4.5.37 · D3Linear Algebra (Full)

Worked examples — Orthogonal matrices — properties, det = ±1

2,865 words13 min readBack to topic

Before we start, three reminders in plain words, so no symbol appears un-earned.


The scenario matrix

Every orthogonal-matrix problem is one (or a mix) of the cells below. The examples that follow are tagged with the cell they hit, so you can see the whole territory is covered.

Cell Case class What's tricky about it Covered by
A Pure rotation (), reading the angle off the matrix Ex 1
B Pure reflection (), axis of the flip, self-inverse Ex 2
C The impostor: but not orthogonal necessary vs sufficient Ex 3
D Degenerate / zero-angle & identity limiting case Ex 4
E Build one from a single vector (Gram–Schmidt feel) normalising, choosing partner, sign choice Ex 5
F rotation about an axis, one fixed eigenvector, block structure Ex 6
G reflection, orientation flip in 3D Ex 7
H Word problem (robot / camera) translating physical rigid motion to Ex 8
I Exam twist: product & "guess the det" closure, of a product Ex 9
J Eigenvalues on the unit circle complex eigenvalues of a rotation Ex 10

The figure below sums up the whole table in one picture: the left panel is the archetype of every "" cell (A, D, F, H, I-rotation — a spin that keeps orientation), and the right panel is the archetype of every "" cell (B, G, I-reflection — a flip across a mirror line). As you work each example, place it mentally on the correct side of this picture.

Figure — Orthogonal matrices — properties, det = ±1

Example 1 — Cell A: pure rotation, read the angle

Steps.

  1. Compare to the rotation template . Matching: , . Why this step? The parent's rotation matrix has a fixed shape; if fits it, the angle is read off directly.
  2. Solve for : and both hold at . Why this step? We need one angle satisfying both equations — one equation alone is ambiguous.
  3. Confirm orthonormal columns. Column 1 is , length ✓. Column 2 is , length ✓. Dot product ✓. Why this step? This is the test done column-by-column.
  4. Determinant: . Why this step? certifies it's a rotation (a spin), consistent with orientation preserved.
Worked example Verify

: unit columns ✓. Apply to : you get , an arrow at exactly above the -axis — matches the claimed angle.


Example 2 — Cell B: pure reflection, find the mirror line

Steps.

  1. Orthonormal columns. : length ✓. : length ✓. Dot ✓. Why this step? Verifies , i.e. it is orthogonal.
  2. Determinant: . Why this step? means it is a flip, not a spin.
  3. Find the fixed direction (mirror line). A reflection fixes vectors on the mirror: solve , i.e. . Row 1: . So the mirror is the line , direction . Why this step? A reflection has eigenvalue along the mirror and perpendicular to it — the parent's eigenvalue fact.
Worked example Verify

Plug : ✓ (fixed). Plug (perpendicular to mirror): ✓ (flipped).


Example 3 — Cell C: the impostor ( but not orthogonal)

Steps.

  1. Don't trust alone. is necessary but not sufficient (parent's mistake box). Run the real test. Why this step? Orientation-preserving maps include shears, which stretch — we must check lengths.
  2. Test the columns. , length. Why this step? Orthogonal columns must be unit length; already kills it.
  3. Compute to be sure. . Why this step? The definitive certificate: ⇒ not orthogonal.
Worked example Verify

, so changes lengths — an isometry never does. Not orthogonal ✓.


Example 4 — Cell D: degenerate / zero-angle limit

Steps.

  1. Plug : . Why this step? The identity is the "do-nothing" rotation — the limiting/degenerate case.
  2. Check: ✓, . It is orthogonal, in (the rotations). Why this step? Confirms the identity sits inside the orthogonal group (it's the group's neutral element).
  3. Plug : . . Why this step? This is a subtle case: flips both axes, yet — it is a rotation (by ), not a reflection, because flipping both axes is two flips = a spin.
Worked example Verify

(even dimension), (odd dimension). Parity of decides the sign of .


Example 5 — Cell E: build one from a single vector

Steps.

