Quick Notes: The Levi-Civita Symbol, some of its properties and

examples of use
Alejandro Jaramillo1
1
Grupo Interacción Micro y Mesoescala, Centro de Ciencias de la Atmósfera, UNAM

June 22, 2020

Abstract
This note is part of the collection of quick notes, corresponding to short writings on diverse topics of everyday use in the
Atmospheric Sciences and Geophysical Fluid Dynamics.
Posted on Authorea 22 Jun 2020 — CC-BY 4.0 — https://doi.org/10.22541/au.159285584.47325232 — This a preprint and has not been peer reviewed. Data may be preliminary.

This note shows some properties of the Levi-Civita Symbol, which is extremely useful in tensor manipulations, greatly simplifying
the vectorial calculations related to the vector product between vectors. These notes are based on chapter 1 of the excellent
book by (Sepúlveda, 2009).

“Half of mathematics is the art of saving space.”

Mokokoma Mokhonoana

The Levi-Civita Symbol

Let {eˆi } be the set of the unitary elements of a orthogonal base for the tridimensional space, where i = 1, 2, 3.
Because the elements are orthonormal we can write:

eˆi · eˆj = δij , (1)

where δij is the Kronecker delta defined as:

(
0, if i 6= j
δij = (2)
1, if i = j

One definition of the Levi-Civita Symbol comes from the cross product between the elements of the orthogonal
unit vectors that forms a base for the tridimensional space:

3
X
eˆi × eˆj = ijk eˆk , (3)
k=1

where ijk is the Levi-Civita symbol and is defined as:


 1,
 if (i.j, k) is an even permutation of (1, 2, 3)
ijk = −1, if (i.j, k) is an odd permutation of (1, 2, 3) (4)

0, if there is a repeated index


1
 An even permutation of the index is defined in the direction of the flow of the arrows in Fig. 1. They
correspond to the permutations 123, 231, and 312. An odd permutation corresponds to flow against the
arrows, i.e., 132, 321, and 213.
One simple algebraic form to express these properties of the the Levi-Civita Symbol is given by:

1
ijk = (i − j)(j − k)(k − i). (5)
2

From the definition of ijk given by Eq. (4), is clear that any change of contiguous index will introduce a
minus sign, i.e, ijk = −ikj or ijk = −jik .
Posted on Authorea 22 Jun 2020 — CC-BY 4.0 — https://doi.org/10.22541/au.159285584.47325232 — This a preprint and has not been peer reviewed. Data may be preliminary.

Figure 1: Even permutation of the index

Some properties of the Levi-Civita Symbol

One of the main applications of the Levi-Civita symbol is in the simplification of vector operations by using
the summation notation and the benefits and simplifications
P3 that this notation brings for vector calculation.
A very useful relationship is obtained from the k=1 ijk lmk . Explicitly writing the terms of the summation
gives:

3
X
ijk lmk = ij1 lm1 + ij2 lm2 + ij3 lm3 (6)
k=1

It is clear that for each one the three elements of the right hand side, the values for the index i, j, l and m
are restricted by the corresponding value of k. For example, for the first term ij1 lm1 : i, j, l and m can
only be 2 or 3, otherwise this term will be zero. This leads us to consider only the next possibilities:



 231 231 = 1

  
321 321 = 1
ij1 lm1 = (7)
 231 321 = −1


321 231 = −1


2
 From Ec. (7), it is evident that ij1 lm1 = 1 if i = l and j = m; ij1 lm1 = −1 if i = m and j = l, and
that ij1 lm1 = 0 if i = j and m = l. It is straightforward to show that the same applies for the case k = 2, 3.
From Eq. (6) and using Eq. (7), it is clear that when i, j, l, and m take fixed values, at most one of the
three terms in Eq. (6) will be different from zero. We can write then:

3
X
ijk lmk = δil δjm − δim δjl , (8)
k=1

Which can also be written in determinant form as:

3  
X δ δim 
ijk lmk =  il (9)
δjl δjm 
k=1

From Eq. (9), it seems that there is a relationship between the multiplication of two Levi-Civita symbols
and the determinant. To find a relationship, let us study the multiplication of ijk lmn . We need to explore
all of the possible non-zero elements of this product. These elements are tabulated in tables 1, 2, 3, 4 and 5.
Posted on Authorea 22 Jun 2020 — CC-BY 4.0 — https://doi.org/10.22541/au.159285584.47325232 — This a preprint and has not been peer reviewed. Data may be preliminary.

