Sei sulla pagina 1di 17

Quantum Field Theory

Physics 253a, Fall 2004


Arthur Jaffe
Harvard University, Cambridge, MA 02138, USA
October 21, 2004

Quantum Field Theory as Quantum Mechanics

In these notes we study quantum field theory in a box, namely we replace Euclidean 3-space R3
by a periodic box or torus T3 . In I.3 and I.4 we show that we can interpret bosonic quantum field
theory as ordinary Schrodinger quantum mechanics with an infinite number of coordinates. Each
coordinate arises from a Fourier component of the field. Another way to describe the situation is
that quantum field theory is quantum mechanics of an infinite number of degrees of freedom.
We call the solution to the Klein-Gordon equation a linear field theory. It gives rise to the
quantum theory of uncoupled harmonic oscillators. On the other hand, solutions to non-linear field
equations correspond with non-linear, coupled oscillators. The potential function that produces
interaction among the oscillators comes from a non-linear equation of motion! The energy density for
a non-linear field ultimately produces the physical forces between physical particles. Understanding
the origin of these forces and calculating their effects in particular problems is a major part of
understanding quantum field theory.
The quantum field theory (after passage to a box of infinite volume) is relativistically covariant.
Yet the Schrodinger equation on field space resembles the ordinary equation in a non-relativistic
problem! At first this may appear surprising. Stated differently, there is a coordinate q (with an
infinite number of components) such that we can write the Hamiltonian in the ordinary Schrodinger
form,
1
H = q + V (q) ,
(I.1)
2
where q denotes the Laplacian of the form
X
2
q =
i 2 ,
with i > 0 .
(I.2)
qi
i
1

Arthur Jaffe

The potential V (q) is a function that describes the energy density of interaction. The oscillators
interact either in a linear or in a non-linear fashion. The main difference between this equation and
the ordinary Schrodinger equation in non-relativistic quantum theory is that the field coordinate
q has an infinite number of coordinate directions, while the usual Schrodinger equation concerns a
finite number of particles and therefore a finite number of coordinates directions. One component
(or degree of freedom) arises from each Fourier mode in the decomposition of the field (~x). We use
this representation in this lecture and the next one to solve a simple example, the mass perturbation.
This will be the basis of our later discussion of mass renormalization. In the simplest case, V (q) is
a quadratic polynomial of the form,

1X
i qi2 1 .
(I.3)
V (q) =
2 i
In this case, the coordinate q is an infinite dimensional oscillator. In the next few classes, we study
this example, and perturbations of this Hamiltonian by other quadratic polynomials. Much of the
rest of the course concerns polynomial (or other simple) choices for V (q) that lead to non-linear
equations of motion.
Remark. Both in quantum mechanics of particles and in the quantum field theory of fields, there
are very few examples of interest that one can solve exactly, in closed form. In other words, there
are few problems for which we do not have to resort to perturbation theory. The quadratic energy
density is one problem that we can both solve exactly, and also using perturbation theory. Even
though the answers are straightforward, they provide great insight into the structure of quantum
field theory. They will serve as a guide for many properties of field theory that we later can only
study approximately. So enjoy the luxury of being able to compare two solutions!
One can calculate the eigenvalues of H, as well as its eigenstates and Greens functions (vacuum
expectation values of time-ordered products of fields) h| T (x1 )(x2 ) (xn ) |i. In this lecture
we study a simple such example, where V is a quadratic polynomial. Such toy models that we can
completely solve form the basis of our understanding in more complicated situations. Take their
properties seriously!

I.1

Interaction Picture Fields on a Cylindrical Space-Time

In order to simplify the presentation, we replace Minkowski 4-space temporarily by a cylindrical


space-time, the 4-dimensional analog T3 R of the familiar cylinder. Here T3 denotes the 3torus with periods {L, L, L}, while the real line R parameterizes time. The momentum lattice
corresponding to T3 is
2
~n} ,
(I.4)
K = {~k =
L
q
where ~n denotes a vector with integer components. We take k0 = (k) = ~k 2 + 2 .

