Uniform convergence of function sequences: Visual walkthrough

Step 1 — What is a "sequence of functions"? (a stack of graphs)

WHY we start here: every symbol later ( $f_{n}$ , $f$ , "the gap") is a feature of this picture. If you cannot see the stack of curves, nothing after makes sense.

PICTURE: below, each coloured curve is one $f_{n} (x) = x^{n}$ on $[0, 1]$ . As the label number $n$ climbs, the curve flattens toward the floor — except it is nailed to height $1$ at the right edge $x = 1$ .

Figure — Uniform convergence of function sequences

Step 2 — "Converges at a point" = one vertical column of numbers settles

WHY this reduction: convergence of a curve is scary; convergence of a column of numbers is Calculus-1. Pointwise convergence is nothing more than "do this column-by-column, every column separately."

PICTURE: two vertical dashed columns. At $x = 0.5$ the heights $0.5, 0.25, 0.125, \dots$ dive to $0$ . At $x = 1$ the heights are $1, 1, 1, \dots$ — stuck at $1$ . Different columns, different destinations.

Step 3 — The GAP: how wrong is curve $n$ , and where is it wrongest?

WHY the absolute value and not the plain difference: convergence cares about closeness, and closeness is a distance. A gap of $- 0.3$ and $+ 0.3$ are the same amount of "off."

PICTURE: the gap $g_{n} (x) = x^{n}$ (since $f = 0$ on $[0, 1)$ ) shaded in. Notice it is tiny on the left and rears up to nearly $1$ as $x$ slides toward the right edge. The tallest bar is the villain of the whole story.

Step 4 — The single number that decides everything: the tallest gap $M_{n}$

WHY $sup$ and not $max$ ? $max$ demands the tallest bar be actually attained by some $x$ . But the tallest gap of $x^{n}$ on $[0, 1)$ is approached as $x \to 1^{-}$ and never reached on $[0, 1)$ . $sup$ ("supremum" = least upper bound) is the honest tool: the smallest ceiling no bar pokes through, whether or not a bar touches it. It answers "how bad can it get?" without needing a champion point.

PICTURE: the gap curve with a red horizontal ceiling pressed down until it just kisses the top of the bars. That ceiling's height is $M_{n}$ . Here it sits at $M_{n} = 1$ for every $n$ — the ceiling never drops.

Step 5 — WHY "tallest gap $\to 0$ " IS uniformity (unpacking the definition)

WHY the tube picture: " $∣ f_{n} (x) - f (x) ∣ < ε$ for all $x$ " says the entire graph of $f_{n}$ lies inside the shaded tube $f (x) \pm ε$ . Uniform convergence = "eventually every curve is trapped in the tube." And a curve is trapped in the tube exactly when its tallest gap $M_{n}$ is below $ε$ . Same statement, two languages.

PICTURE: the limit curve with an $ε$ -tube; a late curve $f_{n}$ swimming safely inside it (uniform-friendly case), contrasted with a curve poking out near the right edge (our $x^{n}$ , which never fits).

Step 6 — The degenerate/edge case: shrink the domain and uniformity returns

WHY this case earns its own step: it proves uniformity is a property of the domain, not the formula. Nothing about $x^{n}$ changed; only where we look did.

PICTURE: the gap curve $x^{n}$ cut off at $x = a$ . Now the tallest bar sits at the right endpoint $x = a$ — and it is attained there, so $sup = max = a^{n}$ . As $n$ grows, $a^{n}$ (a fraction to a big power) collapses to $0$ . Ceiling drops; convergence is uniform.

Step 7 — A fully uniform example, so the tube snaps shut everywhere at once

WHY show a positive case: to see uniformity we need the ceiling to drop without any domain surgery — a clean $M_{n} \to 0$ over an infinite domain.

PICTURE: the flat tube $0 \pm \frac{1}{n}$ around the limit $f = 0$ . However wildly $sin (n x)$ oscillates, the whole curve is boxed between $\pm \frac{1}{n}$ . The box shrinks as $n$ grows — uniform on all of $R$ .

The one-picture summary

Recall Feynman retelling — the walkthrough in plain words

Picture a slowly-changing family of hills, one hill per step number $n$ , all trying to become the same "target" hill. Step 1–2: to check if they arrive, stand on one spot and watch the ground rise or fall there over the steps — that's checking one point. Step 3: measure how far off each hill is from the target at every spot — the "gap." Step 4: find the single worst gap anywhere, call its height $M_{n}$ ; use a supremum, not a max, because the worst spot might be a place you approach but never quite land on. Step 5: draw a fat ribbon (a tube) of thickness $ε$ around the target; the hill has "arrived uniformly" when it fits entirely inside the ribbon — which happens exactly when the worst gap $M_{n}$ is under $ε$ . Uniform convergence is just "eventually the whole hill fits in an ever-thinner ribbon," i.e. $M_{n} \to 0$ . Step 6: if one nasty corner keeps a hill sticking out, chop that corner off the map and the rest fits fine — uniformity depends on where you look. Step 7: the friendliest hills are the ones you can box in a shrinking ribbon of height $\frac{1}{n}$ no matter where you stand. That's the whole subject: watch the tallest gap; if it dies, everything nice comes along for free.

Connections

Pointwise convergence — the column-by-column check of Step 2.
Weierstrass M-test — how the ceiling $M_{n}$ idea powers series convergence.
Continuity preserved under uniform limits — why a curve trapped in a shrinking tube can't hide a jump.
Interchange of limit and integral — the tube's thickness $M_{n}$ bounds the area error.
Cauchy sequences in metric spaces — the tube language is the sup-norm distance.
Equicontinuity and Arzelà–Ascoli — the next chapter of curve-convergence.