WHY this object exists: A system of linear equations like
2x+3yx−y=5=1
has all its structure in the coefficients. Strip the variables away and you get [213−1]. The matrix is the system. This compression is the entire point.
WHY same order only? A matrix is a packaged list of mn numbers. To add two packages you must pair up corresponding numbers; that pairing only exists when shapes match. There is no "natural" way to add a 2×3 to a 3×2.
We do not define AB entrywise. Instead we force it to match composition of transformations.
Let A act on vectors, and B act first. We want (AB)x=A(Bx).
Take Bx: its k-th entry is ∑jbkjxj.
Now apply A: row i of A dotted with the vector Bx:
[A(Bx)]i=∑kaik(Bx)k=∑kaik∑jbkjxj=∑j((AB)ijk∑aikbkj)xj
WHY inner dimensions must match (p=p): the dot product needs the same number of terms in the row and the column. (m×p)(p×n)=(m×n).
∑kaikbkj — row i of A dotted with column j of B.
Condition for AB to exist?
Columns of A = rows of B (inner dimensions match).
Why is matrix multiplication non-commutative?
It encodes composition of transformations; order of actions matters.
Reversal law for transpose of a product?
(AB)⊤=B⊤A⊤.
Definition of symmetric matrix?
A⊤=A, i.e. aij=aji.
Definition of skew-symmetric matrix?
A⊤=−A; main diagonal is all zeros.
What makes a matrix orthogonal?
A⊤A=I (preserves lengths/angles).
When is a square matrix singular?
When detA=0 (no inverse exists).
Does AB=0 imply A=0 or B=0?
No — matrices have zero divisors.
How to split any square matrix into symmetric + skew parts?
A=21(A+A⊤)+21(A−A⊤).
What does the identity matrix do?
Acts as "do nothing": IA=AI=A.
Recall Feynman: explain to a 12-year-old
A matrix is like a recipe table for changing arrows on graph paper. The numbers tell each arrow where to move. Adding two tables means combining recipes step-by-step (easy). But multiplying tables means "first follow recipe B, then recipe A" — and just like putting on socks then shoes, swapping the order gives a different result. That's why AB and BA aren't the same. A square table that's a perfect mirror across its diagonal is symmetric; the magic "do nothing" table is the identity.
Dekho, ek matrix bas numbers ka ek rectangular table hai — par asli baat ye hai ki ye table ek transformation ko represent karta hai, matlab arrows ko stretch, rotate ya shear karne ka rule. Jab tum linear equations likhte ho, unke coefficients ko table mein pack kar do — bas wahi matrix hai. Isi liye matrices itni powerful hain: poora equation system ek chhote grid mein simat jaata hai.
Addition aur scalar multiply simple hai — entry-by-entry, lekin sirf jab dono ka shape (order) same ho. Multiplication thoda tricky hai aur entrywise NAHI hota. Yaad rakho: row of A ko column of B ke saath dot product karo, aur inner dimensions match hone chahiye — (m×p)(p×n). Iska reason ye hai ki AB ka matlab hai "pehle B apply karo, phir A" — yani composition. Isi liye AB=BA generally — jaise socks pehle ya shoes pehle, order matter karta hai!
Types bhi important hain exam ke liye: identityI kuch nahi badalta, symmetric matrix mirror jaisi hoti hai (A⊤=A), skew-symmetric ka diagonal hamesha zero, orthogonal matrix lengths preserve karti hai (A⊤A=I), aur singular matrix wo jiska det=0 — uska inverse nahi hota. Ek pyaara trick: koi bhi square matrix ko symmetric + skew part mein tod sakte ho — 21(A+A⊤)+21(A−A⊤).
Sabse common galti: students sochte hain (AB)⊤=A⊤B⊤ — galat! Sahi hai (AB)⊤=B⊤A⊤, order ulta ho jaata hai (socks-and-shoes rule). Aur AB=0 ka matlab ye nahi ki A ya B zero ho — matrices mein zero divisors hote hain. In do baaton ko pakka yaad rakhna.
