Chapter VIII · The convergence of infinite series and infinite integrals

Infinite integrals

§§184–186 40 min

Integrals over an infinite range, defined as limits of ordinary integrals; their comparison tests; and substitution and integration by parts carried to infinity.

Why this matters

The integral test compares ∑ϕ(n)\sum\phi(n) with ∫1ξϕ(x) dx\int_1^\xi \phi(x)\,dx as ξ→∞\xi \to \infty. That limit deserves a name of its own: it is the integral of ϕ\phi over the infinite range [1,∞)[1, \infty). §§184–186 define such infinite integrals, show that their theory runs parallel to that of series (convergence, comparison, the general principle), point out where the parallel breaks, and extend the rules of substitution and integration by parts to them. Infinite integrals are everywhere in later analysis, from ∫0∞e−x2dx\int_0^\infty e^{-x^2}dx to the Gamma function and Fourier transforms.

The idea

An ordinary integral ∫aXϕ(x) dx\int_a^X \phi(x)\,dx is (for positive ϕ\phi) the area under the graph between aa and XX. Let the right-hand end run off to infinity. The region becomes infinitely long, but that does not make its area infinite: if the graph comes down to the axis fast enough, the areas ∫aXϕ\int_a^X \phi approach a finite limit, and we call that limit ∫a∞ϕ\int_a^\infty \phi. This is exactly how the sum of a series is the limit of its partial sums.

Explore An area that runs off to infinity

Try this. With x−sx^{-s} and s=2s = 2, move XX to the right, then show xx up to 100100 and 10001000. Now lower ss to 11 and below. Then try cos⁡x\cos x, xcos⁡xx\cos x and Hardy’s peaks.

∫1Xx−2 dx\int_{1}^{X} x^{-2}\,dx
xφ(x)
XΦ(X)
  • the integrand; shaded: area from 11 to XX (below the axis counts negative)
  • Φ(X)=∫1Xϕ\Phi(X) = \int_{1}^X \phi
  • its limit
Φ(5.5)=0.81818\Phi(5.5) = 0.81818. As X→∞X \to \infty the integral converges: its value is 11. For s>1s > 1, Φ(X)=(1−X1−s)/(s−1)→1/(s−1)\Phi(X) = (1 - X^{1-s})/(s-1) \to 1/(s-1).
What to notice

For s=2s = 2 the area creeps up to 11 and never passes it; for s≤1s \le 1 it grows without bound, just like the partial sums of ∑n−s\sum n^{-s}. ∫0Xcos⁡x dx=sin⁡X\int_0^X \cos x\,dx = \sin X swings for ever between −1-1 and 11, and ∫0Xxcos⁡x dx\int_0^X x\cos x\,dx swings ever more widely: integrals can oscillate, as series can. The peaks never get lower, yet their total area is finite.

The last example is the surprise. For a convergent series the terms must tend to 00. For a convergent integral the integrand need not tend to 00, even when it is positive and continuous: it can have taller and taller (or, here, equally tall) spikes, provided they become thin fast enough.

Precise statements

Definition Infinite integral (of the first kind) Hardy: §184

Let ϕ\phi be continuous for x≥ax \ge a. If

lim⁡X→∞∫aXϕ(x) dx=l\lim_{X \to \infty} \int_a^X \phi(x)\,dx = l

exists, we say that the infinite integral ∫a∞ϕ(x) dx\displaystyle\int_a^\infty \phi(x)\,dx converges and has the value ll. If ∫aXϕ→∞\int_a^X \phi \to \infty (or −∞-\infty) we say it diverges to ∞\infty (or −∞-\infty); otherwise it oscillates, finitely or infinitely, according as ∫aXϕ\int_a^X\phi stays bounded or not.

