Real analysis — rigorous epsilon-delta, metric spaces: Visual walkthrough

4.10.22 · D2 · Maths › Advanced Topics (Elite Level) › Real analysis — rigorous epsilon-delta, metric spaces

Hum sirf yeh assume karte hain ki tum ek graph padh sakte ho: ek horizontal axis (input $x$ ) aur ek vertical axis (output $y$ ). Baaki sab kuch neeche earn kiya jayega.

Step 1 — Woh vague word jo hume khatam karna hai: "approaches"

KYA. Humare paas ek rule hai jo ek number $x$ ko ek number $y$ mein badalta hai. Hum likhte hain $y = f (x)$ ; symbol $f$ sirf rule ka naam hai, aur $f (x)$ ka matlab hai "rule ka output jab tum use $x$ feed karo." Is poore page ka rule hoga $f (x) = x^{2}$ (input ko square karo).

Hum is sentence ko samajhna chahte hain: "jaise $x$ , $3$ ke paas aata hai, $f (x)$ , $9$ ke paas aata hai." "Paas aata hai" words fisalane wale hain — kitna paas matlab paas? Yahi vagueness hum destroy karte hain.

KYUN. Picture dekho: jaise input marble $3$ ki taraf dono taraf se slide karta hai, uski height $9$ ki taraf slide karti hai. Lekin "slide karna" ek verb hai, koi cheez nahi jo hum check kar saken. Hume ek test chahiye jo ek doubter actually perform kar sake.

PICTURE. Curve $y = x^{2}$ , input point $3$ neeche mark kiya aur target height $9$ bayi taraf mark ki. Dashed lines pairing $3 \mapsto 9$ dikhati hain.

Step 2 — "Close" ko measurable distance mein badlo

KYA. " $f (x)$ , $L$ ke close hai" ban jaata hai: $f (x)$ aur $L$ ke beech ki distance chhoti hai. Do numbers $p$ aur $q$ ke beech ki distance likhi jaati hai $∣ p - q ∣$ — unhe subtract karo, sign hato. Toh output ki closeness exactly ek number $∣ f (x) - L ∣$ hai.

KYUN. Ek distance ek non-negative number hoti hai jise hum ek threshold se compare kar sakte hain. Yahi poora trick hai: ek vibe ("close") ko ek inequality se replace karo (" $∣ f (x) - L ∣$ kuch se less hai").

PICTURE. Vertical output axis par, $L = 9$ beech mein baitha hai. Hum uske around half-height $ε$ ki ek tube draw karte hain. Symbol == $ε$ == (Greek "epsilon") sirf ek chosen positive number hai — allowed output error. "Tube ke andar" literally matlab hai $∣ f (x) - L ∣ < ε$ .

Step 3 — Challenge–response game

KYA. Doubter pehle tube pick karta hai — koi bhi $ε > 0$ , chahe kitna bhi tiny. Phir tera kaam hai ki $a$ ke around ek matching input window dhundo jo itni narrow ho ki uske andar har input tube ke andar land kare. Us input window ki half-width == $δ$ == (Greek "delta") kehlati hai, ek aur positive number — allowed input error.

KYUN. Order matter karta hai. Doubter ko $ε$ tumse kuch sune bina shrink karne ki ijazat honi chahiye; phir tum $δ$ se react karo. Agar tumhe pehle $δ$ fix karna padta, toh baad ka tiny $ε$ tumhe pakad sakta tha. Toh $ε$ challenge hai, $δ$ tumhara response hai, aur $δ$ ko $ε$ par depend karne ki ijazat hai.

PICTURE. Ab do tubes: $L$ ke around half-height $ε$ ki green output tube (horizontal band), aur $a$ ke around half-width $δ$ ki plum input window (vertical band). Woh claim jise tumhe defend karna hai: vertical band mein har $x$ curve ka ek point horizontal band ke andar produce karta hai.

Step 4 — Poori definition, assembled

KYA. Steps 2 aur 3 ko promise "agar input window mein hai, toh output tube mein hai" se glue karo, aur demand karo ki yeh har challenge ke liye ho.

KYUN. Single arrow $⟹$ ("agar … toh …") promise hai. " $\forall ε$ " ("for all $ε$ ") woh cheez hai jo output ko arbitrarily close pin karta hai — ek lucky $ε$ kuch prove nahi karta.

PICTURE. Step 3 ke do bands logical flow ke saath arrow ke roop mein draw kiye gaye: plum window ke andar $\Rightarrow$ green tube ke andar.

Limits and Continuity se compare karo: continuity wahi sentence hai lekin $L = f (a)$ ke saath aur puncture hata ke.

Step 5 — Ab actually JEETNA: input distance ko factor out karo

KYA. Humare concrete claim ke liye, output error hai $∣ x^{2} - 9∣$ . Hum ise rewrite karte hain taaki controllable quantity $∣ x - 3∣$ explicitly appear ho: $∣ x^{2} - 9∣ = ∣ (x - 3) (x + 3) ∣ = ∣ x - 3∣ \cdot ∣ x + 3∣.$

KYUN. Humara ek hi knob hai $δ$ , aur $δ$ control karta hai $∣ x - 3∣$ ko. Toh hume $∣ x - 3∣$ ko ek factor ke roop mein expose karna hi padega. Jo bacha, $∣ x + 3∣$ , woh "steepness multiplier" hai — wo amount jis se input error output error mein magnify hoti hai.