Physics 253a, Fall 2004

On this space we define interaction picture fields (~x, t). In order to be very clear, and also in
order not to have too many subscripts, we use the convention: Denote the interaction picture
field built with a mass by (~x, t). Of course, we should not confuse the mass with a vector
index. This field is

1
1 X
q
(I.5)
(~x, t) = 3/2
a~k eikx + a~k eikx ,
L
kK
2 (~k)
and the time derivative
q
(~x, t) =

(~x, t)
i X
= 3/2
t
L
kK

(~k) ikx


a~k e a~k eikx .


2

(I.6)

Then for any integer vector ~n, the field satisfies the periodicity condition
(~x + ~nL, t) = (~x, t) .

(I.7)

We assume that each of the creation and annihilation operators is independent,


[a~k , a~k0 ] = 0 ,

and [a~k , a~k0 ] = ~k~k0 .

(I.8)

Then as a consequence, the interaction picture fields satisfy the canonical commutation relations
[ (~x, t), (~x0 , t)] = i(~x ~x0 ) .

I.2

(I.9)

The Free Field as a Harmonic Oscillator

Here we show that the free field can be interpreted as a harmonic oscillator with a coordinate q
that has an infinite number of components. Let us write the time-zero field (~x) = (~x, 0)
given by (I.5) and (~x) = (~x, 0) given by (I.6) in terms of cosines and sines,


1 X
1
i~k~
x
i~k~
x
q
(~x) =
a
e
+
a
e
~k
~k
L3/2 kK
2 (~k)




1 X
1

~
~
q
=
a
+
a
cos(
k

~
x
)

i
a

a
sin(
k

~
x
)
,
~k
~k
~k
~k
L3/2 kK
~
2 (k)
(I.10)
and
(~x) =

i
L3/2
i
L3/2


(~k)  i~k~x
i~k~
x
a~k e
a~k e
2




(~k) 
a~k a~k cos(~k ~x) i a~k + a~k sin(~k ~x) .
2

X
kK

X
kK

(I.11)

Arthur Jaffe

We would like to identify the coefficients in the expansion of as an infinite number of oscillator
coordinates q, and the coefficients in the expansion of as their conjugate momentum coordinates
p. However, the expansions (I.10) and (I.11) are not what we need: the cosine-coefficients of
do not commute with the sine-coefficients of , etc. Therefore these coefficients do not satisfy the
desired canonical commutation relations for qs and ps of independent oscillators.
We can solve this problem by combining the pairs of modes labelled by ~k and ~k. For 0 6=
~k K 0 , define the following dimensionless coordinates that are even or odd under inversion of ~k
(denoted by subscripts e or o):

1
a~k + a~k + a(~k) + a(~k) ,
qe (~k) =
2

i 
q0 (~k) =
a~k a~k a(~k) + a(~k) ,
2


i

~
~
~
pe (k) =
a a~k + a(k) a(k) ,
2 ~k


1

~
~
~
a + a~k a(k) a(k) .
(I.12)
po (k) =
2 ~k
For ~k = 0, define only one canonical pair,
1
qe (0) = (a(0) + a(0)) ,
2

and

i
pe (0) = (a(0) a(0)) .
2

(I.13)

Note that I have chosen a dimensionless set of qs and ps in order to simplify the description of
fields with different masses.
In order to simplify the final notation slightly (avoiding having to modify the canonical relations
to take into account the fact that we have twice the number of needed qs and ps), we introduce a
momentum space sum restricted to the set K 0 so that in our final expressions we do not sum over
both ~k and ~k.
K 0 = {~k : ~k K , and k1 0} .
(I.14)
If we restrict attention to ~k, ~k 0 K 0 and j, j 0 = e, o, then the coordinates (I.12) are canonical. In
fact,
[qj (~k), qj 0 (~k 0 )] = [pj (~k), pj 0 (~k 0 )] = 0 ,
and [pj (~k), qj 0 (~k 0 )] = ijj 0 ~k~k0 .
(I.15)
Exercise: Verify the commutation relations (I.15).
Now we rewrite the fields , in (I.10) and (I.11) in terms of these canonical coordinates.

