Series — partial sums, convergence definition: Visual walkthrough

4.3.3 · D2 · Maths › Calculus III — Sequences & Series › Series — partial sums, convergence definition

Hum ek hi sawaal ka jawab denge, visually: jab tum $a + a r + a r^{2} + \dots$ forever add karte ho, toh pile kaahan jaake rukti hai — aur kab refuse karti hai settle hone se?

Step 1 — "Running total" kya hota hai? Dots draw karo.

WHAT. Ek list of numbers se shuru karo jinhe hum add karna chahte hain. List ko $(a_{n})$ bol lo: pehla number $a_{1}$ hai, doosra $a_{2}$ , aur aise aage. Neeche ka chhota $n$ bas ek counter hai — yeh kehta hai "list mein kaun sa number hai". Toh $a_{5}$ matlab hai "paanchwa number".

WHY. Hum kabhi physically infinitely many numbers nahi add kar sakte — addition kabhi bhi sirf do cheezein combine karti hai ek waqt mein. Toh forever add karne ke bajaye, hum ek finite chunk add karte hain aur dekhte hain ki chunk-total kya karta hai. Us finite total ka ek naam hai:

PICTURE. Neeche, har number $a_{n}$ ek coloured bar hai. Bar $N$ ke upar dot $s_{N}$ ki height par baithta hai = uss bar tak ke saare bars ka total. Jaise hum zyada bars include karte hain, dots ko climb karte dekho.

Convergence bas yahi hoga: kya woh climbing dots ek fixed height par level off karte hain?

Step 2 — Hamari specific list: geometric wali.

WHAT. Hum ek particular list choose karte hain jahan har number pichle wale ka ek fixed factor $r$ guna hota hai: $a_{1} = a, a_{2} = a r, a_{3} = a r^{2}, \dots, a_{n} = a r^{n - 1} .$ Yahan $a$ starting number hai aur $r$ common ratio hai — woh fixed multiplier jo tumhe ek term se agle tak le jaata hai.

WHY. Yeh woh ek series hai jo hum haath se poori tarah conquer kar sakte hain, aur chapter ki baaki sab cheezein issi se compare hoti hain (dekho Geometric series, Ratio & Root Tests, Comparison Test). Yeh samjho aur tum model ke maalik ban jaate ho.

PICTURE. $a = 1$ aur $r = \frac{1}{2}$ ke saath, har bar pichle wale ki aadhi height hai — ek shrinking staircase. Exponent $n - 1$ har bar par annotate hai taaki tum dekho kyun yeh $r^{0} = 1$ se shuru hota hai.

Step 3 — Multiply-and-slide trick.

WHAT. Poore partial sum ko $r$ se multiply karo. Har term ko $r$ ka ek extra factor mil jaata hai, yaani uska exponent ek badh jaata hai: $r s_{N} = was a a r + was a r a r^{2} + \dots + new last term a r^{N} .$

WHY. $s_{N}$ aur $r s_{N}$ dono lists dekho: yeh almost identical hain, bas ek slot shift hain. Jab do lists ek lamba middle share karti hain, toh unhe subtract karne se woh middle annihilate ho jaata hai. Yeh wahi annihilation idea hai jo Telescoping series ke peeche hai.

PICTURE. Top row = $s_{N}$ ke bars. Bottom row = wahi $r s_{N}$ ke bars, ek right shift hue. Overlapping (shared) bars grey hain; sirf do un-matched ends bachte hain.

Step 4 — Subtract karo, aur middle gayab ho jaata hai.

WHAT. $s_{N}$ ko $r s_{N}$ ke upar line up karo aur subtract karo: $s_{N} - r s_{N} = (a + a r + a r^{2} + \dots + a r^{N - 1}) - (a r + a r^{2} + \dots + a r^{N - 1} + a r^{N}) = a - a r^{N} .$ Har interior term ek baar $+$ ke saath aur ek baar $-$ ke saath aata hai: cancel ho jaata hai. Sirf top ki pehli term ( $a$ ) aur bottom ki last term ( $a r^{N}$ ) khadi rehti hain.

WHY. Humne $N$ pieces ki sum ko sirf do pieces ki sum mein badal diya. Yahi fayda hai — koi " $\dots$ " nahi bachta, toh hum actually compute kar sakte hain.

PICTURE. Cancelled bars fade out ho jaate hain; ek pale-yellow $+ a$ aur chalk-pink $- a r^{N}$ hi bachte hain.

Ab dono sides ko factor karo: $s_{N} (1 - r) = a (1 - r^{N}) .$ Left side par, $s_{N} - r s_{N} = s_{N} (1 - r)$ — common $s_{N}$ factor out karo. Right side par, $a$ factor out karo.

Step 5 — Running total ke liye solve karo.