Integrals ∫−∞a\int_{-\infty}^a are defined in the same way, and ∫−∞∞ϕ\int_{-\infty}^\infty \phi is defined as ∫−∞0ϕ+∫0∞ϕ\int_{-\infty}^0 \phi + \int_0^\infty \phi when both converge (Examples LXXIII, 5–6). Hardy’s remarks on the definition (§184, (i)–(vii)), restated:

  1. If Φ\Phi is any antiderivative of ϕ\phi, then ∫a∞ϕ=Φ(∞)−Φ(a)\int_a^\infty \phi = \Phi(\infty) - \Phi(a), where Φ(∞)\Phi(\infty) means lim⁡x→∞Φ(x)\lim_{x\to\infty}\Phi(x), whenever this limit exists.
  2. If ϕ≥0\phi \ge 0, then ∫aXϕ\int_a^X \phi increases with XX, so the integral either converges or diverges to ∞\infty.
  3. General principle. ∫a∞ϕ\int_a^\infty \phi converges if and only if for every δ>0\delta > 0 there is an X(δ)X(\delta) with ∣∫x1x2ϕ∣<δ\left|\int_{x_1}^{x_2} \phi\right| < \delta whenever x2>x1≥X(δ)x_2 > x_1 \ge X(\delta).
  4. Integral test restated. If ϕ\phi is positive and decreasing, then ∑ϕ(n)\sum \phi(n) and ∫1∞ϕ(x) dx\int_1^\infty \phi(x)\,dx converge or diverge together.
  5. If ∫a∞ϕ\int_a^\infty \phi converges and b>ab > a, then ∫b∞ϕ\int_b^\infty \phi converges and ∫a∞ϕ=∫abϕ+∫b∞ϕ\int_a^\infty \phi = \int_a^b \phi + \int_b^\infty \phi.

For positive integrands, the analogues of the theorems for positive series hold (§185).

Theorem Comparison for positive integrands Hardy: §185

Let ϕ\phi and ψ\psi be continuous and positive for x≥ax \ge a.

  1. ∫a∞ϕ\int_a^\infty \phi converges if and only if there is a constant KK with ∫aXϕ<K\int_a^X \phi < K for all X>aX > a.
  2. If ∫a∞ϕ\int_a^\infty \phi converges and ψ(x)≤Kϕ(x)\psi(x) \le K\phi(x) for x≥ax \ge a, then ∫a∞ψ\int_a^\infty \psi converges and ∫a∞ψ≤K∫a∞ϕ\int_a^\infty \psi \le K\int_a^\infty\phi.
  3. In particular: if ϕ(x)≤Kx−s\phi(x) \le Kx^{-s} with s>1s > 1, the integral converges; if ϕ(x)≥Kx−s\phi(x) \ge Kx^{-s} with K>0K > 0 and s≤1s \le 1, it diverges; and if xsϕ(x)→l>0x^s\phi(x) \to l > 0, it converges or diverges according as s>1s > 1 or s≤1s \le 1 (here a>0a > 0).

There is no analogue of d’Alembert’s test, since an integrand has no “next term”. And the analogy with series breaks at one point:

Warning The integrand need not tend to 0 Hardy: §185, Fig. 47

There are positive continuous functions ϕ\phi with ∫0∞ϕ\int_0^\infty \phi convergent but ϕ(x)↛0\phi(x) \not\to 0.

Hardy’s example: let ϕ\phi be 00 except for a triangular peak of height 11 and base 2/(n+1)22/(n+1)^2 centred at each integer n≥1n \ge 1. The peak at nn has area 1/(n+1)21/(n+1)^2, so ∫0Xϕ<∑n≥11/(n+1)2\int_0^X \phi < \sum_{n\ge1} 1/(n+1)^2 for every XX, and the integral converges, while ϕ(n)=1\phi(n) = 1 for every nn.

Finally, the rules of calculation carry over (§186).

Theorem Substitution in an infinite integral Hardy: §186 (1)

Let ∫a∞ϕ(x) dx\int_a^\infty \phi(x)\,dx converge. Let x=f(t)x = f(t), where ff has a continuous derivative, f(b)=af(b) = a, the substitution rule of §166 holds on every interval [b,τ][b, \tau], and f(t)→∞f(t) \to \infty as t→∞t \to \infty. Then

∫a∞ϕ(x) dx=∫b∞ϕ{f(t)} f′(t) dt,\int_a^\infty \phi(x)\,dx = \int_b^\infty \phi\{f(t)\}\,f'(t)\,dt ,

the integral on the right being convergent.

