Let's build the whole system from three guiding questions.
Q1: How do I count the type of object?
Count the free indices.
0 free indices → scalar (e.g. aibi).
1 free index → vector / column of numbers (e.g. Aijbj has free index i).
2 free indices → matrix / rank-2 tensor (e.g. Aij).
WHY? Each free index ranges over 1…n, so one free index gives you n separate components.
Q2: Why must a summed index appear exactly twice?
Because summation pairs things: ∑i(thisi)(thati). If an index appeared three times you couldn't tell which pair to sum — the notation would be ambiguous. An index appearing 3+ times in one term is illegal.
Q3: Why can I rename a dummy index freely?
Because ∑iaibi=∑kakbk — the name of the summation variable doesn't matter. So aibi=akbk. This trick is essential for avoiding clashes.
Imagine a recipe that says "for every pair of matching socks, add their sizes." You don't have to write "add up" each time — the matching is the signal to add. Einstein's trick is the same: whenever a label shows up twice in the same chunk, that's a secret "add them all up" command. A label that shows up only once is just telling you "this is component number such-and-such," like a house number — you keep it. So by glancing at which letters are doubled, you instantly know what to add and what kind of answer (number, list, or grid) you'll get.
Dekho, idea bahut simple hai. Jab hum bar-bar ∑ (summation sign) likhte hain, toh notation bhari ho jaati hai. Einstein ne notice kiya ki jis index par hum sum karte hain, wo hamesha ek hi term mein do baar aata hai. Toh unhone rule bana diya: agar koi index ek term mein exactly do baar aaye, toh uske upar automatically sum karo aur ∑ likhna chhod do. Jaise aibi ka matlab seedha ∑iaibi, yaani dot product. Bas yaad rakho: "Twice = sum, Once = run" — do baar wala index sum hokar gaayab, aur ek baar wala (free index) survive karke answer ka shape decide karta hai.
Free indices count karke aap bata sakte ho answer kya hai: zero free index matlab scalar (number), ek free index matlab vector, do free index matlab matrix. Isiliye matrix multiply Cik=AijBjk mein j repeat hua (sum ho gaya), aur i,k free reh gaye — toh answer matrix hai. Yeh ek built-in galti-pakadne wala tool bhi hai: equation ke dono taraf free index same hone chahiye, warna kuch galat hai.
Do special cheezein zaroor yaad rakho. Kronecker deltaδij identity matrix hai aur uska kaam hai index badalna: δijaj=ai. Levi-Civitaεijk orientation/sign sambhalta hai, isliye cross product (a×b)i=εijkajbk banta hai. Aur ek jaadui identity εijkεilm=δjlδkm−δjmδkl se to lagbhag saari vector calculus identities, jaise BAC-CAB rule, do line mein nikal aati hain.
Yeh matter karta hai kyunki physics aur higher maths (tensors, relativity, electromagnetism) mein har formula index notation mein chhota aur clean ho jaata hai. Ek baar yeh aadat ho gayi, toh aap lambe-lambe vector proofs ko mechanically index manipulate karke solve kar loge — bina geometry yaad kiye.