The fundamental theorem of calculus, proved
Calculus says that integrating a derivative gives back the function, and differentiating an integral gives back the integrand. Once the integral is defined by upper and lower sums, exactly which functions does each half hold for?
Where you are 6 in Integration and sequences of functions
Why this exists
The Riemann integral, defined defined the integral through upper and lower sums, and Which functions are integrable showed which functions have one, but nothing yet ties the integral to derivatives. Calculus did so in two statements: The second fundamental theorem of calculus evaluates \int_a^b f as F(b) - F(a) for an antiderivative F, and The first fundamental theorem of calculus proves \displaystyle \frac{d}{dx}\int_a^x f = f(x) for continuous f.
This lesson proves both under the weakest hypotheses this integral allows. In the evaluation formula the derivative need only be integrable. In the differentiation formula the integrand need only be integrable, and the derivative is found at each point where it is continuous. Books number the two halves differently, so here they are named by what they do.
Definition
Throughout, a < b. For f \in \mathcal{R}[a, b] and a \le c < d \le b, f \in \mathcal{R}[c, d] (Which functions are integrable, Exercise 10). To integrate over a piece in either direction, set
Integral function. For f \in \mathcal{R}[a, b] and a base point d \in [a, b], the integral function of f from d is F : [a, b] \to \mathbb{R}, F(x) = \int_d^x f.
Antiderivative. F : [a, b] \to \mathbb{R} is an antiderivative of f if F is continuous on [a, b] and F'(x) = f(x) for every x \in (a, b). Nothing is asked at a and b.
Lipschitz. F is Lipschitz with constant B > 0 if \lvert F(x) - F(y) \rvert \le B\lvert x - y \rvert for all x, y. Such an F is continuous, with \displaystyle \delta = \frac{\varepsilon}{B}.
The one-sided derivatives F'_-(c) and F'_+(c) are as in Differentiability and the derivative rules, proved; at a or b the derivative of a function on [a, b] is one-sided because of the domain.
Intuition
For the evaluation formula, cut [a, b] by a partition. On each piece the mean value theorem finds c_i where the tangent is parallel to the chord (Figure 1), so the rise of F over the piece is f(c_i)\,\Delta x_i, a rectangle between the lower one, m_i\,\Delta x_i, and the upper one, M_i\,\Delta x_i. Adding, the total rise F(b) - F(a) lies between L(P, f) and U(P, f) for every partition P, and only one number does that.
For the differentiation formula: an integral function changes over [x, y] by \int_x^y f, at most B per unit of length when \lvert f \rvert \le B, so it is continuous whatever f does. Near a point c where f is continuous, f is nearly the constant f(c), so F(x) - F(c) is nearly f(c)(x - c). A jump in f becomes a corner in F, while a single stray value of f changes no integral (Example 1, Figure 2).
Results
In Results 1 to 4, f \in \mathcal{R}[a, b].
1. Oriented additivity. For all c, d, e \in [a, b],
and if \lvert f(t) \rvert \le B for every t in the closed interval with ends c and d, then \left\lvert \int_c^d f \right\rvert \le B\lvert d - c \rvert.
2. Evaluation theorem. Let F : [a, b] \to \mathbb{R} be continuous on [a, b] and differentiable on (a, b), with F'(x) = f(x) for every x \in (a, b). Then
3. Finitely many exceptions. Result 2 still holds if, for a finite set S, F is continuous on [a, b] and differentiable with F'(x) = f(x) only at the points of (a, b) \setminus S.
4. Differentiation theorem. Let d \in [a, b] and F(x) = \int_d^x f for x \in [a, b].
(a) If \lvert f \rvert \le B on [a, b], then \lvert F(x) - F(y) \rvert \le B\lvert x - y \rvert for all x, y \in [a, b], so F is continuous.
(b) If f is continuous at c \in [a, b], then F is differentiable at c and F'(c) = f(c).
If f is continuous on [a, b], then F is an antiderivative of f with F' = f at every point of [a, b], and every antiderivative of f is F + C for a constant C.