From Table 1 for i = l, we get 1 if j = m and k = n, and we get −1 if j = n and k = m, i.e, we will have a
term of the form δil δjm δkn − δil δjn δkn .

Table 1: Non-zero elements of ijk lmn with i = l

123 123 = 1 231 231 = 1 312 312 = 1
132 132 = 1 213 213 = 1 321 321 = 1
132 123 = −1 231 213 = −1 312 321 = −1
123 132 = −1 213 231 = −1 321 312 = −1

From Table 2 for j = m, we get 1 if i = l and k = n, and we get −1 if i = n and k = l, i.e, we will have a
term of the form δil δjm δkn − δin δjm δkl .

Table 2: Non-zero elements of ijk lmn with j = m

312 312 = 1 123 123 = 1 231 231 = 1
213 213 = 1 321 321 = 1 132 132 = 1
312 213 = −1 123 321 = −1 231 132 = −1
213 312 = −1 321 123 = −1 132 231 = −1

From Table 3 for k = n, we get 1 if i = l and j = m, and we get −1 if i = m and j = l, i.e, we will have a
term of the form δil δjm δkn − δim δjl δkn .

Table 3: Non-zero elements of ijk lmn with k = n

231 231 = 1 132 132 = 1 123 123 = 1
321 321 = 1 321 312 = 1 213 213 = 1
231 321 = −1 132 312 = −1 123 213 = −1
321 231 = −1 312 132 = −1 213 123 = −1

From Table 4 for i = m, we get 1 if j = n and k = l, and we get −1 if j = l and k = n, i.e, we will have a
term of the form δim δjn δkl − δim δjl δkn .

3
 Table 4: Non-zero elements of ijk lmn with i = m
123 312 = 1 231 123 = 1 312 231 = 1
132 213 = 1 213 321 = 1 321 132 = 1
123 213 = −1 231 321 = −1 312 132 = −1
132 312 = −1 213 123 = −1 321 231 = −1

Finally, From Table 5 for i = n, we get 1 if j = l and k = m, and we get −1 if j = m and k = l, i.e, we will
have a term of the form δin δjl δkm − δin δjm δkl .

Table 5: Non-zero elements of ijk lmn with i = n

123 231 = 1 231 312 = 1 312 123 = 1
132 321 = 1 213 132 = 1 321 213 = 1
123 321 = −1 231 132 = −1 312 213 = −1
132 231 = −1 213 312 = −1 321 123 = −1
Posted on Authorea 22 Jun 2020 — CC-BY 4.0 — https://doi.org/10.22541/au.159285584.47325232 — This a preprint and has not been peer reviewed. Data may be preliminary.

Note that we can summarize the results obtained from Tables 1 to 5, by writing the multiplication of ijk lmn
as:

ijk lmn = δil δjm δkn + δim δjn δkl + δin δjl δkm − δil δjn δkn − δin δjm δkl − δim δjl δkn
(10)
= δil (δjm δkn − δjn δkn ) − δim (δjl δkn − δjn δkl ) + δin (δjl δkm − δjm δkl ) ,

Which can be written using the determinant as:

 
 δil δim δin 

ijk lmn =  δjl δjm δjn  (11)
δkl δkm δkn 

From Eq. (11), we can also use the Levi-Civita symbol to define the determinant of a matrix in summation
form. As we know, the determinant of a matrix is given by:
 
a11 a12 a13 

a21
 a22 a23  =a11 (a22 a33 − a23 a32 ) + a12 (a23 a31 − a21 a33 ) + a13 (a21 a32 − a22 a31 ) (12)
a31 a32 a33 

