4.3.13Calculus III — Sequences & Series

Root test

1,660 words8 min readdifficulty · medium2 backlinks

WHAT it says

We use lim sup\limsup so the test always has a value, but for most exam problems the ordinary limit limann\lim \sqrt[n]{|a_n|} exists and you use that.


WHY it works (derivation from scratch)

We build it entirely from the geometric series rn\sum r^n, which converges iff r<1|r|<1.


Figure — Root test

HOW to use it (worked examples)


Steel-manned mistakes


When to reach for it (80/20)


Root test limit formula
L=lim supnannL=\limsup_{n\to\infty}\sqrt[n]{|a_n|}
Conclusion when L<1L<1
Series converges absolutely.
Conclusion when L>1L>1 or L=L=\infty
Series diverges (terms don't go to 0).
Conclusion when L=1L=1
Inconclusive — use another test.
Why does root test work for L<1L<1?
Pick rr with L<r<1L<r<1; eventually an<rn|a_n|<r^n, and rn\sum r^n converges, so comparison gives convergence.
Why does root test work for L>1L>1?
an>1|a_n|>1 infinitely often, so an↛0a_n\not\to0; the nn-th term test forces divergence.
What is limnnn\lim_{n\to\infty}\sqrt[n]{n}?
11 (since n1/n=elnn/ne0=1n^{1/n}=e^{\ln n/n}\to e^0=1).
When should you prefer the root test?
When the general term contains an nn-th power ()n(\cdots)^n or basen^n.
For (2n+1)n/(3n+5)n\sum (2n+1)^n/(3n+5)^n, find LL
L=2/3<1L=2/3<1, converges.
Why is L=1L=1 for every 1/np\sum 1/n^p?
1/npn=eplnn/n1\sqrt[n]{1/n^p}=e^{-p\ln n/n}\to 1 as lnn/n0\ln n/n\to0.

Recall Feynman: explain to a 12-year-old

Imagine each term in the sum is like the height you fall after each bounce of a ball. If every bounce keeps a fixed fraction rr of the previous height (r<1r<1), the total distance is finite — the ball stops. The root test is a clever trick to find that fraction: if a term looks like rr multiplied by itself nn times (rnr^n), you "undo" the multiplying by taking the nn-th root, and out pops rr. If r<1r<1 the bounces die out (sum is finite = converges). If r>1r>1 the bounces grow forever (diverges). If rr comes out exactly 11, the trick can't tell — you need another tool.

Concept Map

converges iff

extract ratio r

built from

L < 1

L > 1

L = 1

proof via

needs L < r < 1

proof via

example

best when

Geometric series sum r^n

|r| < 1

Root test idea

L = limsup nth-root |a_n|

Converges absolutely

Diverges

Inconclusive

Comparison with r^n

nth-term test a_n not to 0

sum 1/n^p all give L=1

Terms of form c_n^n

Hinglish (regional understanding)

Intuition Hinglish mein samjho

Dekho, root test ka basic idea bohot simple hai. Hum jaante hain ki geometric series rn\sum r^n tabhi converge karti hai jab r<1|r|<1. Ab kisi bhi series an\sum a_n ke baare mein hum poochhte hain: "Kya iske terms eventually rnr^n jaise behave karte hain?" Agar haan, to us chhupe hue rr ko nikaalne ke liye hum nn-th root le lete hain — kyunki rnn=r\sqrt[n]{r^n}=r. Isliye L=limannL=\lim \sqrt[n]{|a_n|} nikaalo: yeh limit hi effective ratio hai.

Rule yaad rakho: L<1L<1 matlab terms tezi se chhote ho rahe hain, series converge. L>1L>1 matlab terms bade ho rahe hain, ana_n zero tak nahi jaa raha, isliye diverge. Aur L=1L=1 matlab test "Lost" — koi conclusion nahi, dusra test (p-series, comparison, integral) lagao. Proof bhi seedha hai: L<1L<1 ho to ek rr choose karo jo LL aur 11 ke beech ho, fir eventually an<rn|a_n|<r^n, aur rn\sum r^n converge karta hai — comparison se kaam khatam.

Root test kab lagana hai? Jab term mein nn-th power dikhe — jaise (kuch)n(\text{kuch})^n ya cnc^n. Tab n\sqrt[n]{\cdot} power ko cleanly cancel kar deta hai. Ek important trap: polynomial factors jaise n2n^2 ka nn-th root n2/n1n^{2/n}\to 1 hota hai, 00 nahi! Sirf exponential ka base bachta hai. Isliye n2/2n\sum n^2/2^n ke liye L=1/2L=1/2, converge karta hai. Bas yahi 20% concept 80% problems solve kar dega.

Go deeper — visual, from zero

Test yourself — Calculus III — Sequences & Series

Connections