s


X
1 1
2

qe (~k) cos(~k ~x) + qo (~k) sin(~k ~x) ,


(~x) =
q
(0)
+

e
3/2
L

(~k)
~
kK 0
~
k6=0

(I.16)

Physics 253a, Fall 2004

and

(~x) =

1
pe (0) +
L3/2
~

Xq

2 (~k) pe (~k) cos(~k ~x) + po (~k) sin(~k ~x) .




kK 0
~
k6=0

(I.17)
Remark. In the case of mass = 0, the representation of the fields and needs to be
modified. This is an infra-red problem that we do not discuss now.
We also verify the canonical commutation relations in terms of this representation of the fields,
a check that we have not introduced extra factors of 2, etc. Clearly the fields (~x) and (~x0 )
commute, as do (~x) and (~x0 ). Also

X
i

cos(~k ~x) cos(~k ~x0 ) + sin(~k ~x) sin(~k ~x0 )


1
+
2
[ (~x), (~x0 )] =

L3
0
~
kK
~
k6=0



X
i

=
1
+
2
cos ~k (~x ~x0 )

3
L
0
~
kK
~
k6=0

i X
cos(~k (~x ~x0 )) = i(~x ~x0 )) ,
3
L kK

(I.18)

as expected.

I.3

The Oscillator Representation for the Hamiltonian

The free field energy is


Z
H0, (~x)d~x ,

H0, =

(I.19)

T3

with the energy density



1
H0, (~x) = : (~x)2 + ( (~x))2 + 2 (~x)2 : .
2

(I.20)

We insert the coordinate representations (I.16) and (I.17) into (I.20) to obtain the representation of the energy in terms of harmonic oscillators. Use the fact that the functions sin(~k ~x), cos(~k ~x)
form the orthogonal Fourier basis for L( T3 , d~x) and have the normalization
Z
Z
1
2 ~
sin (k ~x)d~x =
cos2 (~k ~x)d~x = L3 .
(I.21)
2
T3
T3

Arthur Jaffe

Let us introduce the convention that

~
kK 0
j=e,o

omits a sum over a possible odd k = 0 mode.

Then the free field Hamiltonian H0, can be written as a sum of oscillator Hamiltonians,
H0,



1 X
2
2
~
~
~
=
(k) pj (k) + qj (k) 1 .
2~ 0

(I.22)

kK
j=e,o

The oscillators have angular frequency (~k), and with zero-point energy 12 (~k), where ~k lies in
the lattice (I.4) of allowed Fourier momenta for the torus T3 . These individual Hamiltonians have
eigenvalues that are positive integer multiples of (~k). This is the representation claimed in (I.1),
in terms of a scaled coordinate function q, with the (infinite) zero-point energy,
1 X ~
1X ~
(k) =
(k) .
2~ 0
2
kK
j=e,o

(I.23)

~kK

This is the zero-point energy constant subtracted in the potential (I.3).

I.4

The Total Hamiltonian when HI is a Function of

Consider a perturbation of the form


Z
H = H0, + HI ,

where HI =

HI ( (~x))d~x ,

(I.24)

T3

with the energy density of the perturbation a function of (~x). Since the field (I.16) depends on the
oscillator coordinates qe (k) and qo (k), but neither on pe (k) nor po (k), we obtain a representation of
the total Hamiltonian H of the form
H=



1 X
(~k) pj (~k)2 + qj (~k)2 1 + VI (q) .
2~ 0

(I.25)

kK
j=e,o

Of course, the form of the potential function VI (q) is dictated by the energy density expression
HI (~x) that we choose. The total Schrodinger potential in (I.1) is
V (q) =



1 X
(~k) qj (~k)2 1 + VI (q) .
2~ 0
kK
j=e,o

(I.26)

Physics 253a, Fall 2004

I.5

Mass Shifts

A very simple form of the energy density that gives something interestingand for which we can
compute everythingis the case that HI is a quadratic function of . Let us suppose that
Z