5. Integration by parts. Let F, G : [a, b] \to \mathbb{R} be continuous, S finite, and f, g \in \mathcal{R}[a, b] with F' = f and G' = g at every point of (a, b) \setminus S. Then
6. Substitution. Let \varphi : [a, b] \to \mathbb{R} be differentiable with \varphi' \in \mathcal{R}[a, b], and let f be continuous on an open interval J containing \varphi([a, b]). Then
Why it is true
Two facts from The Riemann integral, defined are used freely: if m \le f \le M on [a, b], then m(b - a) \le \int_a^b f \le M(b - a) (Result 3 there), and \int_a^b f is the only number between every lower sum and every upper sum (Result 4 there).
Result 1. Proof. Let I(c, d, e) = \int_c^d f + \int_d^e f + \int_e^c f; the claim is I(c, d, e) = 0. By the convention, swapping two of the points reverses every term (for example I(d, c, e) = -I(c, d, e)), and every arrangement is reached by swaps, so it is enough to take c \le d \le e. If two points coincide, two terms cancel and the third is 0. If c < d < e, then I(c, d, e) = \int_c^d f + \int_d^e f - \int_c^e f = 0 by additivity, Result 3 of Which functions are integrable. For the bound with c < d, -B \le f \le B on [c, d] gives \left\lvert \int_c^d f \right\rvert \le B(d - c); the convention handles d \le c. \square
Result 2. Proof. Let P = \{x_0, x_1, \dots, x_n\} be any partition of [a, b]. On [x_{i-1}, x_i], F is continuous and is differentiable inside, so the mean value theorem (Theorem 3 of The mean value theorem and its consequences) gives c_i \in (x_{i-1}, x_i) with
As m_i \le f(c_i) \le M_i and \Delta x_i > 0, m_i\,\Delta x_i \le F(x_i) - F(x_{i-1}) \le M_i\,\Delta x_i. Add over i; the middle sum telescopes to F(x_n) - F(x_0), so
This holds for every P, and \int_a^b f is the only number that does so. \square
Result 3. Proof. List a, b and the points of S \cap (a, b) in increasing order, a = s_0 < s_1 < \cdots < s_k = b. On each [s_{j-1}, s_j], F is continuous, differentiable inside with F' = f, and f is integrable, so Result 2 gives \int_{s_{j-1}}^{s_j} f = F(s_j) - F(s_{j-1}). Add over j: the left side is \int_a^b f by Result 1, and the right side telescopes to F(b) - F(a). \square
Result 4. Proof. (a) By Result 1, F(x) - F(y) = \int_d^x f - \int_d^y f = \int_y^x f, and the bound of Result 1 gives \left\lvert \int_y^x f \right\rvert \le B\lvert x - y \rvert.
(b) Let \varepsilon > 0, and choose \delta > 0 with \displaystyle \lvert f(t) - f(c) \rvert < \frac{\varepsilon}{2} for t \in [a, b], \lvert t - c \rvert < \delta. Take x \in [a, b] with 0 < \lvert x - c \rvert < \delta. A constant integrates to \int_c^x f(c)\,dt = f(c)(x - c) in either direction, so Result 1 and the sum rule (Result 4 of Which functions are integrable) give
Every t between c and x has \lvert t - c \rvert \le \lvert x - c \rvert < \delta, so Result 1, applied to the integrable function f - f(c) with \displaystyle B = \frac{\varepsilon}{2}, bounds the right side by \displaystyle \frac{\varepsilon}{2}\lvert x - c \rvert. Dividing by \lvert x - c \rvert,
So F'(c) = f(c), one-sided if c is a or b. If f is continuous on [a, b], (b) holds everywhere, and for another antiderivative G, G - F has derivative 0 on (a, b), so it is constant by Theorem 4(c) of The mean value theorem and its consequences. \square
Result 5. Proof. F and G are continuous, so integrable (Result 1 of Which functions are integrable), and products of integrable functions are integrable (Exercise 12 there), so h = fG + Fg is integrable by the sum rule. FG is continuous on [a, b], and at each x \in (a, b) \setminus S the product rule gives (FG)'(x) = f(x)G(x) + F(x)g(x) = h(x). Result 3 gives \int_a^b h = F(b)G(b) - F(a)G(a), and the sum rule splits \int_a^b h into \int_a^b fG + \int_a^b Fg. \square
Result 6. Proof. \varphi is continuous, so \varphi([a, b]) is a closed bounded interval [p, q], possibly a single point (Extreme and intermediate value theorems, proved). As J is open, it contains some [p', q'] with p' < p and q < q'. Let H(u) = \int_{\varphi(a)}^{u} f for u \in [p', q']. As f is continuous there, Result 4 gives H'(u) = f(u) at every u \in (p', q'), which includes every value of \varphi. By the chain rule (Differentiability and the derivative rules, proved), H \circ \varphi is differentiable on [a, b] with derivative f(\varphi(x))\,\varphi'(x), and this function is integrable: f \circ \varphi is continuous (Continuity, defined and proved), so integrable, \varphi' \in \mathcal{R}[a, b], and products of integrable functions are integrable (Which functions are integrable, Exercise 12). Result 2 gives
Worked examples
Example 1: a stray value and a jump
Let f : [0, 3] \to \mathbb{R} be f(t) = 2t for 0 \le t < 1, f(1) = 4, f(t) = 2 for 1 < t < 2, and f(t) = t - 1 for 2 \le t \le 3 (Figure 2). Find F(x) = \int_0^x f, decide where F is differentiable, and compare F' with f.
