NORDITA NORDISK INSTITUT FOR TEORETISK ATOMFYSIK

Danmark . Finland . Island . Norge . Sverige

FIELD THEORY

Predrag Cvitanovie
(lecture notes prepared by Ejnar Gyldenkerne)

NORD ITA LECTURE NOTES

January 1983
TO MY GIRLS
 - 1 -

CONTENTS:

PREFACE 3
1. INTRODUCTION 4
A. Land of Quefithe 6
2. GENERATING FUNCTIONALS 9
A. Propagators and vertices 9
B. Green functions 10
C. Dyson-Schwinger equations 12
D. Combinatoric factors 13
E. Generating functionals 15
F. Connected Green functions 18
G. Free field theory 20
H. 1PI Green functions 20
1. Vacuum bubbles 25
J. Summary 30

3. PATH INTEGRALS 33
A. A Fourier transform 33
B. Gaussian integrals 34
C. Wick expansion 36
D. Tree expansion 38
E. Legendre transformations 40
F. Saddlepoint expansion 40
G. Point transformations 44
H. Summary 45

4. FERMIONS 47
A. Pauli princi pl e 47
B. Anticommuting sources 49
C. Fermion arrows 51
D. Ferm;on;c path integrals 54
E. Fermionic determinants 55
F. Fermionic jacobians 57
G. Summary 58
 - 2 -

5. SPACETIME PROPAGATION 61
A. Free propagation 61
B. A leap of faith 66
C. Scattering matrix 67

6. FROM GHOULIES TO GHOSTIES 71

A. Massless vector particles 71
B. Photon propagator 74
C. Colored quarks 76
D. Compton scattering 77
E. Color algebra 78
F. Three-gluon vertex 79
G. Ghosts 84
H. Four-gluon vertex 85
I. QCD action 88
J. SUl1lllary 91

7. QCD WARD IDENTITIES 93

A. Ward identities for full Green functions 93
B. Examples of Ward identities 96
C. Becchi-Rouet-Stora identities 97

APPENDICES 101
A. 2-particle irreducibility 101
B. Solution to the ufind 7 errors II problem 109
C. Some popular gauges 113
D. Feynman rules for QCD 115
 &

- 3 -

PREFACE
In the fall of 1979, Benny Lautrup and I set out to write
the ultimate Quantum Chromodynamics review. The report was
going to consist of four parts, one for each line of

From Ghoulies and Ghosties

and Long-leggetg Beasties
and Things that go bump in the Night
Good Lord, deliver us!

Ghoulies are body-snatchers and grave robbers; they are those

revel in that which is revolting.~.

Benny had previously described ghosties in a very nice set

of QCD lectures t which we were going to use as the first part.
Green functions resemble long-leggety beasties; they, and the
general formalism of field theory, were to be developed in the
second part. The things that go bump in the night are clearly
the many unpleasant surprises of field theory; divergences, to-
gether with the regularization and computation techniques, were
to be covered in the third part. Finally, good Lord deliver us,
we were gOing to actually calculate a few basic QCD integrals.
W~ll, while I was lecturing about the long-leggety beasties,
Benny deserted me for lattice, and the ultimate QeD review was
never written. That these lectures appear at all is largely due
to tireless work by Ejnar Gyldenkerne and to the criticisms of
the QeD study group at the Niels Bohr Institute. In writing
these lectures I have profited much from discussions with Benny
Lautrup, to whom I direct my thanks.

t B• Lautrup, "Of ghoulies and ghosties - an introduction to QCD", Basko Polje

1976 lectures, available as Niels Bohr Institute preprint NBI-HE-76-14.
 - 9 -

2. GENERATING FUNCTIONALS
A. Propagators and vertices
A particle (an elementary excitation of a theory) is speci-
fied by a list of attributes; its name, its state (spin up, in-
going, ••. ), its spacetime location, etc. To develop the form-
~~a~li~inOffield theory, onedoes~ not'need any specific part o f - - - - -
this information, so we hide it in a single collective index:

i = {q,a,a, !l,X , •••• }

].I

q particle type
a colour
~ spin
].I Hinkowski indices
x spacetime coordinates (2.1)
].I

