# How does computing the determinant of a matrix with unit vectors give you the Cross Product?

Say you had $(a_x,a_y,a_z)\times(b_x,b_y,b_z)$, you would set up a matrix like the following:

And the resulting would be your Cross Product or the coordinates of an orthogonal vector. My question is why? Why does forming it that way give you the magnitude of an orthogonal vector and how is it related to the $\sin(\theta)$ definition of Cross Product.

• See also math.stackexchange.com/questions/22/…, although that question and its answers touch only on the orthogonal-vector part of this question, not the $\sin(\theta)$ part. Feb 5, 2016 at 18:42
• To me, there are really three questions that tell us why the cross product is what it is: 1. what line should the cross product be on? 2. what should be the length of the cross product be? 3. what sign convention should we use? The first one is the most important fact about the cross product: it is on the line perpendicular to both input vectors. Given the first answer, the second answer is determined by requiring that the cross product of two perpendicular unit vectors be a unit vector. The third answer is the right hand rule, which is a lot more arbitrary than the first two answers.
– Ian
Feb 5, 2016 at 18:49
• This definition of the cross product is definitely degrading the notion of determinant, and should be abolished. Outside of "Schaum's Outline" (which was outdated even when I was a student in the fifties of the last century), and similar texts, all elements of a matrix should be of the same type. Feb 5, 2016 at 18:58
• @ChristianBlatter Thanks for pointing out the problem with this notation. When I first saw it, I thought it was trying to take the determinant of a $5\times3$ matrix (the first three rows of which are $I_3$). Feb 21, 2016 at 22:32
• We have some good answers below, and I don't see any further requests for clarification, so is it fair to say the question has been answered? If so, I recommend "accepting" one of the answers. (You choose which one.) Feb 21, 2016 at 22:34

The determinant of a $3\times3$ matrix can be viewed as the triple product of its columns (or rows): \begin{align} \det\begin{bmatrix} x_1&y_1&z_1\\ x_2&y_2&z_2\\ x_3&y_3&z_3 \end{bmatrix} &= \begin{bmatrix} x_1\\ x_2\\ x_3 \end{bmatrix} \times \begin{bmatrix} y_1\\ y_2\\ y_3 \end{bmatrix} \cdot \begin{bmatrix} z_1\\ z_2\\ z_3 \end{bmatrix}\\ &= \begin{bmatrix} (x\times y)_1\\ (x\times y)_2\\ (x\times y)_3 \end{bmatrix} \cdot \begin{bmatrix} z_1\\ z_2\\ z_3 \end{bmatrix}\tag{1} \end{align} If we replace $\begin{bmatrix} z_1\\ z_2\\ z_3 \end{bmatrix}$ in $(1)$ by $\begin{bmatrix} \boldsymbol{i}\\ \boldsymbol{j}\\ \boldsymbol{k} \end{bmatrix}$, we get \begin{align} \det\begin{bmatrix} x_1&y_1&\boldsymbol{i}\\ x_2&y_2&\boldsymbol{j}\\ x_3&y_3&\boldsymbol{k} \end{bmatrix} &= \begin{bmatrix} (x\times y)_1\\ (x\times y)_2\\ (x\times y)_3 \end{bmatrix} \cdot \begin{bmatrix} \boldsymbol{i}\\ \boldsymbol{j}\\ \boldsymbol{k} \end{bmatrix}\\[6pt] &=(x\times y)_1\boldsymbol{i}+(x\times y)_2\boldsymbol{j}+(x\times y)_3\boldsymbol{k}\\[18pt] &=x\times y\tag{2} \end{align}

And the resulting would be your Cross Product or the coordinates of an orthogonal vector. My question is why? Why does forming it that way give you the magnitude of an orthogonal vector

Your last equation can be written as $$(a \times b)_i = \epsilon_{ijk} a_j b_k \quad (1)$$ where $\epsilon_{ijk}$ is the skew-symmetric or Levi-Civita tensor and the Einstein summation convention is used (we sum over same indices, here: $j$ and $k$, each from $1$ to $3$).