f is bounded by 4 and continuous except at 1 and 2, so f \in \mathcal{R}[0, 3] by Result 6 of Which functions are integrable. For 0 \le x \le 1, f agrees with 2t on [0, x] except possibly at t = 1, so Result 7 there and Result 2 here give F(x) = x^2. For 1 \le x \le 2, Result 1 adds \int_1^x 2\,dt, so F(x) = 1 + 2(x - 1) = 2x - 1. For 2 \le x \le 3,
so \displaystyle F(3) = \frac{9}{2}, and F is Lipschitz with constant 4 by Result 4(a).
By Result 4(b), F'(x) = f(x) at every x \ne 1, 2. At 1 the left quotient is \displaystyle \frac{x^2 - 1}{x - 1} = x + 1 \to 2 and the right quotient is \displaystyle \frac{(2x - 1) - 1}{x - 1} = 2, so F'(1) = 2 although f(1) = 4: Result 4(b) has no converse. At 2 the left quotient is \displaystyle \frac{(2x - 1) - 3}{x - 2} = 2, and the right quotient is
so F'_-(2) = 2 and F'_+(2) = 1: F has a corner. At both points the one-sided derivatives of F are the one-sided limits of f (Exercise 12).
Example 2: Lucas’s annealer
Lucas runs an annealer, an oven that heats glass and cools it slowly. Its controller sets the rate of temperature change r(t), in degrees Celsius per minute: \displaystyle r(t) = \frac{t}{20} for 0 \le t < 100, r(t) = 5 for 100 \le t < 150, r(t) = 0 for 150 \le t < 180, and r(t) = -2 for 180 \le t \le 240. The temperature T is continuous with T(0) = 20, and T'(t) = r(t) for 0 < t < 240 except at t = 150 and t = 180, where T has corners. Find T(60) and T(240).
Solutionr is bounded by 5 and continuous except at 150 and 180 (at t = 100 both formulas give 5), so r \in \mathcal{R}[0, 240]. On [0, 60] there is no exception and \displaystyle T' = \frac{t}{20}, so Result 2 gives \displaystyle T(60) = 20 + \int_0^{60} \frac{t}{20}\,dt = 20 + \frac{3600}{40} = 110. On [0, 240], T is continuous with T' = r off S = \{150, 180\}, so Result 3, split by Result 1, gives
so T(240) = 400 degrees.
Example 3: a derivative that is integrable but not continuous
Let \displaystyle g(x) = 2x\cos\frac{1}{x} + \sin\frac{1}{x} for \displaystyle 0 < x \le \frac{1}{\pi} and g(0) = 0. Prove that \displaystyle g \in \mathcal{R}[0, \frac{1}{\pi}] and \displaystyle \int_0^{1/\pi} g = -\frac{1}{\pi^2}.