A particle is an interesting particle only if it can do

something. The simplest thing it can do is to change its posi-
tion, its spin or some other attribute. The probability (ampli-
tude) that this happens is described by the (bare) propagators:

I'.ij = ••- - " (2.2)

i j

Beyond this, many things can happen; a particle can split in-
to two, or three, or many other particles. The probability (am-
plitude) that this happens is described by (bare) vertices:

Y kim = (2.3)

A particle can also be created (or removed from the system).

This is described by a source (or a sink):

J =~ (2.4)
i
 - 10 -

The concept of a particle makes sense only if its persist-

ence probability (2.2) is appreciable, i.e. if (2.3), the prob-
ability of its disintegration, is relatively small. In that
~ ~~~ -~ ...... -~. --..................... - : - - -•......: ............. -~.-

case the interactions (2.3) may be treated as small corrections,

and the perturbation. theory applies. If the "particle" de-
scribed by attributes (2.1) has a negligible persistence proba-
bility, the theory should be reformulated in terms of another
set of "elementary excitations" which are a better approxima-
tion to the physical spectrum of the theory (an easy thing to
say) •
How many identical particles (particles with all the same
labels) can coexist? We shall consider two extremes; infinity
(bosons) or at most one (fermions). Other more perverse possi-
bilities cannot be excluded. Assumption of additivity of proba-
bilities/amplitudes then implies that the bosonic propagators
and vertices must be symmetric under interchange of indices
l1ij =l\ji' Yijk=Yjik=Yikj = .•.• (The argument is similar to the
one we shall use for fermions in chapter 4). For the time be-
ing, we assume that the vertices (2.3) are symmetric.

B. Green functions
A typical experiment consists of a setup of the initial
particle configuration, followed by a measurement of the final
configuration. The theoretical prediction is expressed in terms
of the Green functions. For example, if we are considering an
experiment in which particles i and j interact, and the outcome
is particles k, ~, and m, we draw the corresponding Green func-
tions

(2.5)

(remember that labels i, j, stand for all variables and in-

dices which specify a particle.)
A Green function is a sum of the probabilities (amplitudes)
 - 11 -

associated with all possible ways in which the final state can
be reached. This is represented by an infinite sum of Feynman
diagrams:

Each Feynman diagram corresponds to a sum (or an integral). For

example, diagram

represents the probability that 1) a particle whose type, loca-

tion, etc. is described by the collective index a reached any
state labeled £; 2) that £ splits into any two particles labeled
c and £, and so forth. The intermediate states are summed over
the entire range of possible index values

Here the summation signs imply sums over discrete indices (such
as spin) and integrals over continuous indices (such as position).
In the future we shall drop the. explicit summation signs, and
use instead Einstein's repeated index convention; if an index
appears twice in a term, it is summed (integrated) over.

Exercise 2.B.l Continuous indices. For QCD the collective index i

stands for:
x lJ spacetime coordinates,
II :::: 1,2, .. . . _1 d Minkowski indices I
j = 1,2, •.• ,N gluon colours.
 - 12 -

If the propagator,is denoted by D&~(X,y) and the three-

gluon vertex by yt~~(x,y,z), write down the complete ex-
pr.ession for the above self-energy diagram.

C. Dyson-Schwinger equations
A Green function consists of an infinity of Feynman dia-
grams. For a theory to be manageable, i t is essential that
these diagrams can be generated systematically, in order of
their relative importance.
Consider (for simplicity) a theory with only cubic and
quartic vertices t . Take a Green function and follow a particle
into the blob. Two things can happen; either the particle sur-
vives

or it interacts at least once:

===>
More precisely, entering the diagram via leg 1, we either reach
leg 2, or leg 3, ••. , or hit a three-vertex, or a four-vertex,
etc. Adding up all the possibilities, we end up with the Dyson-
Schwinger equations:

+ + (2.6)

tRemember that the different particle types are covered by a single collect-
ive index, so QCD is also this type.

;~b
.," ..~ ...... Hi.
 - 13 -

Iteration of the Dyson-Schwinger (DSI equations yields all

Feynman diagrams contributing to a given process, ordered by
the number of vertices (the order in perturbation theory).
A few words about the diagrammatic notation; a diagrammatic
equation like (2.6) contains precisely the same information as
its algebraic transcription

G., k,;/}"G, k+lI'kG , ,+ ••• +lI .. G k'

~).. N ~x, J.. ~ J •• " ~J .. k

Indices can always be omitted. An internal line implies a sum-

mation/integration over the corresponding indices, and for ex-
ternal lines the equivalent points on each diagram represent
the same index in all terms of a diagrammatic equation. The ad-
vantages of the diagrammatic notation are obvious to all those
who prefer the comic strip editions of "The greatest story ever
t
told" to the unwieldy, fully indexed version . Two of the prin-
cipal benefits are that it eliminates "dummy indices" and that
it does not force Feynman integrals into one-dimensional format
(both being means whereby identical integrals can be made to
look totally different).

D. Combinatoric factors
For a three-leg Green function the DS equations yield

.. -
j ,;~ J?~
= (disconnected) + +

= (disconnected) +2 -< - - _.....-

+ (more vertices)
---~ --.-...... -~

It is rather unnatural that an expansion of a three-leg Green

function does not start with the bare three-vertex, but twice

t e • Itzykson and J.-B. Zuber, Quantum Field Theory (McGraw-Hill, N.Y., 1980).
 -14 -

the bare three-vertex. This is easily fixed-up by including

compensating combinatorial factors into DS equations:

+ -..

... +_ 1
(k-l)\
_-<:,
..
. --
-fO:it}. #

. (2.7)

To illustrate how the DS equations generate the perturba-

tion expansion, we expand a two-leg Green function up to one
loop:

+ (more loops)

The one-loop tadpole is given by

= i· c@ + (more loops)
+ (more loops)
(2.8)

