Noether Theorem, Noether Charge and All That

Created for PF by Samalkhaiat

1.0 Transformations

Let G be a Lie group whose action on Minkowski space-time M(1,3) , is formally

realized by coordinate transformations
x x = g x,

g G .

(1.1)

Infinitesimally, we may write this as

x = x + x ,

(1.2)

where x = f ( x; ) are smooth functions of the coordinates and 1 ,

= 1, 2,..., dim(G) , are the infinitesimal group parameters.
We assume that our dynamical variables (the fields) are described by continuous
functions r {1, 2,.., n},r ( x) 2 (M(1,3) , n ) and transform by finite-dimensional

( n n matrix) representation of the group G on M(1,3) : ( (G ), M(1,3) ),

r ( x) r ( x ) = Dr s ( g ) s ( x) ,

(1.3)

Dr s ( g1 g 2 ) = Dr t ( g1 ) Dt s ( g 2 ) .

(1.4)

Close to the identity element of the group, i.e. infinitesimally, we can write
Dr s ( g ) = r s + ( ) r s ,

(1.5)

where the n n matrices s form a representation of the Lie algebra of the group

[ , ] = iC .

(1.6)

Using (1.5), we can rewrite (1.3) as

*r ( x) = r ( x ) r ( x) ,

(1.7)

*r ( x) ( ) r s s ( x) .

(1.8)

where

For later use, let us now derive some of the important properties of the variation
symbol * . From (1.6) and (1.8), we see that

[* , * ] r ( x) = iC * r ( x) .

(1.9)

Using its definitions (1.7), we can easily show that * acts as derivation on functions

* ( f ( x) g ( x)) = ( * f ( x)) g ( x) + f ( x) ( * g ( x)) .

(1.10)

Since *r is written entirely in terms of the primary fields without time derivatives,
it (anti)commutes with s ( y ) at equal time
[ *r (t , x ) , s (t , y )] = 0 .

(1.11)

To first order in x we can write

r ( x ) = r ( x ) + x r ( x ) .

(1.12)

Using this expansion, we can rewrite (1.7) as

*r ( x) = r ( x) + x r ( x) ,

(1.13)

r ( x) r ( x) r ( x) ,

(1.14)

where

is the infinitesimal change in the functional form of fields. If the field index r is a
space-time index taking values in the set {, , , ,...} , in 4-dimensional spacetime r ( x) is nothing but the Lie derivative of a space-time tensor r ( x) with
respect to the infinitesimal transformation (1.2). The transformation considered is
called space-time (symmetry) transformation if x 0 and internal (symmetry)
transformation if x = 0 ; ( * ) is called the orbital part of the (symmetry)
transformation.
Commuting both sides of (1.13) with s (t , y ) (at equal time) and using (1.11), we
find
[ r (t , x ) , s (t , y )] = x 0 [r (t , x ) , s (t , y )] .
(1.15)
And finally, we note that it will be useful to write (1.13) as an operator equation

* = + x ,

(1.16)

with ( * , ) are given by their usual definitions on functions (1.7) and (1.14). Notice
that, while commutes with derivatives, * does not commute with :

[ , * ] = ( x ) .

(1.17)

Exercise (1.1) In D-dimensional Minkowski space-time, show that

x = + x + x + c ( 2 x x x 2 ) ,
is the most general solution to the conformal Killing equation

( x ) + ( x ) =

2
( x ) .
D

Solution is given in [1].

Exercise (1.2) Let f ( x) be a solution to the conformal Killing equation

f + f = 2 ( x) g .
Show that f x is a constant along the geodesic.
Solution:

d
1
f x ) = x ( f x ) = ( f + f ) x x = g x x = 0 .
(
d
2

2.0 Invariance of the Action and Noether Theorem

Emmy Noether [1882-1935]

The action [ S :( (G ), ) ] is given by the integral

S[r ] = d 4 x L(r ( x), r ( x))

where 4 is an arbitrary, contractible and bounded region in (M(1,3) , ) . The

Lagrangian density L( x) is a local real function of fundamental (primary)
fields r ( x) , that is, it is constructed from r ( x) and their derivatives, r , at the
same space-time point x . In order for the canonical formalism to make sense, L( x)
is assumed to contain no second or higher derivatives of r ( x) and to be at most
quadratic in r ( L : n 4 n ). It is also assumed that the fields are wellbehaved and vanish sufficiently rapidly at infinity.
Under the transformation (1.1), a region D M(1,3) is mapped into a new region D by
point-to-point correspondence. Therefore, infinitesimally g ( ) induces an
infinitesimal change in the region of integration
g ( ) : = +
and maps the action integral into
g ( ) : S [r ] S [ r ] =

d 4 x L(x )

(2.1)

where

L ( x ) L( r ( x ), r ( x )) .
If the action integral is invariant (i.e., S [r ] = S [r ], and r ) under certain group
of transformations (with parameters ), the theory is said to have a symmetry
corresponding to those transformations. Therefore, in order for the group G to be a
symmetry group of our theory, we must have

( d x L( x ) ) =
4

d 4 x L ( x ) d 4 x L( x ) = 0 .