  1. Normalize. , so . Why this step? Columns of must be unit vectors (Gram–Schmidt step 1).
  2. Choose a perpendicular unit partner. Rotating by gives ; by gives . Why this step? In the plane exactly two unit vectors are perpendicular to (opposite directions) — that's the sign freedom.
  3. Assemble and pick the rotation. , → rotation. , → reflection. Why this step? The sign of the partner controls whether orientation is kept () or flipped ().
Worked example Verify

✓. , ✓.


Example 6 — Cell F: a rotation about an axis

Steps.

  1. Insert : , so . Why this step? Concrete numbers make the checks unambiguous.
  2. Orthonormal columns: , , : each unit length, pairwise dot products ✓. Why this step? Direct verification of in 3D.
  3. Determinant via block structure: the top-left multiplies the rotation block whose det is . So . Why this step? ⇒ a genuine 3D rotation, so (the rotations in space).
  4. Axis = fixed direction: , so the -axis is fixed; the rotation happens in the plane. Why this step? A 3D rotation always fixes a line (its eigenvector with eigenvalue ) — that line is the axis.
Worked example Verify

and : the plane vectors rotate , while stays put ✓. ✓.


Example 7 — Cell G: a reflection

Steps.

  1. Orthogonal check: columns are — orthonormal ✓, and (flip twice returns home). Why this step? Confirms .
  2. Determinant: diagonal matrix ⇒ . Why this step? reflection (orientation flips), so is not in — exactly one axis flipped is an odd count.
Worked example Verify

Take the right-handed triple . After : — a left-handed triple. Handedness flipped, matching ✓.


Example 8 — Cell H: word problem (camera gimbal)

Steps.

  1. Rotation about leaves fixed, acts on : . Why this step? Spinning about the vertical axis is the -axis rotation block, mirroring Example 6.
  2. New forward vector: . Why this step? This is the physical answer — where the lens now points.
  3. No scaling check: , same as . Why this step? An orthogonal map is an isometry — the drone's "zoom" (length) can't change from a pure rotation. Units: a unit direction stays a unit direction.
Worked example Verify

✓. ✓.


Example 9 — Cell I: exam twist (product & guess the det)

Steps.

  1. Orthogonality by closure: . So is orthogonal. Why this step? The orthogonal group is closed under products — products stay orthogonal (parent's closure box).
  2. Determinant is multiplicative: . Why this step? — no multiplication of the matrices needed.
  3. Interpret: ⇒ the combined map is a reflection (it leaves ). A rotation followed by a reflection is a reflection (odd total number of flips). Why this step? The sign of is the "flip parity"; even = rotation, odd = reflection.
Worked example Verify

Concretely: (), (). , ✓ — the swap/reflection matrix from the parent.


Example 10 — Cell J: eigenvalues on the unit circle

Steps.

  1. Characteristic polynomial: . Why this step? Eigenvalues are roots of (see Eigenvalues and eigenvectors).
  2. Solve: . Why this step? A rotation by has no real eigenvector (nothing is left pointing the same way), so the eigenvalues are complex — exactly as the parent predicted, with .
  3. Check magnitude: . Why this step? The parent proved every eigenvalue of an orthogonal matrix has ; here , on the unit circle.
Worked example Verify

Product of eigenvalues ✓; sum with ✓.


Recall One-line summary of every cell

Spin = ; flip = ; impostor = right but ; identity/ are rotations; is a reflection; product's det multiplies; eigenvalues live on the unit circle.

Active recall

How do you tell a orthogonal matrix is a rotation vs reflection?
rotation (spin, in ), reflection (flip).
Why is a rotation but a reflection?
: even gives (rotation), odd gives (reflection).
has but columns of length — orthogonal?
No; orthogonal needs unit orthonormal columns, i.e. , not just .
of (rotation)(reflection)?
, so the product is a reflection.
Eigenvalues of the plane rotation?
, both with .
How do you find a reflection's mirror line?
Solve ; the fixed direction is the mirror.
What does denote?
The special orthogonal group — orthogonal matrices with , i.e. the pure rotations.