Calculus3Integration Techniques5 of 5
Integration Techniques · 5

Improper Integrals

1

When a region goes on forever

Let's find the area under y=1x2y = \frac{1}{x^2} to the right of x=1x = 1. The region never stops, so we write ∞\infty for the upper bound. We can't plug ∞\infty into an antiderivative, because infinity isn't a number. But we can stop at a right edge b>1\edge{b} > 1, where the area is an ordinary integral that we already know how to compute.

11
55
1010
11
xx
yy
b\htmlClass{lc-green}{b}
1x2\htmlClass{lc-blue}{\dfrac{1}{x^2}}
1−1b\htmlClass{lc-blue}{1 - \dfrac{1}{\htmlClass{lc-green}{b}}}
∫1b1x2 dx=[−1x]1b=1−1b\displaystyle \begin{aligned}\int_1^{\edge{b}} \fin{\frac{1}{x^2}}\,dx &= \Big[-\frac1x\Big]_1^{\edge{b}}\\ &= 1 - \frac{1}{\edge{b}}\end{aligned}

Then we watch what that area does as b\edge{b} moves out, and try the same thing on y=1xy = \frac1x.

Find the area under y=1x2y = \frac{1}{x^2} and y=1xy = \frac1x for x≥1x \ge 1
11
55
1010
11
xx
yy
1x2\dfrac{1}{x^2}
Text solution
Let b\edge{b} go to infinitySince 1b\frac{1}{\edge{b}} is positive, the area out to b\edge{b} stays below 11: 0.90.9, then 0.990.99, then 0.9990.999. And as 1b\frac{1}{\edge{b}} goes to 00, the area approaches 11.
∫1∞1x2 dx\displaystyle \int_1^\infty \fin{\frac{1}{x^2}}\,dx =lim⁡b→∞(1−1b)\displaystyle = \lim_{\edge{b}\to\infty}\Big(1 - \frac{1}{\edge{b}}\Big) =1\displaystyle = 1
The same steps for 1x\infa{\frac1x}The antiderivative is ln⁡x\ln x, and ln⁡1=0\ln 1 = 0, so the area out to b\edge{b} is ln⁡b\ln \edge{b}. It grows without bound, so the integral diverges.
∫1∞1x dx\displaystyle \int_1^\infty \infa{\frac1x}\,dx =lim⁡b→∞ln⁡b\displaystyle = \lim_{\edge{b}\to\infty}\ln \edge{b} =∞\displaystyle = \infty

Both curves approach 00, yet one area is 11 and the other is infinite. So the curve approaching 00 doesn't tell us whether the area is finite. We have to take the limit, and that limit is how we define any integral with ∞\infty as a bound.

Integral to infinity
∫a∞f(x) dx=lim⁡b→∞∫abf(x) dx\displaystyle \begin{aligned}\int_a^\infty f(x)\,dx &= \lim_{\edge{b}\to\infty}\int_a^{\edge{b}} f(x)\,dx\end{aligned}
∫a∞f(x) dx=lim⁡b→∞∫abf(x) dx\displaystyle \begin{aligned}&\int_a^\infty f(x)\,dx\\ &\quad = \lim_{\edge{b}\to\infty}\int_a^{\edge{b}} f(x)\,dx\end{aligned}

When the limit is a finite number, we say the integral converges. If the limit is infinite, or if there's no limit at all, we say the integral diverges.

2

Why 1x2\frac{1}{x^2} has finite area and 1x\frac1x doesn't

So why is one area 11 and the other infinite? Let's cut both regions at x=1,2,4,8x = 1, 2, 4, 8, and so on, so that each piece is twice as wide as the one before it, and see what happens to the heights.

11
22
44
88
xx
ln⁡2\htmlClass{lc-pink}{\ln 2}
ln⁡2\htmlClass{lc-pink}{\ln 2}
ln⁡2\htmlClass{lc-pink}{\ln 2}
12\htmlClass{lc-blue}{\dfrac12}
14\htmlClass{lc-blue}{\dfrac14}
18\htmlClass{lc-blue}{\dfrac18}
1x\htmlClass{lc-pink}{\dfrac{1}{x}}
1x2\htmlClass{lc-blue}{\dfrac{1}{x^2}}
Why ∫1∞dxx2\int_1^\infty \frac{dx}{x^2} is finite and ∫1∞dxx\int_1^\infty \frac{dx}{x} is infinite
11
11
xx
yy
1x2\htmlClass{lc-blue}{\dfrac{1}{x^2}}
1x\htmlClass{lc-pink}{\dfrac{1}{x}}
Text solution
For 1x\infa{\frac1x}, every piece is the sameDoubling xx cuts the height in half, while the width doubles. So every piece has area ln⁡2\infa{\ln 2}, and adding ln⁡2\infa{\ln 2} forever gives infinity.
12x\displaystyle \infa{\frac{1}{2x}} =12⋅1x\displaystyle = \frac12 \cdot \infa{\frac1x}
∫121x dx\displaystyle \int_1^2 \infa{\frac1x}\,dx =∫241x dx\displaystyle = \int_2^4 \infa{\frac1x}\,dx =ln⁡2\displaystyle = \infa{\ln 2}
For 1x2\fin{\frac{1}{x^2}}, each piece is half the one beforeDoubling xx divides the height by 44, while the width only doubles. The running totals are 1−1b1 - \frac{1}{\edge{b}}, which approach 11.
1(2x)2\displaystyle \fin{\frac{1}{(2x)^2}} =14⋅1x2\displaystyle = \frac14 \cdot \fin{\frac{1}{x^2}}
12+14+18+⋯\displaystyle \fin{\frac12} + \fin{\frac14} + \fin{\frac18} + \cdots =1\displaystyle = 1
Any power ppEach time xx doubles, the area of a piece is multiplied by 21−p2^{1-p}, which is less than 11 only when p>1p > 1. Integrating out to b\edge{b} gives the same rule.
2⋅12p\displaystyle 2 \cdot \frac{1}{2^p} =21−p\displaystyle = 2^{1-p}
∫1∞dxxp\displaystyle \int_1^\infty \frac{dx}{x^p} =1p−1(p>1)\displaystyle = \frac{1}{p-1} \quad (p > 1)

So the power of xx decides it. The area is finite only when the height falls faster than the pieces widen, and that happens when p>1p > 1.

∫1∞1xp dx:p>1 → convergesp≤1 → diverges\displaystyle \begin{aligned}\int_1^\infty \frac{1}{x^p}\,dx:\quad & p > 1 \ \to\ \text{converges}\\ & p \le 1 \ \to\ \text{diverges}\end{aligned}
3

An infinitely tall region

A region can also be infinitely tall. As xx approaches 00 from the right, 1x\fin{\frac{1}{\sqrt x}} goes to infinity, so the region under it from 00 to 11 has a vertical asymptote at x=0x = 0. Just as we set a right edge at b\edge{b} when a region stretched infinitely far to the right, we set a left edge at a\edge{a}, between 00 and 11, and let it approach 00. Let's do it for 1x\fin{\frac{1}{\sqrt x}} and for 1x2\infa{\frac{1}{x^2}}.

∫a1x−12 dx=2−2a\displaystyle \begin{aligned}\int_{\edge{a}}^1 \fin{x^{-\frac12}}\,dx &= 2 - 2\sqrt{\edge{a}}\end{aligned}
∫a1x−2 dx=1a−1\displaystyle \begin{aligned}\int_{\edge{a}}^1 \infa{x^{-2}}\,dx &= \frac{1}{\edge{a}} - 1\end{aligned}
11
xx
yy
a\htmlClass{lc-green}{a}
1x\htmlClass{lc-blue}{\dfrac{1}{\sqrt{x}}}
2−2a\htmlClass{lc-blue}{2 - 2\sqrt{\htmlClass{lc-green}{a}}}
a    2−2a    1a−114130.011.8990.00011.989999\displaystyle \begin{array}{ccc}\edge{a} & \;\; 2 - 2\sqrt{\edge{a}} \;\; & \dfrac{1}{\edge{a}} - 1\\[10pt] \dfrac14 & 1 & 3\\[8pt] 0.01 & 1.8 & 99\\[4pt] 0.0001 & 1.98 & 9999\end{array}
Find the area under y=1xy = \frac{1}{\sqrt x} from x=0x = 0 to x=1x = 1, and compare it with 1x2\frac{1}{x^2}
11
22
11
22
33
44
xx
1x\dfrac{1}{\sqrt{x}}
Text solution
Let a\edge{a} approach 00As a→0+\edge{a} \to 0^+, 2a2\sqrt{\edge{a}} goes to 00, so this infinitely tall region has area 22. For 1x2\infa{\frac{1}{x^2}}, 1a\frac{1}{\edge{a}} goes to infinity.
∫011x dx\displaystyle \int_0^1 \fin{\frac{1}{\sqrt x}}\,dx =lim⁡a→0+(2−2a)\displaystyle = \lim_{\edge{a}\to0^+}\big(2 - 2\sqrt{\edge{a}}\big) =2\displaystyle = 2
∫011x2 dx\displaystyle \int_0^1 \infa{\frac{1}{x^2}}\,dx =lim⁡a→0+(1a−1)\displaystyle = \lim_{\edge{a}\to0^+}\Big(\frac{1}{\edge{a}} - 1\Big) =∞\displaystyle = \infty
Cut toward 00Each piece is half as wide as the one to its right, and halving xx multiplies the height of 1xp\frac{1}{x^p} by 2p2^p. For 1x\fin{\frac{1}{\sqrt x}}, each piece is about 0.710.71 times the one to its right. For 1x2\infa{\frac{1}{x^2}}, it's twice as big.
12⋅212≈0.71\displaystyle \frac12 \cdot 2^{\frac12} \approx 0.71
12⋅22\displaystyle \frac12 \cdot 2^2 =2\displaystyle = 2
Any power ppAs a→0+\edge{a} \to 0^+, a1−p\edge{a}^{1-p} goes to 00 only when p<1p < 1. At p=1p = 1, −ln⁡a-\ln \edge{a} goes to infinity.
∫01dxxp\displaystyle \int_0^1 \frac{dx}{x^p} =11−p(p<1)\displaystyle = \frac{1}{1-p} \quad (p < 1)

Earlier, on [1,∞)[1, \infty), the integral of 1x2\fin{\frac{1}{x^2}} converged and the integral of 1x\infa{\frac{1}{\sqrt x}} diverged. Between 00 and 11, it's the other way round. Cutting toward 00, the pieces under 1x\fin{\frac{1}{\sqrt x}} get smaller, and their totals approach 22.

11
12\dfrac12
14\dfrac14
18\dfrac18
xx
yy
0.59\htmlClass{lc-blue}{0.59}
0.41\htmlClass{lc-blue}{0.41}
0.29\htmlClass{lc-blue}{0.29}
1x\htmlClass{lc-blue}{\dfrac{1}{\sqrt{x}}}
0.59+0.41=1=2−2141+0.29=1.29≈2−218\displaystyle \begin{aligned}\fin{0.59} + \fin{0.41} &= 1 = 2 - 2\sqrt{\tfrac14}\\[6pt] 1 + \fin{0.29} &= 1.29 \approx 2 - 2\sqrt{\tfrac18}\end{aligned}

Every curve 1xp\frac{1}{x^p} passes through (1,1)(1, 1). To the right of 11, a larger power pp makes the curve lower. Between 00 and 11, a larger pp makes it taller.

11
22
12\dfrac12
11
xx
2\htmlClass{lc-blue}{2}
∞\htmlClass{lc-pink}{\infty}
∞\htmlClass{lc-pink}{\infty}
∞\htmlClass{lc-pink}{\infty}
∞\htmlClass{lc-pink}{\infty}
1\htmlClass{lc-blue}{1}
44
22
≈1.41\approx 1.41
≈0.71\approx 0.71
0.50.5
0.250.25
1x\htmlClass{lc-blue}{\dfrac{1}{\sqrt{x}}}
1x\htmlClass{lc-pink}{\dfrac{1}{x}}
1x2\htmlClass{lc-pink}{\dfrac{1}{x^2}}
1x\htmlClass{lc-pink}{\dfrac{1}{\sqrt{x}}}
1x2\htmlClass{lc-blue}{\dfrac{1}{x^2}}

So the tail converges when p>1p > 1, the part from 00 to 11 converges when p<1p < 1, and at p=1p = 1, both diverge.

