Convergence tests for improper integrals — comparison: Visual walkthrough

This is the visual companion to the parent topic. Read the steps in order — each picture carries the next line of the argument. We will meet the two star players — a curve we'll call $f$ and a taller "lid" curve we'll call $g$ — properly in Steps 1 and 4; for now just know $f$ is our curve and $g$ is a helper sitting above it.

Step 1 — What "area under a curve" even means

WHAT. Pick a function and call it $f$ — this is our curve, the one whose area we ultimately care about. On a window from $x = a$ to $x = t$ , the symbol $\int_{a}^{t} f (x) d x$ means the area sandwiched between the curve $y = f (x)$ and the flat floor $y = 0$ , from the left wall $x = a$ to the right wall $x = t$ .

Term by term:

$f (x)$ — the height of our curve at horizontal position $x$ .
$d x$ — an infinitely thin slice of width, one sliver of the region.
$\int_{a}^{t}$ — "add up every sliver" as $x$ sweeps from $a$ to $t$ .

WHY start here. We can only trust an argument if the words underneath the symbols are honest. Everything later is just "this area is smaller than that area," so we must first agree that the integral is an area.

PICTURE. The blue sliver is one $f (x) d x$ . Stack all the slivers between the two walls and you get the shaded region.

Figure — Convergence tests for improper integrals — comparison

Step 2 — Why $\infty$ forces us to use a limit

WHAT. We want the area all the way to the right forever: $\int_{a}^{\infty} f d x$ . But $\infty$ is not a number — you cannot make a wall stand at infinity. So we do the only legal thing: keep the right wall at a finite spot $t$ , measure the area, then slide $t$ farther and farther right and watch what the area does: $\int_{a}^{\infty} f (x) d x = lim_{t \to \infty} \int_{a}^{t} f (x) d x .$

$t$ — the movable right wall (a real, finite number).
$lim_{t \to \infty}$ — "let the wall run off to the right; what number does the area approach?"

If that running area settles onto a finite number, we say the integral converges. If it keeps growing without bound, it diverges.

WHY a limit and not just "plug in $\infty$ ." You physically cannot integrate over an infinite bar in one shot; you can only integrate over finite bars. The limit is the bridge from "finite window I can compute" to "infinite window I want."

PICTURE. Watch the right wall march right in three snapshots; the running area $F (t)$ climbs.

Step 3 — Name the running area, and notice it only goes UP

WHAT. Give the running area a name: $F (t) = \int_{a}^{t} f (x) d x .$ Now suppose $f (x) \geq 0$ everywhere (the curve never dips below the floor). Then when you push the wall right by a little, you can only add area — never subtract it, because every new sliver has height $\geq 0$ . So: $t_{2} > t_{1} ⟹ F (t_{2}) \geq F (t_{1}) .$ A function that never decreases is called monotone increasing.

$F (t)$ — total area collected up to the wall $t$ .
" $\geq 0$ " on $f$ — the promise that no sliver ever cuts area away.

WHY this matters so much. An increasing quantity has only two possible destinies: either it climbs forever (diverges), or it bumps into a ceiling and levels off (converges). There is no oscillating, no wandering.

This is the Monotone Convergence Theorem at work, and here is its statement so you needn't leave the page: a quantity that is increasing and stays below some fixed ceiling must approach a finite limit.

Definition Why "increasing + capped ⇒ has a limit" is even

true (completeness of $R$ ) Picture all the values $F (t)$ collects. They keep rising but every one sits below the ceiling $M$ . The real numbers have a special property — completeness — which guarantees that any set of numbers with an upper bound has a least upper bound $L$ (a lowest possible ceiling, no gaps allowed). Now $F$ climbs toward that $L$ : it can't pass $L$ (that's a ceiling), and it can't stall below $L$ forever (else a slightly lower number would be an even tighter ceiling, contradicting $L$ being the least). Squeezed between "can't pass" and "can't stall short," $F (t)$ has no choice but to home in on $L$ — that is its finite limit. Without completeness (imagine only fractions, with gaps) the values could crowd toward a hole that isn't a number at all; it is the gap-free real line that makes the limit genuinely exist. This is the machine we cash in at Step 5 — and it is why we demand $f \geq 0$ , so that $F$ is genuinely increasing.

PICTURE. The staircase of $F (t)$ can only step up as $t$ grows.

Step 4 — Stack a second curve on top and read the inequality

WHAT. Bring in a second, taller function $g$ — the helper "lid" — with $0 \leq f (x) \leq g (x) for all x \geq a .$ This says: at every horizontal position, the $g$ -curve is at least as high as the $f$ -curve, and both stay above the floor.

Now give the lid's running area its own name, exactly as we named $F$ in Step 3: $Let G (t) = \int_{a}^{t} g (x) d x .$

$0 \leq f$ — $f$ never dips below the floor (from Step 3, so $F$ increases).
$f \leq g$ — $g$ is a lid sitting on top of $f$ everywhere.
$G (t)$ — total area under the lid $g$ up to the wall $t$ (and, since $g \geq 0$ , it too is increasing).

Because $g$ is taller at every $x$ , its area up to any wall is at least as big: $F (t) = \int_{a}^{t} f \leq \int_{a}^{t} g = G (t) .$

WHY. Comparing two hard areas directly is painful. But comparing heights is trivial — you just look at which curve is on top. Height-by-height " $f \leq g$ " instantly gives area-by-area " $F \leq G$ ." We traded a hard question (an integral) for an easy one (an inequality of formulas).

PICTURE. $g$ is the plum lid; $f$ is the orange region living underneath it.

Step 5 — The finite lid becomes a ceiling for our area

WHAT. Now suppose we already know the lid's total area is finite: $\int_{a}^{\infty} g d x = M (some finite number) .$ Since $G (t)$ is also increasing and approaches $M$ , we have $G (t) \leq M$ for every $t$ . Chain it with Step 4: $F (t) \leq G (t) \leq M .$ So $F (t)$ is an increasing quantity (Step 3) that never exceeds $M$ — it is bounded above by a ceiling.

$M$ — the finite total area under the lid.
$F (t) \leq M$ — our area can never poke through the ceiling $M$ .

WHY this closes the argument. By the Monotone Convergence Theorem (which we powered with completeness in Step 3), an increasing quantity trapped under a ceiling cannot run off to infinity — it has a least upper bound and homes in on it. Therefore $lim_{t \to \infty} F (t)$ exists and is finite: $0 \leq f \leq g and \int_{a}^{\infty} g converges ⟹ \int_{a}^{\infty} f converges .$ That is the Direct Comparison Test, built from area, from a limit, from "increasing," and from "trapped under a ceiling" — nothing else.

PICTURE. The orange staircase $F (t)$ rises but is pinned under the dashed ceiling $M$ ; it must level off.

Step 6 — The mirror case: divergence pushes upward

WHAT. Flip the roles. Suppose the bottom curve already has infinite area: $\int_{a}^{\infty} f d x = \infty, 0 \leq f \leq g .$ Since $F (t) \leq G (t)$ (Step 4) and $F (t) \to \infty$ , the bigger quantity $G (t)$ is dragged up with it: $G (t) \geq F (t) \to \infty ⟹ \int_{a}^{\infty} g diverges .$

WHY we need both directions. Convergence proofs squeeze from above (find a finite lid). Divergence proofs push from below (find an infinite floor beneath). Same picture, read in opposite directions — which is exactly the mnemonic "squeeze to please, push to die."

PICTURE. A divergent $f$ underneath forces the taller $g$ to blow up too.

Step 7 — The degenerate trap: why "smaller than infinite" proves NOTHING

WHAT. The tempting wrong move: " $f \leq g$ and $\int g = \infty$ , so $\int f = \infty$ too." Watch it fail. Take $f = 1/ x^{2}$ and $g = 1/ x$ on $[1, \infty)$ . Indeed $f \leq g$ . But $\int_{1}^{\infty} \frac{d x}{x ^{2}} = 1 (finite), \int_{1}^{\infty} \frac{d x}{x} = \infty.$ A finite area is sitting underneath an infinite one. Being below a bottomless well tells you nothing about your own depth.

$f \leq g$ here is TRUE — but useless, because it points the wrong way.

WHY. The only two arrows that carry information are: $(small \leq convergent) \Rightarrow convergent, (big \geq divergent) \Rightarrow divergent .$ The other two combinations ("small $\leq$ divergent", "big $\geq$ convergent") give no conclusion. Always check the arrow points from the known fate to the unknown one in the correct direction.

PICTURE. The finite orange area $1/ x^{2}$ nestles under the infinite teal area $1/ x$ — proof that the naive implication is false.

The one-picture summary

WHAT. One diagram holds all seven steps at once: our nonnegative curve $f$ (orange, area shaded), a lid $g$ above it (plum), and a floor curve below it (teal, dashed). Each pictorial element is a step made visible.

Orange region — $F (t) = \int_{a}^{t} f$ , the increasing running area from Steps 1–3.
Plum lid $g$ — the finite ceiling from Steps 4–5: squeeze $f$ under it ⇒ converges.
Teal dashed floor — the infinite floor from Step 6: push $f$ over it ⇒ diverges.

WHY these three elements and no others. Every legal comparison is one of exactly two moves, and the picture shows both simultaneously: bound above by a finite lid to win convergence, or bound below by an infinite floor to force divergence. The trap of Step 7 is exactly what happens if you mix the arrows up — a finite curve happily lives beneath an infinite one, proving nothing.

Recall Feynman retelling — the whole walkthrough in plain words

An integral is just the area trapped under a curve (Step 1). We can't build a wall at infinity, so we build one at a finite spot $t$ and slide it right, watching the area — that sliding is the limit (Step 2); the very same sliding, aimed at a blow-up point instead, handles vertical-asymptote integrals too. If the curve never dips below the floor, the area can only grow as the wall moves; it never shrinks (Step 3). A rising quantity that stays under a ceiling has a least ceiling it homes in on — that's completeness of the real line, and it's what makes the limit truly exist. Now lay a taller curve $g$ over our curve $f$ like a lid: since $g$ is higher everywhere, its area is bigger everywhere (Step 4). If that lid has a finite total area $M$ , our growing area is stuck below $M$ forever — so it must settle to a finite value. That's convergence (Step 5). Read the picture upside-down and you get divergence: if a curve underneath us already has infinite area, we're dragged up to infinity too (Step 6). The one trap: being smaller than something infinite proves nothing — a tidy finite area like $1/ x^{2}$ happily lives under the bottomless $1/ x$ (Step 7). Only two directions carry news: squeeze under a finite lid to prove finite, push over an infinite floor to prove infinite. And it all works only because both curves stay above the floor.

Connections

Parent topic
The p-integral and p-series — the lid $g$ is almost always a power $x^{- p}$ .
Monotone Convergence Theorem — the engine behind Step 3 and Step 5.
Improper integrals — infinite discontinuities (type 2) — the same argument covers vertical asymptotes.
Comparison test for infinite series — the exact same picture for sums.
Absolute vs conditional convergence — what to do when $f$ changes sign.
Integral Test for series