(2.2)

Changing the dummy integration variable in the second integral, we get

d 4 x L ( x ) d 4 x L( x ) = 0 .

(2.3)

L ( x ) = L( x ) + L( x ) ,

(2.4)

Using the relation

we can rewrite (2.3) in the form

d 4 x L( x ) d 4 x L( x ) + d 4 x L( x ) = 0 .

(2.5)

We know from elementary calculus that

b + b

a + a

dx f ( x) dx f ( x) f (b) b f ( a) a = dx

d
( f ( x) x ) .
dx

(2.6)

Generalizing this to the 4-dimensional case in (2.5), we find (to first order in x )
that

d 4 x ( L x ) + L( x ) = 0 .

(2.7)

The same result can be obtained if we apply the derivation property of * on the lefthand side of (2.2)

*
Using the relation

( d x L( x ) ) = ( d x ) L( x ) + d x L ( x ) = 0 .
4

(2.8)

det |1+ | = exp ( Tr In |1+ |) = exp ( Tr ) ,

we can expand the Jacobian of the transformations to first order as

x
det = det ( + x ) = 1 + Tr ( x ) 1 + x .
x
Thus
x
d 4 x = det d 4 x = d 4 x + ( x )d 4 x, * (d 4 x) = d 4 x ( x ) .
x

(2.9)

And applying the operator equation (1.16) to the Lagrangian, gives us

*L = L + x L .

(2.10)

Putting (2.9) and (2.10) in equation (2.8) gives us back equation (2.7).
Since 4 is an arbitrary contractible region, the integrand in equation (2.7) must
vanishes identically

L + (L x ) 0 .

(2.11)

This is an identity with respect to all its arguments, if the group G is an invariance
group of the action integral. It holds for all functions r ( x) , it does not matter whether

r ( x) s are solutions of the Euler-Lagrange equations or not. Since the divergence

term is not discarded, the behaviour of r ( x) on the boundary, , is also irrelevant.
Since [ , ] = 0 , we can write

 L( x) =

L
L
r +
(r ) .
r
( r )

(2.12)

Introducing the Euler derivatives

L
L L

r
r

(2.13)

into equation (2.12) and inserting the resulting equation back into equation (2.11), we
arrive at the oether Identity

L
L
r + L x +
r 0 .

( r )

(2.14)

Let us assume the fields r ( x) are solutions of the Euler-Lagrange equations

L
= 0, r = 1, 2,..., n .
( x)

(2.15)

Thus, the Noether identity implies that the object (Noether current)
J ( x; ) L x +

L
r ,
( r )

(2.16)

satisfies the conservation law

J = 0 .

(2.17)

The existence of the conserved (symmetry) current J ( x) is known as the first

oether theorem. Of course, physical currents do not depend on the parameters of
the symmetry group. Indeed, they can be factored out of equation (2.16) leaving a
group index on the physical current
J ( ) ( x) =

L
r + L x , = 1, 2,..., dim(G) .
( r )

(2.18)

Thus, there are as many conserved currents as there are parameters. Notice that the
conserved current is not unique. In fact, adding a total divergence of antisymmetric
tensor to the Noether current does not spoil the conservation law,

J ( ) = J ( ) + F ( ) , F ( ) = F ( ) .

(2.19)

Quantities like these F are called super-potentials. They play important role in the
construction of the conserved charges in general relativity and other generally
covariant theories. Below, we will use this freedom to construct the symmetric
energy-momentum tensor out of the non-symmetric canonical tensor.
6

It will be instructive to give another derivation of the Noether current which is

ideologically different from the one we have just done. It consists in comparing the
change L without using the equations of motion (i.e. the off-shell variation) with
that when the Euler-Lagrange equations are satisfied (i.e. with the on-shell variation
of L ). In this method, the explicit form of L as well as r are assumed known.
According to (2.12), an arbitrary infinitesimal change in the fields induces the
following change in the Lagrangian
L
L
L

r .
r +

( )
( )

r
r
r

=
L

Using the equations of motion (i.e. on-shell), we find

L
r .
( r )

L =

(2.20)

This is always true, whether or not r is a symmetry transformation.

