Find the bad points
List every infinite bound and every point where f is undefined or infinite. Here: the upper bound ∞ only.

This improper integral calculator evaluates integrals with an infinite limit or a vertical asymptote: it writes the limit form, finds asymptotes inside the interval, decides whether the integral converges or diverges, and gives the value or the reason it diverges.
Results
Your improper integral, as a limit and step by step
Drag the cutoff toward the infinite or singular end. The partial integral settles on the limit when the integral converges and runs away when it diverges.
Interactive diagram
Partial integral and its limit
Improper integral
Antiderivative
Step-by-step solution
—Step by step
To use the improper integral calculator, follow these 4 steps. The result and the graph update as you type.
Type the integrand
Use one variable, for example 1/x^2, exp(-x) or 1/sqrt(x).
Enter the limits
Type inf or -inf for infinity, or use the quick buttons. A limit where f blows up counts too.
Read the verdict
The result says converges or diverges, with the value or the reason, such as “grows without bound”.
Watch the partial integral
Drag the cutoff toward the improper end and watch the partial integral settle or run away.
Definition
An improper integral is a definite integral with an infinite limit of integration, or with an integrand that goes to infinity somewhere on the interval. The Fundamental Theorem of Calculus does not apply to it directly, so the integral is defined as a limit of ordinary definite integrals.
There are 2 types of improper integrals:
An improper integral converges when the limit exists and is finite, and diverges when the limit is infinite or does not exist. The area under the curve can be finite over an infinitely long interval: the region under y = 1/x² from x = 1 to ∞ has area exactly 1, while the region under y = 1/x has infinite area.
Improper integrals appear across 4 fields: probability and statistics (the normal distribution, expected value, variance and moments), physics (gravitational potential, quantum mechanics, heat conduction), engineering (the Laplace transform and the Fourier transform in signal processing) and calculus itself (series analysis through the integral test).
The slider runs t from 1 to 10⁸. Green area: under 1/x², which approaches 1. Orange band: the extra area under 1/x, which grows like ln t without any bound.
Type I
A Type I improper integral has an infinite limit of integration, and it is evaluated by replacing the infinite bound with a variable t and letting t go to infinity. The finite bound stays where it is.
The integral converges when the integrand shrinks fast enough in the tail. Exponential decay is always fast enough: ∫₀^∞ e−x dx = 1 and ∫₋∞⁰ eˣ dx = 1. Tails of the form 1/xᵖ depend on the exponent: 1/x² gives 1, 1/x diverges. An oscillating integrand such as sin x has partial integrals 1 − cos t that swing between 0 and 2 forever, so ∫₀^∞ sin x dx diverges.
| t | ∫₀ᵗ e⁻ˣ dx | ∫₁ᵗ 1/x² dx | ∫₁ᵗ 1/x dx | ∫₀ᵗ sin x dx |
|---|---|---|---|---|
| 10 | 0.99995 | 0.9 | 2.3026 | 1.8391 |
| 100 | 1.00000 | 0.99 | 4.6052 | 0.1377 |
| 1,000 | 1.00000 | 0.999 | 6.9078 | 0.4376 |
| 10,000 | 1.00000 | 0.9999 | 9.2103 | 1.9522 |
| t → ∞ | 1 (converges) | 1 (converges) | ∞ (diverges) | no limit (diverges) |
Type II
A Type II improper integral has an integrand that blows up at a point of a finite interval: a vertical asymptote at an endpoint or inside the interval. The bad point is approached with a one-sided limit.
An asymptote at an endpoint is handled by stopping a small distance ε short of it and letting ε → 0⁺. ∫₀¹ 1/√x dx = lim (2 − 2√ε) = 2, so the region beside the y-axis has finite area 2 while its height is infinite. ∫₀¹ 1/x dx = lim (−ln ε) = ∞ diverges.
