3 < x < 3, \displaystyle f(x) = \frac{9 - x^2}{x - 3} = -(x + 3). So \displaystyle x_n = 3 + \frac{1}{n} gives f(x_n) \to 6 and \displaystyle y_n = 3 - \frac{1}{n} gives f(y_n) \to -6 \ne 6 = f(3); Theorem 3 gives the discontinuity. No value works: it differs from 6 or from -6.
- At every non-integer and at 0, and nowhere else. On (m, m + 1), f(x) = mx, a polynomial near each point. At an integer m, f(m) = m^2, while \displaystyle x_n = m - \frac{1}{2n} gives \displaystyle f(x_n) = (m - 1)\left(m - \frac{1}{2n}\right) \to m^2 - m, which differs from m^2 unless m = 0 (Theorem 3). At 0: f(x) = -x on [-1, 0) and 0 on [0, 1), so \lvert f(x) \rvert \le \lvert x \rvert and \delta = \min\{1, \varepsilon\} works.
- a = 116, b = 81, T(12) = 84.6. The first piece gives T(4) = 116, and the middle formula equals a at t = 4, so a = 116; at t = 9 it gives 116 - 35 = 81, so b = 81. Then T(12) = 81 + 3.6.
- \displaystyle \delta = \sqrt{2} - \frac{4}{3} \approx 0.0809. At rational x, \displaystyle P(x) < \frac{1}{4} exactly when the denominator is at least 5. Of the fractions with denominator at most 4, the nearest to \displaystyle \sqrt{2} \approx 1.4142 is \displaystyle \frac{4}{3}; next is \displaystyle \frac{3}{2}, at distance about 0.0858. Any wider \delta tests \displaystyle \frac{4}{3}, where \displaystyle P = \frac{1}{3} \ge \frac{1}{4}.
- \displaystyle [-2, 2] \setminus \left\{-\frac{\pi}{3}, \frac{\pi}{3}\right\}. The root needs \lvert x \rvert \le 2; the denominator vanishes when \displaystyle \cos x = \frac{1}{2}, which in [-2, 2] (as \displaystyle 2 < \frac{5\pi}{3}) happens only at \displaystyle \pm\frac{\pi}{3}. The numerator is the square root (Example 2) composed with a polynomial (Theorems 5 and 6); the denominator is continuous by Theorems 4 and 6; the quotient by Theorem 4.
- [-2, -1] \cup [1, 2], closed, and (-\infty, -2) \cup (-1, 1) \cup (2, \infty), open: they are the preimages of the closed (-\infty, 0] and the open (0, \infty) under a polynomial (Theorem 7). Then x^4 - 5x^2 + 4 = (x^2 - 1)(x^2 - 4) \le 0 exactly when 1 \le x^2 \le 4.
- Proof. By the reverse triangle inequality (The real numbers as an ordered field), \big\lvert \lvert f(x) \rvert - \lvert f(c) \rvert \big\rvert \le \lvert f(x) - f(c) \rvert, so a \delta serving \varepsilon for f serves it for \lvert f \rvert. Converse: let s(x) = 1 on \mathbb{Q} and -1 elsewhere. Then \lvert s \rvert = 1 is continuous, but rationals and irrationals tending to any c give the limits 1 and -1, so by Theorem 3 s is continuous nowhere. \square
- Proof. For real u and v, \displaystyle \max\{u, v\} = \frac{u + v + \lvert u - v \rvert}{2}: if u \ge v the right side is u, otherwise v. So \displaystyle \max\{f, g\} = \frac{1}{2}\big(f + g + \lvert f - g \rvert\big), continuous at c by Theorem 4 and Exercise 10. \square
- Proof. Let c \in [a, d] and \varepsilon > 0. If c < b, let \delta_f serve \varepsilon for f at c and \delta = \min\{\delta_f, b - c\}: the tested points lie in [a, b), where h = f. The case c > b is symmetric. At c = b, where h(b) = f(b) = g(b), let \delta = \min\{\delta_f, \delta_g\} for \delta_f, \delta_g serving \varepsilon at b: a tested point of [a, b] has \lvert f(x) - f(b) \rvert < \varepsilon, one of (b, d] has \lvert g(x) - g(b) \rvert < \varepsilon. \square
- Proof. At 0, \delta = \varepsilon^3 works, since the cube root is increasing and odd, so \displaystyle \lvert \sqrt[3]{x} \rvert < \varepsilon when \lvert x \rvert < \varepsilon^3. At c \ne 0, let \displaystyle b = \sqrt[3]{c} and \displaystyle a = \sqrt[3]{x}. Then x - c = a^3 - b^3 = (a - b)(a^2 + ab + b^2), and
So \displaystyle \lvert a - b \rvert \le \frac{4\lvert x - c \rvert}{3b^2}, and \displaystyle \delta = \frac{3}{4}b^2\varepsilon gives \displaystyle \lvert \sqrt[3]{x} - \sqrt[3]{c} \rvert < \varepsilon. \square 14. Proof. D|_{\mathbb{Q}} is the constant 1 on \mathbb{Q}, continuous with any \delta. At a rational c, the irrationals \displaystyle c + \frac{\sqrt{2}}{n} tend to c with D = 0 \ne 1; at an irrational c, rationals \displaystyle r_n \in \left(c, c + \frac{1}{n}\right) tend to c with D = 1 \ne 0. By Theorem 3, D is discontinuous everywhere. So “D is continuous on \mathbb{Q}” is false although D|_{\mathbb{Q}} is continuous. \square 15. Proof. f(0) = f(0 + 0) = 2f(0), so f(0) = 0. Given c and \varepsilon, let \delta serve \varepsilon at 0. Since f(x) = f(x - c) + f(c), every x with \lvert x - c \rvert < \delta has \lvert f(x) - f(c) \rvert = \lvert f(x - c) - f(0) \rvert < \varepsilon. By induction f(nx) = nf(x) for n \ge 1, and f(-x) = -f(x) because f(x) + f(-x) = f(0) = 0; so f(m) = mf(1) for every integer m. For \displaystyle r = \frac{m}{q} with q \ge 1, qf(r) = f(qr) = f(m) = mf(1), so f(r) = rf(1). Thus f and the continuous x \mapsto f(1)x agree on \mathbb{Q}, and Example 4 gives f(x) = f(1)x for every x. \square 16. Proof. u = g - f is continuous (Theorem 4). \{f < g\} = u^{-1}((0, \infty)), the preimage of an open set (Theorem 1 of Open, closed and compact sets in R), so it is open by Theorem 7(a). \{f = g\} = u^{-1}(\{0\}), and \{0\} is closed (Theorem 5 of Open, closed and compact sets in R), so it is closed by Theorem 7(b). \square