On the other hand, if without the use of equations of motion (i.e. just by substituting
r in (2.12)) the Lagrangian changes by a total divergence of some object ,

L = ,

(2.21)

then the action remains unchanged, and the transformation, r , is a symmetry

transformation of the theory.
From (2.19) and (2.20), it follows that the conserved Noether current is

J ( x) =

L
r ( x) .
( r )

(2.22)

Notice that using this method to derive the current, you do not need to know how x
transforms under the symmetry group in question. So, how can you distinguish
between internal and space-time symmetries, if you dont have x ? To answer this
question we use the fact that adding total divergences to a Lagrangian does not affect
the dynamics. Indeed, it is always possible to find a dynamically equivalent
Lagrangian L ( x) such that
L ( x) = L( x) + .
(2.23)

Thus

L = L + = ( + ) .

(2.24)

So, if = 0 , or it is possible to remove it by a suitable choice of , then you are

dealing with an internal symmetry. Otherwise, it is space-time symmetry.

2.1 Translation Invariance and the Stress Energy-Momentum

Tensor
Using the relation

r ( x) = *r ( x) x r ( x) ,

(2.25)

which follows from (1.16) when we factor out the infinitesimal parameters , we
can rewrite (2.18) in the form
J ( ) ( x ) =

L
*r ( x) x ,
( r )

(2.26)

L
r L( x) .
( r )

(2.27)

where

( x)

Introducing the canonical conjugate field

r ( x)

L( x)
,
r ( x)

(2.28)

into the 00 component, we find

H( x) 00 ( x) = r ( x) r ( x) L( x) .

(2.29)

In the classical mechanics of continuous system, H( x) is nothing but the Hamiltonian

density of the system. Therefore, it follows from considerations of covariance that
represents the canonical stress energy-momentum tensor of the field. Let us
calculate its divergence on shell, i.e. by using the Euler-Lagrange equations of
motion,

L
L
L
r +
r L( , , x) = .
r
( r )
x

Clearly, this vanishes (on shell) if the Lagrangian does not depend explicitly on x .
Thus, the energy-momentum tensor is the conserved Noether current associated with
the symmetry group of space-time translations. Indeed, under space-time translations

x = , x = .

(2.30)

And all fields, regardless of their tensorial character, are invariant,

*r ( x) = 0 .

(2.31)

Putting (2.24) and (2.25) into the defining equation (2.20) of J ( ) ( x) , we find

J ( x) ( x) .

Thus, it follows trivially from (2.17) that the energy-momentum tensor is conserved

( x) = 0 .

(2.32)

2.2 Lorentz Invariance and the fields Moment Tensor

Under infinitesimal Lorentz transformations

x = x ,

(2.33)

the field transforms according to

i
2

*r ( x) = ( )r s ( x) ,
s

(2.34)

where s are the appropriate (spin) transformation matrices for the field r . Using
these relations, we can write the conserved current, (2.26), in the form

1
L
J = i
( ) r s s ( x x ) .
( )

2
r

This implies that the moment tensor

M = ( x x ) i
= M

L
( )r s s
( r )

(2.35)

is conserved. Thus, a field theory is Poincare invariant if and only if

M = + S = 0 .

(2.36)

The object

L
( ) r s s ( x) ,
(2.37)
( r )
depends entirely on the intrinsic properties (tensorial nature) of the field. In classical
field theory, it characterizes the polarization properties of the field. Thus, it
corresponds to the spin (angular momentum) of particles described by the quantized
field.
For a single component field (scalar), the (spin) matrix vanishes and (2.35)
becomes
M = x x .
S ( x) i

This relation is very muck like the relation between momentum and orbital angular
momentum in the classical mechanics of particle xi p j x j pi =ijk Lk . Therefore, the
first term in (2.35),
L = x x ,
(2.38)
should correspond to some intrinsic orbital angular momentum of the field. Thus
the moment tensor
M = L + S

(2.39)

must be related the total angular momentum tensor of the field.

From (2.35) and (2.36) we can clearly see the connection between the symmetry
properties of the canonical energy-momentum tensor and the structure of the total
moment tensor M . For a scalar field, since S = 0 , we find that = . On the
other hand, for arbitrary multi-component field, is not necessarily symmetric.
However, the Poincare invariance condition, (2.36), together with the freedom of
adding super-potential to the current, (2.19), allow us to construct a symmetric
energy-momentum tensor. Since S = S , let us seek for a quantity F such
that
S = F F .
(2.40)
Putting this in the Poincare invariance condition (2.36), we find

+ F
= + F .
Therefore, if we set
T + F ,

(2.41)

T = T .

