This will delete your progress and chat data for all chapters in this course, and cannot be undone!

Glossary

Select one of the keywords on the left…

Linear AlgebraDeterminants

Reading time: ~25 min

The determinant of a square matrix is a single number which captures some important information about how the transformation behaves. In this section, we will develop a geometrically-motivated definition of the determinant.

Exercise Suppose that is a region in and that is an matrix. Consider the singular value decomposition .

Let .By what factor does transform volumes?

Let .In terms of the entries of , by what factor does transform volumes?

Let .By what factor does transform volumes?

Solution.Since and are orthogonal, and both preserve volumes. So they multiply volumes by a factor of 1. Since scales volumes by a factor of along the first axis, along the second, and so on, it scales volumes by a factor of .

Volume scale factor

We see from this exercise that a linear transformation from to scales the volume of any -dimensional region by the same factor: the volume scale factor of .

Exercise Find the volume scale factor of the matrix by describing how the matrix transforms a region in \mathbb{R}^3.

Solution.Since A[x,y,z] = [x,z,ky], we see that A stretches (or compresses) regions in \mathbb{R}^3 by a factor k along the y-axis and then reflects across the plane y = z.For example, the unit cube is mapped to a 1 \times 1 \times k box Since such a box has volume k, the volume scale factor of S is k.

Orientation factor

Another geometrically relevant piece of information about T is whether it preserves or reverses orientations. For example, rotations in \mathbb{R}^2 are orientation preserving, while reflections are orientation reversing. Let's define the orientation factor of T to be +1 if T is orientation preserving and -1 if T is orientation reversing.

Definition We define the determinant of a transformation T to be the product of its orientation factor and its volume scale factor.

We define the determinant of a matrix A to be the determinant of the corresponding linear transformation \mathbf{x}\mapsto A\mathbf{x}.

Exercise Interpret A = \begin{bmatrix} 0 & -1 \\\ -1 & 0 \end{bmatrix} geometrically and use this interpretation to find \det A, the determinant of A.

Solution.Since A\begin{bmatrix} x \\\ y \end{bmatrix} = \begin{bmatrix} -y \\\ -x \end{bmatrix},A reflects points in \mathbb{R}^2 across the line y = -x.Therefore, it preserves areas and reverses orientations. So its determinant is -1.

There is relatively simple formula for \det A in terms of the entries of A.For example,

\begin{align*}\left|\begin{array}{cc} a & b \\\ c & d \end{array}\right| = ad - bc\end{align*}

is the determinant of a 2 \times 2 matrix. However this formula is terribly inefficient if A has many entries (it has n!terms for an n\times n matrix), and all scientific computing environments have a det function which uses much faster methods.

Exercise For various values of n, use the expression np.linalg.det(np.random.randint(-9,10,(n,n)))det(rand(-9:9, n, n)) to find the determinant of an n\times n matrix filled with random single-digit numbers. How large does n have to be for the determinant to be large enough to consistently overflow?

import numpy as np
np.linalg.det(np.random.randint(-9,10,(n,n)))

using LinearAlgebra
det(rand(-9:9, n, n))

Solution.Trial and error reveals that this determinant starts to consistently return infInf at n = 187.

Exercise Suppose that A and B are 3 \times 3 matrices, with determinant 5 and \frac{1}{2} respectively. Suppose that R \subset \mathbb{R}^3 is a 3D region modeling a fish whose volume is 14. What is the volume of the transformed fish BA(R)?

19.5

35

12

16.5

Solution.The volume of A(R) is 5 \cdot 14 = 70.The volume of BA(R) = B(A(R)) is \tfrac{1}{2} \cdot 70 = 35.

Exercise Let R \subset \mathbb{R}^3 be 3D region modeling a fish, and suppose A an invertible 3 \times 3 matrix. If R has volume 15 and A^{-1}(R) has volume 5, then the determinant of A is equal to ?

Solution.We can see that the matrix A^{-1} scales volumes by \frac{1}{3}, and hence \det A^{-1} = \frac{1}{3}.This implies that \det A = 3.

Determinants can be used to check whether a matrix is invertible, since A is noninvertible if and only if it maps \mathbb{R}^n to a lower-dimensional subspace of \mathbb{R}^n, and in that case A squishes positive-volume regions down to zero-volume regions.

Exercise Let A = \begin{bmatrix} 2 & -2 \\\ -4 & 0 \end{bmatrix}.Find the values of \lambda \in \mathbb{R} for which the equation A \mathbf{v} = \lambda \mathbf{v} has nonzero solutions for \mathbf{v}.

Solution.We can rewrite A\mathbf{v} = \lambda \mathbf{v} as A\mathbf{v} = (\lambda I) \mathbf{v}, where I is the identity matrix. We can rearrange this to give the equation (A - \lambda I)\mathbf{v} = 0.This has a nontrivial solution if (A - \lambda I) has a nonzero nullspace. Since A - \lambda I is a square matrix, this is equivalent to it having determinant zero.

Exercise For an n \times n square matrix, which of the following is the relationship between \det A and \det (3A)?

XEQUATIONX3239XEQUATIONX.

XEQUATIONX3240XEQUATIONX.

XEQUATIONX3241XEQUATIONX.

\det(3A) = 3^{n} \det (A).

Solution.The answer is (4) \det(3A) = 3^{n} \det (A).There are two ways to see this, algebraically and geometrically.

To check that this is the right answer using algebra, let A = I_{n} be the n \times n identity matrix, with determinant 1.The matrix 3A is diagonal, with threes on the diagonal. Its determinant is the product of the entries on its diagonal, 3 \times 3 \times \cdots \times 3 = 3^{n}.

Geometrically, we know that the determinant of A measures how much A scales volume. The matrix 3A scales by a factor of three more in each dimension. Since there are n dimensions, the total scaling of volume is multiplied by a factor 3^n.

Exercise Is every matrix with positive determinant positive definite?

Solution.No. Consider the negation of the 2 \times 2 identity matrix. It has determinant 1, yet its eigenvalues are both negative.

Congratulations! You have completed the Data Gymnasia Linear Algebra course.