Improper integrals extend definite integration to cases where the interval is infinite or the integrand blows up. You handle them by replacing the problematic bound with a variable, evaluating the resulting proper integral, and then taking a limit. Whether that limit exists (convergent) or not (divergent) is the central question, and the comparison theorem lets you answer it without computing the integral itself.
Difficulty: Intermediate to Advanced. Prerequisites: all integration techniques from Part 1 (u-substitution, polynomial division, completing the square, trig substitution, partial fractions), L'Hopital's Rule, and a solid understanding of limits at infinity and one-sided limits.
Not every definite integral fits neatly between two finite bounds with a well-behaved integrand. In practice, you often need to integrate over an infinite interval (the area under a curve that extends forever) or handle a function that becomes unbounded within the interval. Improper integrals give you a rigorous framework for deciding whether such areas are finite and, if so, computing them.
This material sits near the end of the integration-techniques portion of most calculus courses and connects forward to series convergence tests, probability density functions, and Laplace transforms. If you are not yet confident with limit evaluation (especially L'Hopital's Rule and limits involving exponentials), review that material before proceeding.
Improper integral
A definite integral where either the interval of integration is infinite (e.g. integral from 1 to infinity) or the integrand has a vertical asymptote within the interval. In simple terms, it is an integral that "should not work" under the standard definition but can sometimes be given a finite value through limits.
Convergent integral
An improper integral whose defining limit exists and is finite. Think of it as: the area under the curve, despite extending to infinity or approaching a blow-up, adds up to a finite number.
Divergent integral
An improper integral whose defining limit does not exist or is infinite. The area is unbounded, so no finite answer can be assigned.
p-integral test (p-test)
The integral of 1/x^p dx from 1 to infinity converges if and only if p > 1. The integral of 1/x^p dx from 0 to 1 converges if and only if p < 1. This is the most commonly used benchmark for convergence arguments.
Comparison theorem
If 0 <= f(x) <= g(x) for all x in the interval, and if the integral of g(x) converges, then the integral of f(x) converges too. Conversely, if the integral of f(x) diverges, then the integral of g(x) diverges. Think of it as: a smaller function trapped under a convergent one must also converge.
L'Hopital's Rule (in the context of improper integrals)
Used to evaluate limits of the form 0/0 or infinity/infinity that arise after integrating. Often needed when the antiderivative contains products like tln(t) or te^(-t) that require careful limit evaluation.
When the integrand blows up at a point in the interval, replace that endpoint with a limit variable and check whether the limit exists.
Type I: integrand blows up at an endpoint
For the integral of x^(-4/3) from 0 to 1 (which arises in a piecewise problem), f(x) tends to infinity as x approaches 0 from the right.
Write it as lim (t -> 0+) of the integral from t to 1 of x^(-4/3) dx = lim (t -> 0+) [-3/x^(1/3)] from t to 1 = lim (t -> 0+) (-3 + 3/t^(1/3)).
Since 3/t^(1/3) tends to infinity as t approaches 0+, this integral diverges. The full piecewise integral is therefore not convergent.
The p-test for integrals from 0 to 1
The integral of 1/x^p from 0 to 1 converges when p < 1 and diverges when p >= 1.
For the integral of (5 + x^(-1/3))/x^(1/2) dx from 0 to 1, split as 5/x^(1/2) + 1/x^(5/6). Both exponents (1/2 and 5/6) are less than 1, so each piece converges.
The value is 5/(1 - 1/2) + 1/(1 - 5/6) = 10 + 6 = 16.
Improper integrals involving inverse trig functions
For the integral of 8 arcsin(x)/sqrt(1 - x^2) dx from 0 to 1, the integrand blows up as x approaches 1 from the left.
Substitute u = arcsin(x), so du = dx/sqrt(1 - x^2). The integral becomes 8 times the integral of u du = 4u^2.
Evaluate as lim (t -> 1-) of 4(arcsin(t))^2 = 4(pi/2)^2 = pi^2. Convergent.
Integrals involving ln(x) near zero
For the integral of 4 ln(6x) dx from 0 to 1, the integrand tends to negative infinity as x approaches 0+.
Integration by parts gives 4x ln(6x) - 4x. The critical limit is lim (t -> 0+) of 4t ln(6t), which equals 0 by L'Hopital's Rule.
Result: 4 ln(6) - 4 = 4(ln(6) - 1). Convergent.
