Ann. Phys.

(Leipzig) 8 (1999) 5, 389Ð404

Perturbative gauge invariance: electroweak theory II

Andreas Aste and GuÈnter Scharf 1

Institut fuÈr Theoretische Physik der UniversitaÈt ZuÈrich,
Winterthurerstraûe 190, CH-8057 ZuÈrich, Switzerland

Michael DuÈtsch 2
II. Institut fuÈr Theoretische Physik, UniversitaÈt Hamburg,
Luruper Chaussee 149, D-22761 Hamburg, Germany

Received 5 March 1999, accepted 8 March 1999

Abstract. A recent construction of the electroweak theory, based on perturbative quantum gauge
invariance alone, is extended to the case of more generations of fermions with arbitrary mixing.
The conditions implied by second order gauge invariance lead to an isolated solution for the ferm-
ionic couplings in agreement with the standard model. Third order gauge invariance determines
the Higgs potential. The resulting massive gauge theory is manifestly gauge invariant, after con-
struction.

1 Introduction

In a recent paper [1] we have described a formulation of the electroweak theory in

terms of asymptotic fields alone. Since in our causal approach the asymptotic
gauge fields are massive from the very beginning (except the photon), spontaneous
symmetry breaking plays no role.
Let us summarize the strategy. The fundamental problem in gauge theories is
the appearance of unphysical fields in the gauge potentials. An additional main
difficulty is that already the quantization of the massive 3 free gauge fields in a
l-gauge requires an indefinite metric space. To solve these two problems in the
free theory one has to select the space of physical states Hphys which must be a
(pre) Hilbert space, i.e. the inner product must be positive definite. There are two
popular methods at hand: the Gupta Bleuler method (which works well for abelian
gauge theories only) and the BRST-formalism [2]. We take the latter in the form
of Kugo and Ojima [3]: the physical (pre) Hilbert space is defined as the cohomol-
Ker Q
ogy of a nilpotent operator Q, i.e. Hphys ˆ Ran Q. To define such an operator we
introduce unphysical fields: each gauge field Ama gets three scalar partners, the

1
Work supported by Swiss National Science Foundation.
2
Work supported by Alexander von Humboldt Foundation.
3
This problem appears also for the massless free gauge fields. The peculiarity of the massive
(free) theory is that we need the unphysical bosonic scalars Fa to restore the nilpotency of Q.
390 Ann. Phys. (Leipzig) 8 (1999) 5

fermionic fields ua ; u~a (ªghost fieldsº) and the bosonic Fa . Then

def „ $
Q ˆ d3 x …@n Ana ‡ ma Fa † @ 0 ua ;
does the job.
To overcome the presence of unphysical fields in the interacting theory one has
additionally to prove that they decouple. The latter means that in the adiabatic
limit (if it exists) the S-matrix induces a well-defined unitary operator. For this
purpose one needs a notion of gauge invariance. The most natural formulation
which yields the decoupling is to require that Q commutes with the S-matrix in the
adiabatic limit (if it exists). This is essentially the content of our formulation of
gauge invariance
def Pn @ n
dQ Tn ˆ ‰Q; Tn Š ˆ i n Tn=l ; …1:0†
lˆ1 @x l

(where we use the notations of [1]), which is independent of the adiabatic limit, i.e.
it makes sense also in theories in which this limit does not exist. This is a pure
quantum formulation of gauge invariance. It has turned out that this symmetry
requirement (1.0) cannot be satisfied with the fields at hand so far. By introducing
an additional scalar field (the ``Higgs fieldº), which is physical, gauge invariance
can be saved. In addition (1.0) is sufficiently strong to determine the couplings of
all fields.
The main steps in this construction are the following. Starting from the pure
Yang-Mills coupling T1A at first order, gauge invariance requires couplings T1u of
the ghost fields ua u~a and fixes them. This step is the same as in the massless theory
[4]. However, the massive gauge fields need couplings to the bosonic scalar part-
ners Fa to compensate mass terms in the first order gauge variation. Gauge invar-
iance then determines the couplings T1F of these unphysical scalars to the gauge
fields and to the ghosts. But the resulting theory would not be gauge invariant at
second order. This requires the ``Higgs fieldº F0 , its coupling T1F0 to all other
fields is again determined by gauge invariance. The coupling to leptons is uniquely
fixed by first and second order gauge invariance if one family of fermions is con-
sidered [1]. However, it is well known that gauge invariance is then violated at
third order by the triangular anomalies. Hence, the theory must still be enlarged
by including at least one complete generation of leptons and quarks. In what fol-
lows we consider the general situation of arbitrarily many generations of fermions
with arbitrary mixing. It is our aim to analyse what gauge invariance has to say
about the couplings in this case.
Assuming that the reader is familiar with the main results of [1], we present in
the following section the general ansatz for the fermionic coupling T1F and deter-
mine the constraints imposed by gauge invariance at first order. At second order
(Sect. 3) we obtain a long list of interesting conditions for the coupling matrices.
At present we cannot control all solutions of this system. But inserting a general
ansatz which is valid in the vicinity of the standard model, we get a unique solu-
tion that agrees with the standard model. In Sect. 4 we briefly discuss gauge invar-
iance of the third order triangular graphs. We recover the well-known fact that the
cancellation of the axial anomalies restricts the charges of the fermions.
A. Aste et al., Perturbative gauge invariance: electroweak theory II 391

In Sect. 5 we show how gauge invariance of third order tree graphs determines
the last free parameters in the scalar coupling. In this way we derive the Gold-
stone-Higgs potential. Other approaches where this quartic double-well potential is
not postulated but deduced, are based on methods of non-commutative geometry
which have been developed by Connes [5] and others [6, 7, 8, 9, 10], or on super-
connections [11]. The scalar potential comes out in the form
V…F0 † ˆ l…F40 ÿ 4aF30 ‡ 4a2 F20 † ; …1:1†
with
2
aˆ mW ; …1:2†
g
where mW is the W-mass and g the universal gauge coupling constant. The physi-
cal `Higgs' field F0 is realized in an ordinary Fock representation with unique
vacuum and vacuum expectation value hF0 i ˆ 0. (In the usual convention ÿF0 is
used instead of F0.) If we introduce the shifted field
j ˆ F0 ÿ a ; …1:3†
the potential assumes the usual symmetric form [12]
V ˆ l…j2 ÿ a2 †2 : …1:4†
Now the vacuum expectation value hji ˆ ÿa is different from 0. If the shift (1.3)
is carried out everywhere in the scalar couplings (see (3.20) of [1]) we get quad-
ratic mass terms (with the wrong sign) for the already massive gauge bosons.
These quadratic terms can simply be resummed which changes the W- and Z-
masses from their finite values into zero! That means, we have inverted the Higgs
mechanism. This gives the connection of our approach with the standard theory.
However, this connection does not enter our construction.

2 Fermionic coupling and first order gauge invariance

We start from the following generalization of the simple leptonic electroweak cou-
pling to more than one family
P ‡ 1
T1F ˆ ig Wm bjk ej gm nk ‡ Wm‡ b01 j gm g5 nk
jk e …2:1:1†
j; k
‡ Wmÿ b2jk 
nj gm ek ‡ Wmÿ b02 nj gm g5 ek
jk  …2:1:2†
‡ Zm b3jk ej gm ek ‡ Zm b03 j gm g5 ek
jk e …2:1:3†
‡ Zm b4jk 
nj gm nk ‡ Zm b04 nj gm g5 nk
jk  …2:1:4†
‡ Am b5jk ej gm ek ‡ Am b05 j gm g5 ek
jk e …2:1:5†
‡ Am b6jk 
nj gm nk ‡ Am b06 nj gm g5 nk
jk  …2:1:6†
‡ F‡ c1jk ej nk ‡ F‡ c01 j g5 nk
jk e ‡ Fÿ c2jk 
nj ek ‡ Fÿ c02 nj g5 ek
jk  …2:1:7†
‡ F3 c3jk ej ek ‡ F3 c03 j g5 ek ‡ F3 c4jk 
jk e nj nk ‡ F3 c04 n j g5 n k
jk  …2:1:8†

‡F0 c0jk ej ek ‡ F0 c00 j g5 ek ‡ F0 c5jk 
jk e nj nk ‡ F0 c05 n j g5 n k :
jk  …2:1:9†
392 Ann. Phys. (Leipzig) 8 (1999) 5

Here we have used the same notation as in [1] (5.1): All products of field opera-
tors throughout are normally ordered p (Wick monomials). W; Z; A denote the
gauge fields and F ˆ …F1  iF2 †= 2, F3 the unphysical scalars. F0 is the physical
scalar, but ej …x† stands for the electron-, muon-, tau-fields, as well as for the quark
fields d, s, b, and nk …x† represents the corresponding neutrini and the other quark
fields u, c, t. In [1] we have only considered the coupling to leptons. There the
terms (2.1.6) are missing because the neutrini have vanishing electric charge, but
here, for the quark couplings, we must include them. We also assume that the
asymptotic Fermi fields fulfil the Dirac equations
=@ ej ˆ ÿimej ej ; ej gm † ˆ imej ej ;
@m …
=@ nk ˆ ÿimnk nk ; nk gm † ˆ imnk 
@m … nk ; …2:2†
with arbitrary non-vanishing unequal masses 6ˆ mej
6ˆ 0, 6ˆmek mnj
6ˆ 0 for all mnk
j 6ˆ k. We do not use further information about the multiplett structure of the
fermions.
According to [1] the gauge structure is introduced as follows. We define a gauge
charge
def „ P
3 $
Q ˆ d3 x …@n Ana ‡ ma Fa † @ 0 ua ;
aˆ0