Theorem Integration by parts to infinity Hardy: §186 (2)

If ff and ϕ\phi have continuous derivatives for x≥ax \ge a, and two of the three limits below exist, then so does the third, and

∫a∞f(x)ϕ′(x) dx=lim⁡ξ→∞f(ξ)ϕ(ξ)−f(a)ϕ(a)−∫a∞f′(x)ϕ(x) dx.\int_a^\infty f(x)\phi'(x)\,dx = \lim_{\xi\to\infty} f(\xi)\phi(\xi) - f(a)\phi(a) - \int_a^\infty f'(x)\phi(x)\,dx .

Proofs

Proof Comparison for positive integrands
Strategy

The function Φ(X)=∫aXϕ\Phi(X) = \int_a^X\phi plays the role of the partial sums. For positive ϕ\phi it increases, and an increasing function of XX tends to a limit exactly when it is bounded. Then compare areas.

Part 1: bounded increasing functions converge

For X2>X1X_2 > X_1, Φ(X2)−Φ(X1)=∫X1X2ϕ≥0\Phi(X_2) - \Phi(X_1) = \int_{X_1}^{X_2}\phi \ge 0, so Φ\Phi increases. An increasing function bounded above tends to a limit as X→∞X \to \infty (the continuous-variable form of §69, proved in Ch. V); an unbounded one tends to ∞\infty.

Part 2: compare the areas

If ψ≤Kϕ\psi \le K\phi, then ∫aXψ≤K∫aXϕ≤K∫a∞ϕ\int_a^X \psi \le K\int_a^X\phi \le K\int_a^\infty\phi for every XX, by the monotonicity of the ordinary integral. So ∫aXψ\int_a^X\psi is bounded, hence converges by Part 1, with the stated inequality in the limit.

Part 3: the yardstick x−sx^{-s}

By §181, ∫1∞x−s dx\int_1^\infty x^{-s}\,dx converges for s>1s > 1 and diverges for s≤1s \le 1 (and likewise from any a>0a > 0). The first two claims are Part 2 and its divergence form. If xsϕ(x)→l>0x^s\phi(x) \to l > 0, then 12l x−s<ϕ(x)<2l x−s\tfrac12 l\, x^{-s} < \phi(x) < 2l\, x^{-s} for large xx, and the first two claims apply from that point on (the finite piece ∫ax0ϕ\int_a^{x_0}\phi does not affect convergence, by remark 5).

Proof Substitution in an infinite integral
Strategy

Apply the ordinary rule on a finite range, then let the upper limit go to infinity on both sides at once.

Finite ranges

For each τ>b\tau > b, with ξ=f(τ)\xi = f(\tau), the rule of §166 gives ∫aξϕ(x) dx=∫bτϕ{f(t)}f′(t) dt\int_a^{\xi}\phi(x)\,dx = \int_b^\tau \phi\{f(t)\}f'(t)\,dt.

Let τ→∞\tau \to \infty

As τ→∞\tau \to \infty, ξ=f(τ)→∞\xi = f(\tau) \to \infty, so the left side tends to ∫a∞ϕ\int_a^\infty\phi (a function with a limit at infinity, composed with a function tending to infinity, has the same limit). Hence the right side has the same limit, which is the statement.

Other endings

If instead ξ→∞\xi \to \infty as τ→−∞\tau \to -\infty, or as τ\tau tends to a finite cc, the same argument gives ∫a∞ϕ=−∫−∞bϕ{f(t)}f′(t) dt\int_a^\infty\phi = -\int_{-\infty}^b \phi\{f(t)\}f'(t)\,dt, or lim⁡τ→c∫bτϕ{f(t)}f′(t) dt\lim_{\tau\to c}\int_b^\tau \phi\{f(t)\}f'(t)\,dt. The second case leads to integrals of the second kind (next lesson).

Proof Integration by parts to infinity
Strategy

Write the finite-range formula with upper limit ξ\xi and let ξ→∞\xi \to \infty.

