CROSS PRODUCTS AND THE PERMUTATION TENSOR

In class we have studied that the vector product between two vectors A and B is written
as:

C=AxB

and has a magnitude equal to |A| |B| sin θ, and a direction determined by application of
the right hand rule.

If we write each vector in component form, and take the term by term vector product, we
obtain for the resulting vector C:

C = (AyBz – AzBy )i + (-AxBz + AzBx)j + (AxBy - AyBx )k

B B B B B B (1)

In class we discussed the patterns observed in (1). First, each component is the
difference of products. Second, each product in a component of the cross product
represents a permutation of the components of the vectors A and B. Finally, we notice
that the i component of the cross product involves no x terms; similarly the j and k
components of the cross product involve no y or z terms.

We will learn to write cross products in summation notation, however, in order to do that
we need a mathematical structure that will allow us to reproduce the patterns we see in
eq. (1).

The mathematical formalism that allows us to write cross products and curls in
summation notation is the Levi-Civita Permutation Tensor. In three dimensions, the
Levi-Civita permutation tensor (called henceforth the permutation tensor) is written as εijk
and has the properties:

0 if any two indices are equal

εijk = +1 if all indices are different and are cyclic
-1 if all indices are different and are anti-cyclic

Cyclic permutations are 123, 231 and 312; anti-cyclic permutations are 132, 213 and 321.

Let’s spend a little time investigating the properties of the permutation tensor before
moving on to writing cross products.

Question: How many different ways can we write the permutation tensor (in 3 space)?

The answer is 27. We have 3 choices for the first index, 3 for the second, and 3 for the
third; 3x3x3 = 27.
What are these 27 different ways of writing εijk? It gets a little tedious, but it is
instructive to write out all these permutations:

The twenty-seven possible permutations are:

ε111 ε211 ε311

ε112 ε212 ε312
ε113 ε213 ε313
ε121 ε221 ε321
ε122 ε222 ε322
ε123 ε223 ε323
ε131 ε231 ε331
ε132 ε232 ε332
ε133 ε233 ε333

We can use the definition of the permutation tensor given above to realize that only six of
these terms are non-zero. These are ε123, ε132, ε213, ε231, ε312 and ε321. You should be able
to determine which of these equal +1 or -1.

Let’s look at equation (1) in a little more detail to see how we can write cross products in
summation notation. In this write-up we will denote x, y, z components by the indices 1,
2, 3 respectively.

As eq. (1) shows, the i component of C can be written as A2B3 – A3B2; the j component
B B

can be written as A3B1 – A1B3; and the k component as A1B2 – A2B1. The complete
B B B B

vector C is of course the sum of all these components.

Let’s consider now the expression:

εijk AjBk B (2)

First, notice that there are two repeated indices, j and k; this means that we will have to
sum over j and k. The i index is not repeated and so is not summed over in this
expression. Of course, i can be any integer between 1-3.

Let’s write explicitly the sum represented by (2). Remember that we will sum over j and
k; i is a dummy index that will run from 1 to 3. Remember also though that once we
choose a value of i, the values of j and k must be chosen so that no indices are the same
(or else the the value of that particular term will be zero). Thus, we can write out the
expression in (2) explicitly:

ε123 A2B3 + ε132 A3B2 + ε231 A3B1 + ε213 A1B3 +ε312 A1B2 + ε321 A2B1 (3)
B B B B B B
Make sure you review the expression in (3) carefully. Notice that these terms contain the
only non-zero terms of the permutation tensor. This pattern of components in (3) should
look very familiar by now.

Using the properties of the permutation tensor described above, we can rewrite and
rearrange(3) as:

(A2B3 - A3B2) + (A3B1 - A1B3) +(A1B2 - A2B1)

B B B B B B

Notice that each parenthesis consists of one of the components of the cross product vector
C from eq. (1). Notice further that the terms in the first parenthesis correspond to i =1;
the terms in the second parenthesis correspond to i = 2; and the final parenthesis
corresponds to i =3.

Thus, we can express the cross product C = A x B in summation notation as:

Ci = εijk AjBk
B (4)

In other words, the ith component of the vector C is given by the expression above, and
the complete vector C results from summing all its components.

A simple proof:

Let’s use this description of the cross product to prove a simple vector result, and also to
get practice in the use of summation notation in deriving and proving vector identities.

We know that the cross product of two vectors is perpendicular to each of the vectors;
that is, we expect C ⊥ Α and also C ⊥ B. In terms of vector multiplication, this means
that A·C = 0 = B·C. But we already know that in summation notation, the dot product
between two vectors can be written as AiCi, since in summation notation you sum over
repeated indices, and the product AiCi = A1C1 + A2C2 + A3C3 = A·C.

