Levi-Civita Symbol

(Dated: 2017-02-21)

1
I. LEVI-CIVITA SYMBOL

To handle cross products and combinations of cross products and dot products of vectors
it is convenient to use the Levi-Civita symbol. Consider a cross product of two vectors a
and b:

c=a b: (1)

The Cartesian components of c can be written as

X
ci = ijk aj bk ; (2)
jk
or

X
ck = ai b j ijk; (3)
ij

where the Levi-Civita symbol is de…ned according to

xyz = yzx = zxy = 1; (4)

xzy = yxz = zyx = 1; (5)

and all other components are zero. Thus, all indices must be di¤erent for ijk to be non-zero.
As can be seen from the de…nition, interchanging any two indices changes the sign of .
A curl r a can also be written using the Levi-Civita symbol as follows:

X
(r a)i = ijk @j bk ;
jk

where

@ @ @
@x = ; @y = ; @z = :
@x @y @z
Let us verify that the Levi-Civita symbol reproduces the de…nition of cross product:

cx = xyz ay bz + xzy az by : (6)

2
These are the only non-zero terms in the summation over the j and k indices in (2). Since
xyz = 1 and xzy = 1 we …nd

c x = ay b z az b y : (7)

Similarly,

cy = yzx az bx + yxz ax bz = az b x ax bz; (8)

To handle vector algebra using the Levi-Civita symbol, the following formula is very
useful:

X X
ijk klm = ijk lmk = il jm im jl : (10)
k k

This formula can be understood by realising that ijk is zero unless all indices are di¤erent.
For example, if i = x and j = y only the following two possibilities give a non-zero result:

xyz zxy = xyz xyz = 1 when l = i = x and m = j = y; (11)

or

xyz zyx = xyz yxz = 1 when l = j = y and m = i = x: (12)

Similar results are obtained for other choices of i and j and therefore we obtain the gen-
eral result in (10). In some books the summation symbol is sometimes left out with the
convention that repeated indices are summed:

ci = ijk aj bk ; (13)

ijk klm = ijk lmk = il jm im jl : (14)

3
II. EXAMPLES

A. Example 1

r=a (b c): (15)

X
ri = ijk aj (b c)k
jk
X X
= ijk aj klm bl cm
jk lm
!
X X
= ijk klm aj b l c m : (16)
jlm k

We now use the formula in (10) and obtain

X
ri = ( il jm im jl ) aj bl cm
jlm
X
= (aj bi cj aj b j c i ) : (17)
j

Since

X X
aj cj = a c; aj bj = a b,
j j

we then …nd

r = b(a c) (a b)c:

B. Example 2

V = a (b c) :

4
 X
V = ai (b c)i
i
X X
= ai ijk bj ck
i jk
!
X X
= ijk ai bj ck
k ij
X
= (a b)k ck
k

= (a b) c:

We can also in the third line of the above equation write

!
X X
V = ijk ai ck bj
j ik
!
X X
= jik ai ck bj
j ik
X
= (a c)j bj
j

= (a c) b

Thus we have

a (b c) = (a b) c = (a c) b:

C. Example 3

r=r (a b) :

5
 X
ri = ijk @j (a b)k
jk
!
X X
= ijk @j klm al bm
jk lm
!
X X
= ijk klm @j (al bm ):
jlm k

Using the formula in (10) we …nd

X
ri = ( il jm im jl ) (bm @j al + al @j bm )
jlm
X
= (bj @j ai + ai @j bj bi @j aj aj @j bi ) :
j

Since

X X X X
bj @j ai = (b r)ai ; ai @j bj = ai (r b); bi @j aj = bi (r a); aj @j bi = (a r)bi ;
j j j j

we obtain

r (a b) = (b r)a + a(r b) b(r a) (a r)b:

D. Example 4

V = r (a b) :

X
V = @i (a b)i
i
X X
= @i ijk aj bk
i jk
X
= ijk @i (aj bk )
ijk
X
= ijk (bk @i aj + aj @i bk ) :
ijk

6
The …rst term is

!
X X X
ijk bk @i aj = bk kij @i aj
ijk k ij
X
= bk (r a)k
k

= b (r a)

and the second term is

!
X X X
ijk aj @i bk = aj jik @i bk
ijk j ik
X
= aj (r b)j
j

= a (r b):

Therefore

r (a b) = b (r a) a (r b):

E. Example 5

We have two useful identities

r=r (r ) = 0

and

r (r a) = 0:

We …rst show the …rst identity,

X
ri = ijk @j (r )k
jk
X
= ijk @j @k :
jk

7
Since

@j @k = @k @j ;

we have for the x component

rx = ( xyz + xzy )@y @z = 0;

because xzy = xyz . Similarly for the other components of r.

X
r (r a) = @i (r a)i
i
X X
= @i ijk @j ak
i jk
X
= ijk @i @j ak :
ijk

For k = x and using @i @j ak = @j @i ak we have on the right-hand-side

( yzx + zyx )@y @z ax = 0:

This is true for any k and therefore

r (r a) = 0: