Visual walkthrough — Epsilon-delta definition of a limit — formal proofs

We will take one concrete limit and squeeze the definition out of it, picture by picture: ${\lim }_{x\to 2}{x}^2=4.$ Everything else (linear case, false limits) is a shadow of this one worked example.

Step 1 — Two number lines, two dots

WHY. A limit is a promise about two places at once: "if the input is near here, the output is near there." You cannot reason about "near" until you can see the two neighbourhoods living on two different rulers. One ruler is not enough — the whole drama is the link between them.

PICTURE. Look at the figure. The bottom violet line is the input; the dot at $x=2$ is where we aim. The top magenta line is the output; the dot at $y=4$ is where we want to land. The orange curve $y=x^2$ is the rule $f$ that carries a point from the bottom line up to the top line.

• $x$ — a chosen input, a point free to slide along the bottom ruler.
• $a=2$ — the input we are approaching (a dot, not a value we plug in).
• $L=4$ — the output we hope to approach.
• The orange arrow shows the rule $f(x)=x^2$: it takes $x=2$ up to $y=4$.

Step 2 — The adversary draws the output band ($\epsilon$)

WHY this symbol, and why first? "Close to $L$" is meaningless until someone names how close. The number $\epsilon$ is that "how close" — it is a measured tolerance, not a mood. And it comes first because the whole point is that you don't get to pick the tolerance — the challenger does, and can make it as cruelly small as they like.

PICTURE. The magenta band around $y=4$ has half-height $\epsilon$. Anything landing inside the shaded strip counts as a win on the output side.

Step 3 — Pull the band back through the curve

WHY. The definition says nothing about wandering along the curve — it only knows inputs and outputs. So we ask: which inputs produce an output inside the band? Geometrically that is exactly "shine the band sideways onto the curve, then read off the shadow on the input ruler." That shadow is the set of inputs guaranteed to win.

PICTURE. The two magenta edges meet the orange curve at two points; dashed navy lines drop to the input ruler, cutting out a little interval. Notice it is not symmetric about $x=2$ — the left gap and right gap differ. That asymmetry is the seed of the word "min" coming in Step 6.

• Left edge of shadow: solves $x^2=4-\epsilon$, i.e. $x=\sqrt{4-\epsilon }$.
• Right edge of shadow: solves $x^2=4+\epsilon$, i.e. $x=\sqrt{4+\epsilon }$.
• Which root? The equation $x^2=4\pm \epsilon$ has two solutions, $\pm \sqrt{4\pm \epsilon }$. We keep only the positive root because we are working near $x=2>0$; the negative root sits over near $x=-2$, far from our target, so it is irrelevant here.
• Because the curve is steeper on the right, the right gap is a little shorter.

Step 4 — Fit the biggest symmetric band that fits ($\delta$)

WHY a symmetric $\delta$-band, when the shadow is lopsided? The definition is written as $0<|x-a|<\delta$ — one single distance $\delta$ in every direction. That is a deliberate simplification: it is easier to state and to reuse. The price is that we must shrink to the narrower side of the shadow so the symmetric band never pokes out. That "shrink to the narrower side" is precisely the $\min$ you will meet in Step 6.

PICTURE. The violet band of half-width $\delta$ sits inside the shadow. The little hollow circle at $x=2$ is the excluded point — we never actually stand on the target.

Step 5 — Why "factor out $|x-a|$" is the honest bookkeeping

WHY this and not brute square-roots? We need a rule that turns "input is within $\delta$" into "output is within $\epsilon$" for every $x$ at once, cleanly. Factoring exposes the machine: the output distance is the input distance multiplied by a stretch factor $|x+2|$. Control the input distance and control the stretch, and the output distance obeys automatically.

PICTURE. The figure shows the curve near $x=2$ and the local "stretch": moving the input a tiny step $|x-2|$ moves the output a step about $|x+2|$ times as big. The steeper the curve, the bigger the stretch — that is what $|x+2|$ measures.

• $|x-2|$ — how far your input strayed from the target; the thing you directly control with $\delta$.
• $|x+2|$ — the amplification: near $x=2$ it is about $4$, so a wobble of size $d$ in $x$ becomes a wobble of about $4d$ in $y$.

Step 6 — Tame the stretch, then choose $\delta =\min$

WHY the $\min$, seen in one glance. Two constraints, two candidate widths. $\delta$ must satisfy both, so it must be no bigger than either — that is exactly what "minimum" means. This is the symmetric-band-inside-the-lopsided-shadow idea from Steps 3–4, now in algebra.

PICTURE. The figure stacks the two candidate half-widths ($1$ and $\epsilon /5$) as two bars; the shorter bar wins and becomes $\delta$. For small $\epsilon$ the $\epsilon /5$ bar is shorter (the $\epsilon$-condition bites); for large $\epsilon$ the constant $1$ bar wins (the cage bites).

Step 7 — The degenerate & edge cases (nothing left unshown)

Case A — huge $\epsilon$ (loose demand). If $\epsilon =100$, then $\epsilon /5=20$, and $\min (1,20)=1$. The cage wins; $\delta =1$. Why fine: a loose demand can't force a tiny $\delta$ — you keep your comfortable margin.

Case B — tiny $\epsilon$ (brutal demand). If $\epsilon =0.05$, then $\epsilon /5=0.01$, and $\min (1,0.01)=0.01$. The $\epsilon$-rule wins; $\delta =0.01$. Why fine: as the output band shrinks, your input band shrinks in step — and it always can, because $\epsilon /5>0$ for every $\epsilon >0$.

Case C — the linear function, where the stretch is constant. For ${\lim }_{x\to 3}(2x+1)=7$ we get $|f(x)-L|=2|x-3|$: the stretch is the constant $2$, no cage needed, and $\delta =\epsilon /2$ exactly. This is why straight lines are the easy case — no $x$ hides inside the stretch, so no $\min$.

PICTURE. Left panel: the parabola's shadow is lopsided (stretch varies) → $\min$ needed. Right panel: the line's shadow is perfectly symmetric (stretch constant) → clean $\delta =\epsilon /2$.

The formal statement, now fully earned

The one-picture summary

Active recall

Connections

• Epsilon-delta definition of a limit — formal proofs — the parent: definition, recipe, and the algebra behind these pictures.
• Definition of the derivative — the derivative is the limit of the stretch factor $|x+2|$-style slopes.
• Continuity — same two-band picture with $L=f(a)$: the shadow now includes the target point.
• One-sided limits — use only the left or right half of the shadow.
• Limit Laws (sum, product, quotient) — each law reuses the "factor and bound" move from Step 5.
• Uniform continuity — demand one $\delta$-band that fits the shadow at every $a$ at once.

Concept Map