Visual walkthrough — Root test

4.3.13 · D2 · Maths › Calculus III — Sequences & Series › Root test

Koi bhi symbol aane se pehle, teen simple words par agree karte hain.

Shuru karne ke liye bas itni hi vocabulary chahiye. Baaki sab hum khud kamaate hain.

Recall Jin tools pe hum lean karenge, unke reminders

Yeh neeche aate hain; yahan har ek ka ek-line meaning hai, ground-up read ke liye.

• $\underset{n\to \infty }{\lim }{b}_n=L$ ("limit $L$ hai") ka matlab hai: numbers $b_n$ $L$ ke utne close aa jaate hain aur rehte hain jitna chahte ho, ek baar $n$ itna bada ho jaaye.
• $\mathrm{limsup}$ ("limit superior") limit ka ek safety-net version hai jo hamesha exist karta hai — yeh track karta hai sequence ki sabse high value jo woh return karta rehta hai, chahe ordinary limit wobble kare.
• $\ln x$ ("natural logarithm") answer karta hai "kis power of $e$ se $x$ milta hai?" Ek key fact jo hum use karte hain: $\frac{\ln n}{n}\to 0$ (upar wala bottom ki comparison mein bahut slow grow karta hai).
• $e^x$ exponential function hai; yahan hume sirf itna chahiye ki yeh continuous hai aur $e^0=1$.

---

Step 1 — Gold-standard shrinker: ek geometric series

KYA. Ek fixed number $r$ lo (ek "ratio") aur woh list banao jahan har term pichhli term ko $r$ se multiply karke milti hai: $\underset{a_0}{\underbrace{1}},\underset{a_1}{\underbrace{r}},\underset{a_2}{\underbrace{r^2}},\underset{a_3}{\underbrace{r^3}},\dots a_n=r^n.$ Yahan $r^n$ ka matlab hai "$r$ ko khud se $n$ baar multiply karo." Yeh ek Geometric series hai.

KYUN. Yeh terms ke shrinking ka sabse simple possible tarika hai ek steady rate par, aur hum jaante hain exactly kab iska total finite hai: ${\sum }_{n=0}^{\infty }{r}^n=\frac{1}{1-r}\text{converges}⟺|r|<1.$ Symbol $|r|$ ("$r$ ki size", sign ignore karke) $<1$ kehta hai: har term genuinely pichhle ki ek fraction hai, toh terms ka pile badhna ruk jaata hai. Agar $|r|\ge 1$ toh har term utni badi ya isse badi hai jitni pichhli, aur total blast ho jaata hai.

PICTURE. Blue bars dekho: jab $r=0.6$ toh har bar apne left neighbour ka $60\%$ hai, aur woh collapse ho kar kuch nahi reh jaate — running total (yellow) flat ho jaata hai. Pink bars $r=1.2$ ke saath aur aur lamba hote jaate hain — total upar ki taraf race karta hai.

---

Step 2 — Clever question: "$a_n$ ke andar kaun sa hidden ratio chhupa hai?"

KYA. Hum ek aisi series dekhte hain jinke terms list mein kaafi neeche jaane par ek geometric series jaisi behave karti hain. Hume exact equality nahi chahiye — sirf itna ki term ka $n$-th root ek single number ki taraf settle ho. Precisely, hum poochh rahe hain ki $\sqrt[n]{|a_n|}⟶r(\text{ek fixed number}),$ jo ki loose phrase "$a_n$, $r^n$ jaisa behave karta hai" ka honest, limit-based meaning hai. Hum $r$ ko $a_n$ se nikalna chahte hain bina pehle se jaane.

YEH TOOL KYUN — $n$-th root. Agar $a_n=r^n$ hai, toh "raise-to-the-$n$" ko hum undo kaise karein? Hum iska inverse apply karte hain, $n$-th root, jo $\sqrt[n]{}$ likha jaata hai. Yeh poochhta hai "woh kaun sa number hai jo khud se $n$ baar multiply hone par yeh deta hai?" $r^n$ pe apply karne par yeh exactly $r$ wapas deta hai: $\sqrt[n]{r^n}=r.$ Yahi poori wajah hai ki root test ek root use karta hai, na ki, koi logarithm ya ratio — ek root woh precise key hai jo ek $n$-th power ko unlock karti hai.

PICTURE. Green curve $a_n=r^n$ hai jo zero ki taraf gir rahi hai. Straight yellow line woh hai jo $\sqrt[n]{a_n}$ wapas deta hai: ek flat, constant $r$. Root us shrinking curve ko us single number mein "flatten" kar deta hai jisne use generate kiya tha.