(2.42)

we see that

Furthermore the conservation of the canonical energy-momentum tensor, = 0 ,

implies
T = F .
(2.43)
Thus, the conservation of the symmetric tensor T demands (super-potential)
F = F

(2.44)

Now, we can use (2.40) and (2.44) to solve for the super-potential
1
F = (S S + S ) .
2

(2.45)

10

The conserved tensor T is called the Belinfante symmetric energy-momentum

tensor. In terms of this symmetric tensor, the total moment tensor becomes

M = ( x T x T ) + (F F S ) + P ,
P = F x F x
The quantity in the middle bracket vanishes when (2.40) and (2.44) are used, and if
we drop the super-potential P , we arrive at the Belinfante moment tensor
M = x T x T .

(2.46)

Notice that the (internal) spin part has disappeared from the total moment tensor
whose conservation now follows from the conservation of the symmetrical energymomentum tensor.
The physical importance of the symmetrical (Belinfante) energy momentum tensor
follows from the belief that gravitons couple to T , and not to . Indeed, if the
matter theory is minimally coupled to gravity and its action is varied with respect to
the metric tensor g , the symmetric energy momentum tensor is obtained in the limit
g

T =

d D x L ( , , g ) .

g g

(2.47)

With the symmetrical energy momentum tensor, the conserved Poincare currents
take on the compact (Bessel-Hagen [2]) form

J = T x .

(2.48)

Since x = + x , we can rewrite the current in the form

1
J = T + x T = T ( x T x T ) ,
2

(2.49)

or
1

J = T M
2

(2.50)

Taking the divergence of the current (2.48), and using the symmetry of the energy
momentum tensor, we find
1
J = ( T ) x + T ( x + x ) = 0 .
2
The first term on the right hand side vanishes because of T = 0 , and the second
term vanishes because the Poincare transformations (Killing vector fields)

x f ( x) = + x ,

(2.51)

11

are the most general solutions to the Killing equation (in any number of dimensions)
f + f = 0 .

(2.52)

J = T x ,

(2.53)

Now consider the current

where T is the conserved and symmetric energy momentum tensor, and x = f is

vector field satisfying the conformal Killing equation in D-dimensional space-time
f + f =

2
f .
D

(2.54)

Taking the divergence of the current, we find

J =

f
T .
D

(2.55)

Thus, the current (2.53) is conserved provided that the conserved and symmetric
energy momentum tensor is traceless T = 0 . Indeed, conformal invariance allows us
to further improve the Belinfante energy-momentum tensor so that it is traceless much
in the same way that Poincare invariance allowed us to make the canonical Energymomentum tensor symmetric [1]. This way, the conserved conformal currents can be
written in the compact Bessel-Hagen form (2.53).

2.3 Poincare and Scale Invariance and the Traceless Energy

Momentum Tensor
Under scale (dilatation) transformation
x x = e x , x = x ,

(2.56)

the fields transform according to

and

r ( x) r ( x ) = ed r ( x), * r ( x) =d r ( x) .

(2.57)

r ( x) = ( d + x ) r ( x) .

(2.58)

where the real number d represents the scaling dimension of the field r (we assume
all components have the same scaling dimension). In D dimensions, its value for
scalar and vector fields, is given by
D2
d=
.
(2.59)
2
Using (2.26), we find the canonical dilatation current
12

D = x +

L
d r .
( r )

(2.60)

Thus
L

d r .
D = +
( )

(2.61)

Therefore, in scale invariant theories, the trace of the canonical energy-momentum

tensor is given by

L
d r .
( r )

(2.62)

We will now show that Poincare and scale invariance allow us to construct
conserved, symmetric and traceless energy momentum tensor. Recall that for any
rank-3 tensor R = R , anti-symmetric in the last two indices, the following
combination is a tensor anti-symmetric in the first two indices
F =

1
(R + R R ) = F .
2

(2.63)

= 0 and the following modified energy-momentum tensor is

Thus, F
conserved

1
T = + (R + R R ) .
2

(2.64)

Taking the trace, T = T , we find

T = + R .

