From Classical to Quantum Field Theory1

D. E. Soper2
University of Oregon
Physics 665, Quantum Field Theory
January 2001

Classical Mechanics

Let J (t), J = 1, 2, 3, be the position of a particle at time t. To describe

the mechanics of such a particle in a potential V , we use the action (with
repeated indices J implying summation)
S[] =

dt

1
m J (t) J (t)
2

V ((t)) .

(1)

The integral runs from an initial time tI to a final time tF . The notation
S[] indicates that S depends on the whole function , as distinct from (t)
at any one time t. We consider a small change in ,
J (t) J (t) + J (t),

(2)

where (t) is suitably smooth and vanishes at the initial time tI and the
final time tF . We define the variation, S, of the action by
S[ + ] = S[] + S[, ] + O(2 ),

(3)

where S is linear in . This is the definition. Calculation gives

)

V ((t))
J (t)
S =
dt m J (t) J (t)
J
)
(
Z
V ((t))

=
dt m J (t)
J (t).
J
Z

(4)

There is no surface term here because vanishes at the integration endpoints.

1
2

Copyright, 2001, D. E. Soper

soper@bovine.uoregon.edu

This illustrates an important mathematical concept, the functional derivative. The functional derivative
S
(5)
J (t)
is a function of t such that
S[, ] =

dt

S
J (t).
J (t)

(6)

Thus S/J (t) is something like the partial derivative of S with respect to
the value of J at time t. However, this doesnt make sense if taken too
literally since we cant change J at just one time. In some applications, the
functional derivative could be a distribution or generalized function that is
not an ordinary function but, for instance, contains delta functions. In the
present case, our calculation shows that
V ((t))
S
= m J (t)
.
J (t)
J

(7)

The principle of stationary action says that the function J (t) actually
chosen by nature is one for which S = 0 for every allowed . That is
S
= 0.
J (t)

(8)

We then obtain a second order differential equation for J (t):

m J (t) =

V ((t))
.
J

That is, F = ma, where F is minus the gradient of the potential.

Exercise. Calculate S/(t) for
R
3
1. S[] = dt ((t))
2

2. S[] = dt cos((t)) ((t))

3. S[] = dt exp[((t))
]

4. S[] = ((tA ))2 , where tA is a fixed time.

A ))2 , where tA is a fixed time.
5. S[] = ((t

(9)

Classical field theory

We consider a field (x) = (t, ~x). This is the simplest kind of field and is
analogous to the vector potential A (x) in electrodynamics.
Just think of as specifying an independent dynamical variable for each
P
value
of
~
x
.
Thus
~
x
replaces
J
in
the
previous
section.
We
replace
J by
R
d~x. The equations of motion for can be specified by giving an action and
saying that we apply the principle of stationary action. A common form of
the action is
Z
S[] = dx L ( (x), (x)) .
(10)
Here L is called the Lagrangian density, or often just the Lagrangian even
though strictly speaking the Lagrangian is the integral of L over ~x. A typical
example is
(11)
L = 21 ( )( ) 12 m2 2 4!1 4 .
We consider variations of that are nice smooth functions of x and
vanish at tF and tI and at |~x| . The functional derivative of the action
with respect to (x) is defined by
S[ + ] = S[ + ] +

dx

S[]
(x) + O(2 ).
(x)

(12)

The principle of stationary action says that

S =

dx

S[]
(x)
(x)

(13)

vanishes for every allowed . That is

S[]
= 0.
(x)

(14)

For an action of the form of Eq. (10) we calculate

S[]
L( (x), (x))
L( (x), (x))
=
+
.
(x)
( )

(15)

For the specific action Eq. (11) we have

S[]

= (x) m2 (x) (x)3 .

(x)
3!
3

(16)

Thus the equation of motion is the partial differential equation

(x) = m2 (x)

(x)3 .
3!

(17)

We can see the analogy with electrodynamics. The parameter m would be

the photon mass, but the photon mass is zero, so set m = 0. Then we have
the wave equation with a source (/3!) (x)3 for the waves instead of the
electromagnetic current J (x).

Symmetries and Noethers Theorem

The theorem is
An invariance of S[] gives a conservation law.
Instead of stating the theorem in its most general form we give three
examples.
1) J (x), J = 1, 3, , . . . , N with
L = 21 ( J )( J ) 12 m2 J J 4!1 (J J )2 .

(18)

Consider an infinitesimal O(N ) rotation,

J (x) = AJK K (x)

(19)

with AJK = AKJ . Then one easily sees that

L = 0.

(20)

Then
0 = L
L( , )
L( , )
J (x) +
J
=
( J )
J
(
)
L( , )
L( , )
L( , )
=
J +
J +
J
( J )
J
( J )
S[]
=
J (x) J (x)
(21)
J (x)
4

where
J (x) =
or
J (x) =

L( , )
J .
( J )

(22)

L( , )
AJK J (x)
( J )

(23)

We calculate
J (x) = ( J (x)) AJK J (x)

(24)

for our particular lagrangian.

When the fields obey their equations of motion, S[]/J (x) = 0, then
J (x) = 0. We are
familiar with the fact that J (x) = 0 is the statement
R
that the quantity d~x J 0 (x) is conserved. We interpret J 0 as the density of
the conserved stuff and J~ as the current of this stuff.
Exercise. Consider J (t) with J = 1, 2, 3 and
S[] =

dt

1
m J (t) J (t)
2

V (J (t)J (t)) .

(25)

The Lagrangian is invariant under a change of with

J = AJK K

(26)

with AJK = AKJ . How does Noethers Theorem work in this case of ordinary mechanics instead of field theory? What are the conserved quantities?
What is their physical interpretation?
1) Consider an infinitesimal translation, 0 (x) = (x a), or
(x) = a (x)

(27)

Then, assuming that L depends on and its derivatives but not directly on
x, we have
L(x) = a L(x).
(28)
Then
0 = L + a L
L( , )
L( , )
=
(x) +
+ a L
( )

(
)
L( , )
L( , )
L( , )

=
+
+
+ a L
( )

( )
S[]
=
(x) a T (x)
(29)
(x)
5

where

L( , )
a g L.
( J )

(30)

L( , )
( (x)) g L.
( )

(31)

a T (x) =
or
T (x) =

There are four conserved quantities,

P =

d~x T 0

(32)

These are the components of the energy and momentum.

Exercise. For
L = 21 ( )( ) 12 m2 2 4!1 4 .

(33)

what is T ? In particular, what is the energy density T 00 ?

1) Consider an infinitesimal Lorentz transformation, 0 (x) = (x) (x
A x ), or
(x) = A x (x)
(34)
where A = A . Then since L depends on and its derivatives but not
directly on x, we have
L(x) = A x L(x).

(35)

Then (using the antisymmetry of A in the third line below)

0 = L + A x L
L(, )
L(, )
(x) +
+ A x L
=
( )

)
(
L(, )
L(, )
L(, )

=
+
+
+ A x L
( )

( )
S[]
1
=
(x) + A M (x)
(36)
(x)
2

where
L(, )
1
A M (x) =
+ A x L
2
( )
L(, )
A x + g A x L
=
( )

L(, )
1
A
[x x ]
=
2
( )

[g x g x ]L

(37)

or
M (x) =

L(, )
[x ( ) x ( )] [g x g x ]L.
( )

(38)

There are six conserved quantities (considering the antisymmetry in {, }),

J =

d~x M 0

(39)

These are the components of the angular momentum.

Exercise. For our sample field theory, what evaluate M 120 and M 030 .

Quantization

We have seen how to get a classical field theory from a lagrangian density L.
But how do we get a quantum field theory? The fields should have the same
equations of motion as in the classical field theory. But they should be operators. The question then is, what sort of operation to they perform. It turns
out that we will learn just about everything about what the operators do if
we know their commutation relations with one another. Thus we investigate
commutation relations.
We base the theory on classical mechanics. Consider the lagrangian
L=

d~x L(~x, t).

(40)

This is really the lagrangian, whereas L is the lagrangian density (even

though we will sometimes be sloppy and call L the lagrangian). In our
7

treatment, we consider space and time in very different roles. At any time t
there is a function of ~x such that (~x, t) = (~x). The function , treated
as a whole, is an element of a function space that represents the dynamical
coordinates of our system, just like the position of a particle in three dimensional space or a particle on the surface of a sphere. For a later time t,
(~x, t) defines a new function of ~x. Thus we have a different value of
for each t, which we can call (t). The function (t), when evaluated at ~x
is (t; ~x) = (~x, t). Mostly, however, we will omit writing the parameter t.
Similarly, at any time t the function (~x, t)/dt defines a function of ~x, which
we can call .
This function represents the rate of change of the dynamical
coordinates, the velocities. In this sense, we can consider L to be a function
in classical mechanics, where Q
L[, ]
of these two functions. (Cf. L(Q, Q)
depends on the time and may have many components, Q (t).)
Following classical mechanics, we define the canonical momenta
(~x) =

L[, ]

.
(~
x)

(41)

This is a functional derivative, defined by

L[, ]

L[, ]

L[, ]

d~x
(~x) +
(~
x) .
(~x)
(~
x)

(42)

This is the general definition, but for our sample lagrangian density
~ ()
~ 1 m2 2 1 4
L = 12 2 12 ()
2
4!