If the tuple $(i,j,k)$ consists of different numbers from $\{1, 2, 3\}$, thus it is a permutation of $(1,2,3)$, it is defined as sign of the permutation $\pm1$, otherwise it vanishes.

So the above is the compact notation for \begin{align} (a \times b)_1 &= \epsilon_{123} a_2 b_3 + \epsilon_{132} a_3 b_2 = a_2 b_3 - a_3 b_2 \\ (a \times b)_2 &= \epsilon_{231} a_3 b_1 + \epsilon_{213} a_1 b_3 = a_3 b_1 - a_1 b_3 \\ (a \times b)_3 &= \epsilon_{312} a_1 b_2 + \epsilon_{321} a_2 b_1 = a_1 b_2 - a_2 b_1 \end{align}

Compare this with the definition of the determinant, which is an alternating multilinear form in its $n$ arguments: $$\det A = \det(a_1, \dotsc, a_n) = \epsilon_{i_1 i_2 \dotsm i_n} a_{1 i_1} a_{2 i_2} \dotsm a_{n i_n} \quad (2)$$ It is the signed sum of all permutations of the components.

So indeed it happens that $$\det(e_i, a, b) = e_i \cdot (a \times b) = (a \times b)_i \quad (3)$$ where $e_i$ is the $i$-th canonical base vector \begin{align} e_1 = (1, 0, 0)^T \\ e_2 = (0, 1, 0)^T \\ e_3 = (0, 0, 1)^T \end{align}

The form you used employs the rule of Sarrus to calculate the determinant, which holds only for three dimensions. Equation $(2)$ holds for arbitrary dimension.

The definition of the vector product, equation $(1)$, consists of signed sums of permutations of the involved vector components and the definition of the determinant $(2)$ also makes uses of signed sum of permutations of its argument components. That way it happens that one can define a vector product as determinant.

and how is it related to the $\sin(\theta)$ definition of Cross Product.

You probably mean $$\lVert a \times b \rVert = \lVert a \rVert \lVert b \rVert \sin\angle(a, b) \quad (4)$$ This derives from the triple product $$a \cdot (b \times c) = \det(a, b, c)$$ which can be expressed as determinant, we used it above with the $i$-th canonical base vector for equation $(3)$. The determinant gives the volume of the parallelepiped (think tower of Pisa for a stack of cards) formed by the vectors $a, b, c$.

From this one can derive equation $(4)$.

Looking at the computation from the right angle, what you compute is a new vector in the dual space, such that $\vec a \times \vec b$ maps any vector $\vec c$ to $\mathbb R$ (or $\mathbb C$), in a way that $\det(\vec c, \vec a,\vec b)=(\vec a\times\vec b)\cdot \vec c$.

You can see this by replacing $(i,j,k)$ by $(c_1, c_2, c_3)$. Everything else follows from the properties of $\det$.

Especially $\vec a\times\vec b$ is orthogonal to $\vec a$ because $\det(\vec a, \vec a, \vec b)=0$, similar for $\vec b$.

If you start with the definition of cross-product as $$\underline{a}\times\underline{b}=|\underline{a}||\underline{b}|\sin \theta \underline{\hat{n}},$$ where $\theta$ is the angle between $\underline{a}$ and $\underline{b}$ and $\underline{\hat{n}}$ is the unit vector perpendicular to $\underline{a}$ and $\underline{b}$ in the sense of a right-hand triad' then it follows from this definition that:

1. $\underline{i}\times\underline{j}=0=\underline{j}\times\underline{j}=\underline{k}\times\underline{k}$

2. $\underline{i}\times\underline{j}=\underline{k}$ and $\underline{j}\times\underline{k}=\underline{i}$ and $\underline{k}\times\underline{i}=\underline{j}$

1. If the letters are in anticyclic order, the result is correspondingly negative, so, for example, $\underline{j}\times\underline{i}=-\underline{k}$ and so on.

Therefore, if we assume the distributivity of cross-product (not proved here), then the cross product of the two vectors gives exactly the same result as you obtain from evaluating the determinant.