An asymptote inside the interval at x = c splits the integral into 2 improper integrals, one on each side of c, and both must converge. Plugging the endpoints straight into the antiderivative skips the asymptote and gives a wrong number. ∫₋₁¹ 1/x² dx is not [−1/x]₋₁¹ = −2: the integrand is positive, and each half diverges to ∞.
The red dashed line is the vertical asymptote x = 0. Green area under 1/√x heads to 2; the extra orange area under 1/x heads to ∞.
Method
To evaluate an improper integral with limits, find every bad point, replace each one with a variable, integrate as usual, and take the limit. Here are the 4 steps for ∫₁^∞ 1/x³ dx.
Find the bad points
List every infinite bound and every point where f is undefined or infinite. Here: the upper bound ∞ only.
Rewrite as a limit
Integrate on the finite piece
Take the limit
| Situation | Limit form | Example |
|---|---|---|
| Upper bound ∞ | ||
| Lower bound −∞ | ||
| Both bounds infinite | ||
| Asymptote at a | ||
| Asymptote at b | ||
| Asymptote inside at c | split at c, 2 one-sided limits |
Tests
An improper integral converges when its limit is a finite number and diverges otherwise. There are 3 standard tests that decide convergence without computing the integral: the p-test, the comparison test and the limit comparison test.
The p-test says ∫₁^∞ 1/xᵖ dx converges to 1/(p − 1) when p > 1 and diverges when p ≤ 1. The rule flips near 0: ∫₀¹ 1/xᵖ dx converges to 1/(1 − p) when p < 1 and diverges when p ≥ 1. p = 1 is the borderline, and it diverges in both places, so ∫₀^∞ 1/xᵖ dx diverges for every p.
| Integral | Converges when | Value |
|---|---|---|
| p > 1 | ||
| p < 1 |
Green: a part that converges. Orange: a part that diverges. No value of p makes both parts green.
The comparison test says: if 0 ≤ f(x) ≤ g(x) and ∫ g converges, then ∫ f converges; if ∫ f diverges, then ∫ g diverges. For x ≥ 1, e−x² ≤ e−x, so ∫₁^∞ e−x² dx converges. (1 + sin² x)/x ≥ 1/x, so ∫₁^∞ (1 + sin² x)/x dx diverges.
The limit comparison test says: if f, g > 0 and lim f(x)/g(x) = L with 0 < L < ∞, then ∫ f and ∫ g both converge or both diverge. x/(x³ + 1) behaves like 1/x² for large x, and the ratio tends to 1, so ∫₁^∞ x/(x³ + 1) dx converges. Its value is about 0.8356.
Logarithms on the borderline
Logarithms decide the cases right at p = 1. ∫₂^∞ 1/(x ln² x) dx converges to 1/ln 2 ≈ 1.4427, while ∫₂^∞ 1/(x ln x) dx diverges. Tails this slow are where numeric tools struggle most.
Two infinite ends
An integral over the whole real line is split at any point, usually 0, into 2 improper integrals, and it converges only when both halves converge separately.
The halves must be taken as separate limits. ∫₋∞^∞ x dx diverges, since ∫₀^∞ x dx = ∞; the symmetric limit lim ∫₋ₜᵗ x dx = 0 does not change that. That symmetric value is the Cauchy principal value; it is a different, weaker idea than convergence.
Probability uses these integrals constantly. Every probability density integrates to 1 over the real line: the normal distribution through the Gaussian integral, and the exponential distribution 2e−2x on [0, ∞). The expected value ∫ x f(x) dx and the variance are improper integrals too, and the Cauchy (Lorentz) distribution has no expected value, since ∫₋∞^∞ |x|/(π(1 + x²)) dx diverges.
| Integral | Value | Name or use |
|---|---|---|
| Gaussian integral, normal distribution | ||
| Cauchy / Lorentz distribution | ||
| Dirichlet integral, signal processing | ||
| Gamma function Γ(n + 1) | ||
| Laplace transform of 1 |
Worked examples
Here are 5 worked examples of improper integrals: 3 that converge and 2 that diverge. Green shading marks a finite area and orange shading an infinite one.