For the integral of 2 ln(x)/x dx from 0 to 5, substitute u = ln(x) to get u^2. The limit as t approaches 0+ of (ln(t))^2 is infinity, so this integral diverges.
When the upper (or lower) limit is infinite, replace it with a finite variable t, evaluate the definite integral, and take the limit as t approaches infinity.
Integration by parts with exponential decay
For the integral of 3x e^(-x/3) dx from 4 to infinity, set I = lim (t -> infinity) of the integral from 4 to t.
Integrate by parts: let u = 3x, dv = e^(-x/3) dx. Then du = 3 dx, v = -3e^(-x/3).
The antiderivative works out to -9(x + 3)e^(-x/3). Evaluate from 4 to t: 63e^(-4/3) - 9(t + 3)e^(-t/3).
By L'Hopital's Rule, lim (t -> infinity) of (t + 3)e^(-t/3) = 0 (exponential decay dominates any polynomial).
Result: I = 63e^(-4/3). Convergent.
The key principle: x^m e^(-x) tends to 0 as x tends to infinity, for any power m. This is why exponential-decay integrals over infinite intervals almost always converge.
Power-function integrals (the p-test for infinity)
The integral of 8/x^4 dx from 1 to infinity: antiderivative is -(8/3)x^(-3). Evaluate: (8/3)(1 - t^(-3)) tends to 8/3 as t tends to infinity.
Area = 8/3. Convergent because the exponent 4 is greater than 1.
Rational-function integrals
For the integral of 8x/(1 + x^2)^2 dx from 1 to infinity, substitute u = 1 + x^2. The integral becomes 4 times the integral of 1/u^2 du.
Evaluate: 4(1/2 - 1/(1 + t^2)) tends to 2 as t tends to infinity. Convergent.
Area under exponential curves
The area under y = e^(-x/4) from x = -1 to infinity: integral = -4e^(-x/4) evaluated from -1 to t = 4e^(1/4) - 4e^(-t/4).
As t tends to infinity, 4e^(-t/4) tends to 0, so the area is 4e^(1/4). Convergent.
The comparison theorem is your main tool for determining convergence without computing the integral. The idea: bound the integrand above or below by a simpler function whose convergence you already know.
How to apply it
Show that 0 <= f(x) <= g(x) for all x in the interval.
If the integral of g(x) converges, then the integral of f(x) converges.
If the integral of f(x) diverges, then the integral of g(x) diverges.
Worked example: integral of sin^2(x)/(3 + x^2) dx from 1 to infinity
Bound: 0 <= sin^2(x)/(3 + x^2) <= 1/(3 + x^2) <= 1/x^2.
The integral of 1/x^2 from 1 to infinity converges (p-test with p = 2 > 1).
Therefore the integral of sin^2(x)/(3 + x^2) converges by comparison.
Why other integrals diverge (from the homework)
The integral of 2/sqrt(7x - 1) dx from 1 to infinity: for large x, 2/sqrt(7x - 1) behaves like 2/sqrt(7x), which is comparable to 1/x^(1/2). Since p = 1/2 < 1, this diverges.
The integral of (1 + sqrt(7x))/x dx from 1 to infinity: for large x, this behaves like sqrt(7)/x^(1/2), which diverges.
The integral of 11/sqrt(x^2 - 1) dx from 1 to infinity: for large x, this behaves like 11/x, and p = 1 gives divergence.
The integral of 11/(x + 4) dx from 0 to infinity: this behaves like 1/x for large x, which diverges.
The p-test as your benchmark
Nearly every comparison argument ends with the p-test. Memorise both forms: the integral of 1/x^p from 1 to infinity converges if and only if p > 1; the integral of 1/x^p from 0 to 1 converges if and only if p < 1.
Misconceptions
Students often assume that any integral with an infinite limit diverges. Many converge, especially those involving exponential decay (e^(-x)) or high powers of x in the denominator.
Students sometimes evaluate an improper integral as if it were proper, plugging in infinity directly without taking a limit. This is not valid; you must set up the limit explicitly.
A common error is forgetting that a piecewise-defined function can create an improper integral even on a finite interval. If f(x) = x^(-4/3) for x <= 1, the integral from 0 to 2 is improper because f blows up at x = 0.
Students mix up the two p-tests. Remember: for the integral from 1 to infinity, convergence requires p > 1. For the integral from 0 to 1, convergence requires p < 1. They go in opposite directions.