SolutionProof. Let \displaystyle F(x) = x^2\cos\frac{1}{x} for x > 0 and F(0) = 0. F is continuous at 0 because \lvert F(x) \rvert \le x^2, and for x > 0 the product and chain rules give
g is bounded, \displaystyle \lvert g \rvert \le \frac{2}{\pi} + 1, and continuous except at 0, where \displaystyle \sin\frac{1}{x} has no limit; so \displaystyle g \in \mathcal{R}[0, \frac{1}{\pi}] by Result 6 of Which functions are integrable. Result 2 gives \displaystyle \int_0^{1/\pi} g = F(\frac{1}{\pi}) - F(0) = \frac{1}{\pi^2}\cos\pi = -\frac{1}{\pi^2}. \square
The value is about -0.1013. A version of the theorem that asks for a continuous integrand says nothing here.
Example 4: a limit of sums
Find \displaystyle \lim_{n \to \infty} \sum_{k=1}^{n} \frac{n}{(n + k)^2}, with an error bound for each n.
SolutionWith \displaystyle f(x) = \frac{1}{(1 + x)^2} and P_n the partition of [0, 1] into n equal pieces, the sum is \displaystyle S_n = \sum_{k=1}^{n} f(\frac{k}{n})\,\frac{1}{n}. As f is decreasing, its infimum on each piece is at the right end, so S_n = L(P_n, f), and the upper sum is \displaystyle U(P_n, f) = S_n + \frac{f(0) - f(1)}{n} = S_n + \frac{3}{4n}. f is continuous with antiderivative \displaystyle -\frac{1}{1 + x}, so Result 2 gives \displaystyle \int_0^1 f = -\frac{1}{2} + 1 = \frac{1}{2}. The integral lies between the two sums, so
and \displaystyle S_n \to \frac{1}{2} by the squeeze theorem.
Example 5: a logarithm from an integral
For x > 0 let \displaystyle \ell(x) = \int_1^x \frac{dt}{t}. Prove that \ell(xy) = \ell(x) + \ell(y) for all x, y > 0.
SolutionProof. On any [p, q] with 0 < p < 1 < q, \displaystyle \frac{1}{t} is continuous, so Result 4 with base point 1 gives \displaystyle \ell'(x) = \frac{1}{x} on (p, q). Every x > 0 lies in such an interval, so \displaystyle \ell'(x) = \frac{1}{x} on (0, \infty). Fix y > 0 and let h(x) = \ell(xy) - \ell(x). By the chain rule,
So h is constant on every [p, q] \subseteq (0, \infty) by Theorem 4(c) of The mean value theorem and its consequences, hence on (0, \infty), and h(x) = h(1) = \ell(y) - \ell(1) = \ell(y). \square
So the natural logarithm can be defined as \ell, with no exponents.
Example 6: a derivative that is not integrable
Let \displaystyle F(x) = x^2\sin\frac{1}{x^2} for 0 < x \le 1 and F(0) = 0. Show that F is differentiable on [0, 1], that F' \notin \mathcal{R}[0, 1], and that \int_c^1 F' \to \sin 1 as c \to 0^+.
SolutionAt 0, \displaystyle \left\lvert \frac{F(x) - F(0)}{x} \right\rvert = \left\lvert x\sin\frac{1}{x^2} \right\rvert \le x \to 0, so F'(0) = 0. For x > 0,
At \displaystyle x_k = \frac{1}{\sqrt{2k\pi}} for k \ge 1 the sine is 0 and the cosine is 1, so \displaystyle F'(x_k) = -2\sqrt{2k\pi}, unbounded below. Integrable functions are bounded, so F' \notin \mathcal{R}[0, 1], and Result 2 does not apply on [0, 1]. On [c, 1] with 0 < c < 1, F' is continuous, so it does:
since \displaystyle \left\lvert c^2\sin\frac{1}{c^2} \right\rvert \le c^2. The limit is an improper integral.
Boundaries
- Integrability of F' is assumed, not proved. Example 6 has an unbounded derivative, and even a bounded derivative can fail to be Riemann integrable (Volterra’s function, built in a later analysis course, whose Lebesgue integral handles every bounded derivative).
- F must be continuous at the exceptional points (Exercise 6). Nor can “finitely many” become “a set of length zero”: the Cantor function, from a later analysis course, rises continuously from 0 to 1 on [0, 1] with derivative 0 off a set of total length zero.
- Result 4(b) runs one way. F can be differentiable where f is not continuous (Example 1 at 1; Exercise 3, at every rational). At a jump of f, F has a corner (Exercise 12).