substituting the tadpole into the above, we finally obtain the

~~~

self-energy expansion up to two vertices with all the correct

combinatoric factors:

~-=~f:_l__ =
_ t_ _ ___ + i...()-.+ i...Q....+ i.2.. + to-{) ()-o + (more loops)

~ (2.9)

This expansion looks like the usual ;pa + $4 theory, but it is

not only that: the combinatoric factors are correct for any
theory with cubic and quartic vertices, such as QeD with its
full particle content.
 - 15

Exercise 2.D.1 Feynman diagrams in.the collective index notation look

diagrams for scalar field theories. Nevertheless, they do
contain the peiturbative expansion for theories with arbitrary
particle content. As an example, consider a QED-type theory
with an Hin" particle (electron), and "out" particle (positron)
and a scalar particle (photon). The collective index (2.1) now
ranges over an array of three sub-collective indices
a¥ in] ....-.....- electron
i:1 a, out := ... positron
[
U ~ photon
Index ~ stands for the charged particle's position and spin,
and index ~ stands for all labels characterizing the neutral
particle. The uin ll - "out JJ labels can be eliminated by taking
~ to be an upper index for lIin ll particles, and a lower index
for "out particles. Diagrammatically they are distinguished
ll

by drawing arrows pointing away from upper indices and down

into lower indices:

Show that if the sources and fields are replaced by J =

(Ija, Ijb' J u)' 1> = ('iJ a , ljib, AU), the combinatoric factors in
(2.9) cancel, and the vertices such as the electron-pcsitron-
photon vertex have no combinatoric weight:

Exercise 2.D.2 Write the Dyson-Schwinger equations for QED-like

t.heories. (We say IlQED-like" because electrons are fermions.
We shall return to the fermion DS equations later.)

Exercise 2.D.3 Determine the one-loop self-energy diagrams (2.9) for

QED-like theories.

E. Generating functionals
The structure of the DS equations is very general~ still,
at present we have to write them separately for two-leg Green
function, three-leg Green function, and so on. To state rela-
tions between Green functions in a more compact way we intro-
duce generating functionals. A generating functional is the
vacuum (legless) Green function for a theory with sources (2.4):

=1+ , (2.10)
 - 16 -

(as J. is a function which depends on both discrete and con-

~

tinuous indices, Z[J].is. a Jurictinnal). The coefficients in

this expansion are the u'sual Green furic·tions. They can be re-
trieved from the generating functional by differentiation:

d
ell Z[J]I ' etc. (2.11 )
k J=O

The DS equations (2.7) can be written as

d
ell
{1 d
+2 Y'kR, ell
Z[J] =1:. .. J.
d
ell
-1
+ 31 Y, u
d d
ell ell ell
d}
Z[J]
i 1) J J. 9., k • J m m 9., k
(2.12)

The bare propagators and vertices can themselves be collect-

ed in a functional called the action:

S[<I>] : - ~ ¢il!.~~¢j + SI[¢] , (2.13)

S [CP] =L y. <l>itPj ••• CPR, (2.14)

I m ljk .. £
'_v_---'
m'•
m legs

Now the Dyson-Schwinger equations can be stated in an even

more elegant way:

(2.15)

where

The action (or the Lagrangian) is just another way of defining

the propagators and vertices for a given theory. Giving the
Lagrangian or listing the Feynman rules is one and the same
thing.
 - 17 -

Exercise 2~E.1 Functional derivatives. For continuous indices the

Kronecker deltas are replaced by Dirac deltas. For example,
check that in d-dimensions
dJ (x) d
6 (x-y) ,
dJ (y)
is the correct definition of the derivative in (2.11).

Exercise 2.E.2 Feynman rules. Consider¢" theory given by the Lagrangian

density
.c (x) 1 3 q, (xl 311 q, (x) _l m2 q, (xl 2 _ q,(x) 3
11 2

S = JddX£ (xl.
Read off the bare propagators and vertices (the Feynman rules)
from the Lagrangian.
d d"
