Orthogonal Matrix Verification – From Intuition to Precision
An orthogonal matrix is a square matrix that, multiplied by its own transpose, gives back the identity. Why care? Think of a rotation in 2D: rotating the coordinate axes preserves the lengths of vectors and the angles between them. A matrix that preserves both lengths and angles is orthogonal. "Verification" just means checking whether a given matrix has this property.
The Intuition: What Does "Orthogonal" Mean Here?
"Orthogonal" means "at right angles." For matrices it refers to the columns:
- Each column vector has length 1 (a unit vector).
- Any two different columns are perpendicular (their dot product is zero).
So the columns form an orthonormal set. The same holds for the rows.
That is why the matrix is called orthogonal: its columns are orthogonal to each other and each is normalized to length 1.
The Precise Definition
A square n×n matrix A is orthogonal if and only if:
where AT is the transpose and I the n×n identity.
Verification: How to Check
Compute ATA and check whether it equals the identity.
Example: Check A=(cosθsinθ−sinθcosθ).
ATA=(cosθ−sinθsinθcosθ)(cosθsinθ−sinθcosθ)=(cos2θ+sin2θ00sin2θ+cos2θ)=(1001)=I
So this rotation matrix is orthogonal.
Why This Works: The Column Interpretation
Let the columns of A be c1,…,cn. The (i,j) entry of ATA is the dot product ci⋅cj.
- When i=j: the entry is ∥ci∥2. For it to equal 1, each column must have length 1.
- When i=j: the entry is ci⋅cj. For it to equal 0, different columns must be orthogonal.
So ATA=I is exactly the condition that the columns are orthonormal.
For an orthogonal A, also AAT=I (rows are orthonormal too), and A−1=AT — the inverse is just the transpose, a huge computational advantage.
Common Mistakes to Avoid …