(2.65)

Using the condition for scale invariance, (2.62), the trace becomes

L
T = R
d r .
( r )

(2.66)

Recall that, in order to symmetrize the energy-momentum tensor, Poincare

invariance allowed us to choose R = S , see (2.45). So, for Poincare and scale
invariant theories, traceless energy-momentum tensor can be obtained by writing
R = S +

1
V V ) ,
(
D 1

(2.67)

13

where the vector field V is determined from (2.66) by setting T = 0 :

L

d r S .
V =
( )

(2.68)

Notice that for a scalar field, the spin matrix vanishes. Thus from (2.37) we find
S = 0 . Therefore, if we consider the scale invariant field theory
2D
1

D 2
L =
,
2

(2.69)

we find
V =

D2 2 2
D2 2
, V =
.
4
4

(2.70)

Thus, the conserved, symmetric and traceless energy-momentum tensor is

T

= T +

D2
2 ) 2 ,
(
4( D 1)

(2.71)

where T is the symmetric energy-momentum tensor

T = L .

(2.72)

Exercise (2.1) Show that the energy momentum tensor (2.71), is conserved,
T = 0 and traceless, T = 0 .
Exercise (2.2) Consider the free Maxwell theory in D-dimensional space-time
1
L = F F .
4
Show that the symmetric, conserved energy momentum tensor is given by
the Belifante expression in any number of dimensions
1
T = F F + F 2 ,
4
but traceless only in 4 dimensions
D
T = 1 F 2 .
4

Exercise (2.3) Show that the above Maxwell theory is scale invariant in any number
of dimensions. Given that

14

A = x +

D2
A .
2

Show that (up to a trivially conserved super potential) the conserved dilation current
is given by
4 D
D = T x +
F A .
2
Show, by explicit calculations, that D = 0 .

2.4 The New Improved CCJ tensor [3] and the Belinfante tensor
from the action: fun with super-potentials
Basically, this subsection is about obtaining the results of the last two subsections
from varying the action integral. Recall that on-shell, the variation of the action can
be written as
L

(2.73)
S = d 4 x
*r x .

To this variation, we will add zero and show that the coefficient of x is the new
improved energy-momentum tensor of Callan, Coleman and Jackiw [3].
Consider the functional
R = d 4 x ( x ) ,
where is a local function of the field and its derivatives to finite order. We vary
this using (1.10),
*R = * ( d 4 x) + d 4 x * ( ) .
If in the first integral we make use of (2.9), and in the second integral we use (1.16),
we find

* R = d 4 x ( x )( ) + d 4 x ( ) + d 4 x ( x )( ) .
Combining the first and the third terms and using [ , ] = 0 in the second term, we get

* R = d 4 x ( + x ) .

(2.74)

Using (1.16) again, this can be rewritten as

* R = d 4 x * + ( ) x .

(2.75)

Integrating the second term by part, we find

15

* R = d 4 x * ( ) x + ( x x ) .
The second term vanishes identically, and we are left with
.

* R = d 4 x * ( ) x .

(2.76)

Now, we add (2.76) and subtract (2.75) from (2.73)

L
* r ( ) x
( r )

* S = d 4 x
d x
4

{(

+ ) x

(2.77)

Let us assume that our symmetry transformations change the action according to

* S = d 4 x ( x ) .

(2.78)

Now, if we can find such that

L
* r ( ) x = ( , ) ,
( r )

(2.79)

holds, then the Noether current will have the Bessel-Hagen form
J = T x ,

(2.80)

with a new improved energy-momentum tensor given by

T = + ( ) .

(2.81)

Exercise (2.4) Consider the free massless scalar field action in four dimensions
1 4
d x .

2
*

(i) Show that S = 0 , i.e. ( x) = 0 , under the Lorentz transformations

S=

x = x , * = 0 .
Then show that Lorentz invariance implies the following conditions on
= .

16

(ii) Show that * S = 0, = 0 , under the scale transformations

x = x , * = .
Use (2.79) to show that scale invariance implies
1
= 2 .
6
Use this to show that (2.81) leads to the symmetric and traceless energy-momentum
tensor.
1
T = + ( 2 ) 2 .
6
(iii) Show that the conformal transformations

x = 2 x (c x ) c x 2 , * = 2(c x ) ,
change the scalar field action by total divergence

* S = d 4 x = d 4 x ( c 2 ) .
Use Lorentz invariance and scale invariance restrictions on to show that (2.79) is
satisfied for the conformal transformations.

3.0 Canonical Quantization

Before introducing the Noether charge and discuss its properties, we need to set up
our quantization scheme. We will work with the canonical quantization rule: Poisson
bracket of functions goes to equal-time (anti)commutator of the corresponding
operators1

i{ f (t , x), g (t , y )}PB [ f (t , x), g (t , y )] .

However, we will not use the hat to indicate operators. Also, we will not write the
time argument explicitly. Unless stated otherwise, all (anti)commutators are
considered to be at equal times. So, for any local functional F , we use the following
rules

F ( x)
= [ F ( x), r ( y )] ,
r ( y )
F ( x)
i {F ( x), r ( y )}PB = i
= [ F ( x), r ( y )] .
r ( y )

i {F ( x),r ( y )}PB = i

(3.1)
(3.2)

The fundamental equal-time commutation relations follow by putting F = r , r :

1

In ref. [4] I have out-lined the proof of

lim
0

i
A B A B
( AB BA) =

x p p x
17

[ r ( x), s ( y )] = i rs 3 (x y ),
[ r ( x), s ( y )] = [ r ( x), s ( y )] = 0.

(3.3)

Notice that

r ( x) =

L( , , ) .
r

(3.4)

In most cases, the Lagrangian is quadratic in r . Thus, r is linear in r . We will

assume that the simultaneous linear equations (3.4) can be solved for r . That is we
assume that it is possible to represent the velocity by some function

r ( x) = f r ( , , ) .

(3.5)

The functional derivative of any local function of the field variables, such as the
Lagrangian, will be calculated using the chain rule:
(x)
L( x) L( x) s ( x) L( x) ( j s ) L( x) s ( x)
,
=
+
+
r ( y ) s ( x) r ( y ) ( j s ) r ( y ) s ( x) r ( y )

( x)
L( x) 3
L( x) ( x ) 3
(x y ) +
j (x y ) + s ( x) s ,
r ( y )
r ( x)
( j r )
( x)
L( x) L( x) s ( x)
=
= s ( x) sr
.
r
r
( y ) s ( x) ( y )
( y )

Thus, using (3.1) and (3.2) in the above two equations, we infer the following equaltimes commutation relations
[L( x), r ( y )] = s ( x)[s ( x), r ( y )] ,

[L( x), r ( y )] = i 3 (x y )

(3.6)

L
L
+ i (jx ) 3 (x y )
+ s ( x)[s ( x), r ( y )] (3.7)
r
( jr )

The last two terms in (3.7) can be combined to give

[L( x), r ( y )] = i

L( x) 3
L( x)
(x y ) +
[ s ( x), r ( y )] .
r ( x)
( s )

(3.8)

We will now use these commutation relations together with the fundamental equaltime commutation relations (3.3) to derive few important equations. Lets start by
deriving the so-called Heisenberg equations
[iH, r ( x)] = r ( x),
[iH, r ( x)] = r ( x).

(3.9)

18

The Hamiltonian H is obtained by integrating the Hamiltonian density which we

defined in (2.29)
H = d 3 x H( x) = d 3 x 00 ( x) ,

(3.10)

whereas the 3-momentum of the field is given by

P j = d 3 x 0 j ( x)

(3.11)

So, let us commute the Hamiltonian density with r ( y ) and r ( y )

[H( x), r ( y )] = [ s ( x), r ( y )]s ( x) + s ( x)[s ( x), r ( y )] [L( x), r ( y )],

[H( x), r ( y )] = s ( x)[s ( x), r ( y )] [L( x), r ( y )].
Inserting (3.6) and (3.7) and using (3.3), we find
[iH( x), r ( y )] = 3 (x y )r ( x) ,
L( x) 3
L( x) ( x ) 3
[iH( x), r ( y )] =
(x y ) +
j (x y ) .
r
( j r )

(3.12)
(3.13)

Integrating over x, we find (after integrating (3.13) by part)

[iH, r ( y )] = r ( y ),
[iH, r ( y )] =

L( y )
L( y )
L( y )
r
j
= 0

= ( y ),

r ( y )
r
( j r )

where the Euler-Lagrange equation and the definition of the conjugate momentum
(2.28) have been used. In QFT these are called Heisenberg equations which,
according to (3.1) and (3.2), correspond to the canonical Hamilton equations in
classical field theory
H
H
= r ( x) ,
= ( x) .
(3.14)
r
( x)
r ( x)
In exactly the same way, we can establish the following more general equal-time
commutation relations
[i 0 ( x), r ( y )] = 3 (x y ) r ( x) ,

(3.15)

L( x)
L( x) 3
[i 0 ( x), r ( y )] = 0
j r ( x ) j 3 ( x y ) + 0
(x y )

j
r
r

(3.16)
Integrating these over x and introducing the Euler derivative (2.13) in (3.16), we find

19

[iP (t ), r ( x)] [i d 3 x 0 ( x) , r ( x)] = r ( x) ,

[iP (t ), r ( x)] [i d 3 x 0 ( x), r ( x)] = r ( x) + 0

L
,

(3.17)
(3.18)

where
P (t ) = d 3 x 0 ( x) ,

(3.19)

is the field energy-momentum 4-vector. We will talk about the physical meaning of
these relations, and show that P is indeed a time-independent 4-vector, when we
introduce the Noether charge. We will also see why the Euler-Lagrange equation of
motion appears in (3.18) but not in (3.17).
Notice that (3.18) shows the equivalence between Lagrangian and Hamiltonian
formalisms
L( x)
P

r
r
= 0 [iP , ( x)] = ( x) =
.
(3.20)
( x)
r ( x)

r
Okay, since we calculated the ETCR of the zeroth component of translation
current, 0 , with the fields, let us do some justice to the Poincare group and do the
same thing with the zeroth-component of Lorentz current M o . From (2.35) this is
given by

M 0 ( x) = x 0 ( x) x 0 ( x) i r ( x) ( )r s s ( x) .

(3.21)

Using (3.15) and the fundamental ETCR (3.3), we find

[iM 0 ( x) , l ( y )] = ( x l x l i ( )l s s ( x) ) 3 (x y ) .

(3.22)

Integrating over x and changing y x , we get our final result

[iM (t ) , r ( x)] [i d 3 x M 0 ( x), r ( x)] = x r x r i ( ) r s s .
(3.23)
In the next section we will show that the integral
M (t ) = L (t ) + S (t ) = d 3 x M 0 ( x) ,

(3.24)

is time-independent Lorentz tensor representing the total angular momentum of the

field, with the orbital angular momentum L and the spin angular momentum S are
given by

L (t ) d 3 x Lo ( x) = d 3 x ( x 0 ( x) x 0 ( x) ) ,
S (t ) d 3 x S 0 ( x) = i d 3 x r ( x)( ) r s s ( x).

(3.25)

20

Before going any further, let us make the following observations: If we contract
(3.17) with the translation parameter a and (3.23) with the Lorentz parameter , we
recover the correct infinitesimal Poincare transformations on the field

r ( x) = [ia P (t ) , r ( x)] = a r ( x) ,
i
2

(3.26)

i
2

r ( x) = [ M (t ) , r ( x)] = x r ( ) r s s ( x)
(3.27)
i
s
= x r ( ) r s ( x) .
2
Remarkably, the only assumption used to derive these relations is (that the fields
satisfy) the fundamental ETCRs (3.3). Indeed, neither dynamical considerations (i.e.,
equations of motion) nor symmetry considerations (i.e., conservation laws) have been
used to obtain (3.26) and (3.27). In fact we will now derive even more general result,
i.e., without specifying the group of the transformations.
Setting = 0 in the Noether current (2.18), we find

J (0 ) ( x) = r ( x) r ( x) x 0 L( x) .

(3.28)

Thus, at x 0 = y 0

[ J (0 ) ( x), s ( y )] = r ( x)[ r ( x), s ( y )] + [ r ( x), s ( y )] r ( x) x 0 [L( x), s ( y )] .

Due to (1.15) and (3.6) the first and the third terms add up to zero, and the second
term, when evaluated from the ETCR (3.3), gives

[iJ (0 ) ( x) , s ( y )] = s ( x) 3 (x y ) .

(3.29)

Hence, a 3-volume integration over x will generate the infinitesimal transformations

on the fields

r ( x) = [iQ (t ), r ( x)] ,

(3.30)

where Q (t ) (the Noether charge) is given by the integral

Q (t ) = d 3 x J (0 ) ( x) .

(3.31)

The Noether charge is going to be the subject of the next section. Again, the important
fact about (3.30) is that it has been derived without reference to a specific form of
L( x) , i.e., without any commitment to symmetry or dynamics. Thus, it seems that
(3.30) is always valid, and that any result that can be derived from it will necessarily
be true. However, this is true only if we ignore the usual difficulties of local quantum
field theory.
The so-called Ward-Takahashi identity in QFT is just an alternative form of the
canonical equal-time relation (3.30) and can be derived from the following timeordered product

21

T ( J ( ) ( x) r ( y ) ) = ( x 0 y 0 ) J ( ) ( x) r ( y ) + ( y 0 x 0 )r ( y ) J ( ) ( x) . (3.32)
Differentiating this with respect to x and using the relations
(x ) ( x 0 y 0 ) = 0 ( x 0 y 0 ),
(x ) ( y 0 x 0 ) = 0 ( x 0 y 0 ) ,

(3.33)

we find
T ( J ( ) ( x) r ( y ) ) = T ( J ( ) ( x) r ( y ) ) + ( x 0 y 0 )[ J (0 ) ( x) , r ( y )] . (3.34)
Using (3.29) in the second term on the right, we get
T ( J ( ) ( x) r ( y ) ) = T ( J ( ) r ( y ) ) i 4 ( x y ) ( )r ( x) .

(3.35)

Now, on integrating (3.35) over a space-time region containing the point y, we

arrive at the Ward-Takahashi identity

i r ( y ) = d 4 x T ( J ( ) ( x) r ( y ) ) d 4 x T ( J ( ) ( x) r ( y ) ) .

(3.36)

It is clear that the first term on the right hand side of (3.36) vanishes if the current is
conserved, i.e., symmetry current.
Let us now return to (3.25) and use it to build up the equal time commutator
[iM , r ( x)] :
[iS , r ( x)] = d 3 x l ( x) ( )l s [ s ( x) , r ( x)] .
Using (3.3), this becomes
[iS , r ( x)] = i s ( x) ( ) s r .

(3.37)

To calculate the commutator with the orbital angular momentum, we multiply (3.16)
by ( ix ) and integrate over x. Integration by parts then gives

L( y )
0 L ( y )
[i d 3 x x 0 ( x) , r ( y )] = j j ( y r ( y ) ) j y 0
+ y

( jr )
r

(3.38)
Using the identities
0 0 = j j ,

0 0 = j j ,

(3.39)

in (3.38), we find after some rearrangement

22


L
[i d 3 xx 0 , r ( y )] = r + y r 0 0 r + j

( j r )

L
L
+ y 0
r j (
) .

(
)

r
j
r

In this equation, we use the definition of the conjugate momentum

r ( y) =

L( y )
L( y )
, r ( y ) = 0
,
( 0r )
( 0 r )

(3.40)

to obtain

[i d 3 x x 0 , r ( y )] = r + y r 0

L
L
+ y 0
. (3.41)

( r )

Thus, the required commutator follows by the anti-symmetric combination

[iL , r ( x)] = x[ ] r [ ]0

L( x)
L( x)
+ x[ ]0
.

( r )

(3.42)

Adding (3.42) to (3.37), we get

[iM , r ( x)] = x r ( x) x r ( x) + i s ( x)( ) s r + F r .

(3.43)

Notice the extra term in the Lorentz transformation law for r ,

F r = F r x[ ]0

L( x) [ ]0 L( x)
.

( r )

(3.44)

This extra piece does not vanish even when the equations of motion are satisfied. This
is because the set{ r ( x)} does not form a covariant manifold under the Poincare
group.
Now, let us use Heisenberg equations (3.9) to prove the following theorem
Theorem 3.1

If the equal-time commutation relations (3.3) are valid at a certain

time x 0 = y 0 = t , they are also valid at t + .
Proof:
The proof will be based on Heisenberg equations and the following Jacobi identity
[ r (x, t ),[iH, s (y , t )]] + [ s (y , t ),[r (x, t ), i H]] + [iH,[ s (y , t ), r (x, t )]] = 0 .
23

Since at x 0 = y 0 = t , we have
[ r (x, t ), s (y , t )] = i rs 3 (x y ) .
Thus the third term in the Jacobi identity vanishes because the deltas are c-numbers.
And we are left with
[ r (x, t ),[iH, s (y , t )]] + [[iH, r (x, t )], s (y , t )] = 0 .

(3.45)

Using Heisenberg equations (3.9) in (3.45), we find

[ r (x, t ) , s (y , t )] + [r (x, t ) , s (y , t )] = 0

(3.46)

Multiplying this by and adding the result to the fundamental equal-time

commutation relation, we find
[ r (x, t ) , s (y , t )] + [r (x, t ) , s (y , t )] + [r (x, t ) , s (y , t )] = i rs 3 (x y ) (3.47)
Thus, to first order in we can write this as
(r (x, t ) + r (x, t ) ),( s (y , t ) + s (y , t ) ) = i rs 3 (x y ) ,

(3.48)

[ r (x, t + ) , s (y , t + )] = i rs 3 (x y ) .

(3.49)

or

And, by exactly the same method, we can establish the remaining commutation
relations
[ r (x, t + ) , s (y , t + )] = [ r (x, t + ) , s (y , t + )] = 0 .

(3.50)
qed

References and Further Reading

[1] Samalkhaiat. Conformal Group, Conformal Algebra and Conformal Invariance in
Field Theory, Physicsforums.com, thread number 172461.
[2] E. Bessel-Hagen, Math. Ann. 84, 258 (1921).
[3] C. G. Callan, Jr., S. Coleman and R. Jackiw, A New Improved Energy
Momentum Tensor, Ann. Phys. 59, 42 (1970).
[4] Samalkhaiat. Physicsforums.com, thread number

24

25