PICTURE. $x = 3$ ke paas curve, right-triangle-style rise/run ke saath: ek chhota horizontal nudge $∣ x - 3∣$ ek vertical jump $∣ x^{2} - 9∣$ produce karta hai, aur dono ka ratio roughly $∣ x + 3∣$ hai — local steepness.

Step 6 — Magnifier ko cage karo (preliminary restriction)

KYA. $∣ x + 3∣$ constant nahi hai — yeh badhta hai jaise $x$ badhta hai, toh hum ise simply ek fixed number nahi keh sakte. Hum temporarily promise karte hain ki window zyada se zyada $1$ wide hai: force karo $∣ x - 3∣ < 1$ . Iska matlab $2 < x < 4$ , hence $5 < x + 3 < 7$ , toh $∣ x + 3∣ < 7$ .

KYUN. Agar hum $∣ x + 3∣$ ko free chodh dete toh woh enormous ho sakta tha, aur koi finite $δ$ use tame nahi kar paata. Pehle window ko width $1$ tak shrink karke, hum magnifier ko ek concrete number $M = 7$ se bound karte hain. Yeh woh "cap" hai jo ek moving multiplier ko ek fixed wale mein badal deti hai.

PICTURE. Curve ke neeche plum strip $2 < x < 4$ ; us strip ke upar steepness $∣ x + 3∣$ $7$ par horizontal line ke neeche rehti hai. Strip ke bahar yeh $7$ se upar chadhti — isliye hum ise cage karte hain.

Step 7 — $δ$ choose karo aur darwaza band karo

KYA. Cage ke andar, $∣ x^{2} - 9∣ < 7 ∣ x - 3∣$ . Ise $ε$ se neeche force karne ke liye, demand karo $7∣ x - 3∣ < ε$ , yaani $∣ x - 3∣ < ε /7$ . Do promises dono ek saath hold karni chahiye — width $< 1$ aur width $< ε /7$ — toh hum chhota lete hain: $δ = min (1, \frac{ε}{7})$

KYUN. $min$ dono restrictions ek saath guarantee karta hai: cage valid rehta hai (toh $M = 7$ legal hai) aur arithmetic $7∣ x - 3∣ < ε$ work karta hai. Agar $ε$ bada hai, $1$ binding cap hai; agar $ε$ tiny hai, $ε /7$ hai.

PICTURE. $ε /7$ window (patli) aur $1$ window (chaudi) saath draw ki gayi; $δ$ narrower shaded overlap hai. Jaise $ε \to 0$ , overlap $ε /7$ line pe collapse karta hai.

Step 8 — Edge aur degenerate cases (koi bhi scenario undikha mat chodho)

Ek-picture summary

KYA. Ek frame: curve $y = x^{2}$ , $9$ ke around half-height $ε$ ki green output tube, $3$ ke around half-width $δ = min (1, ε /7)$ ki plum input window, aur guarantee arrow — window mein har $x$ tube mein land karta hai.

Recall Feynman retelling (plain words, koi symbols nahi)

Ek doubter $9$ height ke around kuch chosen thinness ki ek horizontal ribbon draw karta hai aur mujhe dare karta hai: "curve ko andar rakho." Main jawab deta hoon input $3$ ke around ek vertical stripe draw karke jo itni narrow ho ki uske upar baitha curve ka piece ribbon se kabhi bahar na nikle. Kitna narrow dhundne ke liye, main notice karta hoon curve ka height error input error times local steepness ke barabar hota hai. $3$ ke paas steepness roughly $6$ hai, aur main ise safely $7$ se bound kar leta hoon ek baar jab main promise karo ki meri stripe $1$ se under wide hai. Toh agar main apni stripe ribbon ki thinness ko $7$ se divide karke (aur $1$ ke under) se patli rakhun, curve trap ho jaata hai. Kyunki main yeh kisi bhi thinness ke liye kar sakta hoon jo doubter naam kare, curve truly $9$ ki taraf jata hai — yahi matlab hai "limit $9$ hai." Aur kyunki maine sirf "distance" use ki, bilkul wahi kahani vectors, functions, ya kisi bhi space ke liye kaam karti hai jahan distance make sense kare.

Recall Self-test

$δ$ do bounds ka minimum kyun hona chahiye? ::: Dono promises ek saath hold karni chahiye — cage $∣ x - 3∣ < 1$ (taaki magnifier $∣ x + 3∣ < 7$ valid rahe) aur $∣ x - 3∣ < ε /7$ (taaki $7∣ x - 3∣ < ε$ ho). Chhota lene se dono guarantee hote hain. Chain $∣ x^{2} - 9∣ = ∣ x - 3∣∣ x + 3∣$ mein, kaun sa factor $δ$ control karta hai? ::: $∣ x - 3∣$ — input error; $∣ x + 3∣$ woh magnifier hai jise hume bound karna hai. Agar doubter ko $ε = 0$ allow kiya jaye toh kya toot jaata hai? ::: Yeh demand karta hai ek negative distance $∣ f (x) - L ∣ < 0$ , impossible; limits arbitrarily close hone ke baare mein hain, exactly equal hone ke baare mein nahi.