So, if we wish to prove that A·C = 0 if C=AxB using only summation notation, let’s
begin with the expression in (4) and realize that:

AiCi = Ai (εijk AjBk)

B (5)

Since we are now working only with scalar quantities, we can reorder the multiplications
on the right hand side any way we wish, and we can rewrite (5) as:

AiCi = Bk (εijk AiAj) (6)

Examine the term in parentheses in (6). This term is simply the cross product of AxA.
However, since the cross product of any vector with itself is zero (since the magnitude is
proportional to sinθ), the expression AiCi is zero, and we have proven that the cross
product is perpendicular to each of the original vectors.
 USE OF SUMMATION NOTATION TO PROVE VECTOR IDENTITIES

THE “BAC-CAB” RULE

Let us consider the triple vector product:

G = A x (B x C) (1)

You can write the cross products out term by term, but this becomes lengthy and messy.
Using summation notation provides an elegant, terse and quick means of proving these
identities.

Here, we will show that:

G = A x (B x C) = B(A·C) – C(A· B) (2)

the so-called “BAC-CAB” rule.

First, we set D = B x C, and we can write the ith component of D as:

Di = ε ijk Bj Ck
B (3)

Then, we can write G = A x D. We can write this cross product in summation notation
as:

Gm = ε mni An Di (4)

It is important to understand why these subscripts are chosen as they are in eq. (4). We
cannot use the same set of subscripts “ijk” again in the permutation tensor in (4); these
subscripts were used in (3). While the exact choice of subscript is often arbitrary, we
should use a different set of subscripts in writing the new cross product.

Notice, however, that we do use the subscript “i” for the D term. This is because D is the
cross product of B and C, and we must use (3) as the expression for D in equation (4).

Notice also the pattern of subscripts in (3) and (4). The first subscript in each Levi-Civita
tensor refers to a component of the vector resulting from the cross product; in other
words, the “i” in (3) means we are computing the “ith” component of D; the “m” in (4)
means we are computing the “mth” component of G. The second subscript refers to a
component of the first vector in the cross product, and the final subscript labels a
component of the second vector in the cross product.

We can now substitute the expression for Di from (3) into (4) and obtain:

Gm = ε mni An (ε ijk Bj Ck)

B (5)
Since all the terms in (5) are scalars, we can rearrange terms and write:

Gm = ε mni ε ijk An Bj Ck B (6)

We realize that ε mni = ε nim = ε imn , so we can rewrite (6) as:

Gm = ε imn ε ijk An Bj Ck
B (7)

Recognize that in (7) we have a product of two permutation tensors, and each has the
same index (“i”) in the same location. This allows us to use the “ε−δ” relationship:

ε imn ε ijk = δjm δkn - δkmδjn (8)

Using the relationship in (8) to expand the product of permutation tensors in (7) yields:

Gm = (δjm δkn - δkmδjn) An Bj Ck B

= δjm δkn An Bj Ck - δkmδjn An Bj Ck (9)

B B

Let’s consider each term on the RHS of (9). In order that the first term be non-zero, we
have the conditions that j = m and k =n. The second term can be non-zero if and only if
k=m and j =n. Making these substations in (9) gives:

Gm = An Bm Cn - An Bn Cm
B B (10)

Since we are dealing with scalar quantities, we can switch order of multiplication as we
please, allowing us to write (10) in the very recognizable form:

Gm = Bm (An Cn) - Cm (An Bn) B (11)

Note that the terms in parentheses are A·C and A·B. So we can readily observe that (11)
is the component form of the vector identity:

G = B (A·C) – C(A·B) (12)

QED
 TRIPLE VECTOR PRODUCT

Proof of ∇ x (∇ x F) = ∇(∇· Α) − ∇2F

The proof of this identity follows the same path of the proof of the “BAC-CAB” identity.

First, we set A =∇ x F, so that G = ∇ x A. We can the write the “ith” component of A

as:

Ai = εijk (∂/∂xj F k) (1)

and G = ∇ x A becomes:

Gm = εmni (∂/∂xn Ai) = εmni ∂/∂xn (εijk ∂/∂xj Fk)

= εmni εijk ∂/∂xn (∂/∂xj Fk ) (2)

As before, we can permute the first Levi-Civita symbol so that (2) becomes:

εimn εijk ∂/∂xn (∂/∂xj Fk ) (3)

We can now apply the “ε – δ” relationship to (3) to obtain:

Gm = (δjmδkn – δjnδkm) (∂/∂xn (∂/∂xj Fk )) (4)

Equation (4) admits two terms; in the first term on the RHS of (4), we see that:

j=m
k=n

In the second term on the RHS:

j=n
k=m

Making these substitutions into (4) we get:

Gm = ∂/∂xn (∂/∂xm Fn ) - ∂/∂xn (∂/∂xn Fm) (5)

We are almost done, believe it or not. We just have to recognize what these terms in (5)
represent. Let’s look at the first term in the RHS of (5). We can rearrange the order of
differentiation (always valid for continuously differentiable functions) and obtain:
 ∂/∂xm (∂/∂xn Fn ) (6)

However, it should be apparent that the term in parentheses, (∂/∂xn Fn) is merely
∇· F, and ∂/∂xm of ∇· F is the “mth” component of the gradient of the scalar function
∇· F.

We can write the second term on the RHS of (5) as (∂/∂xn ∂/∂xn )Fm . This is just the
“mth” component of ∇2F, and putting all this together yields the desired identity:

∇ x (∇ x F) = ∇(∇· F) − ∇2F