- Integral functions are Lipschitz, so \displaystyle \sqrt{x} on [0, 1] is not \int_0^x f for any f \in \mathcal{R}[0, 1] (Exercise 13).
Where this goes next
Pointwise and uniform convergence passes limits through integrals, and with Result 2 through derivatives. Power series, proved differentiates a power series term by term by integrating the derived series and applying Result 4. Integration by parts applied n times gives the integral form of the remainder that Taylor's theorem with remainder left for this lesson; Exercise 15 is the case n = 1.
Exercises
Integral functions
- Let f(t) = 3t^2 for 0 \le t \le 1 and f(t) = 5 - t for 1 < t \le 4, and let F(x) = \int_0^x f. Find F(x) on each piece, F(4), and the one-sided derivatives of F at 1.
- Let f(t) = t^2 - 2t on [-1, 3] and F(x) = \int_1^x f. Find the least B with \lvert f \rvert \le B on [-1, 3], and a \delta > 0 with \displaystyle \lvert F(x) - F(y) \rvert < \frac{1}{100} whenever \lvert x - y \rvert < \delta.
- Let \tau be Thomae’s function (Which functions are integrable, Example 6) and F(x) = \int_0^x \tau on [0, 1]. Find F and F', and every point where F'(x) \ne \tau(x).
Evaluation
- Let \displaystyle g(x) = 3x^2\cos\frac{1}{x^2} + 2\sin\frac{1}{x^2} for x \ne 0 and g(0) = 0. Prove that \displaystyle g \in \mathcal{R}[0, \frac{1}{\sqrt{\pi}}] and find \displaystyle \int_0^{1/\sqrt{\pi}} g.
- Jenny logs the annealer’s power: P(t) = 3t kilowatts for 0 \le t < 2 hours, 6 for 2 \le t < 5, and 2 for 5 \le t \le 8. The energy meter reading E(t), in kilowatt-hours, is continuous with E(0) = 0 and E'(t) = P(t) for 0 < t < 8 except at t = 5. Find E(1), E(8), and the cost of the run at \$0.15 per kilowatt-hour.
- Refute: “If F is differentiable at all but finitely many points of [a, b], and f \in \mathcal{R}[a, b] equals F' wherever F' exists, then \int_a^b f = F(b) - F(a).”
Limits of sums
- Find \displaystyle \lim_{n \to \infty} \sum_{k=1}^{n} \frac{k}{n^2 + k^2}, with bounds for each n.
- Find \displaystyle \lim_{n \to \infty} \sum_{k=1}^{n} \frac{1}{\sqrt{n^2 + kn}}.
Parts and substitution
- Let g(x) = -1 for x < 1 and g(x) = 1 for x \ge 1. Use Result 5 with G(x) = \lvert x - 1 \rvert to find \int_0^3 x\,g(x)\,dx, and check by splitting at 1.
- Use Result 6 with \varphi = \sin to find \int_0^{3\pi/2} \sin^2 x\cos x\,dx, and explain why \varphi need not be one-to-one.
- Let A(x) = \int_x^1 e^{-t^2}\,dt. Find \int_0^1 A(x)\,dx exactly.
Proofs
- Let f \in \mathcal{R}[a, b], c \in [a, b), and \lim_{t \to c^+} f(t) = L. Prove that F(x) = \int_a^x f has F'_+(c) = L.
- Prove that no f \in \mathcal{R}[0, 1] has \displaystyle \sqrt{x} = \int_0^x f for every x \in [0, 1].
- Let f be continuous on \mathbb{R}, and suppose \int_x^{x+1} f has the same value for every x. Prove that f(x + 1) = f(x) for every x.
- Let f and f' be differentiable on [a, b] with f'' \in \mathcal{R}[a, b]. Prove that f(b) = f(a) + f'(a)(b - a) + \int_a^b f''(t)(b - t)\,dt.
- Let f \in \mathcal{R}[a, b] with \int_p^q f = 0 for all rationals p < q in [a, b]. Prove that \int_c^d f = 0 for all c, d \in [a, b], and that f(x) = 0 at every point where f is continuous.
- Let f be continuous on [a, b] with f(x) = \int_a^x f for every x \in [a, b]. Prove that f(x) = 0 for every x.