(Parent $\mathrm{limsup}$ use karta hai — limit ka ek version jo hamesha exist karta hai chahe ordinary limit wobble kare; reminder box dekho. Yahan har example mein ordinary $\lim$ exist karta hai, toh $L$ ko ordinary limit ke roop mein padho.)

---

Step 3 — Case $L<1$: terms ko ek convergent geometric series ke neeche trap karo

KYA. Maano $L<1$ hai. Hum ek helper number $r$ choose karte hain jo strictly $L$ aur $1$ ke beech ho: $L<r<1.$

KYUN. Hume ek fixed ratio $r$ chahiye jo do kaam kare: (1) yeh $L$ se upar baithe taki terms eventually iske neeche aa jaayein, aur (2) yeh $1$ se neeche rahe taki $\sum r^n$ abhi bhi converge kare (Step 1). Kyunki $L<1$ hai, $L$ aur $1$ ke beech empty room hai — toh aisa $r$ hamesha exist karta hai. (Koi bhi kaam karega; yeh matter nahi karta kaunsa.)

PICTURE. Number line: $L$ left pe baitha hai, $1$ right pe, aur hum $r$ ko unke beech ke gap mein daalte hain. Yellow band $[L,1)$ woh "room" hai jo $L<1$ guarantee karta hai.

---

Step 4 — "Eventually" ek real inequality mein kyun badalta hai

KYA. Kyunki $\sqrt[n]{|a_n|}\to L$ aur $L<r$ hai, roots eventually $r$ ke neeche slip kar jaate hain: ek cutoff index $N$ hota hai aisa ki $\sqrt[n]{|a_n|}<r\text{har }n\ge N\text{ ke liye.}$ Dono sides ko $n$-th power mein raise karo (dono sides positive hain, toh inequality direction safe hai): $|a_n|<r^n(n\ge N).$

KYUN. "$\to L$" ka matlab hai roots $L$ ke arbitrarily close aa jaate hain aur rehte hain (reminder box dekho). Kyunki $r$ $L$ ke upar ek fixed gap hai, kisi point $N$ ke baad roots kabhi $r$ se upar nahi jaate. Dono sides ko un-rooting karna roots ke baare mein ek fact ko terms ke baare mein ek fact mein convert karta hai: $N$ ke baad, har term matching geometric term se chhoti hai.

PICTURE. Blue root-dots $L$ ki taraf march karte hain aur $n=N$ par dashed line $y=r$ ke neeche cross kar jaate hain. Wahan se, pink term-bars $|a_n|$ green geometric bars $r^n$ ke andar completely aa jaate hain.

Ab Comparison test isko finish karta hai: ${\sum }_{n\ge N}{r}^n$ converge karta hai (yeh geometric hai with $0\le r<1$), aur $|a_n|$ term by term iske neeche baitha hai, toh $\sum |a_n|$ converge karta hai — series absolutely converge karti hai. $■$

---

Step 5 — Case $L>1$: terms marne se refuse karti hain, toh sum settle nahi ho sakta

KYA. Maano $L>1$ hai. Hum Step 3–4 ka mirror image chalate hain. Ek helper number $r$ choose karo jo strictly $1$ aur $L$ ke beech ho: $1<r<L.$ Kyunki $L>1$ hai, $L$ ke neeche aur $1$ ke upar room hai, toh aisa $r$ exist karta hai.

KYUN (precise limit argument). Kyunki $\sqrt[n]{|a_n|}\to L$ aur $r<L$ hai, roots eventually $r$ se upar rise kar jaate hain: ek cutoff $N$ hota hai aisa ki $\sqrt[n]{|a_n|}>r>1\text{sabhi }n\ge N\text{ ke liye.}$ Dono sides ko un-root karo (raise to the $n$-th power, dono sides positive): $|a_n|>r^n>1(n\ge N).$ Toh $N$ ke baad har term $1$ se zyada hai — particularly terms nahi shrink kar sakti zero ki taraf: $a_n\nrightarrow 0.$ n-th term divergence test kehta hai ki ek convergent series mein zaroor $a_n\to 0$ hona chahiye. Hamari mein nahi hota, toh yeh diverge karti hai. $■$

(Parent ki chhoti wording "$|a_n|>1$ infinitely often" $\mathrm{limsup}$ version hai; ordinary limit $>1$ ke saath hume upar wala cleaner "for all $n\ge N$" milta hai.)

PICTURE. Blue root-dots $n=N$ par dashed line $y=r>1$ ke upar chadh jaate hain; wahan se pink term-bars $|a_n|$ hamesha ke liye $1$ se upar rehte hain, toh running total (yellow) bina bound ke upar step karta rehta hai.

---

Step 6 — Degenerate case $L=1$: trick blind ho jaati hai

KYA. Jab $L=1$ exactly hota hai, toh $L$ aur $1$ ke beech $r$ choose karne ki koi room nahi hoti (Step 3 collapse ho jaata hai), aur terms zero ki taraf jaaye ya na jaaye bhi ho sakta hai (Step 5 bhi fail ho jaata hai). Test kuch nahi kehta.

KYUN yeh genuinely silent hai. Do famous series dono $L=1$ dete hain phir bhi ulta behave karti hain — p-series dekho: $\sqrt[n]{\frac{1}{n^p}}=n^{-p/n}=e^{-\frac{p\ln n}{n}}⟶e^0=1,$ kyunki $\frac{\ln n}{n}\to 0$ (reminder box). Har symbol: $n^{-p/n}$ root ko ek power ke roop mein rewrite karta hai; $e^{()}$ us power ko ek exponential mein badalta hai jiska exponent $-\frac{p\ln n}{n}$ vanish ho jaata hai — kisi bhi $p$ ke liye. Toh root diverging $\sum \frac{1}{n}$ ($p=1$) aur converging $\sum \frac{1}{n^2}$ ($p=2$) ke beech ka fark feel nahi kar sakta.

PICTURE. Dono root-curves (ek $p=1$ ke liye, ek $p=2$ ke liye) same line $y=1$ par crush ho jaate hain — root test literally inhe separate nahi kar sakta. Verdict ek alag tool se aana chahiye.

---

Ek-picture summary

Upar sab kuch ek decision diagram hai. $L=\lim \sqrt[n]{|a_n|}$ compute karo, phir ise number line par locate karo:

• $1$ ke left mein ($L<1$): $\sum r^n$ ke neeche trap karo ⇒ converges.
• $1$ ke right mein ($L>1$): terms $0$ tak nahi pahunchti ⇒ diverges.
• Exactly $1$: blind spot ⇒ inconclusive.

Recall Feynman retelling — simple words mein poora walkthrough

Ek bouncing ball ki picture karo. Har bounce ke baad yeh apni last height ka kuch fraction rakhti hai. Agar woh fraction ek se kam hai, toh bounces shrink karti hain aur ball ek finite total distance travel karti hai — yeh settle ho jaati hai. Agar fraction ek se zyada hai, toh bounces hamesha ke liye badhti rehti hain. Ab maano tumhe sirf heights $a_n$ dikhayi deti hain aur pucha jaata hai "isme kaun sa fraction chhupa hai?" Agar heights woh fraction hain jo khud se $n$ baar multiply hua ($r^n$), toh tum multiplying ko $n$-th root lekar undo karte ho — fraction $r$ nikal aata hai, jise hum $L$ kehte hain. Agar $L<1$ hai toh tum hamesha ek thoda bada fraction $r$ (abhi bhi ek se kam) apni heights ke upar slide kar sakte ho, inhe sab ek geometric pile ke neeche cap kar sakte ho jo tum jaante ho finite hai, aur conclude kar sakte ho: finite. Agar $L>1$ hai toh tum ek fraction $r$ thoda ek se upar apni heights ke neeche slide kar sakte ho, toh heights kabhi zero nahi girti aur sum kabhi settle nahi ho sakta: infinite. Aur agar $L$ exactly ek pe land kare, toh root ne har clue flat kar diya hai — ek fast-shrinking series aur ek barely-shrinking series iske liye ek jaisi dikhti hain — toh tumhe bilkul alag tool lena padega.

Recall Quick self-check

$r^n$ ka root kya wapas deta hai? ::: $r$ — $n$-th root $n$-th power ko exactly undo karta hai. Case $L<1$ mein, ek accha $r$ kyun zaroor exist karta hai? ::: Kyunki $L<1$ ek gap $(L,1)$ chhodta hai, aur isme koi bhi $r$ $L$ se upar baithta hai (terms iske neeche aa jaati hain) phir bhi $1$ se neeche rehta hai (iska geometric series converge karta hai). Case $L>1$ mein, kaun sa helper $r$ aur kaun sa test kaam khatam karta hai? ::: $1<r<L$ choose karo; eventually $|a_n|>r^n>1$ toh $a_n\nrightarrow 0$, aur n-th term divergence test divergence force karta hai. $L=1$ inconclusive kyun hai? ::: $\sqrt[n]{1/n^p}=e^{-p\ln n/n}\to 1$ har $p$ ke liye, toh root converging aur diverging p-series mein fark nahi kar sakta.