The tail
∫1∞1xp dx\displaystyle \int_1^\infty \frac{1}{x^p}\,dx
p>1 → convergesp≤1 → diverges\displaystyle \begin{aligned}p > 1 &\ \to\ \text{converges}\\ p \le 1 &\ \to\ \text{diverges}\end{aligned}
∫1∞dxx2\displaystyle \int_1^\infty \fin{\frac{dx}{x^2}} =1\displaystyle = 1
The part from 0 to 1
∫011xp dx\displaystyle \int_0^1 \frac{1}{x^p}\,dx
p<1 → convergesp≥1 → diverges\displaystyle \begin{aligned}p < 1 &\ \to\ \text{converges}\\ p \ge 1 &\ \to\ \text{diverges}\end{aligned}
∫01dxx\displaystyle \int_0^1 \fin{\frac{dx}{\sqrt x}} =2\displaystyle = 2

Do we use the same rule near zero and at infinity?

Compare the two areas
4

A vertical asymptote inside the interval

Let's find the integral of 1x2\frac{1}{x^2} from −1-1 to 11. Both bounds are ordinary numbers, with no infinity in sight, so you might try the Fundamental Theorem of Calculus straight away.

∫−111x2 dx=[−1x]−11=−1−1=−2\displaystyle \begin{aligned}\int_{-1}^1 \frac{1}{x^2}\,dx &= \Big[-\frac1x\Big]_{-1}^1\\ &= -1 - 1 = -2\end{aligned}
−1-1
11
xx
yy
1x2\htmlClass{lc-pink}{\dfrac{1}{x^2}}

Wait. Negative 22? For every x≠0x \ne 0, 1x2\infa{\frac{1}{x^2}} is positive, so the area can't be negative. The trouble is at x=0x = 0, right between the bounds.

Evaluate ∫−111x2 dx\int_{-1}^1 \frac{1}{x^2}\,dx and ∫03dx(x−1)23\int_0^3 \frac{dx}{(x-1)^{\frac23}}, or show that they diverge
∫−111x2 dx=  ?∫03dx(x−1)23=  ?\displaystyle \begin{aligned}\int_{-1}^{1} \frac{1}{x^2}\,dx &= \;?\\[4pt] \int_0^3 \frac{dx}{(x-1)^{\frac23}} &= \;?\end{aligned}
Text solution
Split at the asymptoteF(b)−F(a)F(b) - F(a) works only when the function is continuous on the whole interval, and 1x2\infa{\frac{1}{x^2}} has a vertical asymptote at x=0x = 0. The piece from 00 to 11 has p=2p = 2, so it diverges, and so does the whole integral.
∫−111x2 dx\displaystyle \int_{-1}^1 \frac{1}{x^2}\,dx =∫−101x2 dx+∫011x2 dx\displaystyle = \int_{-1}^0 \frac{1}{x^2}\,dx + \int_0^1 \infa{\frac{1}{x^2}}\,dx
Split ∫03\int_0^3 at x=1x = 1The denominator of 1(x−1)23\frac{1}{(x-1)^{\frac23}} is 00 at x=1x = 1. With the antiderivative 3(x−1)133(x-1)^{\frac13}, each piece takes its limit from inside, and both are finite.
∫01dx(x−1)23\displaystyle \int_0^1 \fin{\frac{dx}{(x-1)^{\frac23}}} =3\displaystyle = 3
∫13dx(x−1)23\displaystyle \int_1^3 \fin{\frac{dx}{(x-1)^{\frac23}}} =323\displaystyle = 3\sqrt[3]{2}
Add the piecesBoth pieces converge because the power 23\frac23 is less than 11.
∫03dx(x−1)23\displaystyle \int_0^3 \fin{\frac{dx}{(x-1)^{\frac23}}} =3+323\displaystyle = 3 + 3\sqrt[3]{2}
5

When both ends are infinite

Now both bounds are infinite. Each infinite end needs its own limit, so just as with an asymptote inside the interval, we split the integral into two pieces, with one infinite end on each. Any split point works, but 00 is the easiest choice.

−5-5
55
xx
11+x2\htmlClass{lc-blue}{\dfrac{1}{1 + x^2}}
π2\dfrac{\pi}{2}
−π2-\dfrac{\pi}{2}
arctan⁡x\arctan x
Find ∫−∞∞11+x2 dx\int_{-\infty}^\infty \frac{1}{1+x^2}\,dx and check whether ∫−∞∞x dx\int_{-\infty}^\infty x\,dx converges
00
xx
11+x2\dfrac{1}{1 + x^2}
Text solution
Split at 00The antiderivative is arctan⁡x\arctan x, which approaches π2\frac\pi2 as x→∞x \to \infty and −π2-\frac\pi2 as x→−∞x \to -\infty. So each half converges to π2\frac\pi2.
∫−∞0dx1+x2\displaystyle \int_{-\infty}^0 \fin{\frac{dx}{1+x^2}} =π2\displaystyle = \frac\pi2
∫0∞dx1+x2\displaystyle \int_0^\infty \fin{\frac{dx}{1+x^2}} =π2\displaystyle = \frac\pi2
∫−∞∞dx1+x2\displaystyle \int_{-\infty}^{\infty} \fin{\frac{dx}{1+x^2}} =π2+π2\displaystyle = \frac\pi2 + \frac\pi2 =π\displaystyle = \pi
Split ∫x dx\int x\,dx at 00 tooThe right half diverges on its own, so ∫−∞∞x dx\int_{-\infty}^{\infty} x\,dx diverges.
∫0∞x dx\displaystyle \int_0^\infty \addp{x}\,dx =lim⁡b→∞b22\displaystyle = \lim_{\edge{b}\to\infty}\frac{\edge{b}^2}{2} =∞\displaystyle = \infty

Why not take one limit, from −b-\edge{b} to b\edge{b}? For y=xy = x, the areas below and above the x-axis cancel for every b\edge{b}. But if the right end moves out twice as fast, the integral goes to infinity instead. The answer depends on how the ends go out, so there's no single value. When both halves converge on their own, as they do for 11+x2\fin{\frac{1}{1+x^2}}, it doesn't matter.

−b\htmlClass{lc-green}{-b}
b\htmlClass{lc-green}{b}
y=xy = x
−b\htmlClass{lc-green}{-b}
2b\htmlClass{lc-green}{2b}
y=xy = x
∫−bbx dx=b2−b22=0\displaystyle \begin{aligned}\int_{-\edge{b}}^{\edge{b}} x\,dx &= \frac{\edge{b}^2 - \edge{b}^2}{2} = 0\end{aligned}
∫−b2bx dx=4b2−b22=3b22\displaystyle \begin{aligned}\int_{-\edge{b}}^{2\edge{b}} x\,dx &= \frac{4\edge{b}^2 - \edge{b}^2}{2} = \frac{3\edge{b}^2}{2}\end{aligned}

If the positive and negative areas cancel, does the integral converge?

Move the two bounds independently
6

Limits after integrating

In these three integrals, integrating up to b\edge{b} is routine. The hard part is the limit at the end: one gives ∞⋅0\infty \cdot 0, one gives ∞−∞\infty - \infty, and one keeps oscillating.

Evaluate ∫0∞xe−x dx\int_0^\infty x e^{-x}\,dx, ∫2∞dxx2−1\int_2^\infty \frac{dx}{x^2-1}, and ∫0∞cos⁡x dx\int_0^\infty \cos x\,dx, or show that they diverge
∫0∞xe−x dx=  ?∫2∞1x2−1 dx=  ?∫0∞cos⁡x dx=  ?\displaystyle \begin{aligned}\int_0^\infty x e^{-x}\,dx &= \;?\\[4pt] \int_2^\infty \frac{1}{x^2-1}\,dx &= \;?\\[4pt] \int_0^\infty \cos x\,dx &= \;?\end{aligned}
Text solution
∞⋅0\infty \cdot 0: write it as a fractionBy parts, the area out to b\edge{b} is 1−be−b−e−b1 - \edge{b}e^{-\edge{b}} - e^{-\edge{b}}. The product be−b\edge{b}e^{-\edge{b}} gives ∞⋅0\infty \cdot 0, so we write it as beb\frac{\edge{b}}{e^{\edge{b}}} and use L'Hôpital's rule.
lim⁡b→∞beb\displaystyle \lim_{\edge{b}\to\infty}\frac{\edge{b}}{e^{\edge{b}}} =lim⁡b→∞1eb\displaystyle = \lim_{\edge{b}\to\infty}\frac{1}{e^{\edge{b}}} =0\displaystyle = 0
∫0∞xe−x dx\displaystyle \int_0^\infty xe^{-x}\,dx =1\displaystyle = 1
∞−∞\infty - \infty: combine the logsPartial fractions gives two logarithms that both go to infinity. With ln⁡A−ln⁡B=ln⁡AB\ln A - \ln B = \ln\frac{A}{B}, they become one logarithm that goes to 12ln⁡1=0\frac12\ln 1 = 0, and the 12ln⁡3\frac12\ln 3 from x=2x = 2 is left.
12ln⁡b−1b+1→0\displaystyle \frac12\ln\frac{\addp{\edge{b}-1}}{\subp{\edge{b}+1}} \to 0
∫2∞dxx2−1\displaystyle \int_2^\infty \frac{dx}{x^2-1} =12ln⁡3\displaystyle = \frac12\ln 3
No limit at allThe integral of cos⁡x\cos x out to b\edge{b} is sin⁡b\sin \edge{b}, which keeps oscillating between −1-1 and 11. The limit doesn't exist, so the integral diverges, even though nothing goes to infinity.
∫0∞cos⁡x dx\displaystyle \int_0^\infty \cos x\,dx =lim⁡b→∞sin⁡b\displaystyle = \lim_{\edge{b}\to\infty}\sin \edge{b}
7

Comparing with known integrals

None of our integration methods give an antiderivative for e−x2e^{-x^2}. But to know whether the integral converges, we only need to know whether the area is finite, and we can find that out by comparing it with a curve we can integrate.

11
22
33
44
xx
yy
e−x\htmlClass{lc-blue}{e^{-x}}
e−x2e^{-x^2}
1e\htmlClass{lc-blue}{\dfrac{1}{e}}
x≥1:x2≥xe−x2≤e−x\displaystyle \begin{gathered}x \ge 1:\quad x^2 \ge x\\ e^{-x^2} \le \fin{e^{-x}}\end{gathered}
∫1∞e−x dx=lim⁡b→∞(e−1−e−b)=1e\displaystyle \begin{aligned}\int_1^\infty \fin{e^{-x}}\,dx &= \lim_{\edge{b}\to\infty}\big(e^{-1} - e^{-\edge{b}}\big)\\ &= \frac1e\end{aligned}

Sometimes two comparisons are possible, and only one of them proves anything.