Answers
- F(x) = x^3 on [0, 1], \displaystyle F(x) = -\frac{x^2}{2} + 5x - \frac{7}{2} on [1, 4], \displaystyle F(4) = \frac{17}{2}, F'_-(1) = 3 and F'_+(1) = 4. For 1 \le x \le 4, F(x) = 1 + \int_1^x (5 - t)\,dt \displaystyle = 1 + 5(x - 1) - \frac{x^2 - 1}{2}, the value f(1) not mattering. Differentiating the two formulas at 1 gives 3 and 4, the one-sided limits of f there (Exercise 12).
- B = 3 and \displaystyle \delta = \frac{1}{300}. f(-1) = 3, f(1) = -1 at the critical point, and f(3) = 3. By Result 4(a), \displaystyle \lvert x - y \rvert < \frac{1}{300} gives \displaystyle \lvert F(x) - F(y) \rvert \le 3\lvert x - y \rvert < \frac{1}{100}.
- F(x) = 0 and F'(x) = 0 for every x \in [0, 1], and F'(x) \ne \tau(x) exactly at the rationals of [0, 1]. Every lower sum of \tau is 0, so \int_0^x \tau and \int_x^1 \tau are \ge 0, and they add to \int_0^1 \tau = 0. At irrational x, \tau(x) = 0; at \displaystyle x = \frac{p}{q} in lowest terms, \displaystyle \tau(x) = \frac{1}{q} \ne 0.
- -\pi^{-3/2} \approx -0.180. Proof. Let \displaystyle F(x) = x^3\cos\frac{1}{x^2} and F(0) = 0; \lvert F(x) \rvert \le x^3, so F is continuous, and for x > 0, \displaystyle F'(x) = 3x^2\cos\frac{1}{x^2} + x^3\sin\frac{1}{x^2} \cdot \frac{2}{x^3} = g(x). g is bounded by \displaystyle \frac{3}{\pi} + 2 and continuous except at 0, so integrable by Result 6 of Which functions are integrable, and Result 2 gives \displaystyle \int_0^{1/\sqrt{\pi}} g = F(\frac{1}{\sqrt{\pi}}) = \pi^{-3/2}\cos\pi. \square
- \displaystyle E(1) = \frac{3}{2} kilowatt-hours, E(8) = 30 kilowatt-hours, and the cost is \$4.50. P is bounded and continuous except at 5, and E' = P off \{5\}, so Result 3 gives E(8) = \int_0^2 3t\,dt + \int_2^5 6\,dt + \int_5^8 2\,dt = 6 + 18 + 6. On [0, 1], Result 2 with antiderivative \displaystyle \frac{3t^2}{2} gives E(1).
- Let F(x) = \lfloor x \rfloor on [0, 3] and f = 0. F is differentiable with F' = 0 except at 1, 2 and 3, and f is integrable, but \int_0^3 f = 0 while F(3) - F(0) = 3. Result 3 also needs F continuous on [a, b].
- \displaystyle \frac{\ln 2}{2}, with \displaystyle \frac{\ln 2}{2} \le S_n \le \frac{\ln 2}{2} + \frac{1}{2n}. The term is \displaystyle \frac{1}{n}f(\frac{k}{n}) with \displaystyle f(x) = \frac{x}{1 + x^2}, increasing on [0, 1] since \displaystyle f'(x) = \frac{1 - x^2}{(1 + x^2)^2} \ge 0. So \displaystyle S_n = U(P_n, f) = L(P_n, f) + \frac{f(1) - f(0)}{n}, and Result 2 with antiderivative \displaystyle \frac{1}{2}\ln(1 + x^2) gives \int_0^1 f.
- \displaystyle 2\sqrt{2} - 2. The term is \displaystyle \frac{1}{n}f(\frac{k}{n}) with f(x) = (1 + x)^{-1/2}, decreasing, so S_n = L(P_n, f) with gap \displaystyle U - L = \frac{1}{n}\left(1 - \frac{1}{\sqrt{2}}\right) \to 0. Result 2 with antiderivative \displaystyle 2\sqrt{1 + x} gives \displaystyle \int_0^1 f = 2\sqrt{2} - 2.