From Eq. (12), we can write the determinant of a matrix as the triple scalar product ~a · (~b × ~c), where ~a, ~b
and ~c are defined as:

3
X
~a = a1i eˆi ,
i=1
3
X
~b = a2j eˆj , (13)
j=1
3
X
~c = a3k eˆk .
k=1

From these definitions and using the triple scalar product we get:

4
 3
X 3
X
= a1i eˆi · a2j a3k eˆj × eˆk
i=1 jk=1

X 3

a11
 = a1i a2j a3k ljk eˆi · eˆl
a12 a13 
ijkl=1
a23  = ~a · ~b × ~c

a21 a22 (14)
 3
a31 a32 a33  X
= a1i a2j a3k ljk δil
ijkl=1
3
X
= ijk a1i a2j a3k
ijk=1

Examples
Posted on Authorea 22 Jun 2020 — CC-BY 4.0 — https://doi.org/10.22541/au.159285584.47325232 — This a preprint and has not been peer reviewed. Data may be preliminary.

Here we solve solve problems related to the use of the Levi-Civita Symbol.

Example 1.
P3
Show that jk=1 ijk ljk = 2δij .

Solution
Using Eq.(8) we can write:

3 X
X 3 3
X
ijk ljk = (δil δjj − δij δjl )
j=1 k=1 j=1
3
X
= (δil − δij δjl )
j=1
3
X 3
X
= (δil ) − (δij δjl )
j=1 j=1

= 3δil − δil
= 2δil

Example 2
P3
Show that ijk=1 ijk ijk = 6.

Solution
Using Eq. (8) we can write:

5
 3 X
X 3 X
3 3 X
X 3
ijk ijk = (δii δjj − δij δji )
i=1 j=1 k=1 i=1 j=1
3 X
X 3 3 X
X 3
= (δii δjj ) − (δij δji )
i=1 j=1 i=1 j=1
3
X 3
X
= (3δii ) − (δii )
i=1 i=1
=9−3
=6

Example 3
P3
Show that j=1 δij ijk = 0.

Solution
Posted on Authorea 22 Jun 2020 — CC-BY 4.0 — https://doi.org/10.22541/au.159285584.47325232 — This a preprint and has not been peer reviewed. Data may be preliminary.

Using the properties of the Kronecker delta we have:

3
X
δij ijk = iik
j=1

=0

Example 4
~×B
Show that A ~ = P eˆi ijk Aj Bk .
ijk

Solution
~ and B
We can write A ~ in term of their components {Ai } and {Bi } in the base {eˆi }. Therefore, we have:

3
X 3
X
~=
A ~ =
Aj eˆj and B Bk eˆk
j=1 k=1

From this we can write:

3 X
X 3
~×B
A ~ = Aj Bk eˆj × eˆk
j=1 k=1
3 X
X 3 3
X
= Aj B k jki eˆi (Using Eq. (3) )
j=1 k=1 i=1
3
X
= eˆi ijk Aj Bk (Reordering the index, two changes)
ijk=1

6
 Example 5
~·B
Show that A ~ ×C
~ = P ijk Ai Bj Ck .
ijk

Solution

3
X 3
X
~·B
A ~ ×C
~ = Ai eˆi · eˆl ljk Bj Ck (Using Eq.(3))
i=1 ljk=1
3
X
= δil ljk Ai Bj Ck
iljk=1
3
X
= ijk Ai Bj Ck
ijk=1

Example 6
Posted on Authorea 22 Jun 2020 — CC-BY 4.0 — https://doi.org/10.22541/au.159285584.47325232 — This a preprint and has not been peer reviewed. Data may be preliminary.