HI =
: (~x)2 : d~x ,
2 T3

(I.27)

where > 2 is a real constant. Looking ahead to the interpretation of H, let us define the
perturbed mass m by
p
(I.28)
m = + 2 + .
Because this interaction is quadratic, the interaction Hamiltonian has a very simple structure,

HI =


X 1  ~ 2
qe (k) + qo (~k)2 1 .
2 ~ 0 (~k)
kK

(I.29)

j=e,o

This shows that in the case of our example, the total Hamiltonian
Z

: (~x)2 : d~x
H = H0, +
2 T3

(I.30)

remains an (infinite) sum of independent Hamiltonians. It has the form

1 X
H =
(~k) pj (~k)2 +
2 ~ 0

1+

kK
j=e,o

X
1
=
(~k) pj (~k)2 +
2 ~ 0
kK
j=e,o

(~k)2

m (~k)
(~k)

We call this the mass perturbation Hamiltonian.

!2

!
qj (~k)2 1

2 (~k)2

2
2
~
~
m (k) + (k)
qj (~k)2
.
2 (~k)2

(I.31)

II
II.1

Arthur Jaffe

Exact (Non-Perturbative) Solution of a Field Theory


The Renormalized Hamiltonian

Let E0 denote the ground-state energy of the mass-perturbation Hamiltonian (I.31). If we subtract
E0 from H, we obtain the renormalized Hamiltonian with zero ground state energy zero,
Hrenormalized = H0, + HI E0
Z

: (~x)2 : d~x E0
= H0, +
2 T3

!2
X
~
~
m (k)
1
m (k)
(~k) pj (~k)2 +
=
.
qj (~k)2
~k)
~k)
2 ~ 0

kK

(II.1)

j=e,o

In fact, once we know the terms in Hrenormalized that are quadratic in the pj (~k)s and qj (~k)s, it must
be the case that the zero-point energy of each oscillator summand is zero.

II.2

The Energy Shift

We claim that the energy shift E0 has a beautiful and simple form,

!1/2
!1/2 2
X
~
~
m (k)
(k)
1
.
m (~k)

E0 =
~
~
4
(k)
m (k)
~

(II.2)

kK

We verify this by reading off the ground state energy on each mode in the expression (I.31). Since
the Hamiltonian p2 + 2 q 2 has zero point energy , the Hamiltonian

!2
2
2
~
~
~
1
m (k)
m (k) + (k)
(~k) pj (~k)2 +
qj (~k)2
(II.3)
2
(~k)
2 (~k)2
has ground-state energy
1
(~k)
2

m (~k) m (~k)2 + (~k)2

(~k)
2 (~k)2

!
2m (~k) (~k) m (~k)2 (~k)2
2m (~k) (~k)

!1/2
!1/2 2
~
~
1
m (k)
(k)
.

= m (~k)
4
(~k)
m (~k)

1
=
m (~k)
2

(II.4)

Physics 253a, Fall 2004

Summing over modes we then obtain


E0

!1/2
X
~k)
1

(
m
=
m (~k)

4 ~ 0
(~k)
kK

(~k)
m (~k)

!1/2 2
.

(II.5)

j=e,o

Using the fact that the sums

~
kK 0
j=e,o

and

~kK

of an even function of ~k agree, we see that (II.5)

equals (II.2).

II.3

Vacuum Energy Renormalization of the Quadratic Perturbation

Let us ask about the size of the shift E0 in the ground state energy that we derived in (II.2) in the
problem with a quadratic perturbation (I.24). In particular, we can ask whether the sum (II.2) is
convergent. Since the sum over ~k K ranges over a 3-dimensional lattice, the sum will converge
depending on whether the strictly negative summand (II.4) goes to zero for large |~k| faster than the
dimension of the sum. In other words, we need it to vanish at large |~k| more rapidly than |~k|3 .
Clearly
!1/2
!1/2
(~k)
m (~k)

