What Are Improper Integrals and When Do They Converge?

Can a shape be infinitely long and still hold a finite amount of paint?

That sounds like a trick question. But it's exactly what happens with a strange object called Gabriel's Horn, and untangling it is the whole point of this post.

A trumpet that breaks your intuition

Take the curve $y = 1/x$, starting at $x = 1$ and stretching out forever. Now spin it around the x-axis. You get a trumpet-shaped horn that gets thinner and thinner, forever, but never quite closes up.

Here's the paradox. If you compute the volume inside that horn, you get a finite number. You could fill it with a bucket of paint. But if you compute the surface area of the horn's outside, it's infinite. You could never have enough paint to coat it, not even close.

Same object. One measurement is finite, the other is infinite. That's not a contradiction, it's a fact about infinite processes, and by the end of this post you'll know exactly why it happens.

What makes an integral "improper"

A regular integral, like $\int_2^5 x^2\,dx$, adds up a function's values over a finite, well-behaved interval. Nothing weird happens. You get a number, done.

An integral turns improper the moment one of two things happens.

Either the interval itself goes on forever, like $\int_1^\infty \frac{1}{x^2}\,dx$. Or the function blows up somewhere inside the interval, like $\int_0^1 \frac{1}{x}\,dx$, where $1/x$ shoots to infinity as $x$ approaches 0.

In both cases, you're trying to add up infinitely many things. The question is always the same: does that infinite sum settle on a number, or does it run away?

Case one: the interval goes on forever

Let's start with the horn's volume. It comes from an integral like $\int_1^\infty \frac{1}{x^p}\,dx$, for some power $p$. We can't literally integrate "to infinity" directly, so here's the trick calculus uses.

Pick some large but finite number $b$. Compute the integral up to $b$. Then ask what happens as $b$ keeps growing.

That's it. That's the whole idea behind an infinite-interval improper integral. You replace infinity with a variable, compute normally, then take a limit.

Play with $p$ above. When $p > 1$, the running area chases down a specific number and never overshoots it, even though $b$ keeps growing forever. When $p \le 1$, no such luck. The area keeps climbing without bound.

The reason comes down to speed. When $p > 1$, the function $1/x^p$ shrinks fast enough that each new sliver of area contributes almost nothing, and those contributions add up to a finite total. When $p \le 1$, the function shrinks too slowly. The slivers keep contributing just enough to push the total higher forever.

Working out the antiderivative confirms it: for $p \ne 1$,

As $b \to \infty$, the term $b^{1-p}$ vanishes if $p > 1$ (leaving a finite limit of $\frac{1}{p-1}$) and explodes if $p < 1$. And at exactly $p = 1$, the antiderivative is $\ln b$, which climbs to infinity no matter how slowly.

Case two: the function blows up

Now flip the problem. What if the interval is perfectly finite, but the function itself shoots to infinity at one edge?

Take $\int_0^1 \frac{1}{x^p}\,dx$. As $x$ approaches 0, $1/x^p$ rockets upward. We can't just plug in $x = 0$. So we use the same trick as before, but mirrored: replace the bad endpoint with a variable, then let it sneak up on the danger zone.

Slide $p$ and watch what happens as $\varepsilon$ creeps toward the wall at $x = 0$. And here's the interesting twist: the rule flips.

For the infinite-interval case, convergence needed $p > 1$. For this blow-up case, convergence needs $p < 1$. Smaller $p$ means the function grows more slowly as it approaches zero, so the sliver of area near the wall stays thin enough to add up to something finite. Larger $p$ means the function shoots up too violently, and the area near the wall is unbounded.

Same formula, same technique, opposite condition. Worth sitting with that for a second.

When you can't compute the integral at all

Real functions are rarely as clean as $1/x^p$. Sometimes you'll face something like

There's no elementary antiderivative for that. You can't just plug in bounds and subtract. So how do you know if $\int_1^\infty f(x)\,dx$ converges?

You don't compute it directly. You trap it.

Notice that $\sin(x)$ never dips below $-1$, so the denominator of $f(x)$ never drops below $x^2 + 2$. That means $f(x) \le \frac{1}{x^2+2}$ for every $x$, always. And $\frac{1}{x^2+2}$ does have a clean antiderivative, one whose integral to infinity we can compute exactly.

Drag $b$ outward and watch the white curve stay pinned beneath the gold one, no matter how far you go. If a function is always smaller than something whose total area is finite, its own total area can't be more than that. It's squeezed from above.

This is called the comparison test, and it's one of the most useful tools in this whole topic. You almost never need to compute an improper integral exactly to know whether it converges. You just need a bound you already understand.

The general rule, in one picture

Let's zoom out. We've seen two mirrored rules:

• $\int_1^\infty \frac{1}{x^p}\,dx$ converges when $p > 1$
• $\int_0^1 \frac{1}{x^p}\,dx$ converges when $p < 1$

Both flip at the exact same value: $p = 1$. That's not a coincidence, it's the same balancing act (shrinking width against growing count, or growing height against shrinking width) viewed from two different angles.

Slide $p$ across that boundary and watch both rules trade places at once. This single number, $p = 1$, is the dividing line between "infinite processes that settle down" and "infinite processes that don't," for every power function you'll ever integrate.

Back to the horn

Now we can finally explain the paradox we opened with.

The horn's volume comes from spinning $y = 1/x$ around the x-axis and integrating cross-sectional disks, which works out to $\pi \int_1^\infty \frac{1}{x^2}\,dx$. Here $p = 2$, which is greater than 1. It converges. The volume is exactly $\pi$.

The horn's surface area involves a similar-looking integral, but one that behaves like $\int_1^\infty \frac{1}{x}\,dx$ at its core, where $p = 1$. That's exactly the borderline case that always diverges. The surface area is infinite.

So the horn really does hold a finite bucket of paint on the inside while needing an infinite amount to coat the outside. Not a contradiction. Just two different powers of $x$, on two different sides of the same threshold.

The short version

An integral becomes improper when its interval stretches to infinity or its function explodes at some point. Either way, you replace the dangerous edge with a variable, compute normally, and take a limit.

Whether that limit settles on a number or runs away depends entirely on speed: how fast the function shrinks as it heads toward infinity, or how fast it's allowed to blow up near a wall. For power functions, that whole question collapses into a single number, $p$, and a single threshold, $p = 1$.

And when you can't compute the integral directly, you don't have to. Trap it between two functions you already understand, and let the comparison do the work.

All areas on this page are computed either from closed-form antiderivatives or, where no elementary antiderivative exists, from a 400-interval Simpson's rule approximation.