where Anastands for An ; W1n ; W2n ; Zn and ua for the corresponding ghosts, and a
gauge variation
def
dQ F ˆ QF ÿ …ÿ1†nF FQ :
Then the gauge variations of the asymptotic gauge fields are given by
dQ Am ˆ i @ m u0 ; dQ W1;m 2 ˆ i @ m u1;2 ; dQ Z m ˆ i @ m u3 ;
and for the Higgs and unphysical scalar fields
d Q F0 ˆ 0 ; dQ F1; 2 ˆ imW u1; 2 ; dQ F3 ˆ imZ u3
and finally for the fermionic ghosts
d Q ua ˆ 0 ; a ˆ 0; 1; 2; 3 ;
m m
dQ u~0 ˆ ÿi @m A ; dQ u~1; 2 ˆ ÿi…@m W1;2 ‡ mW F1; 2 † ;
dQ u~3 ˆ ÿi…@m Zm ‡ mZ F3 † : …2:3†
The gauge variations of the fermionic matter fields vanish.
First order gauge invariance means that the gauge variation of

T1 ˆ T1A ‡ T1u ‡ T1F ‡ T1F0 ‡ T1F

has divergence form
m
dQ T1 …x† ˆ i @m T1=1 : …2:4†

To verify this for the fermionic coupling T1F we calculate the gauge variation of
(2.1) and take out the derivatives of the ghost fields. In the additional terms with
A. Aste et al., Perturbative gauge invariance: electroweak theory II 393

derivatives on the matter fields we use the Dirac equations (2.2):

P  ‡ 1
dQ T1F ˆ ÿg @m u …bjk ej gm nk ‡ b01 j gm g5 nk † ‡ uÿ …b2jk 
jk e nj gm ek ‡ b02 nj gm g5 ek †
jk  (a)
j; k

‡ u3 …b3jk ej gm ek ‡ b03 j gm g5 ek ‡ b4jk 

jk e nj gm nk ‡ b04 nj gm g5 nk †
jk  (b)

‡ u0 …b5jk ej gm ek ‡ b05 j gm g5 ek ‡ b6jk 
jk e nj gm nk ‡ b06 nj gm g5 nk †
jk  (c)

‡ g iu‡ ‰b1jk …mej ÿ mnk † ej nk ‡ b01 e
jk …mj ‡ mk † e
n
j g5 nk Š (d)
‡ iuÿ ‰b2jk …mnj ÿ mek † 
nj ek ‡ b02 n e
nj g5 ek Š
jk …mj ‡ mk † 

‡ iu3 ‰b3jk …mej ÿ mek † ej ek ‡ b03 e e

j g5 ek
jk …mj ‡ mk † e

‡ b4jk …mnj ÿ mnk † 

nj nk ‡ b04 n n
nj g5 nk Š
jk …mj ‡ mk † 

‡ iu0 ‰b5jk …mej ÿ mek † ej ek ‡ b05 e e

j g5 ek
jk …mj ‡ mk † e

‡ b6jk …mnj ÿ mnk † 

nj nk ‡ b06 n n
nj g5 nk Šg
jk …mj ‡ mk † 

‡ mW ‰u‡ c1jk ej nk ‡ u‡ c01 j g5 nk ‡ uÿ c2jk 

jk e nj ek ‡ uÿ c02 n j g5 e k Š
jk 

‡ mZ u3 ‰c3jk ej ek ‡ c03 j g5 ek ‡ c4jk 
jk e nj nk ‡ c04 nj g5 nk Š :
jk  …2:5†
Now, to have first order gauge invariance, the terms (d) until the end of (2.5)
which are not of divergence form must cancel. This implies
b05 06
jk ˆ 0 ˆ bjk ; 8j; k ; b5jk ˆ 0 ˆ b6jk for j 6ˆ k …2:6†
and
i i
c1jk ˆ …mej ÿ mnk † b1jk ; c01
jk ˆ …mej ‡ mnk † b01
jk ;
mW mW
i i
c2jk ˆ …mnj ÿ mek † b2jk ; c02
jk ˆ …mnj ‡ mek † b02
jk ;
mW mW
i i
c3jk ˆ …mej ÿ mek † b3jk ; c03
jk ˆ …mej ‡ mek † b03
jk ;
mZ mZ
i i
c4jk ˆ …mnj ÿ mnk † b4jk ; c04
jk ˆ …mnj ‡ mnk † b04
jk : …2:7†
mZ mZ
The result (2.6) means that the photon has no axial-vector coupling and no mixing
in the vector coupling. This is due to the fact that it has no scalar partner because
it is massless.

