GraphMath

QRF, Gram-Schmidt, part 1

From Gram-Schmidt to Q: orthonormal structure, rigid motion and projection

What does the Q part of QR factorization represent?

Gram-Schmidt builds orthogonal directions one by one, then normalizes them to create Q. This chapter develops that process, shows why Q preserves lengths and angles, explains the determinant and geometric meaning of square Q, extends the picture to tall Q and derives QQᵀ as the orthogonal projection onto col(Q).

Key ideas

Gram-Schmidt removes components along earlier directions so that each new vector contributes only its genuinely new direction. Normalizing those directions produces the columns of Q.

  • The columns of Q are orthonormal, so QᵀQ = I and multiplication by Q preserves lengths and angles
  • For square Q, determinant is ±1, matching the geometric role of rotation or reflection
  • For tall Q, multiplication is an isometric embedding from a lower-dimensional space into a higher-dimensional one
  • Left-multiplication by Qᵀ reverses that embedding on the image, while QQᵀ projects onto the embedded subspace
  • When the columns are orthonormal, QQᵀ can be written as a sum of outer products and acts as orthogonal projection onto col(Q)

Part 2 then turns from Q to R and shows how QR factorization separates deformation from rotation or reflection.

Why is Q geometrically special?

Because Q preserves lengths and angles. In the square case it acts as a rigid motion, and in the tall case it embeds the domain into a higher-dimensional space without distortion.

Related chapters

Chapter contents

The chapter is available as a PDF. Page links below are best-effort: most browsers support them but some viewers may ignore the page hint.

Topic Pages
Gram-Schmidt orthogonalization 1–2
Structure of matrix Q 3–4
Properties of Q defined as QᵀQ = I 4–5
Determinant of square Q 5–6
Combined rotation and reflection 6–8
Transformation by a 'tall' Q (m>n) 8–10
Matrix Q Qᵀ 10–11
Q Qᵀ as sum of outer products 11–14

Why does Q Qᵀ become a projection matrix?

Because when the columns of Q are orthonormal, QᵀQ = I and the general projection formula simplifies to Q(QᵀQ)⁻¹Qᵀ = QQᵀ. This makes QQᵀ the orthogonal projection onto the space spanned by the columns of Q.

Was this chapter helpful?

Quick feedback helps us improve the site.