Integrals over an infinite interval, or of a function that blows up, are defined as limits of ordinary integrals. The question is always the same: does that limit exist?
Many of the most important integrals run to infinity or blow up somewhere — distributions over all of space, decaying signals, probabilities. Improper integrals define them carefully as limits.
The Gaussian integral $\int_{-\infty}^{\infty} e^{-x^2}\,dx=\sqrt{\pi}$ is what normalizes the bell curve. Laplace and Fourier transforms — over infinite domains — are the backbone of signal processing and control engineering. Expected values in probability are improper integrals.
Torricelli's "Gabriel's Horn" (1640s) has finite volume but infinite surface area — you could fill it with paint but never paint it. The paradox shocked mathematicians and is a perfect demonstration that infinity does not behave the way intuition expects.
An improper integral is not an ordinary integral with a strange endpoint. It is a limit. The first step is always to identify every place where ordinary Riemann integration fails.
When an antiderivative is hard or impossible, compare the integrand to a known benchmark. The benchmark is usually a $p$-integral, an exponential, or a simpler rational function.
$\displaystyle\int_a^\infty f=\lim_{t\to\infty}\int_a^t f$, when the limit exists (then the integral converges). Integrals over $(-\infty,\infty)$ must be split at a finite point; both pieces must converge.
$\displaystyle\int_1^\infty\frac{dx}{x^2}=\lim_{t\to\infty}\Big[-\frac1x \Big]_1^t=\lim_{t\to\infty}\Big(1-\frac1t\Big)=1.$
If $f$ blows up at $b$, $\displaystyle\int_a^b f=\lim_{t\to b^-}\int_a^t f$; if it blows up at $a$, use $\lim_{t\to a^+}\int_t^b f$. At an interior singularity, split the integral there; every piece must converge.
$\displaystyle\int_0^1\frac{dx}{\sqrt{x}}=\lim_{t\to0^+}\big[2\sqrt x\big]_t^1=2.$ Converges.
$\displaystyle\int_1^\infty\frac{dx}{x^p}$ converges iff $p>1$. $\displaystyle\int_0^1\frac{dx}{x^p}$ converges iff $p<1$.
If $0\le f\le g$ near the same improper endpoint and $\int g$ converges, so does $\int f$. If $\int f$ diverges, so does $\int g$.
$\displaystyle\int_1^\infty\frac{dx}{x^2+1}$ converges, since $\dfrac{1}{x^2+1}\le\dfrac{1}{x^2}$ and $\int_1^\infty x^{-2}$ converges.
• Always rewrite as a limit; don't plug $\infty$ in directly.
• Watch for hidden type-II behaviour (a singularity inside the interval).
• For the comparison test the functions must be nonnegative.
Problems adapted from the MAT137 question bank by Alfonso Gracia-Saz, used under CC BY-SA 4.0. Write full solutions before opening the answer.
Evaluate $\displaystyle\int_0^\infty e^{-x}\,dx$.
Does $\displaystyle\int_1^\infty\frac{dx}{x}$ converge?
Determine convergence of $\displaystyle\int_1^\infty\frac{2+\sin x}{x^2} \,dx$.
Evaluate $\displaystyle\int_1^\infty\frac{dx}{x^2+x}$.
For which $p$ do $\displaystyle\int_1^\infty\frac{dx}{x^p}$ and $\displaystyle\int_0^1\frac{dx}{x^p}$ converge?
What is $\displaystyle\int_{-1}^{1}\frac{dx}{x}$?
Determine convergence of $\displaystyle\int_1^\infty\frac{dx}{x+e^x}$.
Which converge: $\displaystyle\int_1^\infty\frac{1+\cos^2x}{x^{2/3}}\,dx$ or $\displaystyle\int_1^\infty\frac{1+\cos^2x}{x^{4/3}}\,dx$?
Course coverage. This unit is part of the complete MAT137 topic map. It should cover the following items before you move on:
Use the Alfonso Gracia-Saz MAT137 materials as the primary local spine, with the current UofT single-variable calculus topic split as the completeness check. The added items are original study prompts, not copied textbook exercises.
The improper-integral spine is: replace the suspect endpoint by a variable endpoint, compute or estimate the proper integral, then take a limit.
Use these after the main practice set. They are designed to force definition-level reasoning, theorem selection, and mixed application rather than routine calculation only.
For which $p$ does $\int_1^\infty x^{-p}dx$ converge?
Determine whether $\int_0^1 x^{-1/2}dx$ converges.
Show $\int_1^\infty e^{-x^2}dx$ converges.
Why must $\int_{-1}^1 1/x\,dx$ not be assigned value $0$ by symmetry?