HOW — derive from scratch:f′(x)=limh→0h(x+h)n−xn.
Expand the numerator:
(x+h)n−xn=xn+nxn−1h+(2n)xn−2h2+⋯+hn−xn.
Divide by h:
h(x+h)n−xn=nxn−1+every term has h(2n)xn−2h+⋯+hn−1.
Take h→0: each term in the brace contains h, so it dies.
f′(x)=nxn−1.
HOW — from first principles:f(x)=x−m=xm1,f′(x)=limh→0h1((x+h)m1−xm1).
Common denominator:
=limh→0h(x+h)mxmxm−(x+h)m.Why this step? Combining fractions exposes xm−(x+h)m, which we already know how to handle.
Use Stage 1: h(x+h)m−xm→mxm−1, so hxm−(x+h)m→−mxm−1. The
denominator factor (x+h)mxm→x2m. Hence
f′(x)=x2m−mxm−1=−mxm−1−2m=−mx−m−1.
With n=−m: f′(x)=nxn−1. ✓
HOW — derive:
Let y=xp/q⇒yq=xp (both integer exponents). Differentiate both sides
w.r.t. x (use chain rule on the left — itself a consequence of Stage 1):
qyq−1dxdy=pxp−1.Why this step? Left side: dxdyq=qyq−1dxdy. Solve:
dxdy=qp⋅yq−1xp−1.
Substitute y=xp/q, so yq−1=xp(q−1)/q=xp−p/q:
dxdy=qpxp−1−(p−p/q)=qpxp/q−1.
With n=p/q: dxdy=nxn−1. ✓
After expanding (x+h)n−xn and dividing by h, which single term survives as h→0?
The nxn−1 term (all others still contain a factor of h).
Why can't the binomial proof handle n=1/2 or n=−3?
Those exponents give an infinite series, not a finite cancellable expansion.
Trick to prove the rule for negative integers?
Write x−m=1/xm, take a common denominator, reuse the positive-integer result.
Trick to prove the rule for rational p/q?
Set y=xp/q, raise to power q to get yq=xp, then implicitly differentiate.
Differentiating yq w.r.t. x gives?
qyq−1dxdy (chain rule).
dxdx?
2x1.
Is dxd2x=x2x−1?
No — that's the wrong rule; dxd2x=2xln2.
dxdx−2?
−2x−3.
Recall Feynman: explain to a 12-year-old
Imagine a tower made of blocks where each level is x times wider, n levels high — that's
xn. If you make x a tiny bit bigger, how fast does the tower grow? The answer is: you get
n copies of one-level-shorter towers (xn−1) added on, because each of the n levels
"feels" the stretch. That's why you slide the n to the front and drop the height by one. For
roots (like x) we play a flip-game: square it to undo the root, use the easy rule, then
flip back.
Power rule ka matlab simple hai: dxdxn=nxn−1 — exponent ko neeche le aao
(multiply karo) aur exponent mein se 1 ghata do. Lekin ratta maarne ki zaroorat nahi, kyunki
ise hum derivative ki definition se derive kar sakte hain. Positive integer ke liye hum
(x+h)n ko binomial theorem se kholte hain. Jab xn cancel ho jaata hai aur h se
divide karte hain, sirf ek hi term bina h ke bachta hai — wahi nxn−1 hai. Baaki sab
terms mein h hota hai, jo h→0 pe gayab ho jaata hai.
Negative integer (jaise x−2) ke liye binomial seedha kaam nahi karta, isliye usse
1/xm likh kar common denominator lete hain, aur positive integer wala result reuse karte
hain. Rational exponent (jaise x=x1/2) ke liye ek mast trick hai: y=xp/q likho,
dono taraf q power lagao to fraction khatam — yq=xp. Ab dono integer powers hain, to
implicit differentiation (chain rule ke saath) lagao aur answer nikal aata hai qpxp/q−1.
Sabse common galti: socho ki binomial proof har n ke liye chalta hai — nahi! Sirf non-negative
integer ke liye finite expansion milta hai. Doosri galti: 2x ko power rule se differentiate
karna — galat, kyunki yahan base constant aur exponent variable hai (uska rule 2xln2 hai).
Yeh rule itna important hai kyunki har dusra derivative isi pe khada hai — ek baar samajh lo,
zindagi bhar kaam aayega.