Answer: , the p-test with p = 2: 1/(2 − 1) = 1.
Answer: the integral diverges. ln t passes 10 at t ≈ 22,026 and 20 at t ≈ 4.85 × 10⁸; it grows slowly, but it never stops.
Answer: , the p-test near 0 with p = 1/2: 1/(1 − 1/2) = 2.
Answer: the integral diverges. A positive integrand cannot have a negative integral, so −2 is a red flag.
Answer: . e−x² has no elementary antiderivative, so this double-integral trick replaces the usual method.
Watch out
There are 8 common mistakes with improper integrals. Each card pairs the mistake with its fix.
Missing an asymptote inside the interval
Check where the denominator is 0 before integrating. ∫₋₁¹ 1/x² dx diverges; it is not −2.
Plugging in ∞ as a number
Write lim t → ∞ and take the limit. "1/∞ = 0" skips the step that decides convergence.
Using the principal value as the answer
∫₋∞^∞ x dx diverges, while lim ∫₋ₜᵗ x dx = 0. Split at 0 and test each half.
Thinking f → 0 means convergence
1/x → 0, but ∫₁^∞ 1/x dx diverges. The integrand must shrink fast enough.
Mixing up the 2 p-test rules
At ∞: p > 1 converges. Near 0: p < 1 converges. p = 1 diverges in both.
Comparing in the wrong direction
A smaller integrand than a divergent one proves nothing. Bound above by a convergent integral, or below by a divergent one.
Trusting a numeric value alone
A truncated sum of a slowly divergent integral looks finite. Check the limit or use a test.
One limit for 2 bad points
Each infinite bound and each asymptote needs its own independent limit.
Questions
An improper integral converges when its limit form gives a finite number. Rewrite each infinite bound or asymptote as a limit, integrate, and take the limit: a finite result means convergence. Without an antiderivative, use the p-test, the comparison test or the limit comparison test. ∫₁^∞ 1/x² dx = 1 converges; ∫₁^∞ 1/x dx = ∞ diverges.
Yes, an improper integral equals a finite number whenever it converges, even over an infinite interval or next to a vertical asymptote. ∫₁^∞ 1/x² dx = 1, ∫₀¹ 1/√x dx = 2 and ∫₋∞^∞ e^(−x²) dx = √π ≈ 1.7725 are all finite.
∫₁^∞ 1/x dx diverges and ∫₁^∞ 1/x² dx converges since 1/x² shrinks fast enough in the tail and 1/x does not. The partial integrals show it: ∫₁ᵗ 1/x² dx = 1 − 1/t approaches 1, while ∫₁ᵗ 1/x dx = ln t grows without bound. p = 1 is exactly the borderline of the p-test.
No, ∫₀^∞ sin x dx does not converge: the partial integrals 1 − cos t swing between 0 and 2 forever and have no limit. Damped oscillations can converge. ∫₀^∞ sin x / x dx = π/2 (the Dirichlet integral) converges, and ∫₀^∞ e^(−x) sin x dx = 1/2 converges absolutely.
The Cauchy principal value is the symmetric limit of an integral, taken with both ends moving together: PV ∫₋∞^∞ f dx = lim ∫₋ₜᵗ f dx. It can be finite when the improper integral diverges. PV ∫₋∞^∞ x dx = 0 and PV ∫₋₁¹ 1/x dx = 0, but both integrals diverge in the ordinary sense.
No, a numeric calculator cannot prove divergence; it can only show strong evidence. Truncated values of a slowly divergent integral, such as ∫₂^t 1/(x ln x) dx = ln ln t − ln ln 2, grow so slowly that they look finite. A proof needs a limit or a test. This calculator works from the antiderivative when one exists and reports the reason, such as "grows without bound" or "oscillates".
This improper integral calculator evaluates integrals with infinite limits or vertical asymptotes, writes the limit form with steps, and tells you whether the improper integral converges or diverges, with the value or the reason.