Hint: y ..
~J..
k= d~.··
~J
d~"'k S[<I> 11<1>=0 '

and the derivatives are in this case functional derivatives.

Exercise 2.E.3. Zero-dimensional field theory. Consider a <1>' theory

defined by trivial Feynman rules
0--0=1 r ,A.,=g.
The value of a graph with k vertices is gk, and k-th order
contribution to Green function is basically the number of
contributing diagrams. More precisely, if
Z[J]=l: G(m)gk Jm
k,m k m!

the Green function

G (m) = l: C
k G G
is the sum of combinatoric factors of all diagrams with m
legs and k vertices. Use the Dyson-Schwinger equation (2.7)
to show that for a free field theory
G(m) = (m-Il!! m even
o
=0 m odd
Diagrammatically
G(2l =_~1

G(4) =, <. + X +~ = 3, etc.

The zero-dimensional field theory is about the only field

theory which is easily computable to all orders. We shall
use it often to illustrate in a concrete way various field-
theoretic relations •
 - 18 -

F. Connected Green functions

Generating functionals are a powerful tool for stating re-
lations between Green functions. For example, we can use them
to derive relations between the ·full and the connected Green
functions:

Pick out a leg and follow it into a full Green function.

This separates all associated Feynman diagrams into two parts
- the part that is connected to the initial leg, and the re-
mainder:

dW(J]
~. Z(J] dJ.
Z[J] (2.16)
'- '-

The generating functional for the connected Green functions

is defined in the same way as (2.10), the generating functional
for the full Green functions:

+-1ft
3!
+ ••••••• (2.17)

The differential equation (2.16) is easily solved

(2.18)

A disconnected Feynman diagram such as

represents a product of two independent processes; one could

take place on the moon, and the other in Aarhus. The physical-
ly interesting processes are described by the connected Green
functions. To obtain a systematic perturbation series which
 - 19 -

includes only the connected Feynman diagrams, we use the identi-

tyt

_1_ ~ Z[J] = dW[J] + ~ (2.19 )

Z[J] dJ. dJ. dJ,
~ 1. L

to rewrite the DS equations (2.15) in terms of the connected

Green functions:

o = dq,.
dS [dW[J]
dJ
+.£.]
dJ
+J i
(2.20)
~

This is very elegant, but possibly not too transparent. To get

a feeling for these equations, take _the <p 3 + q, 4 DS equations
(2.12) and substitute Z[J] = exp(W[J]). The result is, in the
functional notation

dW[J]
dJ. &ij
1.

(2.21)

and in the longleggedy notation

After reaching a vertex, one continues into diagrams that are

_..... --~- - - -

either mutually disconnected, or connected - that is the reason

that there are extra terms in the connected DS equations, com-
pared with the full Green functions equations (2.12).

t more exp l'l.Cl.t

. 1y
 - 20 -

Exercise 2. 1 Use DS equations (2.2.1) to compute self-eJ;lergy to one

How does the result differ .from (2.9)7

Exercise 2 F 2 Expand some full Green functions in terms of the con-

ones;

~ {~+ .....}@)
)( ~ )ff(+X+'~+"
+~ + (10 tenn!»

Hint: iterating (2.19) is probably the fastest way.

G. Free field theory

The connected generating functional for a free field theory
is trivial: there are no interactions, so the only connected
Feynman diagram is the propagator:

(2.22)

For the free field theory (2.18) gives an explicit expression

for the generating functional:

~l}() + •••• (2.23)

H. One-particle irreducible Green functions

A one-particle irreducible (1PI) diagram cannot be cut into
two disconnected parts by cutting a single internal line. An
arbitrary connected diagram has in general a number of such lines.
The connected and the 1PI Green functions can be related by our
usual diagrammatic trick:

Pick out a leg of a connected diagram. This pulls out a 1PI

 - 21 -

piece, which ends in 0, 1, 2, ... lines whose cutting would dis-

connect the diagram. Those lines continue into further connected
pieces:

~ =.-.."c+ - . +~ +t-< +.....

(2.24)

Here the "field" q, is defined by

= <P = dW[J] (2.25 )

i dJ
i

We draw the lPI Green functions as cross-hatched blobs

r"~J . • k =i.
k
Unlike the full and the connected Green functions, the 1PI ones
do not have propagators on external legs - the external indices
always belong to a vertex of an 1PI diagram. This is indicated
by drawing dots on the edges of 1PI Green functions. Any con-
nected diagram belongs to one and only one term in the expansion
(2.24). For example, going into connected diagram

we pullout a lPI

followed by connected bits

Multiplying both sides of (2.24) by the inverse of the bare

propagator we obtain

i!"..
 - 22 -
(For reasons which should soon be clear, i t is convenient to
define the two-leg r as f ij =-l\~~ + 1fij , where 1fij is the 1PI two-
leg Green function, or the proper self-energy.)

Collecting all 1PI Green functions into th~~ffectiv~ ac-

tion functional

r[¢] = L r, , h' "JJj</>i , (2.26)

m=l lJ .. k m!

we can write (2.24), the relation between the connected and the
1PI Green functions, as:

O=J +dIH]
i d</>.l. '

0= --iC:+ ~. (2.27)

ThiS, together with (2.25), can be summarized by a Leg.endre trans-

formation

W[J] = r[<I>] + ¢.J. • (2.28)

l. l

(2.27) guarantees that W is independent of ¢, and (2.25) guaran-

tees that r is independent of J:

dW[J]
d¢
=0, dIHl = o.
dJ

This is elegant, but how does it help us to get 1PI Green func-
tions? The point is that we are not interested in extracting 1PI
Green functions from the connected ones; what we need are the
lPI Dyson-Schwinger equations, i.e. the systematics of generating
1PI diagrams (and only 1PI diagrams). To achieve this, we must
first eliminate J-derivatives in favour of ¢-derivatives (cf.
(2.25)):

2
...2...
= d¢j ...2... = d W[J] ...2...
dJ. dJ. dtj>. dJ.dJ. d¢.
l. l. J l. J J

(2.29)
 - 23 -

This accounts for all self-energy insertions. The. right-hand

side can be expressed in terms of 1PI Green functions by taking
a derivative of (2.27):

(2.30)

In order to understand this relation diagrammatically, we sepa-

rate out the bare propagator in (2.26) by defining the "inter-
action" part of f:

(2.31)

Now (2.30) can be written as

d 2 W[J]
d2 [<Pl
dJ.dJ. = "'ij + ilik d<P d <P 2 lI.Q,j + ....
1. J k

.=-+...-.@- +........-.- + ...

1
W[J]" = - - - ' - - - (2.32)

Diagrammatically W' is a complete propagator which sums up all

proper self-energies.

We can use (2.25) and (2.27) to eliminate source-dependent

functionals in favour of field-dependent functionals, and (2.2.9)
to replace J-derivatives by <p-derivatives, in order to rewrite
(2.20) as the lPI Dyson-Schwinger equation:

The form of this equation is one of the reasons why the

generating functional for lPI Green functions is called the
effective action. If the derivatives are dropped, the effective
action reduces to the classical action. The role of the deriva-
tives is to generate loops, i.e. quantum corrections (or sta-
tistical fluctuations). We shall return to this in our discussion
 - "....
-
_---
24 -

of path integrals.
DS equations. (2.33) are again so elegant that one is proba-
bly at a loss as to what to do with them. To get a feeling for
their utility, we write them out for theq,3 +q," example (2.21):

=-

d 2 wlJ) d 2 W[J) d 2 wIJJ d'r[CPl

dJjdJm dJkdJn dJ 2dJa dq, d<p dcp
m n a

(2.34)

Such equations are used iteratively. For example, to obtain the

DS equation for the proper self-energyt, take a derivative of
(2.34):

=-

+!~
6~2.35)
Exercise 2.H.l Use (2.32) to show that

dt--- ~