(II.6)
(~k)
m (~k)
tends to zero for large |~k|, and the question is at what rate. Since (~k) and m (~k) are both
functions of ~k 2 and analytic functions of this parameter for large |~k|, the difference (II.6) must tend
to zero at least as fast as ~k12 . Since m (~k) grows for large ~k as |~k|, this means that (II.4) decays at
least as fast as O(|~k|3 ).
(If one were interested, one could check that in fact there is no further cancellation and this
is the exact rate of decay. To do so, utilize the series expansions for small x of (1 + x)1/2 =
1 + 21 x 81 x2 + , and also of (1 + a)/(1 + b) = 1 + a b b(a b) + . Thus one can expand
for small a, b up to order 2 to obtain,
1/2 
1/2 !2

1+b
1+a

= (a b)2 + O(a2 + b2 ) .
(II.7)
1+b
1+a
In the case in question this gives

!1/2
~k)
1

(
m
m (~k)

2
(~k)
as |~k| .)

(~k)
m (~k)

!1/2 2

2
= 1 m (~k) + O(|~k|5 ) |~k|3 ,
2
|~k|4

(II.8)

10

Arthur Jaffe

As a consequence of these asymptotics, we conclude that the sum (II.2) is logarithmically divergent in our 4-dimensional space-time. This means that we must eliminate high momentum degrees
of freedom, then subtract E0 for the regularized problem, and finally remove the regularization on
the renormalized Hamiltonian.

II.4

The Ground State

In our notation from class, we denote the ground state (or vacuum state) for the total Hamiltonian
Hrenormalized by |i, and the ground state of the unperturbed Hamiltonian H0, by |0 i or else |0i.
Then
Hrenormalized |i = 0 ,
and H0, |0i = 0 .
(II.9)
Using the oscillator representations (I.22) and (I.31), we can compute the Schrodinger wave
function for both vectors. We choose the ground state to be positive. Then the normalized ground
wave function for one mode of the free field Hamiltonian


1
(~k) pj (~k)2 + qj (~k)2 1 ,
(II.10)
2
is just
 1/4
1
~ 2
eqj (k) /2 .

(II.11)

Thus the wave function for the free field Hamiltonian is


Y  1 1/4
~ 2
0 (q) =
eqj (k) /2 .

0
~

(II.12)

kK
j=e,o

Of course one should ask in what sense this infinite product converges, but we will finesse this point.
is

Likewise, the corresponding mode in the perturbed Hamiltonian Hrenormalized defined in (II.1)

!2
X
~
~
m (k)
m (k)
1
(~k) pj (~k)2 +
qj (~k)2
.
(II.13)
~k)
~k)
2 ~ 0

kK
j=e,o

It has a ground state wave function equal to


m (~k)
(~k)

!1/4
~

em (k)/qj (k)

2 /2 (~
k)

(II.14)

11

Physics 253a, Fall 2004

leading to the ground state for Hrenormalized


(q) =

m (~k)
(~k)

Y
~
kK 0
j=e,o

II.5

!1/4
~

em (k)qj (k)

2 /2 (~
k)

(II.15)

Wave-Function Overlap and Renormalization

It is of interest to calculate the overlap between the unperturbed and the perturbed ground state.
Using the fact that both the ground states (II.12) and (II.15) are product wave functions, we
compute
Z 1  m (~k)  2 !
1/4
Y
~

m (k)
~ +1 q
dq
e 2 (k)
h0|i =
(~k)1/4 1/2
0
~
kK
j=e,o

Y 21/2 m (~k)1/4 (~k)1/4



1/2
~
~
~
kK 0

(
k)
+

(
k)
m

j=e,o

Y
~kK

1
2

(~k)1/2
m (~k)1/2
+
(~k)1/2
m (~k)1/2

!!1/2
.