~·B
Show that A ~ ×C
~ =B
~ ·C
~ ×A
~=C
~ ·A
~ × B.
~

Solution
Using the results from Example 5 we have:

3
X
~·B
A ~ ×C
~ = ijk Ai Bj Ck
ijk=1
3
X
= δjl ilk Ai Bj Ck
ijkl=1
3
X 3
X
= Bj eˆj · eˆl lki Ck Ai
j=1 jkl=1
~ ·C
=B ~ ×A
~
3
X
= δkl ijl Ai Bj Ck
ijkl=1
3
X 3
X
= Ck eˆk · eˆl lij Ai Bj
k=1 ijl=1
~ ·A
=C ~×B
~

Example 7
~ × (B
Show that A ~ × C)
~ = (A
~ · C)
~ B~ − (A
~ · B)
~ C.
~ This is also known as the BAC-CAB rule.

7
 Solution

3
X 3
X
~·B
A ~ ×C
~ = Ai eˆi × eˆl ljk Bj Ck
i=1 ljk=1
3
X
= Ai Bj Ck ljk eˆi × eˆl
iljk=1
3
X
= Ai Bj Ck ljk ilm eˆm
iljkm=1
3
X 3
X
= (−1)Ai Bj Ck eˆm jkl iml
ijkm=1 l=1
3
X
= (−1)Ai Bj Ck eˆm (δji δkm − δjm δki )
ijkm=1
3 3
Posted on Authorea 22 Jun 2020 — CC-BY 4.0 — https://doi.org/10.22541/au.159285584.47325232 — This a preprint and has not been peer reviewed. Data may be preliminary.

X X
= Ai Bj Ck eˆm δjm δki − Ai Bj Ck eˆm δji δkm
ijkm=1 ijkm=1
3
X 3
X 3
X 3
X
= Ai Ck eˆi · eˆk Bj eˆj − Ai Bj eˆi · eˆj Ck eˆk
ik=1 j=1 ij=1 k=1
~ · C)
= (A ~ B~ − (A
~ · B)
~ C~

Example 8
~ · (B
Show that A ~ × A)
~ = 0.

Solution
Using Example 6 we have:

~·B
A ~ ×A
~=B
~ ·A
~×A
~
=0

Example 9
~ × B)
Show that (A ~ · (C
~ × D)
~ = (A
~ · C)(
~ B ~ · D)
~ − (A
~ · D)(
~ B ~ · C).
~

8
 Solution
  !
X X
~ × B)
(A ~ · (C
~ × D)
~ = eˆi ijk Aj Bk  · eˆl lmn Cm Dn
ijk lmn
X
= Aj Bk Cm Dn ijk lmn δil
ijklmn
X
= Aj Bk Cm Dn ijk imn
ijkmn
X X
= Aj Bk Cm Dn jki mni
jkmn i
X
= Aj Bk Cm Dn (δjm δkn − δjn δkm )
jkmn
X X
= Aj Bk Cm Dn δjm δkn − Aj Bk Cm Dn δjn δkm
jkmn jkmn
X X X X
Posted on Authorea 22 Jun 2020 — CC-BY 4.0 — https://doi.org/10.22541/au.159285584.47325232 — This a preprint and has not been peer reviewed. Data may be preliminary.

= Aj Cm δjm Bk Dn δkn − Aj Dn δjn Bk Cm δkm

jm jkmn jn km
~ · C)(
= (A ~ B ~ · D)
~ − (A
~ · D)(
~ B ~ · C)
~

Example 10
~ × B)
Show that (A ~ · (A
~ × B)
~ = A2 B 2 − (A
~ · B)
~ 2.

Solution
Using Example 9 we have

~ × B)
(A ~ · (A
~ × B)
~ = (A~ · A)(
~ B ~ · B)
~ − (A~ · B)(
~ B ~ · A)
~
= A2 B 2 − (A~ · B)
~ 2

References
Fı́sica matemática. (2009). Universidad de Antioquia.