~Th::-:i-s--:i-s-a:usefuii-"dccen=-t;CCl'--;'ty-for ~der"i~v'1CCn~g-r-e~11';:hoiis
(2.36)

such as (2.34) ...~.~~~-~. -~ - ~

and (2.35).

Exercise 2.H.2 Take successive derivatives of (2.30) to show that the

connected Green functions can be expanded in terms of 1PI Green
functions as

tHere the slash stands for inverse propagator; diagrammatically it is a two-

leg vertex. Other vertices are denoted by dots, and a line connecting two
vertices is always a propagator, so that tl;:j/ljk~i""'k=i-k=Oik'
 25 -

Exercise 2.H.3 Jens J. Jensen, a serious young student of field theory,

is getting set to compute the two-loop QCD beta-function. He
has dra'V'ln up a list of gluon corrections to the three-gIllon
vertex. Use the IPI Dyson-Schwinger equations to check this
list and make Jens aware of 7 (seven) errors before he rushes

1
+- +1.
2 2

1 1
+- +-
2 2

1
+-
2

1 1
+- +-
4 4

1
+-
4

I. Vacuum bubbles
The Green function formalism we have developed so far is
tailored to scattering problems; all the Green functions we
 - 26 -

have c.onsidered had external legs. ;Processes without external

particles (the corresponding legless diagrams are called. vacuum
bubbles) are also physically interesting. For example, if a
particle is propagating through a hot-, dense soup t r a particle-
particle scattering experiment would be a hopeless and messy
undertaking. Such systems are probed by varying bulk parameters,
-~.--

-----such -as--temperature. Indeed, the generating funct.i.onals~d6 not -

depend only on the single-particle sources Ji , but on all inter-
action parameters

(2.38)

Any of these, or any combination of these, can be varied. Dia~

grammatically we view an n-vertex as an n-particle source. For

example, if we rescale Yi') • • k 4GY1)
.. •• k and vary infinitesimally
the coupling constant g, we "touch" each y. _ k vertex in a
1) ••
Green function:

d .1
g -Z[J] =_ =..'Ly ~ (2.39)
dg k! k! ij .. k dJ
k

We can use such generalizations of the Dyson-Schwinger

equations (from varying single-particle sources J i to varying
many-particle sources YIjk .. ~) to compute hosts of physically
significant quantities. One such quantity is the expectation
value of the action. We rescale the entire action (2.13)

(2.40)

and vary fi (depending on the context, fi could be the Planck

constant, coupling constant, inverse temperature or something
else) :

n
1 ~._.'.,
d Z[J] = - v1 ( - ~
fi -dfi 2 ,,,-,.,,
.... "
+_1_
3!
~."'. +4\~+"')
=-~S[~]Z[J] • (2.41 )

tminestrone, to be specific.
 - 27 -

To normalize the expectation value properly, we divide byZ[J]:

--_ .. _-
(2.42)

That this is really an expectation value will perhaps be easier

to grasp in the path-integral formalism, cf. (3.11) in the next
chapter. Anyway, we can use (2.19) to rewrite the above in terms
of connected Green functions:

+-1~-
31
+_1
4!
~
~

=~ +t~ +_I~
31
+1.
4

~ ~
1 +_1_ +_1
2 31 41

(2.43)

(the diagrammatic expansion is for the </>3 + ¢4 theories). Even

better, we can use (2.25) and (2.29) together with the identity
(follows from (2.28)

dW[J] _ dr[p] (2.44)

(ffi-dl\.

to relate the <8[</>]> to the effective action:

(2.45)

The above expansions can be used to compute the perturbative

 - 28 -

expansions for the connected and 1PI. vacuum bubbles (see. exer-
cises). Their physical significance will become clearer in the
next chapter.
•
Exercise 2.I.l Loop expansion. Show that with action (2.40) the ex-
pansion in powers of l'i is the locp expansion, 1. e. that each
loop··ih a Feynman diagram carries a factor l'i. Hence the loop
expansion offers a systematic way of computing quantum correc-
tions (or thermal fluctuations in statistical mechanics). Hint:
each propagator carries a factor Ii, while each vertex carries