(II.16)

The Schwarz inequality shows that h0|i 1. There is a potential divergence in this infinite
product. A divergent product means that h0|i = 0. In fact, each term in the product is less than
one, so if the product diverges, it diverges to zero.
In ordinary quantum mechanics, two strictly positive wave functions 0 (q) and (q) must have
a strictly positive overlap h0|i > 0. Here we see instead that two strictly positive wave functions
may have zero inner product!
Q
P
An infinite product i xi of terms xi (1, 1) converges only if i (1 xi ) converges. We
observe that
!
(~k)1/2
2
1 m (~k)1/2
+
1
,
(II.17)
2 (~k)1/2
m (~k)1/2
|~k|4
as |~k| . This can be summed over ~k K, and therefore the ground state has a non-vanishing
inner product h0|i > 0.
If we are in 5 or more space-time dimensions, then our calculation shows that the two positive
wave functions |0i and |i have vanishing inner product, namely h0|i = 0. In that case, we say

12

Arthur Jaffe

that the state vectors would require an infinite wave-function renormalization, and the (physical)
eigenstates of H would actually lie in a different Hilbert space from those of H0 . Of course, our
conclusion depends on the form of the interaction. For a mass perturbation the vacuum-energy E0
diverges in space-time dimension d 4 and the wave function requires renormalization in d 5.
However, non-linear wave equations give rise to similar, but more singular results.

II.6

Mass Change as a Canonical Scaling Transformation

A canonical transformation in quantum theory is a unitary operator W that takes canonical fields
into new canonical fields. In other words, we replace (~x) and (~x) by new fields
(~x) = W (~x)W ,

and

(~x) = W (~x)W .

(II.18)

A simple example of a canonical transformation is a scale transformation. Choose a real number


and scale each canonical coordinate qj (~k) Qj (~k) = e qj (~k) and do the inverse scaling to the
corresponding momentum pj (~k) Pj (~k) = e pj (~k). In this case, we can write down the unitary
W in closed form as the exponential of its self-adjoint generator,
~

Wj,~k () = ei(qj (k)pj (k)+pj (k)qj (k))/2 .

(II.19)

Note that for ~k, ~k 0 K 0 , the scaling of different coordinates commute. In other words
Wj,~k ()Wj 0 ,~k0 (0 ) = Wj 0 ,~k0 (0 )Wj,~k (); ,

for {j, ~k} =


6 {j 0 , ~k 0 } .

(II.20)

One could write out the generator of this transformation, if one desired, in terms of creation and
annihilation operators, using the expressions (I.12).
We obtain something interesting by letting the scaling parameter depend on the momentum
in a particular way. We set = (~k), where
!1/2
!
~
~

(
k)
1

(
k)
~

.
(II.21)
,
or (~k) = ln
e(k) =
~
2
m (k)
m (~k)
We now define W as the product of commuting unitaries
~

Wj,~k = Wj,~k ((~k)) = ei(qj (k)pj (k)+pj (k)qj (k))(k)/2 ,

and

W =

Wj,~k .

(II.22)

kK 0
j=e,o

We call (II.22) the mass scaling transformation. This transformation implements the following
scaling of coordinates,
!1/2
!1/2
~k)
~k)

m
W qj (~k)W =
qj (~k) ,
and W pj (~k)W =
pj (~k) .
(II.23)
m (~k)
(~k)

13

Physics 253a, Fall 2004

Applied to the interaction picture time-zero field, this transformation lives up to its name.
Taking into account the definitions of the time-zero fields in terms of the q and p coordinates,
namely (I.16) and (I.17), we infer that for the transformation W ,
W (~x)W = m (~x) ,

and W (~x)W = m (~x) .

(II.24)

In other words, the unitary W intertwines time-zero interaction picture fields of different masses!

II.7

Scaling the Hamiltonian

The unitary scale transformation W gives us the relation between H0,m and the Hrenormalized Hamiltonian (II.1). These Hamiltonians are not the same, but they are unitarily equivalent, with W
implementing this equivalence. In fact


1 X

2
2
~
~
~
W H0,m W =
m (k)W pj (k) + qj (k) 1 W
2 ~
kK
j=e,o

1 X
=
m (~k)
2 ~
kK
j=e,o

!
(~k) ~ 2 m (~k) ~ 2
pj (k) +
qj (k) 1
m (~k)
(~k)

X
1
=
(~k) pj (~k)2 +
2 ~
kK
j=e,o

m (~k)
(~k)

!2

~
m (k)
qj (~k)2
(~k)

= Hrenormalized .