WHAT. Provided $1 - r \neq = 0$ (yaani $r \neq = 1$ ), dono sides ko $(1 - r)$ se divide karo:

WHY. Ab humare paas ek formula hai $N$ mein jisme sirf ek moving part hai, $r^{N}$ . Convergence (Step 1 ka "kya dots level off karte hain?") ab ek sharp sawaal ban jaata hai: kya $r^{N}$ settle down karta hai jab $N \to \infty$ ? (Dekho Sequences — limits and convergence ki " $\to$ " ka kya matlab hai.)

PICTURE. Formula ek target height $\frac{a}{1 - r}$ (dashed line) ke roop mein draw hua minus ek shrinking correction $\frac{a r ^{N}}{1 - r}$ . Dots dashed line ki taraf approach karte hain jaise correction melt hoti jaati hai.

Symbol $lim_{N \to \infty}$ ka matlab bas: woh height jis taraf dots head karte hain jaise $N$ without bound badhta hai. Ab hum har possible $r$ ke liye split karte hain.

Step 6 — Case $0 < r < 1$ : pile climb karti hai aur settle ho jaati hai.

WHAT. Jab $r$ ek positive fraction ho ( $0$ aur $1$ ke beech), har term $a r^{n - 1}$ positive hoti hai, toh dots sirf climb hi karte hain. Aur $r^{N}$ steadily $0$ ki taraf march karta hai: $r^{N} \to 0 ⟹ s_{N} = \frac{a ( 1 - r ^{N} )}{1 - r} \to \frac{a ( 1 - 0 )}{1 - r} = \frac{a}{1 - r} .$

WHY. Har naya bar pichle wale ka ek shrunken positive copy hai, toh $N$ terms ke baad remaining tail neeche se nothing mein shrink ho jaati hai — correction $r^{N}$ die kar jaata hai aur dots ek definite ceiling ke against press ho jaate hain.

PICTURE. $a = 1, r = \frac{1}{2}$ ke liye dots $1, 1.5, 1.75, 1.875, \dots$ climb karte hain aur ceiling $\frac{1}{1 - \frac{1}{2}} = 2$ ke against neeche se press karte hain. Pale-yellow error band — kuch chhoti half-width ki strip jise hum tolerance $ε$ kahenge (ek chosen small positive number: "kitna close kaafi close hai?") — eventually har dot ko trap kar leta hai. Yeh $ε$ -band exactly wahi picture hai jo parent note ki convergence definition mein thi.

Step 7 — Case $- 1 < r < 0$ : pile phir bhi settle hoti hai, lekin oscillate karte hue.

WHAT. Ab $r$ ko ek negative fraction lo, jaise $r = - \frac{1}{2}$ . Uski size $∣ r ∣$ abhi bhi $1$ se kam hai, toh WAHI limit hold karti hai — $r^{N} \to 0$ aur $s_{N} \to \frac{a}{1 - r}$ . Lekin $r^{N}$ ki sign ab har step mein flip karti hai ( $r^{1} < 0, r^{2} > 0, r^{3} < 0, \dots$ ), toh correction $\frac{a r ^{N}}{1 - r}$ sign alternate karta hai.

WHY. Kyunki terms positive/negative alternate karte hain, dots ceiling overshoot karte hain, phir undershoot, phir kam overshoot — limit par zig-zagging karte hue neeche se climb karne ki jagah. Yeh phir bhi converge karta hai (swings nothing mein shrink ho jaati hain), bas dono sides se.

PICTURE. $a = 1, r = - \frac{1}{2}$ ke liye: dots $1, 0.5, 0.75, 0.625, \dots$ jaate hain ceiling $\frac{1}{1 - ( - \frac{1}{2} )} = \frac{2}{3}$ ke upar aur neeche bounce karte hue, aur bounces usi $ε$ -band mein shrink hote jaate hain. Yeh woh case hai jo " $∣ r ∣ < 1$ " ke andar chhupa tha aur Step 6 ne nahi dikhaya.

Step 8 — Case $∣ r ∣ > 1$ : pile explode ho jaati hai.

WHAT. Agar $r$ ki size $1$ se zyada ho, toh $r$ se multiply karna cheezein grow karta hai, toh $∣ r^{N} ∣ \to \infty$ . Correction term dominate karti hai aur $s_{N}$ ka koi finite ceiling nahi hota — yeh diverge karta hai.

WHY. Har bar pichle wale se bada hai, toh running total upar accelerate karta hai (ya $r < - 1$ ho toh wider aur wider swing karta hai). Level off hone ke liye kuch nahi hai.

PICTURE. $r = 2$ ke saath: bars $1, 2, 4, 8, \dots$ aur dots $1, 3, 7, 15, \dots$ top se shoot off karte hain — koi ceiling exist nahi karti. Ise directly Steps 6–7 se contrast karo.

Step 9 — Exact-boundary cases $r = 1$ aur $r = - 1$ .