Exam flags
Expect at least one problem asking you to determine convergence or divergence without computing the integral (comparison theorem). You need to produce a clean inequality chain.
Integration by parts combined with L'Hopital's Rule is a classic exam combination. The integral of x e^(-x) or x ln(x) type problems test both skills.
Multi-part problems (express as a limit, compute, evaluate the limit) are common and test whether you understand the structure of an improper integral, not just the computation.
Know the area-under-a-curve interpretation: the area under y = 8/x^4 for x >= 1 is 8/3, while the area under y = e^(-x/4) for x >= -1 is 4e^(1/4). These are standard exam questions.
True or False: The integral of 1/x^(0.5) dx from 1 to infinity converges. (Answer: False. p = 0.5 < 1, so it diverges.)
True or False: The integral of 1/x^3 dx from 0 to 1 converges. (Answer: False. p = 3 > 1, so it diverges on this interval.)
Fill in the blank: To make the integral from 4 to infinity of 3x e^(-x/3) dx proper, replace the upper limit with ____ and take the limit as it approaches ____. (Answer: t; infinity.)
True or False: If 0 <= f(x) <= g(x) and the integral of f(x) converges, then the integral of g(x) converges. (Answer: False. Convergence of the smaller function tells you nothing about the larger one.)
True or False: lim (t -> infinity) of t^5 e^(-t) = 0. (Answer: True. Exponential decay beats any polynomial.)
Q: Determine whether the integral of x^(-4/3) from 0 to 1 converges, and if so find its value.
A: The antiderivative is -3/x^(1/3). As t approaches 0+, 3/t^(1/3) tends to infinity. Divergent.
Q: Find the total area under y = 8/x^4 for x >= 1.
A: Integral from 1 to infinity of 8x^(-4) dx = -(8/3)x^(-3) evaluated from 1 to infinity = 8/3.
Q: Evaluate the integral of 3x e^(-x/3) dx from 4 to infinity.
A: Integration by parts gives -9(x + 3)e^(-x/3). Evaluate: 63e^(-4/3) - 9(t + 3)e^(-t/3). As t tends to infinity, the second term vanishes by L'Hopital's Rule. Answer: 63e^(-4/3).
Q: Does the integral of sin^2(x)/(3 + x^2) dx from 1 to infinity converge?
A: Yes. Bound by 1/x^2, which converges (p = 2 > 1). By the comparison theorem, the original integral converges.
Q: Evaluate the integral of 8 arcsin(x)/sqrt(1 - x^2) dx from 0 to 1.
A: Substitute u = arcsin(x). The integral becomes 8 times the integral of u du = 4u^2 = 4(arcsin(t))^2. As t approaches 1-, arcsin(t) approaches pi/2, so the answer is 4(pi/2)^2 = pi^2.
Q: Determine whether the integral of 4 ln(6x) dx from 0 to 1 converges, and find its value.
A: Integration by parts gives 4x ln(6x) - 4x. The limit of 4t ln(6t) as t approaches 0+ is 0 (by L'Hopital). Answer: 4 ln(6) - 4 = 4(ln(6) - 1).
Q: Determine whether the integral of (5 + x^(-1/3))/x^(1/2) dx from 0 to 1 converges, and find its value.
A: Split as 5/x^(1/2) + 1/x^(5/6). Both exponents are less than 1, so both converge. Value: 5/(1/2) + 1/(1/6) = 10 + 6 = 16.
The comparison theorem for improper integrals is closely related to the comparison test for infinite series. The intuition is identical: if a smaller positive series converges, a larger one may or may not; if a larger one converges, the smaller one must.
Improper integrals appear in probability: every probability density function integrates to 1 over its entire domain, which is often an infinite interval. The convergence of these integrals is what makes probability distributions well-defined.
The integral test for series convergence works by comparing a series to an improper integral. Fluency with improper integrals is therefore a prerequisite for the series chapter that follows in most courses.
improper integral, convergent integral, divergent integral, comparison theorem, comparison test for integrals, p-integral test, p-test, infinite limits of integration, vertical asymptote integral, integration by parts improper, L'Hopital's Rule limits, area under curve to infinity, exponential decay integral, integral calculus, UT Austin integral calculus, HW 07, Wolesensky, Type I improper integral, Type II improper integral