3 Gauge Invariance at Second Order

As discussed in Sect. 4 and 5 of [1], the essential problem in second order gauge
invariance is whether the anomalies in the tree graphs cancel out. These anomalies
n 0 n 0
are the local terms in @nx T2=1 jtree …x; y† ‡ @ny T2=2 jtree …x; y† and come from two sources.
First, if the terms (a)Ð(c) in (2.5) are combined with the terms in (2.1) by a
394 Ann. Phys. (Leipzig) 8 (1999) 5

n 0
fermionic contraction we get the Feynman propagator SFm …x ÿ y† in T2=1 jtree …x; y†
n 0
and T2=2 jtree …x; y†. Taking the divergence with respect to the Q-vertex a d-distribu-
tion is generated due to

i @mx gm SFm …x ÿ y† ˆ mSFm …x ÿ y† ‡ d…x ÿ y† : …3:1†

This d-term is the anomaly. Secondly, we can perform a bosonic contraction be-
tween the terms in (2.1) and the terms (4.5) in [1] which are the anomaly-produc-
AF m
ing part in T1=1 , coming from the Yang-Mills and scalar couplings:

AF m
T1=1 jan ˆ ig sin Q…u1 W2n ÿ u2 W1n † @ m An …3:2:1†

‡ sin Q…u2 An ÿ u0 W2n † @ m W1n ‡ sin Q…u0 W1n ÿ u1 An † @ m W2n …3:2:2†


‡ cos Q …u2 Zn ÿ u3 W2n † @ m W1n ‡ …u3 W1n ÿ u1 Zn † @ m W2n …3:2:3†

‡ …u1 W2n ÿ u2 W1n † @ m Zn …3:2:4†
m m m
‡ sin Q…u0 u1 @ u~2 ‡ u2 u0 @ u~1 ‡ u1 u2 @ u~0 † …3:2:5†
m m m
‡ cos Q…u2 u3 @ u~1 ‡ u3 u1 @ u~2 ‡ u1 u2 @ u~3 † …3:2:6†
 
m2
‡ sin Qu0 …F2 @ m F1 ÿ F1 @ m F2 † ‡ 1 ÿ Z2 cos Qu3 …F2 @ m F1 ÿ F1 @ m F2 †
2mW
(3.2.7)
1 mZ  
‡ cos Q …u2 F1 ÿ u1 F2 † @ m F3 ‡ u1 F3 @ m F2 ÿ u2 F3 @ m F1 …3:2:8†
2 mW
1 1
‡ u1 …F0 @ m F1 ÿ F1 @ m F0 † ‡ u2 …F0 @ m F2 ÿ F2 @ m F0 † …3:2:9†
2 2

1 m m
‡ u3 …F0 @ F3 ÿ F3 @ F0 † : …3:2:10†
2cos Q
To give a representative example, we calculate the anomalies with external field
operators u3 F3 ej g5 ek. Combining the first term in (2.5) (b) with the second term in
(2.1.8) we get an anomaly from the contraction of ek0 …x† with ej0 …y† (SF ‰ek0 …x†; ej0 …y†Š
denotes the corresponding Feynman propagator)

ÿiu3 …x† b3jk0 ej …x† gm SF ‰ek0 …x†; ej0 …y†Š g5 ek …y† F3 …y† c03
j0 k :

It results the anomaly  ÿib3 c03 d…x ÿ y† involving the matrix product of b3 with c03 .
Combining the two terms with reversed order, we obtain c03 b3 with a different sign
so that both terms together yield the commutator i‰c03 ; b3 Š. Similarly, the second
term in (2.5) (b) together with the first term in (2.1.8) gives the anticommutator
fc3 ; b03 g, because the g5 is at a different place. An anomaly of the second source
comes from the last term in (3.2.10) contracted by the two F0 -fields with the sec-
ond term in (2.1.9):

i
ÿ u3 …x† F3 …x† DF ‰@ m F0 …x†; F0 …y†Š ej …y† g5 ek …y† c00
jk :
2cos Q
A. