The mean value theorem and its consequences
Calculus proved the mean value theorem from two results it took on trust. With those proved, what more does the theorem give? Which functions can be derivatives, what does the limit of a derivative at an endpoint say, and why does L’Hopital’s rule hold?
Where you are 3 in Derivatives
Why this exists
The mean value theorem proved Rolle’s theorem, the mean value theorem and their first consequences, taking on trust the extreme value theorem and Fermat’s theorem that f'(c) = 0 at an interior extremum. L'Hopital's rule proved Cauchy’s theorem when g' \ne 0 and L’Hopital’s rule for \displaystyle \frac{0}{0} only. Building on Extreme and intermediate value theorems, proved and Differentiability and the derivative rules, proved, this lesson proves Fermat’s lemma and Cauchy’s theorem in full, and three facts calculus never states: a limit of f' at an endpoint is the derivative there, every derivative has the intermediate value property (Darboux’s theorem), and L’Hopital’s rule holds whenever the denominator alone grows without bound.
Definition
Throughout, I is an interval with more than one point, and an interior point of I is a point of I that is not an endpoint. Derivatives are those of Differentiability and the derivative rules, proved, one-sided at an endpoint of the domain.
Relative extrema. Let S \subseteq \mathbb{R}, f : S \to \mathbb{R} and c \in S. f has a relative maximum at c if there is \delta > 0 with f(x) \le f(c) for every x \in S with \lvert x - c \rvert < \delta, and a relative minimum if instead f(x) \ge f(c) for all such x. An absolute maximum or minimum is a relative one.
Monotone functions. f : I \to \mathbb{R} is increasing if x < y in I gives f(x) \le f(y), and strictly increasing if it gives f(x) < f(y); decreasing and strictly decreasing reverse the inequalities. (The mean value theorem says “increasing” for the strict version, and so does The monotone convergence theorem for sequences, where the non-strict kind is “nondecreasing”.)
Lipschitz functions. f is Lipschitz on I with constant M \ge 0 if \lvert f(x) - f(y) \rvert \le M\lvert x - y \rvert for all x, y \in I.
Intuition
At a relative maximum c, secants from (c, f(c)) to nearby points on the right have slope \le 0, and those to the left have slope \ge 0. Their limit, if it exists, is caught between: 0. At an endpoint only one side exists.
Cauchy’s theorem is the mean value theorem for a path. As t runs over [a, b], the point (g(t), f(t)) moves with direction (g'(t), f'(t)), and the theorem says that at some time this direction is parallel to the chord (Figure 1). Cross-multiplied, the statement survives an instant where the path stops; a quotient would not.
A derivative need not be continuous: the derivative of \displaystyle x^2 \sin\frac{1}{x} in Differentiability and the derivative rules, proved oscillates near 0. Darboux’s theorem says a derivative never jumps: slopes 0 just left of c and 1 just right would make a corner, where there is no derivative at all.
Results
Theorem 1 (Fermat’s interior extremum lemma). Let f : I \to \mathbb{R} have a relative maximum or minimum at an interior point c of I, and let f be differentiable at c. Then f'(c) = 0.
Theorem 2 (Rolle’s theorem). Let f : [a, b] \to \mathbb{R} be continuous, differentiable on (a, b), with f(a) = f(b). Then f'(c) = 0 for some c \in (a, b).
Theorem 3 (Cauchy’s mean value theorem). Let f, g : [a, b] \to \mathbb{R} be continuous and differentiable on (a, b). Then some c \in (a, b) has
If also g'(x) \ne 0 for every x \in (a, b), then g(b) \ne g(a) and \dfrac{f(b) - f(a)}{g(b) - g(a)} = \dfrac{f'(c)}{g'(c)}. With g(x) = x this is the mean value theorem: f(b) - f(a) = f'(c)(b - a) for some c \in (a, b).
Theorem 4 (derivative and monotonicity). Let f : I \to \mathbb{R} be continuous, and differentiable at every interior point of I; below, “every x” means every interior point. (a) f is increasing if and only if f'(x) \ge 0 for every x, and decreasing if and only if f'(x) \le 0 for every x. (b) If f'(x) > 0 for every x, then f is strictly increasing; if f'(x) < 0 for every x, strictly decreasing. (c) f is constant if and only if f'(x) = 0 for every x. (d) If \lvert f'(x) \rvert \le M for every x, then f is Lipschitz on I with constant M.
Theorem 5 (derivative limit at an endpoint). Let a < b and let f : [a, b) \to \mathbb{R} be continuous, differentiable on (a, b), with \lim_{x \to a} f'(x) = L. Then f'(a) = L. The same holds at the right end of (a, b]. Consequently, if f : I \to \mathbb{R} is continuous at an interior point c, differentiable at every other point of I, and f' has both one-sided limits at c, then f is differentiable at c if and only if they are equal, and f'(c) is then their common value. So a derivative has no jump.
Theorem 6 (Darboux’s theorem). Let f : I \to \mathbb{R} be differentiable on I, let a < b be points of I, and let y lie strictly between f'(a) and f'(b). Then f'(c) = y for some c \in (a, b). So f'(I) is an interval.
Theorem 7 (L’Hopital’s rule). Let a < b, let f, g : (a, b) \to \mathbb{R} be differentiable with g'(x) \ne 0 for every x \in (a, b), and let \displaystyle \lim_{x \to a} \frac{f'(x)}{g'(x)} = L, a real number. (a) If \lim_{x \to a} f(x) = \lim_{x \to a} g(x) = 0, then \displaystyle \lim_{x \to a} \frac{f(x)}{g(x)} = L. (b) If \lvert g(x) \rvert \to \infty as x \to a, then \displaystyle \lim_{x \to a} \frac{f(x)}{g(x)} = L. The same holds as x \to b, at an interior point from both sides (Theorem 7 of Limits of functions, epsilon and delta), and as x \to \infty for f, g on (A, \infty).
Why it is true
Theorem 1. Proof. Let f have a relative maximum at c, with \delta as in the definition, and \displaystyle Q(x) = \frac{f(x) - f(c)}{x - c} on I \setminus \{c\}. As c is interior, [c - \eta, c + \eta] \subseteq I for some \eta \in (0, \delta). For n \ge 1 let \displaystyle x_n = c + \frac{\eta}{n} and \displaystyle y_n = c - \frac{\eta}{n}. Then f(x_n) \le f(c) and x_n > c, so Q(x_n) \le 0; likewise Q(y_n) \ge 0. By the sequential criterion (Theorem 2 of Limits of functions, epsilon and delta) both sequences of quotients tend to f'(c), and the order limit theorem (Theorem 8 of Limits of sequences, epsilon and N) gives f'(c) \le 0 \le f'(c). For a minimum, apply this to -f. \square
Theorem 2. Proof. By the extreme value theorem (Theorem 2 of Extreme and intermediate value theorems, proved), f attains a maximum and a minimum on [a, b]. If both equal f(a), f is constant and f' = 0 on (a, b). Otherwise one of them differs from f(a) = f(b), so it is attained at some c \in (a, b), and Theorem 1 gives f'(c) = 0. \square This is the proof in The mean value theorem, with both of its inputs now proved.
Theorem 3. Proof. Let
continuous on [a, b] and differentiable on (a, b) (Theorem 4 of Differentiability and the derivative rules, proved). Expanding, h(a) = f(b)g(a) - g(b)f(a) = h(b), so Theorem 2 gives c with h'(c) = 0, which is the equation. If g' \ne 0 on (a, b), then g(b) \ne g(a), or Theorem 2 would give a zero of g'; and g'(c) \ne 0, so we may divide. \square
Theorem 4. Proof. Let x < y in I. The interior points of [x, y] are interior points of I, so the mean value theorem gives c \in (x, y) with
The sign or size of f' passes to f(y) - f(x), which gives (b), (d) and the “if” parts of (a) and (c). Conversely, if f is increasing and c is interior, f(x) - f(c) has the sign of x - c or is 0, so the difference quotient is \ge 0 and so is its limit f'(c) (Theorem 5(a) of Limits of functions, epsilon and delta). A constant has quotient 0. For the decreasing cases, use -f. \square
Theorem 5. Proof. Let \varepsilon > 0 and choose \delta \in (0, b - a] with \lvert f'(t) - L \rvert < \varepsilon for t \in (a, a + \delta). For a < x < a + \delta, the mean value theorem on [a, x] gives c_x \in (a, x) with
which is within \varepsilon of L. So the quotient tends to L. The right end is the same with [x, b]. For the consequence, apply this to f on [c, c + \eta) and (c - \eta, c] inside I: the one-sided derivatives at c are the one-sided limits of f', and Theorem 6(c) of Differentiability and the derivative rules, proved finishes. \square
Theorem 6. Proof. Suppose f'(a) < y < f'(b) and let g(x) = f(x) - yx on [a, b], continuous since differentiable, with g'(a) < 0 < g'(b). By the extreme value theorem g has a minimum on [a, b] at some c. The difference quotient of g at a tends to g'(a) < 0, so it is negative at some x \in (a, b), where g(x) < g(a). The quotient at b tends to g'(b) > 0, so it is positive at some x \in (a, b), where x - b < 0 forces g(x) < g(b). So c \in (a, b), and Theorem 1 gives g'(c) = 0, that is, f'(c) = y (Figure 2). If f'(a) > y > f'(b), apply this to -f and -y. \square
Theorem 7. Proof. Let \varepsilon > 0 and choose y \in (a, b) with \displaystyle \left\lvert \frac{f'(t)}{g'(t)} - L \right\rvert < \frac{\varepsilon}{2} for every t \in (a, y). Let a < x < y.
(a) Put f(a) = g(a) = 0, making f and g continuous on [a, x]. Theorem 3 on [a, x] gives g(x) \ne 0 and c \in (a, x) with \displaystyle \frac{f(x)}{g(x)} = \frac{f'(c)}{g'(c)}, within \displaystyle \frac{\varepsilon}{2} of L. So \delta = y - a serves \varepsilon.
(b) Theorem 3 on [x, y] gives g(x) \ne g(y) and \displaystyle q_x = \frac{f(x) - f(y)}{g(x) - g(y)} = \frac{f'(c)}{g'(c)} for some c \in (x, y), so \displaystyle \lvert q_x - L \rvert < \frac{\varepsilon}{2}. Solving for f(x), where g(x) \ne 0,
With y fixed, K = \lvert f(y) \rvert + (\lvert L \rvert + \varepsilon)\lvert g(y) \rvert bounds the last numerator. As \lvert g(x) \rvert \to \infty, some \delta \in (0, y - a] gives \displaystyle \lvert g(x) \rvert > \max\left\{\frac{2K}{\varepsilon}, 1\right\} for a < x < a + \delta, and then both terms are below \displaystyle \frac{\varepsilon}{2}, so \displaystyle \left\lvert \frac{f(x)}{g(x)} - L \right\rvert < \varepsilon.
The end b is the mirror image. At \infty, taking A > 0, \displaystyle F(t) = f\left(\frac{1}{t}\right) and \displaystyle G(t) = g\left(\frac{1}{t}\right) have \displaystyle \frac{F'(t)}{G'(t)} = \frac{f'(1/t)}{g'(1/t)} by the chain rule, with G' \ne 0, and the case t \to 0^+ applies. \square
Worked examples
Cauchy’s point on a path
For f(t) = t^3 and g(t) = t^2 on [1, 2], find every c of Theorem 3, and compare it with the points the mean value theorem gives for f and g separately.
Solutionf(2) - f(1) = 7 and g(2) - g(1) = 3, so the equation is 7 \cdot 2c = 3 \cdot 3c^2, that is, c(9c - 14) = 0. The root 0 is outside (1, 2), so \displaystyle c = \frac{14}{9} \approx 1.556. As g' > 0 on (1, 2), the quotient form holds: \displaystyle \frac{f'(c)}{g'(c)} = \frac{3c}{2} = \frac{7}{3}, the chord’s slope. At \displaystyle t = \frac{14}{9} the path is at \displaystyle \left(\frac{196}{81}, \frac{2744}{729}\right) \approx (2.420, 3.764), moving parallel to the chord from (1, 1) to (4, 8) (Figure 1).
Separately, 3c_1^2 = 7 gives \displaystyle c_1 = \sqrt{7/3} \approx 1.528 for f, and 2c_2 = 3 gives c_2 = 1.5 for g: two different times, so dividing two mean value theorems does not give Cauchy’s.
Priya’s two tanks
At a distillery, Priya logs two holding tanks that drain into the stills during a 45-minute run. The wash tank falls from 800 to 350 gallons and the low-wines tank from 500 to 350 gallons, draining at every moment. (a) Show that at some instant the wash tank drained exactly three times as fast as the low-wines tank. (b) Ryan models the volumes, in gallons, as W(t) = 800 - 450u^2 and V(t) = 500 - 150u^3 with \displaystyle u = \frac{t}{45}. Find the instant of (a), and compare the rates there with the average rates.
Solution(a) The volumes are continuous on [0, 45] and differentiable inside, with V'(t) < 0 on (0, 45). Theorem 3 gives c with (350 - 800)V'(c) = (350 - 500)W'(c), that is, W'(c) = 3V'(c), where V'(c) \ne 0. Both rates are negative, so the wash tank drains three times as fast.
(b) By the chain rule, W'(t) = -20u and V'(t) = -10u^2 gallons per minute. W' = 3V' means -20u = -30u^2, so \displaystyle u = \frac{2}{3} and t = 30 minutes, with rates \displaystyle \frac{40}{3} \approx 13.3 and \displaystyle \frac{40}{9} \approx 4.44 gallons per minute. The averages 10 and \displaystyle \frac{10}{3} are reached at t = 22.5 and \displaystyle t = 15\sqrt{3} \approx 26.0: Cauchy’s theorem fixes the ratio, not the rates.
An increasing function whose derivative vanishes
Prove that f(x) = x - \sin x is strictly increasing on \mathbb{R}, and deduce that \sin x < x for every x > 0.
SolutionProof. f'(x) = 1 - \cos x \ge 0, so f is increasing by Theorem 4(a); Theorem 4(b) does not apply, as f'(2k\pi) = 0. Suppose x < y and f(x) = f(y). For x \le t \le y, f(x) \le f(t) \le f(y) = f(x), so f is constant on [x, y] and f' = 0 on (x, y) by Theorem 4(c). But the bounded interval (x, y) holds only finitely many multiples of 2\pi, and f' > 0 at its other points. So f(x) < f(y); with x = 0, y - \sin y > 0 for every y > 0. \square
A derivative at an endpoint from the limit of the derivative
Let f(x) = x^2 \ln x for x > 0 and f(0) = 0, on [0, \infty). Show that f is continuously differentiable on [0, \infty) and that f''(0) does not exist.
SolutionFirst, x \ln x \to 0 as x \to 0^+, by Theorem 7(b) on (0, 1) with u(x) = \ln x over \displaystyle v(x) = \frac{1}{x}: \lvert v \rvert \to \infty, v' \ne 0, and \displaystyle \frac{1/x}{-1/x^2} = -x \to 0. So f(x) = x \cdot x\ln x \to 0 = f(0), and f is continuous. For x > 0,
so Theorem 5 gives f'(0) = 0 with no new limit to compute, and f' is continuous on [0, \infty). For x > 0, \displaystyle \frac{f'(x) - f'(0)}{x} = 2\ln x + 1 \to -\infty, so f''(0) does not exist.
Integer-valued derivatives
Let f be differentiable on an interval I with f'(x) \in \mathbb{Z} for every x \in I. Prove that f(x) = kx + m on I for some k \in \mathbb{Z} and m \in \mathbb{R}.
SolutionProof. Suppose f'(a) \ne f'(b) for some a < b in I, and let p be the smaller of these two integers. Then \displaystyle p + \frac{1}{2} lies strictly between them and is not an integer, yet Theorem 6 gives c \in (a, b) with \displaystyle f'(c) = p + \frac{1}{2}. So f' is a constant integer k, and f(x) - kx has derivative 0 on I, so it is a constant m by Theorem 4(c). \square
L’Hopital’s rule when only the denominator blows up
Find \lim_{x \to 0^+} \dfrac{\ln(\sin x)}{\ln x}, and confirm the answer by algebra.
SolutionOn \displaystyle \left(0, \frac{\pi}{2}\right) let f(x) = \ln(\sin x) and g(x) = \ln x: both are differentiable, \displaystyle g'(x) = \frac{1}{x} \ne 0, and \lvert g(x) \rvert \to \infty. The quotient of derivatives is
by The squeeze theorem and the limit of sine x over x, so Theorem 7(b) gives the limit 1; that f \to -\infty too was not needed. By algebra, \displaystyle \ln(\sin x) = \ln x + \ln\frac{\sin x}{x}, so the quotient is \displaystyle 1 + \frac{\ln(\sin x / x)}{\ln x} \to 1 + 0.
Boundaries
- Interior points. At an endpoint only the sign of the derivative is forced (Exercise 12). And f'(c) = 0 does not give an extremum: x^3 at 0.
- Completeness. Over \mathbb{Q}, Rolle’s theorem fails: (x^2 - 2)^2 on \mathbb{Q} \cap [0, 2] is 4 at both ends, but its derivative 4x(x^2 - 2) vanishes in (0, 2) only at \displaystyle \sqrt{2}. The minimum the proof needs comes from the least-upper-bound property.
- One-way results. \displaystyle x^2 \sin\frac{1}{x} has derivative 0 at 0, but its derivative has no limit there: Theorem 5 has no converse, and Darboux’s theorem rules out only jumps. L’Hopital’s rule has no converse either (L'Hopital's rule).
Where this goes next
Taylor's theorem with remainder applies Cauchy’s mean value theorem to the remainder. The fundamental theorem of calculus, proved applies the mean value theorem on each subinterval of a partition to prove \int_a^b f' = f(b) - f(a). A Lipschitz function, as in Theorem 4(d), is uniformly continuous (Uniform continuity), and Theorem 4 underlies the The first and second derivative tests.
Exercises
Cauchy’s mean value theorem
- Find every c of Theorem 3 for f(t) = \sin t and g(t) = \cos t on \displaystyle \left[0, \frac{\pi}{2}\right], and describe the path (g(t), f(t)).
- Find every c of Theorem 3 for f(t) = e^{2t} and g(t) = e^t on [0, 1], to three decimal places.
- For f(t) = t^3 and g(t) = t^2 on [-1, 2], show that the only c of Theorem 3 is c = 0, and that no c \in (-1, 2) satisfies \displaystyle \frac{f'(c)}{g'(c)} = \frac{f(2) - f(-1)}{g(2) - g(-1)}. Which hypothesis fails?
Monotonicity and endpoints
- Find the smallest M with \displaystyle \lvert \sqrt{x} - \sqrt{y} \rvert \le M\lvert x - y \rvert for all x, y \ge 4.
- Show that 2x + \sin 2x = 1 has exactly one real solution, and that it lies in \displaystyle \left(\frac{1}{4}, \frac{1}{2}\right).
- Let \displaystyle f(x) = \sqrt{x}\,\sin x on [0, \infty). Use Theorem 5 to find f'(0).
- Let f(x) = e^x for x \le 0 and f(x) = 1 + x + ax^2 for x > 0. Show that f'(0) exists for every a, and find the a for which f''(0) exists.
Darboux’s theorem
- Show that no function differentiable on \mathbb{R} has f'(x) = 1 for x \le 0 and f'(x) = x^2 + 2 for x > 0.
- Let \displaystyle s(x) = x^2\sin\frac{1}{x} for x \ne 0 and s(0) = 0. Prove that for every \eta > 0, s' takes every value in [-1, 1] on (0, \eta).
L’Hopital’s rule
- Find \lim_{x \to 1} \dfrac{x^x - x}{1 - x + \ln x}, checking the hypotheses at each use of Theorem 7.
- Find \lim_{x \to 0^+} \dfrac{\ln(1 - \cos x)}{\ln x}.
Proofs
- Let f : [a, b] \to \mathbb{R} have a relative maximum at a and be differentiable at a. Prove that f'(a) \le 0, and that a relative maximum at b, with f differentiable at b, gives f'(b) \ge 0.
- Let f be differentiable on (0, 1] with \lvert f'(x) \rvert \le M for every x. Prove that \lim_{x \to 0^+} f(x) exists.
- Let f be differentiable on I with f' monotone on I. Prove that f' is continuous on I.
- Let f be differentiable on \mathbb{R} with f(0) = 0 and \lvert f'(x) \rvert \le \lvert f(x) \rvert for every x. Prove that f(x) = 0 for every x.
- Let f be differentiable on (A, \infty). Prove that if \lim_{x \to \infty} f(x) and \lim_{x \to \infty} f'(x) both exist, the second is 0. Then refute “if \lim_{x \to \infty} f(x) exists, then \lim_{x \to \infty} f'(x) = 0” with \displaystyle f(x) = \frac{\sin(x^2)}{x} on (1, \infty).
- Let f be differentiable on I with f'(x) \ne 0 for every x \in I. Prove that f is strictly monotone on I.
Answers
- \displaystyle c = \frac{\pi}{4}. The equation (1 - 0)(-\sin c) = (0 - 1)\cos c says \sin c = \cos c. The path is the quarter unit circle from (1, 0) to (0, 1), whose tangent at the midpoint of the arc is parallel to the chord.
- \displaystyle c = \ln\frac{e + 1}{2} \approx 0.620. The equation (e^2 - 1)e^c = (e - 1) \cdot 2e^{2c}, divided by (e - 1)e^c, is e + 1 = 2e^c; and \displaystyle 1 < \frac{e + 1}{2} < e.
- 9 \cdot 2c = 3 \cdot 3c^2 gives 9c(c - 2) = 0, and only c = 0 is in (-1, 2), where f'(0) = g'(0) = 0. The chord ratio is \displaystyle \frac{9}{3} = 3, but \displaystyle \frac{f'(c)}{g'(c)} = \frac{3c}{2} for c \ne 0 is 3 only at the end c = 2. The hypothesis g' \ne 0 fails at 0.
- \displaystyle M = \frac{1}{4}. For x > 4, \displaystyle \frac{1}{2\sqrt{x}} < \frac{1}{4}, so Theorem 4(d) applies. No smaller M works: \displaystyle \frac{\sqrt{x} - 2}{x - 4} = \frac{1}{\sqrt{x} + 2} \to \frac{1}{4} as x \to 4.
- f(x) = 2x + \sin 2x has f'(x) = 2 + 2\cos 2x \ge 0, zero only at \displaystyle \frac{\pi}{2} + k\pi, so as in Example 3 it is strictly increasing and takes the value 1 at most once. As \displaystyle f\left(\frac{1}{4}\right) = \frac{1}{2} + \sin\frac{1}{2} \approx 0.979 and \displaystyle f\left(\frac{1}{2}\right) = 1 + \sin 1 \approx 1.841, Bolzano’s theorem (Extreme and intermediate value theorems, proved) gives a solution in \displaystyle \left(\frac{1}{4}, \frac{1}{2}\right).
- f'(0) = 0. f is continuous on [0, \infty), and for x > 0 \begin{aligned} f'(x) &= \frac{\sin x}{2\sqrt{x}} + \sqrt{x}\cos x \\ &= \frac{\sqrt{x}}{2} \cdot \frac{\sin x}{x} + \sqrt{x}\cos x \to 0 \end{aligned}
- f'(0) = 1 for every a; f''(0) exists exactly when \displaystyle a = \frac{1}{2}, and then f''(0) = 1. f is continuous at 0, and f'(x) is e^x \to 1 on the left and 1 + 2ax \to 1 on the right, so Theorem 5 gives f'(0) = 1. Then Theorem 5 for f', with f'' = e^x \to 1 on the left and f'' = 2a on the right, says f''(0) exists if and only if 2a = 1.
- Such an f has f'(-1) = 1 and f'(1) = 3, so Theorem 6 gives \displaystyle f'(c) = \frac{3}{2} for some c \in (-1, 1). But f' = 1 for x \le 0 and f'(x) = x^2 + 2 > 2 for x > 0.
- Proof. For x \ne 0, \displaystyle s'(x) = 2x\sin\frac{1}{x} - \cos\frac{1}{x} (Differentiability and the derivative rules, proved). With \displaystyle x_k = \frac{1}{2k\pi} and \displaystyle y_k = \frac{1}{(2k + 1)\pi}, s'(x_k) = -1 and s'(y_k) = 1. Choose k with x_k < \eta; then 0 < y_k < x_k < \eta, and Theorem 6 on [y_k, x_k] gives every value strictly between -1 and 1, the ends giving \pm 1. \square
- -2. Near 1, top and bottom tend to 0 and \displaystyle g'(x) = \frac{1}{x} - 1 \ne 0 for x \ne 1. The quotient \displaystyle \frac{x^x(\ln x + 1) - 1}{1/x - 1} is again \displaystyle \frac{0}{0}, with \displaystyle \left(\frac{1}{x} - 1\right)' = -\frac{1}{x^2} \ne 0, and its quotient of derivatives is \displaystyle \frac{x^x(\ln x + 1)^2 + x^{x - 1}}{-1/x^2} \to \frac{2}{-1}. Theorem 7(a) twice, on each side, gives -2.
- 2. On (0, \pi), \lvert \ln x \rvert \to \infty and (\ln x)' \ne 0; the quotient of derivatives is \displaystyle \frac{x\sin x}{1 - \cos x}, of the form \displaystyle \frac{0}{0} with (1 - \cos x)' = \sin x \ne 0, and its own quotient of derivatives is \displaystyle 1 + \cos x \cdot \frac{x}{\sin x} \to 2. Theorem 7(a), then 7(b), give 2.
- Proof. Let \eta = \min\{\delta, b - a\}, with \delta from the definition, and \displaystyle x_n = a + \frac{\eta}{n + 1}. Then f(x_n) \le f(a), so \displaystyle Q(x_n) = \frac{f(x_n) - f(a)}{x_n - a} \le 0. As x_n \to a with x_n \ne a, Q(x_n) \to f'(a) (Theorem 2 of Limits of functions, epsilon and delta), and the order limit theorem gives f'(a) \le 0. At b, with \displaystyle y_n = b - \frac{\eta}{n + 1}, a numerator \le 0 over y_n - b < 0 gives quotients \ge 0, so f'(b) \ge 0. \square
- Proof. By Theorem 4(d), \lvert f(x) - f(y) \rvert \le M\lvert x - y \rvert on (0, 1]. Let x_n \in (0, 1] tend to 0. As \{x_n\} is Cauchy, for \varepsilon > 0 some N gives \displaystyle \lvert x_n - x_m \rvert < \frac{\varepsilon}{M + 1}, hence \lvert f(x_n) - f(x_m) \rvert < \varepsilon, for n, m > N; so \{f(x_n)\} converges (Cauchy sequences). For a second such sequence \{y_n\}, the images of x_1, y_1, x_2, y_2, \dots converge, so all such limits are one number L, and Theorem 2 of Limits of functions, epsilon and delta gives \lim_{x \to 0^+} f(x) = L. \square
- Proof. Using -f if needed, let f' be increasing. Let c \in I not be the right end, and d > c in I. On (c, d), f' \ge f'(c), so L = \inf\{f'(x) : c < x < d\} exists (Least upper bounds and completeness). Given \varepsilon > 0, some x_0 \in (c, d) has f'(x_0) < L + \varepsilon, and for c < x < x_0, L \le f'(x) < L + \varepsilon. So f'(x) \to L as x \to c^+, and Theorem 5 on [c, d) gives f'(c) = L. Likewise on the left. By Theorem 7 of Limits of functions, epsilon and delta, f' is continuous at c. \square
- Proof. Let f(p) = 0, and let M be the maximum of \lvert f \rvert on \displaystyle \left[p, p + \frac{1}{2}\right], attained at x_0. If x_0 \ne p, the mean value theorem on [p, x_0] gives c with \displaystyle M = \lvert f'(c) \rvert (x_0 - p) \le \frac{\lvert f(c) \rvert}{2} \le \frac{M}{2}. Either way M = 0, so f = 0 on \displaystyle \left[p, p + \frac{1}{2}\right], and likewise on \displaystyle \left[p - \frac{1}{2}, p\right]. Starting at p = 0, Proof by induction gives f = 0 on \displaystyle \left[-\frac{k}{2}, \frac{k}{2}\right] for every k \ge 1, and these cover \mathbb{R} (The Archimedean property and density of the rationals). \square
- Proof. Let f \to L and f' \to M > 0. Some X > A has \displaystyle f'(x) > \frac{M}{2} and \lvert f(x) - L \rvert < 1 for x > X. Let p = X + 1. For \displaystyle x > p + \frac{4}{M} the mean value theorem gives f(x) - f(p) = f'(c)(x - p) > 2, while \lvert f(x) - f(p) \rvert < 2. For M < 0 use -f. So M = 0. For the refutation, \displaystyle \lvert f(x) \rvert \le \frac{1}{x} \to 0, but \displaystyle f'(x) = 2\cos(x^2) - \frac{\sin(x^2)}{x^2} is 2 at \displaystyle \sqrt{2k\pi} and -2 at \displaystyle \sqrt{(2k + 1)\pi}, so f' has no limit. \square
- Proof. If f'(p) < 0 < f'(q) for some p, q \in I, Theorem 6 on the interval between them gives a zero of f', which is excluded. So f' > 0 on all of I or f' < 0 on all of I, and Theorem 4(b) finishes, f being continuous. \square