l'i- 1 •

Exercise 2.I.2 Free energy W[O]. Compute

..1:. w[O]
II
~ °ii2 2.nJ'i + _1_
12
e + •••

for </>3 +.p4 theory. Hint: use (2.43) and the DS equations (2.21).

Exercise 2.I.3 Gibbs free energy r[o]. Compute

1.
l'i
r[0] ~ °ii2 9.nl'i,: {_I
12
e 1
+-8 00 }
J 1
+ '24
+ ... (2.46)
for .pS + </>4 theory. Hint; use (2.45) and the DS equations (2.34).
Note that the one-particle reducible diagrams from w[O] are in-
deed missing. The vacuum-bubble combinatoric weights are not al-
ways obvious - equation (2.45) provides the fastest way of com-
puting them, as far as I know.

Exercise 2.I.4 Show that for the zero-dimensional </>3 theory (continua-
tion of exercise 2.E.3)
G(m) ~ (m-! + 3g ~)G(m-2)
dg
Hint: use (2.39) together with the Dyson-Schwinger equations
(2.12).
Show also that
G(I) ~!I G(2)
2
Hence all Green functions can be computed from Z '" G (0) , the
vacuum bubbles. Show that these satisfy
d 5 9 d z(
3 Zd )
~ g 12 + "4 g dg + "4 g dg2 Z
2

Compute the first few terms of the expansion in powers of g.

The complete solution is given in exercise 3.C.l.
 - 29 -

Exercise 2.1.5 Zero-dimensional field theory. Show that the connected

"ac;;uum bubbles 1'1 - w[oJ satisfy
d 25
[ 9dW 3 2 dW dW
~ = g 1:2 + "4 ~ + "4 g dg 2 +dg ( 2 ()2)]
Use this equation to derive recursion relations for connected
m-leg Green functions. Compute the exact propagator D = GC ('2)
2 25 4 390 6
D = 1 + g + 8 g + 32- g + .....

and check that this agrees with

D2=~-o-d..f- =1,
D 1+ = ! -<It- + !<3-- + t ..00-
+ rQ:2- + ~~ +~-6-
+!..st-1+
+!~
8
+!
1+
~ +!
1+
n =~
8
Hint: establish first that

b[Jl =~+ J+~Jd~ + 3~)w[Jl

That relates Gc(m) to the vacuum bubbles W.

Exercise 2.1.6 Zero-dimensional ~3 theory. Combine the DS equation

(2.34) and the previous results to relate IP1 Green fUnctions
with different numbers of legs:

:q,rlq,l =~- q, +*4> ddq, + 3~)rlq,l ,

satisfy

I ••••• •

compute the proper self-energy

1 2
TI=-g+g+-g+4 35 6
2 8

and the proper three-vertex r " r (3 )

r=g+g3+5g5+35g7+

r =A
r:=L~, +A +A A +
d~ dA d,A =5 .
Compare 7T with the preceeding exercise, D = (1 - 7T) -1.

Exercise 2.1.7 Check (2.44).

 - 30 -

J. Summary of the generating functional formalism

generating
functionals

Full Green functions!

Connected Green functions: G(c) k W[J]=

ij .. k

iPI Green functions: r[<I>]=

Full ++ connected relation:

1 d
ZTJT dJ. Z[J]
~

Connected ++ 1PI relations:

¢ =dW[J]
i dJ.
~
1~

---'~~":~';~ + ..........

Dyson-Schwinger equations:

full (i!£..[~]
d¢. dJ
~
+ J.)z [J] = 0
~
,

connected dS [dW[J]
d¢. dJ
+~]+J=O
dJ i
~

lPI

By now we are thoroughly fed up with longleggedy beasties, and

the diagrammatic manipulations:
 - 31 -

,-yiNGTRE NUDO DEL. DI~8LO

ota.
OEVII..:S KNOT

A uU:==========~4 B .

1.

'3, B 4-.
A

s,

(CoNTINueo ",eXT WEE\'.)

Let us now see whether the crow's vision of Quefithe is any more
fun than the mole's version.
Critics say:
... Seen in [Cvitanovic's] framework. field theory books
are like every other form in the universe: they are
generated by changing intervals of tension between a
dominant system and a competing system in a space-time
continuum that is dependent on the process of competition
between these two stabilities and not on any General
Concept of Space and Time ... [Cvi tanovi CiS ]metbod thus
valorizes the microcosm. which illuminates macro~mic form
by the hi gh tendency of m; crocosmi c patterns to~peat
themselves and so greatly 1 imit structural varia~10n in
the macrocosm ... But on another 1eve1. as ; n~ the sagas.
the Song of Rolland. the I11iad. the Odyssey. the
NibelU"¥enlied. the Aeneid and Beowulf. the real dynamic
focus 0 the book is the!power of anger.

Patricia Harris Stablein

A distinguished reviewer says: