Intuition The big pictureAn orthogonal matrix is the matrix version of a rigid motion (rotation or reflection) of space.
It moves vectors around but never stretches, shrinks, or skews them. Lengths stay the same,
angles stay the same. Because volume is preserved, the determinant must measure "no scaling" —
which forces det = ± 1 \det = \pm 1 det = ± 1 orientation : + 1 +1 + 1 − 1 -1 − 1
Definition Orthogonal matrixA real square matrix Q ∈ R n × n Q \in \mathbb{R}^{n\times n} Q ∈ R n × n orthogonal if
Q T Q = Q Q T = I . Q^{\mathsf T} Q = Q Q^{\mathsf T} = I. Q T Q = Q Q T = I . Q Q Q Q − 1 = Q T Q^{-1} = Q^{\mathsf T} Q − 1 = Q T orthogonal group O ( n ) O(n) O ( n )
WHY this definition? The condition Q T Q = I Q^{\mathsf T}Q = I Q T Q = I Q Q Q orthonormal set (mutually perpendicular unit vectors). Let's see why.
Q T Q = I Q^{\mathsf T}Q = I Q T Q = I Q = [ q 1 ∣ q 2 ∣ ⋯ ∣ q n ] Q = [\,q_1 \mid q_2 \mid \dots \mid q_n\,] Q = [ q 1 ∣ q 2 ∣ ⋯ ∣ q n ] q i q_i q i ( i , j ) (i,j) ( i , j ) Q T Q Q^{\mathsf T}Q Q T Q dot product q i T q j q_i^{\mathsf T} q_j q i T q j Q T Q = I Q^{\mathsf T}Q = I Q T Q = I q i T q j = δ i j = { 1 i = j 0 i ≠ j q_i^{\mathsf T} q_j = \delta_{ij} = \begin{cases}1 & i=j \\ 0 & i\neq j\end{cases} q i T q j = δ ij = { 1 0 i = j i = j i = j i=j i = j i ≠ j i\neq j i = j
This is the defining geometric property — derive it from scratch.
Consequences (HOW it gives geometry):
Length: put y = x y=x y = x ∥ Q x ∥ 2 = ( Q x ) T ( Q x ) = x T x = ∥ x ∥ 2 \;\|Qx\|^2 = (Qx)^{\mathsf T}(Qx) = x^{\mathsf T}x = \|x\|^2 ∥ Q x ∥ 2 = ( Q x ) T ( Q x ) = x T x = ∥ x ∥ 2 ∥ Q x ∥ = ∥ x ∥ \|Qx\| = \|x\| ∥ Q x ∥ = ∥ x ∥ Angle: since cos θ = x T y ∥ x ∥ ∥ y ∥ \cos\theta = \dfrac{x^{\mathsf T}y}{\|x\|\,\|y\|} cos θ = ∥ x ∥ ∥ y ∥ x T y
So Q Q Q isometry : a distance-preserving linear map.
Intuition What the sign meansdet Q \det Q det Q ∣ det Q ∣ = 1 |\det Q| = 1 ∣ det Q ∣ = 1 sign records orientation:
det Q = + 1 \det Q = +1 det Q = + 1 special orthogonal Q ∈ S O ( n ) \;Q\in SO(n) Q ∈ S O ( n ) rotation (right hand stays right hand).det Q = − 1 \det Q = -1 det Q = − 1 reflection (or rotation-times-reflection) — handedness flips.
det Q = ± 1 \det Q=\pm1 det Q = ± 1 every matrix with det = ± 1 \det=\pm1 det = ± 1 Why it feels right: the property det = ± 1 \det=\pm1 det = ± 1 Why it's wrong: det = ± 1 \det=\pm1 det = ± 1 necessary but not sufficient . Example
A = ( 1 5 0 1 ) A=\begin{pmatrix}1&5\\0&1\end{pmatrix} A = ( 1 0 5 1 ) det = 1 \det=1 det = 1 A T A ≠ I A^{\mathsf T}A\neq I A T A = I Fix: orthogonality is the full condition Q T Q = I Q^{\mathsf T}Q=I Q T Q = I det = ± 1 \det=\pm1 det = ± 1 byproduct .
Intuition Rows are orthonormal tooBecause Q Q T = I QQ^{\mathsf T}=I Q Q T = I rows of Q Q Q ⇔ \Leftrightarrow ⇔
2 × 2 2\times2 2 × 2 Q = ( cos θ − sin θ sin θ cos θ ) Q=\begin{pmatrix}\cos\theta & -\sin\theta\\ \sin\theta & \cos\theta\end{pmatrix} Q = ( cos θ sin θ − sin θ cos θ )
Check orthonormal columns: ∥ c 1 ∥ 2 = cos 2 θ + sin 2 θ = 1 \|c_1\|^2=\cos^2\theta+\sin^2\theta=1 ∥ c 1 ∥ 2 = cos 2 θ + sin 2 θ = 1 c 1 ⋅ c 2 = − cos θ sin θ + sin θ cos θ = 0 c_1\cdot c_2 = -\cos\theta\sin\theta+\sin\theta\cos\theta=0 c 1 ⋅ c 2 = − cos θ sin θ + sin θ cos θ = 0 Why? Confirms Q T Q = I Q^{\mathsf T}Q=I Q T Q = I Determinant: cos 2 θ + sin 2 θ = 1 ⇒ det = + 1 \cos^2\theta+\sin^2\theta = 1 \Rightarrow \det=+1 cos 2 θ + sin 2 θ = 1 ⇒ det = + 1 Why + 1 +1 + 1 It's a rotation, orientation kept — it's in S O ( 2 ) SO(2) S O ( 2 )
Worked example 4 — Forecast-then-VerifyForecast: is P = ( 0 1 1 0 ) P=\begin{pmatrix}0&1\\1&0\end{pmatrix} P = ( 0 1 1 0 ) det \det det Verify: columns are e 2 , e 1 e_2,e_1 e 2 , e 1 det = 0 ⋅ 0 − 1 ⋅ 1 = − 1 \det=0\cdot0-1\cdot1=-1 det = 0 ⋅ 0 − 1 ⋅ 1 = − 1 swaps axes = a reflection across y = x y=x y = x − 1 -1 − 1
Recall Feynman: explain to a 12-year-oldImagine a flat sheet of stickers. An orthogonal matrix is a way to spin the sheet or flip it
over like a pancake — but never stretch it or squish it. Every sticker stays the same size, and
the distance between any two stickers never changes. If you just spin it, you can lay it back down
the same way — that's "+1". If you have to flip it over to make it match, that's "−1". Since
nothing grows or shrinks, the "size-change number" (determinant) can only be the do-nothing values
+ 1 +1 + 1 − 1 -1 − 1
"O for Orthonormal, T for Transpose-is-inverse, ±1 because flip-or-spin."
Q T Q = I Q^{\mathsf T}Q=I Q T Q = I R otation = R ight-hand kept = + 1 +1 + 1 L ection fL ips = − 1 -1 − 1
Common mistake "Orthogonal matrix means entries are orthogonal-looking / has lots of zeros."Feels right: the swap and identity matrices are sparse.
Wrong: 1 5 3 − 4 4 3 \tfrac15\begin{smallmatrix}3&-4\\4&3\end{smallmatrix} 5 1 3 4 − 4 3 Fix: "orthogonal" refers to columns being orthonormal vectors , not zero entries.
Q − 1 = Q T Q^{-1}=Q^{\mathsf T} Q − 1 = Q T Feels right: transpose is cheap, looks like a shortcut inverse.
Wrong: only holds exactly when Q Q Q Fix: test Q T Q = ? I Q^{\mathsf T}Q\overset{?}{=}I Q T Q = ? I
Definition of an orthogonal matrix? A real square
Q Q Q with
Q T Q = Q Q T = I Q^{\mathsf T}Q=QQ^{\mathsf T}=I Q T Q = Q Q T = I , i.e.
Q − 1 = Q T Q^{-1}=Q^{\mathsf T} Q − 1 = Q T .
What does Q T Q = I Q^{\mathsf T}Q=I Q T Q = I They are orthonormal: unit length and mutually perpendicular (
q i T q j = δ i j q_i^{\mathsf T}q_j=\delta_{ij} q i T q j = δ ij ).
Prove det Q = ± 1 \det Q=\pm1 det Q = ± 1 det ( Q T Q ) = det I ⇒ ( det Q ) 2 = 1 ⇒ det Q = ± 1 \det(Q^{\mathsf T}Q)=\det I\Rightarrow(\det Q)^2=1\Rightarrow\det Q=\pm1 det ( Q T Q ) = det I ⇒ ( det Q ) 2 = 1 ⇒ det Q = ± 1 .
Geometric meaning of det Q = + 1 \det Q=+1 det Q = + 1 − 1 -1 − 1 + 1 +1 + 1 rotation (orientation kept,
S O ( n ) SO(n) S O ( n ) );
− 1 -1 − 1 reflection (orientation flipped).
Why does ∥ Q x ∥ = ∥ x ∥ \|Qx\|=\|x\| ∥ Q x ∥ = ∥ x ∥ ∥ Q x ∥ 2 = x T Q T Q x = x T x = ∥ x ∥ 2 \|Qx\|^2=x^{\mathsf T}Q^{\mathsf T}Qx=x^{\mathsf T}x=\|x\|^2 ∥ Q x ∥ 2 = x T Q T Q x = x T x = ∥ x ∥ 2 .
Is every matrix with det = ± 1 \det=\pm1 det = ± 1 No — necessary not sufficient (e.g. a shear). Need full
Q T Q = I Q^{\mathsf T}Q=I Q T Q = I .
What are the possible eigenvalue magnitudes of Q Q Q All eigenvalues satisfy
∣ λ ∣ = 1 |\lambda|=1 ∣ λ ∣ = 1 (real ones are
± 1 \pm1 ± 1 , complex are
e ± i θ e^{\pm i\theta} e ± i θ ).
Inverse of an orthogonal matrix? Q − 1 = Q T Q^{-1}=Q^{\mathsf T} Q − 1 = Q T , and it is also orthogonal.
Is the product of two orthogonal matrices orthogonal? Yes —
O ( n ) O(n) O ( n ) is a group; closure follows from
( Q 1 Q 2 ) T Q 1 Q 2 = I (Q_1Q_2)^{\mathsf T}Q_1Q_2=I ( Q 1 Q 2 ) T Q 1 Q 2 = I .
det \det det 2 × 2 2\times2 2 × 2 cos 2 θ + sin 2 θ = + 1 \cos^2\theta+\sin^2\theta=+1 cos 2 θ + sin 2 θ = + 1 .
Orthonormal bases & Gram–Schmidt — how to build the columns of Q Q Q Determinants — properties — supplies det ( A B ) = det A det B \det(AB)=\det A\det B det ( A B ) = det A det B det A T = det A \det A^{\mathsf T}=\det A det A T = det A Eigenvalues and eigenvectors — ∣ λ ∣ = 1 |\lambda|=1 ∣ λ ∣ = 1 Rotations and reflections in $\mathbb{R}^2$ and $\mathbb{R}^3$ — the geometric instances.QR decomposition — A = Q R A=QR A = QR Q Q Q Spectral theorem — symmetric matrices are diagonalized by orthogonal Q Q Q Inner product spaces — isometry / length preservation generalized.
Intuition Hinglish mein samjhoDekho, orthogonal matrix ka matlab simple hai: ye ek aisi square matrix Q Q Q orthonormal hote hain — yaani har column ki length 1 1 1 Q T Q = I Q^{\mathsf T}Q=I Q T Q = I Q − 1 = Q T Q^{-1}=Q^{\mathsf T} Q − 1 = Q T
Geometry me ye ek rigid motion hai — rotation ya reflection. Vector ko ghumata hai ya palatata
hai, par length aur angle kabhi change nahi karta. Proof bhi easy: ∥ Q x ∥ 2 = x T Q T Q x = x T x = ∥ x ∥ 2 \|Qx\|^2 = x^{\mathsf T}Q^{\mathsf T}Qx
= x^{\mathsf T}x = \|x\|^2 ∥ Q x ∥ 2 = x T Q T Q x = x T x = ∥ x ∥ 2 determinant ka size 1 1 1 hona hi padega.
Ab sign ki baat. det ( Q T Q ) = det I = 1 \det(Q^{\mathsf T}Q)=\det I=1 det ( Q T Q ) = det I = 1 ( det Q ) 2 = 1 (\det Q)^2=1 ( det Q ) 2 = 1 det Q = ± 1 \det Q=\pm1 det Q = ± 1 + 1 +1 + 1 pure rotation (orientation same, S O ( n ) SO(n) S O ( n ) − 1 -1 − 1 reflection (handedness ulta
ho gaya). Ek important trap: det = ± 1 \det=\pm1 det = ± 1 nahi ban jaati — shear matrix ka
bhi det = 1 \det=1 det = 1 Q T Q = I Q^{\mathsf T}Q=I Q T Q = I