WHAT. Yeh exactly edge $∣ r ∣ = 1$ par baithte hain, jahan hamara Step 5 division banned tha ( $r = 1$ ) ya jahan $r^{N}$ kabhi settle nahi karta ( $r = - 1$ ). Har ek ko haath se handle karo:

$r = 1$ : har term $a$ ke equal hai, toh $s_{N} = a + a + \dots + a = a N$ . Jaise $N$ badhta hai yeh $\pm \infty$ ki taraf march karta hai ( $a$ ki sign). Diverges. (Isliye Step 5 ko $r \neq = 1$ chahiye tha — $1 - r = 0$ se divide karna illegal hai.)
$r = - 1$ : terms hain $a, - a, a, - a, \dots$ , toh running total flip-flop karta hai $a, 0, a, 0, \dots$ Yeh bounded rehta hai lekin kabhi ek value nahi choose karta. Koi limit nahi ⇒ diverges.

WHY. "Diverges" ka matlab hamesha "infinity ki taraf shoot karna" nahi — $r = - 1$ case bounded hai phir bhi koi sum nahi hai. Ek sequence ko converge karne ke liye ek number par settle karna padta hai; forever bouncing karna fail karta hai.

PICTURE. Left: $r = 1$ dots ek straight ramp $a, 2 a, 3 a, \dots$ climb karte hain. Right: $r = - 1$ dots forever do heights ke beech ping-pong karte hain.

Ek picture mein poora summary

Sab kuch $r$ ke liye ek number line par: ek convergence window $(- 1, 1)$ (chalk-blue, woh akela zone jahan finite sum $\frac{a}{1 - r}$ milta hai) — right half $0 < r < 1$ par climbing-in, left half $- 1 < r < 0$ par oscillating-in — aur uske bahar, divergence: $∣ r ∣ > 1$ ke liye explosion, $r = 1$ par straight ramp, $r = - 1$ par flip-flop.

Recall Feynman retelling — plain words mein poora walkthrough

Mujhe $a$ ka total chahiye, phir $a$ times $r$ , phir woh times $r$ again, forever. Main forever add nahi kar sakta, toh main bas pehle $N$ add karta hoon aur ise $s_{N}$ kehta hoon — $N$ steps baad meri pile ki height. Clever trick: main pile ki ek copy banata hoon aur poori copy ko $r$ se multiply karta hoon, jo har block ko ek slot aage slide kar deta hai. Jab main original se copy subtract karta hoon, saare middle blocks match up ho jaate hain aur cancel — sirf pehla block $a$ aur ek leftover block $a r^{N}$ bachte hain. Isse mujhe ek clean formula milta hai $s_{N} = \frac{a ( 1 - r ^{N} )}{1 - r}$ sirf ek moving piece ke saath, $r^{N}$ . Ab main poochta hoon $r^{N}$ kya karta hai. Agar $r$ ek chhota positive fraction hai, toh dots upar climb karte hain aur $\frac{a}{1 - r}$ par settle ho jaate hain. Agar $r$ ek chhota negative fraction hai, toh woh phir bhi wahan settle karte hain lekin dono sides se zig-zag karte hue. Agar $r$ bada hai ( $∣ r ∣ > 1$ ), har block pichle wale se bada hota hai, pile explode kar jaati hai. Exactly edge par, $r = 1$ bas equal blocks ko infinity ki ramp mein stack karta hai, aur $r = - 1$ pile ko do heights ke beech forever flip karta hai, kabhi ek choose nahi karta. Toh pile ka real total sirf tab hota hai jab $r$ strictly $(- 1, 1)$ ke andar ho.

Active recall

Rebuild the closed form

s_{N}

r

se multiply karo, middle cancel karne ke liye subtract karo,

s_{N} (1 - r) = a (1 - r^{N})

milta hai,

1 - r

se divide karo.

Why does

∣ r ∣ < 1

give convergence

Tab

r^{N} \to 0

toh

s_{N} \to \frac{a}{1 - r}

; correction melt ho jaata hai.

How does

- 1 < r < 0

approach the limit

Yeh oscillate karta hai — dots

\frac{a}{1 - r}

ke upar aur neeche zig-zag karte hain shrinking swings ke saath.

Why exclude

r = 1

when dividing

Kyunki wahan

1 - r = 0

hai — division illegal hai; alag se

s_{N} = a N \to \pm \infty

What happens at

r = - 1

Partial sums flip-flop karte hain

a, 0, a, 0, \dots

— bounded lekin koi limit nahi, toh diverge karta hai.

Connections

4.3.03 Series — partial sums, convergence definition (Hinglish) — parent topic
Geometric series — yahan poori tarah derive hua result
Sequences — limits and convergence — " $s_{N} \to S$ " ka kya matlab hai
Telescoping series — Steps 3–4 wali same "middle cancels" idea
n-th Term Test for Divergence — $r = \pm 1, ∣ r ∣ > 1$ cases yeh fail karte hain
Ratio & Root Tests — ek general series ka " $r$ " measure karo
Power series and radius of convergence — $r = x$ ke saath geometric series