Finite range

∫aξfϕ′=f(ξ)ϕ(ξ)−f(a)ϕ(a)−∫aξf′ϕ\int_a^\xi f\phi' = f(\xi)\phi(\xi) - f(a)\phi(a) - \int_a^\xi f'\phi, by §166.

Two limits give the third

The three quantities depending on ξ\xi satisfy a linear relation with the constant f(a)ϕ(a)f(a)\phi(a). If two of them tend to limits as ξ→∞\xi \to \infty, the third, being a sum of the others and a constant, does too, and the relation holds in the limit.

Worked examples and exercises

Exercise ●○○ Hardy: Examples LXXIII, 1 and 2

Let PP and QQ be polynomials of degrees rr and ss with positive leading coefficients, and aa beyond the largest root of QQ. Show that ∫a∞P(x)/Q(x) dx\int_a^\infty P(x)/Q(x)\,dx converges if s>r+1s > r + 1 and diverges otherwise. Which of these converge (a>0a > 0, c≠0c \ne 0)?

∫a∞dxx,∫a∞dxx4/3,∫a∞dxc2+x2,∫a∞x dxc2+x2,∫a∞x2 dxc2+x2.\int_a^\infty \frac{dx}{\sqrt x}, \quad \int_a^\infty \frac{dx}{x^{4/3}}, \quad \int_a^\infty \frac{dx}{c^2 + x^2}, \quad \int_a^\infty \frac{x\,dx}{c^2 + x^2}, \quad \int_a^\infty \frac{x^2\,dx}{c^2 + x^2}.
Hint
xs−rP(x)/Q(x)x^{s-r}P(x)/Q(x) tends to the ratio of the leading coefficients. Use part 3 of the comparison theorem.
Solution

P(x)/Q(x)P(x)/Q(x) is continuous and positive for large xx (beyond the roots of PP and QQ, and the finite piece before that does not matter). xs−rP(x)/Q(x)→A/α>0x^{s-r}P(x)/Q(x) \to A/\alpha > 0, so the integral behaves like ∫∞x−(s−r)dx\int^\infty x^{-(s-r)}dx: convergent if s−r>1s - r > 1, divergent if s−r≤1s - r \le 1.

So: ∫dx/x\int dx/\sqrt x diverges (power 12\tfrac12); ∫dx/x4/3\int dx/x^{4/3} converges; ∫dx/(c2+x2)\int dx/(c^2+x^2) converges (s−r=2s - r = 2); ∫x dx/(c2+x2)\int x\,dx/(c^2 + x^2) diverges (s−r=1s - r = 1; indeed it equals 12log⁡c2+X2c2+a2→∞\tfrac12\log\frac{c^2+X^2}{c^2+a^2} \to \infty); ∫x2dx/(c2+x2)\int x^2dx/(c^2+x^2) diverges (the integrand tends to 11).

Exercise ●○○ Hardy: Examples LXXIII, 7

Show that ∫−∞0dx1+x2=∫0∞dx1+x2=12∫−∞∞dx1+x2=12π\displaystyle\int_{-\infty}^0 \frac{dx}{1+x^2} = \int_0^\infty \frac{dx}{1+x^2} = \frac12\int_{-\infty}^\infty \frac{dx}{1+x^2} = \frac12\pi.

Solution

∫0Xdx/(1+x2)=arctan⁡X→12π\int_0^X dx/(1+x^2) = \arctan X \to \tfrac12\pi as X→∞X \to \infty, and ∫−X0dx/(1+x2)=−arctan⁡(−X)=arctan⁡X→12π\int_{-X}^0 dx/(1+x^2) = -\arctan(-X) = \arctan X \to \tfrac12\pi. Both halves converge, so by definition ∫−∞∞=12π+12π=π\int_{-\infty}^\infty = \tfrac12\pi + \tfrac12\pi = \pi.

Exercise ●○○ Hardy: Examples LXXIV, 1

Let s>1s > 1 and α>0\alpha > 0. Show by the substitution x=tαx = t^\alpha that ∫1∞x−s dx=α∫1∞tα(1−s)−1 dt\int_1^\infty x^{-s}\,dx = \alpha\int_1^\infty t^{\alpha(1-s)-1}\,dt, and check that both sides equal 1/(s−1)1/(s-1).