(43)

we have the simple result

(~x) = (~
x).

(44)

In classical mechanics in the hamiltonian formulation, we solve for in

terms of and and then consider everything to be a function of and
insteat of and .
Then if A is a function of and , we define variational
derivatives by
A[, ]

A[, ]
A[, ]
(~x) +
(~x) .
d~x
(~x)
(~x)

(45)

Note that (~y ) defines a perfectly nice function of and , so we should be

able to write its functional derivatives with respect to (~x) and (~x). These
8

are, according to the definition

(~y )
= (~y ~x)
(~x)
(~y )
= 0.
(~x)

(46)

Similarly
(~y )
= 0
(~x)
(~y )
= (~y ~x).
(~x)

(47)

Again following classical mechanics we define the Poisson bracket of any

two functions A and B of and ,
)

A[, ] B[, ] B[, ] A[, ]

.
{A[, ], B[, ]}PB d~x
(~x) (~x)
(~x) (~x)
(48)
Especially important are the elementary Poisson brackets
Z

{(~y ), (~z)}PB = (~y ~z)

{(~y ), (~z)}PB = 0
{(~y ), (~z)}PB = 0.

(49)

Now comes the essential observation. The algebra of Poisson brackets is

essentially the same as the algebra of commutators. For example
{A, B}PB = {B, A}PB

(50)

{A, B + C}PB = {A, B}PB + {A, C}PB

(51)

{A, BC}PB = B {A, C}PB + {A, B}PB C.

(52)

and
and also
These hold also for commutators of operators. The only difference is that
in the case of commutators, in the last relation the operator ordering for
products of operators has to be as I wrote it, while for Poisson brackets we
9

are talking about multiplication of numbers and the ordering of factors does
not matter. Based on this observation, one replaces {A, B}PB by i [A, B].
In particular, for quantum field theory, the operators (~x) and (~x) obey the
commutation relations
[(~y ), (~z)] = i(~y ~z)
[(~y ), (~z)] = 0
[(~y ), (~z)] = 0.

(53)

The hamiltonian equations of motion

We can construct the hamiltonian according to the usual prescription in

classical mechanics:
H=

d~x

L[, ]

(~
x) L.
(~
x)

This is
H=

d~x H(~x),

(54)

(55)

where the hamiltonian density is

H=

L[, ]

(~
x) L.
(~
x)

(56)

Exercise. Assuming that L = d~x L where L is a function of (x),

~
(x)/t and (x),
what is the relation between H and the energy density
00
T that is conserved because of Noethers Theorem?
R

Now to do classical mechanics in the hamiltonian formulation, we eliminate in favor of and and write H in terms of these variables. Thus,
for instance, with
~ ()
~ 1 m2 2 1 4
L = 12 2 12 ()
2
4!

(57)

~ ()
~ + 1 m2 2 + 1 4
H = 2 L = 21 2 + 12 ()
2
4!

(58)

we have

10

and
=
so
H[, ] =

d~x

1 2

(59)
o

~ ()
~ + 1 m2 2 + 1 4 .
+ 12 ()
2
4!

(60)

The classical equations of motion in the hamiltonian formulation are

(t; ~x) = {H[, ], (t; ~x)}PB

t

(t; ~x) = {H[, ], (t; ~x)}PB .

t

(61)

With our rule about turning Poisson brackets into commutators, these become the quantum equations of motion,

(t; ~x) = i [H[, ], (t; ~x)]

t

(t; ~x) = i [H[, ], (t; ~x)] .

t

(62)

Lets try this, using our sample theory and the elementary commutation
relations. We have
[H[, ], (t; ~x)] =

[H[, ], (t; ~x)] =

=
=

Z
Z
Z

d~y

1
2

(~y )2 , (t; ~x)

(63)

d~y

1
2

d~y

1
2

{(~y ) [(~y ), (t; ~x)] + [(~y ), (t; ~x)] (~y )}

d~y

1
2

{(~y )(i)(~y ~x) + (i)(~y ~x)(~y )}

(~y )2 , (t; ~x)

= i(~x).

(64)

so we get

(t; ~x) = (t; ~x).

t
which, we recall, was the definition of in terms of .

11

(65)

Exercise. Use the commutation relations to find the equation of motion

for (t; ~x) in our example theory. Compare to the equation of motion from
the principle of stationary action.

Exercise. In the previous exercise we found that

[P 0 , (t; ~x)] = i (t; ~x)/t
[P 0 , (t; ~x)] = i (t; ~x)/t

(66)

where P 0 = d~x T 00 . Use the commutation relations for our sample theory
to
R
j
j
j
j0
find something similar for [P , (t; ~x)] and [P , (t; ~x)] where P = d~x T .
R

12