Visual walkthrough — Root test

Before any symbol appears, let us agree on three plain words.

That is all the vocabulary we need to start. Everything else we earn.

Recall Reminders on the tools we'll lean on

These appear below; here is the one-line meaning of each, for a ground-up read.

• $\underset{n\to \infty }{\lim }{b}_n=L$ ("the limit is $L$") means: the numbers $b_n$ get and stay as close to $L$ as you like, once $n$ is big enough.
• $\mathrm{limsup}$ ("limit superior") is a safety-net version of the limit that always exists — it tracks the highest value the sequence keeps returning to, even if the ordinary limit wobbles.
• $\ln x$ ("natural logarithm") answers "what power of $e$ gives $x$?" Key fact we use: $\frac{\ln n}{n}\to 0$ (the top grows far slower than the bottom).
• $e^x$ is the exponential function; here we only need that it is continuous and $e^0=1$.

---

Step 1 — The gold-standard shrinker: a geometric series

WHAT. Take a fixed number $r$ (a "ratio") and build the list where each term is the previous one multiplied by $r$: $\underset{a_0}{\underbrace{1}},\underset{a_1}{\underbrace{r}},\underset{a_2}{\underbrace{r^2}},\underset{a_3}{\underbrace{r^3}},\dots a_n=r^n.$ Here $r^n$ means "$r$ multiplied by itself $n$ times." This is a Geometric series.

WHY. This is the simplest possible way for terms to shrink at a steady rate, and we know exactly when its total is finite: ${\sum }_{n=0}^{\infty }{r}^n=\frac{1}{1-r}\text{converges}⟺|r|<1.$ The symbol $|r|$ ("size of $r$", ignoring its sign) $<1$ says: each term is a genuine fraction of the one before, so the pile of terms stops growing. If $|r|\ge 1$ each term is as big or bigger than the last, and the total explodes.

PICTURE. Look at the blue bars: when $r=0.6$ each bar is $60\%$ of its left neighbour, and they collapse to nothing — the running total (yellow) flattens. The pink bars with $r=1.2$ grow taller and taller — the total races upward.

---

Step 2 — The clever question: "what hidden ratio does $a_n$ hide?"

WHAT. We look at a series whose terms behave like a geometric series far down the list. We do not need exact equality — only that the $n$-th root of the term settles toward a single number. Precisely, we ask for $\sqrt[n]{|a_n|}⟶r(\text{a fixed number}),$ which is the honest, limit-based meaning of the loose phrase "$a_n$ acts like $r^n$." We want to dig $r$ out of $a_n$ without knowing it in advance.

WHY this tool — the $n$-th root. If $a_n=r^n$, how do we undo the "raise-to-the-$n$"? We apply its inverse, the $n$-th root, written $\sqrt[n]{}$. It answers the question "which number, multiplied by itself $n$ times, gives this?" Applied to $r^n$ it hands back exactly $r$: $\sqrt[n]{r^n}=r.$ That is the entire reason the root test uses a root and not, say, a logarithm or a ratio — a root is the precise key that unlocks an $n$-th power.

PICTURE. The green curve is $a_n=r^n$ dropping toward zero. The straight yellow line is what $\sqrt[n]{a_n}$ gives back: a flat, constant $r$. The root "flattens" the shrinking curve into the single number that generated it.

(The parent uses $\mathrm{limsup}$ — a version of the limit that always exists even when the ordinary limit wobbles; see the reminder box. For every example here the ordinary $\lim$ exists, so read $L$ as an ordinary limit.)

---

Step 3 — Case $L<1$: trap the terms under a convergent geometric series

WHAT. Assume $L<1$. We choose a helper number $r$ strictly between $L$ and $1$: $L<r<1.$

WHY. We want a fixed ratio $r$ with two jobs: (1) it must sit above $L$ so the terms eventually fall below it, and (2) it must stay below $1$ so that $\sum r^n$ still converges (Step 1). Because $L<1$, there is empty room between $L$ and $1$ — so such an $r$ always exists. (It doesn't matter which one; any works.)

PICTURE. The number line: $L$ sits on the left, $1$ on the right, and we drop $r$ into the gap between them. The yellow band $[L,1)$ is the "room" that $L<1$ guarantees.

---

Step 4 — Why "eventually" turns into a real inequality

WHAT. Since $\sqrt[n]{|a_n|}\to L$ and $L<r$, the roots eventually slip below $r$: there is a cutoff index $N$ so that $\sqrt[n]{|a_n|}<r\text{for every }n\ge N.$ Raise both sides to the $n$-th power (both sides positive, so the inequality direction is safe): $|a_n|<r^n(n\ge N).$

WHY. "$\to L$" means the roots get and stay arbitrarily close to $L$ (see the reminder box). Since $r$ is a fixed gap above $L$, from some point $N$ onward the roots can never climb back above $r$. Un-rooting both sides converts a fact about roots into a fact about the terms themselves: past $N$, every term is smaller than the matching geometric term.

PICTURE. The blue root-dots march down toward $L$ and cross below the dashed line $y=r$ at $n=N$. From there on, the pink term-bars $|a_n|$ tuck entirely under the green geometric bars $r^n$.

Now the Comparison test finishes it: ${\sum }_{n\ge N}{r}^n$ converges (it's geometric with $0\le r<1$), and $|a_n|$ sits under it term by term, so $\sum |a_n|$ converges — the series converges absolutely. $■$

---

Step 5 — Case $L>1$: the terms refuse to die, so the sum can't settle

WHAT. Assume $L>1$. We run the mirror image of Step 3–4. Pick a helper number $r$ strictly between $1$ and $L$: $1<r<L.$ Because $L>1$, there is room below $L$ and above $1$, so such an $r$ exists.

WHY (the precise limit argument). Since $\sqrt[n]{|a_n|}\to L$ and $r<L$, the roots eventually rise above $r$: there is a cutoff $N$ with $\sqrt[n]{|a_n|}>r>1\text{for all }n\ge N.$ Un-root both sides (raise to the $n$-th power, both sides positive): $|a_n|>r^n>1(n\ge N).$ So past $N$ every term exceeds $1$ — in particular the terms cannot shrink to zero: $a_n\nrightarrow 0.$ The n-th term divergence test says a convergent series must have $a_n\to 0$. Ours doesn't, so it diverges. $■$

(The parent's shorter phrasing "$|a_n|>1$ infinitely often" is the $\mathrm{limsup}$ version; with an ordinary limit $>1$ we get the cleaner "for all $n\ge N$" above.)

PICTURE. The blue root-dots climb above the dashed line $y=r>1$ at $n=N$; from there the pink term-bars $|a_n|$ stay above $1$ forever, so the running total (yellow) keeps stepping upward without bound.

---

Step 6 — The degenerate case $L=1$: the trick goes blind

WHAT. When $L=1$ exactly, there is no room to pick an $r$ between $L$ and $1$ (Step 3 collapses), and the terms may or may not go to zero (Step 5 fails too). The test says nothing.

WHY it's genuinely silent. Two famous series both give $L=1$ yet behave oppositely — see p-series: $\sqrt[n]{\frac{1}{n^p}}=n^{-p/n}=e^{-\frac{p\ln n}{n}}⟶e^0=1,$ because $\frac{\ln n}{n}\to 0$ (the reminder box). Each symbol: $n^{-p/n}$ rewrites the root as a power; $e^{()}$ turns that power into an exponential whose exponent $-\frac{p\ln n}{n}$ vanishes — for any $p$. So the root can't feel the difference between the diverging $\sum \frac{1}{n}$ ($p=1$) and the converging $\sum \frac{1}{n^2}$ ($p=2$).

PICTURE. Both root-curves (one for $p=1$, one for $p=2$) crush down onto the same line $y=1$ — the root test literally cannot separate them. The verdict must come from a different tool.

---

The one-picture summary

Everything above is one decision diagram. Compute $L=\lim \sqrt[n]{|a_n|}$, then locate it on the number line:

• Left of $1$ ($L<1$): trap under $\sum r^n$ ⇒ converges.
• Right of $1$ ($L>1$): terms don't reach $0$ ⇒ diverges.
• Exactly $1$: blind spot ⇒ inconclusive.

Recall Feynman retelling — the whole walkthrough in plain words

Picture a ball bouncing. After each bounce it keeps some fraction of its last height. If that fraction is under one, the bounces shrink and the ball travels a finite total distance — it settles. If the fraction is over one, the bounces grow forever. Now suppose you're only shown the heights $a_n$ and asked "what fraction is hiding in here?" If the heights are that fraction multiplied by itself $n$ times ($r^n$), you undo the multiplying by taking the $n$-th root — out pops the fraction $r$, which we call $L$. If $L<1$ you can always slide a slightly bigger fraction $r$ (still under one) over your heights, cap them all under a geometric pile you know is finite, and conclude: finite. If $L>1$ you can slide a fraction $r$ just above one under your heights, so the heights never fall to zero and the sum can't possibly settle: infinite. And if $L$ lands exactly on one, the root has flattened away every clue — a fast-shrinking series and a barely-shrinking one look identical to it — so you must grab a different tool entirely.

Recall Quick self-check

Root of $r^n$ gives back what? ::: $r$ — the $n$-th root undoes the $n$-th power exactly. In case $L<1$, why must a good $r$ exist? ::: Because $L<1$ leaves a gap $(L,1)$, and any $r$ in it sits above $L$ (terms fall below it) yet below $1$ (its geometric series converges). In case $L>1$, which helper $r$ and which test finish the job? ::: Pick $1<r<L$; eventually $|a_n|>r^n>1$ so $a_n\nrightarrow 0$, and the n-th term divergence test forces divergence. Why is $L=1$ inconclusive? ::: $\sqrt[n]{1/n^p}=e^{-p\ln n/n}\to 1$ for every $p$, so the root can't distinguish converging from diverging p-series.