Solution

x=tαx = t^\alpha maps [1,∞)[1, \infty) onto [1,∞)[1, \infty), increasingly, with dx=αtα−1dtdx = \alpha t^{\alpha-1}dt, and x−s=t−αsx^{-s} = t^{-\alpha s}. So the right-hand integrand is αt−αstα−1=αtα(1−s)−1\alpha t^{-\alpha s}t^{\alpha-1} = \alpha t^{\alpha(1-s)-1}, and the theorem applies because the left side converges (s>1s > 1). Directly: the left is 1/(s−1)1/(s-1); on the right the exponent α(1−s)−1=−(1+α(s−1))\alpha(1-s) - 1 = -(1 + \alpha(s-1)) is less than −1-1, so α∫1∞t−(1+α(s−1))dt=α⋅1α(s−1)=1s−1\alpha\int_1^\infty t^{-(1 + \alpha(s-1))}dt = \alpha\cdot\dfrac{1}{\alpha(s-1)} = \dfrac{1}{s-1}.

Exercise ●●○ Hardy: Examples LXXIV, 4

Show that ∫1∞dx(1+x)x=12π\displaystyle\int_1^\infty \frac{dx}{(1+x)\sqrt x} = \frac12\pi.

Hint
Put x=t2x = t^2.

Solution

The integrand is positive and below x−3/2x^{-3/2}, so the integral converges. With x=t2x = t^2 (t≥1t \ge 1, increasing, t→∞t \to \infty as x→∞x \to \infty), dx=2t dtdx = 2t\,dt and x=t\sqrt x = t:

∫1∞2t dt(1+t2) t=2∫1∞dt1+t2=2(π2−π4)=π2.\int_1^\infty \frac{2t\,dt}{(1+t^2)\,t} = 2\int_1^\infty \frac{dt}{1+t^2} = 2\left(\frac\pi2 - \frac\pi4\right) = \frac\pi2 .
Exercise ●●○ Hardy: Examples LXXV, 1

Show by integration by parts that ∫0∞x dx(1+x)3=12∫0∞dx(1+x)2=12\displaystyle\int_0^\infty \frac{x\,dx}{(1+x)^3} = \frac12\int_0^\infty\frac{dx}{(1+x)^2} = \frac12.

Hint
Take f(x)=xf(x) = x and ϕ(x)=−12(1+x)−2\phi(x) = -\tfrac12(1+x)^{-2}, so that ϕ′(x)=(1+x)−3\phi'(x) = (1+x)^{-3}.

Solution

With f(x)=xf(x) = x, ϕ(x)=−12(1+x)−2\phi(x) = -\tfrac12(1+x)^{-2}: f(ξ)ϕ(ξ)=−ξ2(1+ξ)2→0f(\xi)\phi(\xi) = -\dfrac{\xi}{2(1+\xi)^2} \to 0, f(0)ϕ(0)=0f(0)\phi(0) = 0, and ∫0∞f′ϕ=−12∫0∞(1+x)−2dx=−12\int_0^\infty f'\phi = -\tfrac12\int_0^\infty (1+x)^{-2}dx = -\tfrac12 converges. Two limits exist, so the third does, and

∫0∞x dx(1+x)3=0−0+12∫0∞dx(1+x)2=12[−11+x]0∞=12.\int_0^\infty \frac{x\,dx}{(1+x)^3} = 0 - 0 + \frac12\int_0^\infty\frac{dx}{(1+x)^2} = \frac12\Big[-\frac{1}{1+x}\Big]_0^\infty = \frac12 .
Exercise ●●○ Hardy: Examples LXXIII, 10

Abel’s theorem for integrals. If ϕ\phi is positive and decreasing and ∫a∞ϕ\int_a^\infty\phi converges, show that xϕ(x)→0x\phi(x) \to 0.

Hint
Imitate §179: compare ∫x/2xϕ\int_{x/2}^{x}\phi with 12x ϕ(x)\tfrac12 x\,\phi(x).

Solution

For x>2ax > 2a, the interval [x/2,x][x/2, x] has length x/2x/2, and on it ϕ≥ϕ(x)\phi \ge \phi(x) because ϕ\phi decreases. So 0≤12xϕ(x)≤∫x/2xϕ0 \le \tfrac12 x\phi(x) \le \int_{x/2}^x \phi. By the general principle (remark 3), ∫x/2xϕ→0\int_{x/2}^x\phi \to 0 as x→∞x \to \infty, since both limits pass X(δ)X(\delta). Hence xϕ(x)→0x\phi(x) \to 0. Note that Hardy’s peaks do not contradict this: they are not decreasing.

Check yourself

Check yourself True or false?

If ϕ\phi is positive and continuous and ∫0∞ϕ\int_0^\infty\phi converges, then ϕ(x)→0\phi(x) \to 0 as x→∞x \to \infty.

Why

False: Hardy’s peaks (Fig. 47) have height 11 at every integer, but total area ∑1/(n+1)2\sum 1/(n+1)^2. For series the analogous statement is true; for integrals, a positive function can hide tall values on very short intervals. If ϕ\phi is also decreasing, then ϕ→0\phi \to 0 (and even xϕ(x)→0x\phi(x) \to 0).

Check yourself Choose one

Does ∫1∞x dxx3+1\displaystyle\int_1^\infty \frac{x\,dx}{x^3 + 1} converge?

Why

x2⋅xx3+1→1x^2\cdot\dfrac{x}{x^3+1} \to 1, so by the comparison theorem the integral converges with ∫1∞x−2dx\int_1^\infty x^{-2}dx. We never need an antiderivative to decide convergence (here one exists, but it is messy).

Check yourself Enter a number

What is ∫0∞dx1+x2\displaystyle\int_0^\infty \frac{dx}{1+x^2}? (Give a decimal.)

Why

arctan⁡X→12π≈1.5708\arctan X \to \tfrac12\pi \approx 1.5708. An infinitely long region with a perfectly finite area, as Hardy’s remark (iv) insists.

Check yourself Choose one

How does ∫0∞cos⁡x dx\int_0^\infty \cos x\,dx behave?

Why

∫0Xcos⁡x dx=sin⁡X\int_0^X\cos x\,dx = \sin X, which stays between −1-1 and 11 but has no limit: finite oscillation. Compare ∫0Xxcos⁡x dx=Xsin⁡X+cos⁡X−1\int_0^X x\cos x\,dx = X\sin X + \cos X - 1, which is unbounded in both directions: infinite oscillation (Examples LXXIII, 3–4).

Check yourself True or false?

∫−∞∞x dx\int_{-\infty}^\infty x\,dx converges to 00, since ∫−XXx dx=0\int_{-X}^{X} x\,dx = 0 for every XX.

Why

False. By definition ∫−∞∞\int_{-\infty}^\infty requires both ∫−∞0\int_{-\infty}^0 and ∫0∞\int_0^\infty to converge, and ∫0Xx dx=12X2→∞\int_0^X x\,dx = \tfrac12X^2 \to \infty. The symmetric limit lim⁡∫−XX\lim\int_{-X}^X (today’s “principal value”) is a different and weaker notion.

Hardy’s language

The reader should not be puzzled by the use of the term infinite integral to denote something which has a definite value such as 2 or ½π.

— G. H. Hardy, §184

Summary

Summary
  • ∫a∞ϕ=lim⁡X→∞∫aXϕ\int_a^\infty\phi = \lim_{X\to\infty}\int_a^X\phi when the limit exists; otherwise the integral diverges or oscillates, just as a series does.
  • For positive integrands: converge iff bounded; comparison; ∫∞x−s\int^\infty x^{-s} converges iff s>1s > 1, and xsϕ(x)→l>0x^s\phi(x) \to l > 0 decides by the same rule.
  • Unlike the terms of a series, a positive integrand of a convergent integral need not tend to 00 (Hardy’s peaks).
  • Substitution and integration by parts extend to infinite ranges by letting the upper limit tend to infinity in the finite formulas.
  • ∫−∞∞\int_{-\infty}^\infty needs both halves to converge separately.