II.8

(II.25)

Scaling Eigenstates of the Hamiltonian

The way in which the Hamiltonian scales, as given by (II.25), is a very interesting result. It allows
us to analyze the eigenvalues and eigenvectors. In particular, W maps eigenstates of H0,m to
eigenstates of Hrenormalized . Suppose that |Ei denotes an eigenstate of H0,m with energy eigenvalue
E. Then we infer from (II.25) that
Hrenormalized W |Ei = W W Hrenormalized W |Ei
= W H0,m |Ei
= E W |Ei .

(II.26)

Thus W |Ei is an eigenstate of Hrenormalized (with the same eigenvalue E). Note that in particular
W |0i = |i .

(II.27)

14

Arthur Jaffe

This is the fundamental solution given by mass scaling.


More generally, we constructed a complete set of eigenstates of H0,m earlier in the course (at
least in the case of Minkowski space-time). In fact they are just the states with n excitations of the
zero-particle state. In our situation, we obtain an orthonormal basis of eigenstates as follows: let
~n = {n(~k)} `2 (K) be a vector with integer coefficients indexed by the allowed momenta ~k K
and with only a finite number of components n(~k) different from zero. Define |~ni as the normalized
state in Fock space containing n(~k) quanta with momenta ~k. In terms of the creation operators a~k
and the zero-particle state |0i,

Y (a(k) )n(~k)
|0i .
q
|~ni =
~kK
n(~k)!

(II.28)

The vectors |~ni form an orthonormal basis for the Fock space H, which in addition are eigenstates of the free field Hamiltonian H0,m . This follows from the representation
H0,m =

m (~k) a~k a~k .

(II.29)

n(~k) m (~k) ,

(II.30)

~kK

The vector |~ni has energy eigenvalue


E~n,m =

X
kK

for the Hamiltonian H0,m , and


H0,m |~ni = E~n, |~ni .

(II.31)

In fact the vectors |~ni are eigenvectors for the free field of any mass, as one sees by substituting
from (II.29) and (II.30).

II.9

The Scaling Parameters Describe Constants

Having introduced the scaling parameters {(~k)} in (II.21), observe that they embody some of the
constants computed above as a consequence of the mass scale transformation. For example, the
vacuum energy shift (II.5) can be written as a simple expression in terms of these parameters; it is
just


2
X
E0 =
m (~k) sinh (~k)
.
(II.32)
~kK

15

Physics 253a, Fall 2004

Furthermore the wave-function overlap (II.16) can be expressed simply, either in terms of W
or the scaling parameters. Since this overlap is real, we have

1/2
Y
~
h0|i = h|0i = h0| W |0i =
cosh (k)
.
(II.33)
~kK

II.10

Evolution of Scaling for the Ground State

We know that the scaling transformation W maps the zero-particle state |0i, namely the ground
state of the free field Hamiltonian H0,m , into the ground state of Hrenormalized . In this section we
look at another way to derive this result by studying the scaling transformation in more detail. Let
0 s 1 denote an interpolation parameter. We study the vectors
|si = W s |0i ,

(II.34)

so |0i is the ground state of H0,m and |1i is the ground state of Hrenormalized .
We derive a differential equation that describes this scaling for a single degree of freedom. Let
0 (q)) denote the ground-state wave function for one mode of |0i as given in (II.11),
 1/4
1
2
eq /2 ,
(II.35)
0 (q) =

and let F (s, q) denote the wave function in the same mode for W s |0i, namely

F (s, q) = eis(qp+pq)/2 0 (q) .

(II.36)

We write for = 12 ln m / that


F (0, q) = 0 (q) ,

and F (1, q) = (q) ,

(II.37)

and for intermediate values we say F (s, q) = s (q).


We claim that F (s, q) satisfies the flow equation


1
F (s, q)
2s 2
= e
q
F (s, q) .
s
2

(II.38)

That being the case, we have the solution


R s  2s0 2 1  0
q 2 ds
0 e

F (s, q) = e

1
2

F (0, q)

q 2 21 s

)
= e (
F (0, q)
 1/4
1
1 2s 2
1
=
e 2 s e 2 e q .

1e2s

(II.39)

16

Arthur Jaffe

Setting equal to its value, we obtain



F (s, q) =

s
m
s

1/4

em q

2 /2 s

(II.40)

If we consider all modes simultaneously, then


!1/4
Y m (~k)s
~ s ~ 2
~ s
s (q) =
em (k) qj (k) /2 (k) .
(~k)s
0
~

(II.41)

kK
j=e,o

While this is an alternative derivation of (II.15), it also illustrates how mass scalingwhen applied
to a wave functiongenerates an overall renormalization of the wave function as well as a scaling
of the field.

II.11

Expectation Values of Products of Fields

Now we calculate expectations in the vacuum state of products of fields, as well as expectations of
time ordered products. We take the field theory for the mass perturbation Hamiltonian Hrenormalized
defined by (II.1). This gives the full solution to the mass-shift quantum field theory model, and
completes the justification of calling this field theory a mass perturbation Hamiltonian. Explicitly
the field is
(x) = (~x, t) = eitHrenormalized (~x)eitHrenormalized .
(II.42)
We consider expectations in the ground state of Hrenormalized , the vacuum state |i. We compute
Gn (x1 , x2 , . . . , xn ) = h| T (x1 )(x2 ) (xn ) |i ,

(II.43)

Wn (x1 , x2 , . . . , xn ) = h| (x1 )(x2 ) (xn ) |i .

(II.44)

and

We find that these expectations are just the expectations for a free field of mass m =
Let us denote the corresponding expectations for free fields as

2 + .

Gn (x1 , x2 , . . . , xn )0,m = h0| T m (x1 )m (x2 ) m (xn ) |0i ,

(II.45)

Wn (x1 , x2 , . . . , xn )0,m = h0| m (x1 )m (x2 ) m (xn ) |0i .

(II.46)

and
They satisfy the same type of Gaussian recursion relation that free fields do,
X
W2n (x1 , . . . , x2n ) =
W2 (xi1 ,xi2 ) W2 (xi2n1 ,xi2n ) ,
(2n1)!! pairings

(II.47)

17

Physics 253a, Fall 2004

and likewise for Gn . Thus we claim that


Wn (x1 , x2 , . . . , xn ) = Wn (x1 , x2 , . . . , xn )0,m ,

and Gn (x1 , x2 , . . . , xn ) = Gn (x1 , x2 , . . . , xn )0,m .


(II.48)

Let us first establish this for Wn . We use four properties of the mass scaling transformation
W defined in (II.22):
(i.) W is unitary, as it is a product of unitaries.
(ii.) W (~x)W = m (~x), proved in (II.24).
(iii.) W Hrenormalized = H0,m W , established in (II.25).
(iv.) |i = W |0i, derived in (II.27).
As a consequence of (iiiii), we infer that
W (x)W =
=
=
=

W (~x, t)W
W eitHrenormalized (~x)eitHrenormalized W
eitH0,m W (~x)W eitH0,m
eitH0,m m (~x)eitH0,m = m (x) .

(II.49)

We therefore infer from (i) and (iv) that


Wn (x1 , x2 , . . . , xn ) =
=
=
=

h| (x1 )(x2 ) (xn ) |i


h0| W (x1 )W W (x2 )W W (xn )W |0i
h0| m (x1 )m (x2 ) m (xn ) |0i
Wn (x1 , x2 , . . . , xn )0,m ,

(II.50)

as desired. Since the time-ordered vacuum expectation values are determined by the non-timeordered vacuum expectation values, we have solved the mass-perturbation quantum field theory!
Our next step will be to solve this problem a second time, using perturbation theoryrather
than methods we use here that lead to an exact, non-perturbative solution. Of course the two
methods must give the same answer.

Potrebbero piacerti anche