11
22
33
44
55
66
11
xx
yy
∞\htmlClass{lc-pink}{\infty}
∞\htmlClass{lc-pink}{\infty}
1\htmlClass{lc-blue}{1}
1x\htmlClass{lc-pink}{\dfrac{1}{\sqrt{x}}}
1x\htmlClass{lc-pink}{\dfrac{1}{x}}
1x2\htmlClass{lc-blue}{\dfrac{1}{x^2}}
1x2+x\dfrac{1}{x^2 + \sqrt{x}}
1x2+x<1x2∫1∞dxx2=1\displaystyle \begin{gathered}\frac{1}{x^2 + \sqrt x} < \fin{\frac{1}{x^2}}\\[6pt] \int_1^\infty \fin{\frac{dx}{x^2}} = 1\end{gathered}
1x2+x<1x∫1∞dxx=∞\displaystyle \begin{gathered}\frac{1}{x^2 + \sqrt x} < \infa{\frac{1}{\sqrt x}}\\[6pt] \int_1^\infty \infa{\frac{dx}{\sqrt x}} = \infty\end{gathered}
Decide whether ∫1∞e−x2 dx\int_1^\infty e^{-x^2}\,dx, ∫1∞dxx2+x\int_1^\infty \frac{dx}{x^2 + \sqrt x}, and ∫1∞2+cos⁡xx dx\int_1^\infty \frac{2 + \cos x}{x}\,dx converge
∫1∞e−x2 dx=  ?∫1∞dxx2+x=  ?∫1∞2+cos⁡xx dx=  ?\displaystyle \begin{aligned}\int_1^\infty e^{-x^2}\,dx &= \;?\\[4pt] \int_1^\infty \frac{dx}{x^2+\sqrt{x}} &= \;?\\[4pt] \int_1^\infty \frac{2+\cos x}{x}\,dx &= \;?\end{aligned}
Text solution
Below a curve with finite areaFor x≥1x \ge 1, x2≥xx^2 \ge x, so e−x2≤e−xe^{-x^2} \le \fin{e^{-x}}. The area under e−x\fin{e^{-x}} is 1e\frac1e, so the integral converges, even though we don't know its exact value.
∫1∞e−x2 dx\displaystyle \int_1^\infty e^{-x^2}\,dx ≤∫1∞e−x dx\displaystyle \le \int_1^\infty \fin{e^{-x}}\,dx =1e\displaystyle = \frac1e
Choose the comparison from the largest termAt x=100x = 100, x2+x=10,010x^2 + \sqrt x = 10{,}010, so the denominator behaves like x2x^2 and we expect convergence. So we compare with 1x2\fin{\frac{1}{x^2}}. Being below 1x\infa{\frac{1}{\sqrt x}}, which has infinite area, proves nothing.
1x2+x<1x2\displaystyle \frac{1}{x^2 + \sqrt x} < \fin{\frac{1}{x^2}}
∫1∞dxx2\displaystyle \int_1^\infty \fin{\frac{dx}{x^2}} =1\displaystyle = 1
Above a curve with infinite areaSince cos⁡x≥−1\cos x \ge -1, the numerator 2+cos⁡x2 + \cos x is at least 11. So the curve stays above 1x\infa{\frac1x}, and the integral diverges.
2+cos⁡xx\displaystyle \frac{2 + \cos x}{x} ≥1x\displaystyle \ge \infa{\frac1x}
∫1∞dxx\displaystyle \int_1^\infty \infa{\frac{dx}{x}} =∞\displaystyle = \infty

A comparison only proves convergence or divergence when the inequality goes the right way.

What works
below a curve with finite area → convergesabove a curve with infinite area → diverges\displaystyle \begin{aligned}\text{below a curve with finite area} &\ \to\ \text{converges}\\ \text{above a curve with infinite area} &\ \to\ \text{diverges}\end{aligned}
below a curve with finite area→ convergesabove a curve with infinite area→ diverges\displaystyle \begin{gathered}\text{below a curve with finite area}\\ \to\ \text{converges}\\[8pt] \text{above a curve with infinite area}\\ \to\ \text{diverges}\end{gathered}
What doesn't work
below a curve with infinite area,or above a curve with finite area→ no conclusion\displaystyle \begin{gathered}\text{below a curve with infinite area,}\\ \text{or above a curve with finite area}\\ \to\ \text{no conclusion}\end{gathered}

Both comparisons are true. Which one tells us whether the integral converges?

Check the direction

Summary

What makes an improper integral converge
An infinite bound
∫1∞1x2 dx=lim⁡b→∞∫1b1x2 dx=lim⁡b→∞(1−1b)=1\displaystyle \begin{aligned}\int_1^\infty \fin{\frac{1}{x^2}}\,dx = \lim_{\edge{b}\to\infty}\int_1^{\edge{b}} \fin{\frac{1}{x^2}}\,dx = \lim_{\edge{b}\to\infty}\Big(1 - \frac{1}{\edge{b}}\Big) = 1\end{aligned}
∫1∞1x2 dx=lim⁡b→∞∫1b1x2 dx=lim⁡b→∞(1−1b)=1\displaystyle \begin{aligned}&\int_1^\infty \fin{\frac{1}{x^2}}\,dx\\ &\quad = \lim_{\edge{b}\to\infty}\int_1^{\edge{b}} \fin{\frac{1}{x^2}}\,dx\\ &\quad = \lim_{\edge{b}\to\infty}\Big(1 - \frac{1}{\edge{b}}\Big) = 1\end{aligned}
A vertical asymptote
∫011x dx=lim⁡a→0+(2−2a)=2\displaystyle \begin{aligned}\int_0^1 \fin{\frac{1}{\sqrt x}}\,dx = \lim_{\edge{a}\to0^+}\big(2 - 2\sqrt{\edge{a}}\big) = 2\end{aligned}
∫011x dx=lim⁡a→0+(2−2a)=2\displaystyle \begin{aligned}\int_0^1 \fin{\frac{1}{\sqrt x}}\,dx &= \lim_{\edge{a}\to0^+}\big(2 - 2\sqrt{\edge{a}}\big)\\ &= 2\end{aligned}

An improper integral is a limit of ordinary definite integrals. Replace an infinite upper bound with b\edge{b}, integrate up to b\edge{b}, and take the limit as b\edge{b} goes to infinity. When a curve has a vertical asymptote at an endpoint, replace that endpoint with a variable and approach it from inside the interval.

If the limit is a finite number, the integral converges to that number. If the limit is infinite or doesn't exist, the integral diverges.

Powers of xx at infinity and near zero
11
11
xx
yy
1x\htmlClass{lc-pink}{\dfrac{1}{\sqrt{x}}}
1x2\htmlClass{lc-pink}{\dfrac{1}{x^2}}
The tail from 1 to infinity
∫1∞1xp dxp>1 → converges\displaystyle \begin{gathered}\int_1^\infty \frac{1}{x^p}\,dx\\ p > 1 \ \to\ \text{converges}\end{gathered}
The part from 0 to 1
∫011xp dxp<1 → converges\displaystyle \begin{gathered}\int_0^1 \frac{1}{x^p}\,dx\\ p < 1 \ \to\ \text{converges}\end{gathered}

Every curve y=1xpy = \frac{1}{x^p} passes through (1,1)(1, 1). To the right of 11, a larger power pp makes 1xp\frac{1}{x^p} smaller, so the tail approaches 00 fast enough to have finite area when p>1p > 1. Between 00 and 11, a larger power makes the curve taller, but the area from 00 to 11 is finite when p<1p < 1, even if the curve goes to infinity near zero. At p=1p = 1, both areas are infinite.

A vertical asymptote inside the bounds
∫−111x2 dx=∫−101x2 dx+∫011x2 dx,∫011x2 dx=∞\displaystyle \begin{aligned}\int_{-1}^1 \frac{1}{x^2}\,dx = \int_{-1}^0 \frac{1}{x^2}\,dx + \int_0^1 \infa{\frac{1}{x^2}}\,dx, \qquad \int_0^1 \infa{\frac{1}{x^2}}\,dx = \infty\end{aligned}
∫−111x2 dx=∫−101x2 dx+∫011x2 dx∫011x2 dx=∞\displaystyle \begin{aligned}\int_{-1}^1 \frac{1}{x^2}\,dx &= \int_{-1}^0 \frac{1}{x^2}\,dx\\ &\quad + \int_0^1 \infa{\frac{1}{x^2}}\,dx\\[12pt] \int_0^1 \infa{\frac{1}{x^2}}\,dx &= \infty\end{aligned}

Check the interval before using an antiderivative. The Fundamental Theorem of Calculus requires the function to be continuous on the entire interval [a,b][a, b]. If there is a vertical asymptote between the bounds, split the integral at that point and evaluate each piece with its own limit. If any piece diverges, the whole integral diverges.

Two infinite ends
∫−bbx dx=0,∫0∞x dx=∞\displaystyle \begin{aligned}\int_{-\edge{b}}^{\edge{b}} x\,dx = 0, \qquad \int_0^\infty \addp{x}\,dx = \infty\end{aligned}
∫−bbx dx=0∫0∞x dx=∞\displaystyle \begin{gathered}\int_{-\edge{b}}^{\edge{b}} x\,dx = 0\\[6pt] \int_0^\infty \addp{x}\,dx = \infty\end{gathered}

When both bounds are infinite, split the integral at any middle point into two pieces, each with one infinite bound. Don't just integrate from −b-\edge{b} to b\edge{b} and let b\edge{b} go to infinity. For f(x)=xf(x) = x, that symmetric limit gives 00, but the right-hand piece ∫0∞x dx\int_0^\infty \addp{x}\,dx is infinite, so ∫−∞∞x dx\int_{-\infty}^\infty x\,dx diverges. Because we can move the two bounds outward at different rates, both pieces must converge on their own.

Finishing the limit
0 times infinity
lim⁡b→∞be−b=lim⁡b→∞beb=lim⁡b→∞1eb=0\displaystyle \begin{aligned}\lim_{\edge{b}\to\infty} \edge{b}e^{-\edge{b}} = \lim_{\edge{b}\to\infty}\frac{\edge{b}}{e^{\edge{b}}} = \lim_{\edge{b}\to\infty}\frac{1}{e^{\edge{b}}} = 0\end{aligned}
lim⁡b→∞be−b=lim⁡b→∞beb=lim⁡b→∞1eb=0\displaystyle \begin{aligned}\lim_{\edge{b}\to\infty} \edge{b}e^{-\edge{b}} &= \lim_{\edge{b}\to\infty}\frac{\edge{b}}{e^{\edge{b}}}\\ &= \lim_{\edge{b}\to\infty}\frac{1}{e^{\edge{b}}} = 0\end{aligned}
Infinity minus infinity
12(ln⁡(b−1)−ln⁡(b+1))=12ln⁡b−1b+1→0\displaystyle \begin{aligned}\frac12\big(\addp{\ln(\edge{b}-1)} - \subp{\ln(\edge{b}+1)}\big) = \frac12\ln\frac{\addp{\edge{b}-1}}{\subp{\edge{b}+1}} \to 0\end{aligned}
12(ln⁡(b−1)−ln⁡(b+1))=12ln⁡b−1b+1→0\displaystyle \begin{aligned}&\frac12\big(\addp{\ln(\edge{b}-1)} - \subp{\ln(\edge{b}+1)}\big)\\ &\quad = \frac12\ln\frac{\addp{\edge{b}-1}}{\subp{\edge{b}+1}} \to 0\end{aligned}

If you get ∞⋅0\infty \cdot 0, rewrite the product as a fraction like beb\frac{\edge{b}}{e^{\edge{b}}} and use L'Hôpital's rule. If you get ∞−∞\infty - \infty between logarithms, combine them with ln⁡A−ln⁡B=ln⁡AB\ln A - \ln B = \ln\frac{A}{B} before taking the limit. If an expression like sin⁡b\sin \edge{b} oscillates without approaching any value, the limit doesn't exist and the integral diverges.

Comparing with known integrals
Below a finite area
e−x2≤e−x,∫1∞e−x dx=1e\displaystyle \begin{aligned}e^{-x^2} \le \fin{e^{-x}}, \qquad \int_1^\infty \fin{e^{-x}}\,dx = \frac1e\end{aligned}
e−x2≤e−x∫1∞e−x dx=1e\displaystyle \begin{gathered}e^{-x^2} \le \fin{e^{-x}}\\[6pt] \int_1^\infty \fin{e^{-x}}\,dx = \frac1e\end{gathered}
Above an infinite area
2+cos⁡xx≥1x,∫1∞dxx=∞\displaystyle \begin{aligned}\frac{2 + \cos x}{x} \ge \infa{\frac1x}, \qquad \int_1^\infty \infa{\frac{dx}{x}} = \infty\end{aligned}
2+cos⁡xx≥1x∫1∞dxx=∞\displaystyle \begin{gathered}\frac{2 + \cos x}{x} \ge \infa{\frac1x}\\[6pt] \int_1^\infty \infa{\frac{dx}{x}} = \infty\end{gathered}

For positive functions, compare your function with a positive function whose integral you know. If your curve lies below a curve with finite area, its area is finite, so its integral converges. If your curve lies above a curve with infinite area, its area is infinite, so its integral diverges.

Being below a curve with infinite area or above a curve with finite area doesn't tell us whether our integral converges. To choose the right curve, look at which term dominates when xx is large to guess whether the integral should converge or diverge.

8

Practice

Enter exact values, or "diverges" (or "infinite" for the volume problem). For the two comparison problems, enter "converges" or "diverges".

1
Evaluate ∫1∞1x3 dx\displaystyle\int_1^\infty \frac{1}{x^3}\,dx, or show that it diverges.

Replace the infinite bound with b\edge{b}. What is ∫1bx−3 dx\int_1^{\edge{b}} x^{-3}\,dx?

The upper bound is infinite, so replace ∞\infty with b\edge{b} and take the limit as b→∞\edge{b} \to \infty.
∫1∞dxx3\displaystyle \int_1^\infty \frac{dx}{x^3} =lim⁡b→∞∫1bx−3 dx\displaystyle = \lim_{\edge{b}\to\infty}\int_1^{\edge{b}} x^{-3}\,dx
Use the power rule to integrate x−3x^{-3}, and keep the minus sign.
∫1bx−3 dx\displaystyle \int_1^{\edge{b}} x^{-3}\,dx =[−12x2]1b\displaystyle = \Big[-\frac{1}{2x^2}\Big]_1^{\edge{b}} =12−12b2\displaystyle = \frac12 - \frac{1}{2\edge{b}^2}
As b\edge{b} goes to infinity, 12b2\frac{1}{2\edge{b}^2} goes to 00, so the integral converges to 12\frac12.
lim⁡b→∞(12−12b2)\displaystyle \lim_{\edge{b}\to\infty}\Big(\frac12 - \frac{1}{2\edge{b}^2}\Big) =12\displaystyle = \frac12
2
Evaluate ∫04dx4−x\displaystyle\int_0^4 \frac{dx}{\sqrt{4-x}}, or show that it diverges.

Where is 14−x\frac{1}{\sqrt{4-x}} undefined on [0,4][0, 4]? Approach that bound from inside the interval.

Set up the limitThe denominator is 00 at x=4x = 4, so we approach 44 from the left to stay inside [0,4][0, 4].
4−x\displaystyle \sqrt{4-x} =0 at x\displaystyle = 0 \ \text{at}\ x =4:∫04dx4−x\displaystyle = 4:\qquad \int_0^4 \frac{dx}{\sqrt{4-x}} =lim⁡b→4−∫0bdx4−x\displaystyle = \lim_{\edge{b}\to4^-}\int_0^{\edge{b}} \frac{dx}{\sqrt{4-x}}
The derivative of 4−x4 - x is −1-1, so the chain rule gives an extra minus sign in the antiderivative.
∫0b(4−x)−12 dx\displaystyle \int_0^{\edge{b}} (4-x)^{-\frac12}\,dx =[−24−x]0b\displaystyle = \Big[-2\sqrt{4-x}\Big]_0^{\edge{b}} =4−24−b\displaystyle = 4 - 2\sqrt{4-\edge{b}}
As b\edge{b} approaches 44 from the left, 4−b\sqrt{4-\edge{b}} goes to 00, so the limit is 44.
lim⁡b→4−(4−24−b)\displaystyle \lim_{\edge{b}\to4^-}\big(4 - 2\sqrt{4-\edge{b}}\big) =4\displaystyle = 4
3
Evaluate ∫0∞1x dx\displaystyle\int_0^\infty \frac{1}{\sqrt x}\,dx, or show that it diverges.

What are the two bad places in ∫0∞1x dx\int_0^\infty \frac{1}{\sqrt x}\,dx?

Split the integralThere's a vertical asymptote at x=0x = 0 and an infinite upper bound, so we split at 11 to give each piece one bad place.
∫0∞dxx\displaystyle \int_0^\infty \frac{dx}{\sqrt x} =∫01dxx+∫1∞dxx\displaystyle = \int_0^1 \fin{\frac{dx}{\sqrt x}} + \int_1^\infty \infa{\frac{dx}{\sqrt x}}
For the piece on [0,1][0, 1], the power of xx is p=12<1p = \frac12 < 1, so this piece converges to 22.
∫01dxx\displaystyle \int_0^1 \fin{\frac{dx}{\sqrt x}} =lim⁡a→0+(2−2a)\displaystyle = \lim_{\edge{a}\to0^+}\big(2 - 2\sqrt{\edge{a}}\big) =2\displaystyle = 2
For the tail on [1,∞)[1, \infty), the power of xx is p=12≤1p = \frac12 \le 1, so this piece diverges to infinity. Because the tail diverges, the entire integral diverges.
∫1∞dxx\displaystyle \int_1^\infty \infa{\frac{dx}{\sqrt x}} =lim⁡b→∞(2b−2)\displaystyle = \lim_{\edge{b}\to\infty}\big(2\sqrt{\edge{b}} - 2\big) =∞\displaystyle = \infty
4
For which values of pp does ∫011x2p dx\displaystyle\int_0^1 \frac{1}{x^{2p}}\,dx converge?

Which rule applies on [0,1][0, 1]? What is the power of xx?

The case p>0p > 0When p>0p > 0, the denominator is 00 at x=0x = 0, so the integrand goes to infinity there, with power 2p2p. It converges only when 2p<12p < 1.
p>0:∫01dxx2p:2p<1 → converges,2p\displaystyle p > 0:\qquad \int_0^1 \frac{dx}{x^{2p}}:\quad 2p < 1 \ \to\ \text{converges}, \qquad 2p ≥1 → diverges\displaystyle \ge 1 \ \to\ \text{diverges}
When p≤0p \le 0, the exponent −2p-2p is greater than or equal to 00, so the function is continuous on [0,1][0, 1] and the integral converges.
p\displaystyle p ≤0:1x2p\displaystyle \le 0:\qquad \frac{1}{x^{2p}} =x−2p,−2p\displaystyle = x^{-2p}, \quad -2p ≥0\displaystyle \ge 0
The condition 2p<12p < 1 gives p<12p < \frac12, which includes all values where p≤0p \le 0.
p<12\displaystyle p < \frac12
5
Evaluate ∫0πsec⁡2x dx\displaystyle\int_0^\pi \sec^2 x\,dx, or show that it diverges.

Where is cos⁡x=0\cos x = 0 on the interval [0,π][0, \pi]?

Find the vertical asymptoteBecause cos⁡π2=0\cos\frac\pi2 = 0, the function sec⁡2x\sec^2 x has a vertical asymptote at x=π2x = \frac\pi2, right inside [0,π][0, \pi].
sec⁡2x\displaystyle \sec^2 x =1cos⁡2x,cos⁡π2\displaystyle = \frac{1}{\cos^2 x}, \qquad \cos\frac\pi2 =0\displaystyle = 0
Split the integral at the asymptote so each piece has only one bad place.
∫0πsec⁡2x dx\displaystyle \int_0^\pi \sec^2 x\,dx =∫0π2sec⁡2x dx+∫π2πsec⁡2x dx\displaystyle = \int_0^{\frac\pi2} \sec^2 x\,dx + \int_{\frac\pi2}^{\pi} \sec^2 x\,dx
As b\edge{b} approaches π2\frac\pi2 from the left, tan⁡b\tan \edge{b} goes to infinity. Since this piece diverges, the whole integral diverges.
∫0π2sec⁡2x dx\displaystyle \int_0^{\frac\pi2} \infa{\sec^2 x}\,dx =lim⁡b→π2−(tan⁡b−tan⁡0)\displaystyle = \lim_{\edge{b}\to\frac\pi2^-}\big(\tan \edge{b} - \tan 0\big) =∞\displaystyle = \infty
6
Evaluate ∫−∞∞xe−x2 dx\displaystyle\int_{-\infty}^{\infty} x e^{-x^2}\,dx, or show that it diverges.

Split the integral at 00. What is ∫0∞xe−x2 dx\int_0^\infty x e^{-x^2}\,dx?

Let u=x2u = x^2, so du=2x dxdu = 2x\,dx, and integrate.
u\displaystyle u =x2, du\displaystyle = x^2,\ du =2x dx:∫xe−x2 dx\displaystyle = 2x\,dx:\qquad \int xe^{-x^2}\,dx =−12e−x2+C\displaystyle = -\frac12 e^{-x^2} + C
For the right half, e−b2e^{-\edge{b}^2} approaches 00 as b→∞\edge{b} \to \infty, so the limit is 12\addp{\frac12}.
∫0∞xe−x2 dx\displaystyle \int_0^\infty xe^{-x^2}\,dx =lim⁡b→∞(12−12e−b2)\displaystyle = \lim_{\edge{b}\to\infty}\Big(\frac12 - \frac12 e^{-\edge{b}^2}\Big) =12\displaystyle = \addp{\frac12}
For the left half, e−a2e^{-\edge{a}^2} approaches 00 as a→−∞\edge{a} \to -\infty, so the limit is −12\subp{-\frac12}.
∫−∞0xe−x2 dx\displaystyle \int_{-\infty}^0 xe^{-x^2}\,dx =lim⁡a→−∞(−12+12e−a2)\displaystyle = \lim_{\edge{a}\to-\infty}\Big(-\frac12 + \frac12 e^{-\edge{a}^2}\Big) =−12\displaystyle = \subp{-\frac12}
Both halves converge, so we add them to get 00.
∫−∞∞xe−x2 dx\displaystyle \int_{-\infty}^{\infty} xe^{-x^2}\,dx =12+(−12)\displaystyle = \addp{\frac12} + \Big(\subp{-\frac12}\Big) =0\displaystyle = 0
7
Determine whether ∫1∞1x+ex dx\displaystyle\int_1^\infty \frac{1}{x + e^x}\,dx converges or diverges.

When xx is large, which term in x+exx + e^x grows faster?

Compare with 1x\infa{\frac1x}Dropping exe^x gives a larger function with infinite area, which gives no conclusion.
1x+ex<1x,∫1∞dxx\displaystyle \frac{1}{x + e^x} < \infa{\frac1x}, \qquad \int_1^\infty \infa{\frac{dx}{x}} =∞\displaystyle = \infty
Compare with e−x\fin{e^{-x}}Dropping xx gives a larger function whose integral converges, so our integral converges too.
1x+ex<1ex\displaystyle \frac{1}{x + e^x} < \frac{1}{e^x} =e−x,∫1∞e−x dx\displaystyle = \fin{e^{-x}}, \qquad \int_1^\infty \fin{e^{-x}}\,dx =1e\displaystyle = \frac1e
8
Determine whether ∫2∞1ln⁡x dx\displaystyle\int_2^\infty \frac{1}{\ln x}\,dx converges or diverges.

For x≥2x \ge 2, is ln⁡x\ln x smaller or larger than xx?

For x≥2x \ge 2, ln⁡x\ln x is smaller than xx, so its reciprocal is larger than 1x\infa{\frac1x}.
x\displaystyle x ≥2:ln⁡x<x,1ln⁡x>1x\displaystyle \ge 2:\quad \ln x < x, \qquad \frac{1}{\ln x} > \infa{\frac1x}
The integral of 1x\infa{\frac1x} diverges to infinity, so our larger integral diverges too.
∫2∞dxx\displaystyle \int_2^\infty \infa{\frac{dx}{x}} =lim⁡b→∞(ln⁡b−ln⁡2)\displaystyle = \lim_{\edge{b}\to\infty}\big(\ln \edge{b} - \ln 2\big) =∞\displaystyle = \infty
9
The region under y=1xy = \frac1x for x≥1x \ge 1 is rotated about the xx-axis. Find the volume of the solid, or show that it is infinite. (Challenge)

Each slice is a disk with radius r=1xr = \frac1x. What is the area of that disk?

A slice at xx is a disk with radius r=1xr = \frac1x, so its area is πr2\pi r^2.
r\displaystyle r =1x,πr2\displaystyle = \frac1x, \qquad \pi r^2 =π(1x)2\displaystyle = \pi\Big(\frac1x\Big)^2 =πx2\displaystyle = \frac{\pi}{x^2}
Integrate the disk areas from 11 to ∞\infty.
V\displaystyle V =π∫1∞dxx2\displaystyle = \pi\int_1^\infty \fin{\frac{dx}{x^2}} =πlim⁡b→∞(1−1b)\displaystyle = \pi\lim_{\edge{b}\to\infty}\Big(1 - \frac{1}{\edge{b}}\Big) =π\displaystyle = \pi
Even though the area under y=1xy = \infa{\frac1x} is infinite, squaring the radius turns 1x\infa{\frac1x} into 1x2\fin{\frac{1}{x^2}}, and that makes the volume finite.
∫1∞dxx\displaystyle \int_1^\infty \infa{\frac{dx}{x}} =∞,π∫1∞dxx2\displaystyle = \infty, \qquad \pi\int_1^\infty \fin{\frac{dx}{x^2}} =π\displaystyle = \pi