Cheatsheet: The determinant

The core idea

The determinant of a matrix is the factor by which it scales area (2D) or volume (3D).

new area = det(M) · old area
det(M) = area of the parallelogram the unit square becomes

Same factor for every region, because a linear transformation scales space uniformly.

The formula (2x2)

M = [ a  b ]      det(M) = ad - bc
    [ c  d ]

Main diagonal product minus off-diagonal product. This is the signed area of the parallelogram spanned by the columns [a, c] and [b, d].

Values worth recognizing

det	Meaning
1	Area preserved (rotation, shear, identity)
> 1	Areas grow
between 0 and 1	Areas shrink, nothing collapses
0	Space collapses onto a line or point; columns linearly dependent; not invertible
< 0	Orientation flips (reflection); magnitude still gives area scaling

The determinant is signed: magnitude = size scaling, sign = orientation kept (+) or flipped (-).

Worked examples

Matrix	det = ad - bc	Geometry
[[3,0],[0,2]]	6 - 0 = 6	Unit square to 3x2 rectangle
[[0,-1],[1,0]]	0 - (-1) = 1	Rotation, area preserved
[[2,4],[1,2]]	4 - 4 = 0	Columns dependent, collapse to a line
[[1,0],[0,-1]]	-1 - 0 = -1	Reflection, area kept, orientation flipped
[[1,1],[0,1]]	1 - 0 = 1	Shear, distorts shape but area preserved

3D and the product rule

• 3D: determinant of a 3x3 matrix scales volume; read it as the volume of the parallelepiped the unit cube becomes. 0 = flattened; negative = inside-out.
• Product rule: det(AB) = det(A) · det(B). Scaling factors compose by multiplying (double then triple = times six).

Why it matters for AI

det != 0 means invertible (the transformation can be run backward; no information lost). det = 0 means a dimension was collapsed and cannot be recovered. The absolute determinant is also the volume-rescaling factor when converting between distributions.

Pitfalls to dodge

• Forgetting it is signed. Negative det = orientation flip, not an error.
• Reading det = 0 as small. Zero means total collapse and not invertible, not “a little bit.”
• Mixing the diagonals. It is ad - bc, main minus off.
• Treating det as the full story. It gives size and orientation, not the direction of the stretch.

The one-line version

The determinant is one signed number, how much bigger and which way up, and when it hits zero, a dimension is gone for good.