- \displaystyle \frac{7}{2}. Take F(x) = x, f = 1, G(x) = \lvert x - 1 \rvert and S = \{1\}; G' = g off S. Result 5 gives \displaystyle 3 \cdot 2 - 0 - \int_0^3 \lvert x - 1 \rvert\,dx = 6 - \frac{5}{2}. Split at 1: \displaystyle -\int_0^1 x\,dx + \int_1^3 x\,dx = -\frac{1}{2} + 4.
- \displaystyle -\frac{1}{3}. With f(u) = u^2 on J = \mathbb{R}, Result 6 gives \displaystyle \int_0^{-1} u^2\,du = -\frac{1}{3}. The proof differentiates H \circ \varphi and never inverts \varphi.
- \displaystyle \frac{1 - e^{-1}}{2} \approx 0.316. By Result 4(b) with base point 1, A'(x) = -e^{-x^2}, and A(1) = 0. Result 5 with F = A and G(x) = x gives \int_0^1 A = A(1) \cdot 1 - A(0) \cdot 0 + \int_0^1 xe^{-x^2}\,dx \displaystyle = \left[-\frac{1}{2}e^{-x^2}\right]_0^1.
- Proof. Let \varepsilon > 0, and choose \delta > 0 with \displaystyle \lvert f(t) - L \rvert < \frac{\varepsilon}{2} for t \in (c, c + \delta) \cap [a, b]. For x in that set, Result 1 gives F(x) - F(c) - L(x - c) = \int_c^x (f - L). Let h = f - L on (c, x] and h(c) = 0. By Result 7 of Which functions are integrable, \int_c^x h = \int_c^x (f - L), and \displaystyle \lvert h \rvert \le \frac{\varepsilon}{2} on [c, x], so the bound of Result 1 gives \displaystyle \left\lvert \frac{F(x) - F(c)}{x - c} - L \right\rvert \le \frac{\varepsilon}{2} < \varepsilon. Hence F'_+(c) = L, whatever f(c) is. \square
- Proof. Suppose such an f exists, and let \lvert f \rvert \le B with B \ge 1. By Result 4(a), \displaystyle \sqrt{x} = \lvert F(x) - F(0) \rvert \le Bx for x \in [0, 1]. At \displaystyle x = \frac{1}{4B^2}, which lies in (0, 1], this says \displaystyle \frac{1}{2B} \le \frac{1}{4B}, which is false. \square
- Proof. Let N \ge 1 and A(x) = \int_0^x f on [-N, N]. As f is continuous, A' = f on (-N, N) by Result 4(b). For x \in (-N, N - 1), Result 1 gives \int_x^{x+1} f = A(x + 1) - A(x), whose derivative by the chain rule is f(x + 1) - f(x). A constant function has derivative 0, so f(x + 1) = f(x) on (-N, N - 1), and every real x lies in such an interval for N large. \square
- Proof. f' is differentiable, hence continuous (Differentiability and the derivative rules, proved) and integrable, so Result 2 gives \int_a^b f' = f(b) - f(a). Apply Result 5 with F(t) = b - t, whose derivative is -1, G = f', g = f'' and S = \varnothing: \int_a^b (b - t)f''(t)\,dt = \big[(b - t)f'(t)\big]_a^b + \int_a^b f' = -(b - a)f'(a) + f(b) - f(a). Rearranging gives the formula. \square
- Proof. Let F(x) = \int_a^x f. For rationals p < q in [a, b], F(q) - F(p) = \int_p^q f = 0 by Result 1, so F has one value K at every rational of [a, b]. Each x \in [a, b] is the limit of rationals \displaystyle r_n \in (x - \frac{1}{n}, x + \frac{1}{n}) \cap (a, b) (The Archimedean property and density of the rationals), and F is continuous by Result 4(a), so F(x) = \lim_{n \to \infty} F(r_n) = K (Continuity, defined and proved). Hence \int_c^d f = F(d) - F(c) = 0, and where f is continuous Result 4(b) gives f(x) = F'(x) = 0. \square
- Proof. By Result 4(b), f' = f on [a, b], and f(a) = \int_a^a f = 0. Let h(x) = e^{-x}f(x); then h'(x) = e^{-x}\big(f'(x) - f(x)\big) = 0, so h is constant by Theorem 4(c) of The mean value theorem and its consequences, and h(a) = 0. Hence f = 0. \square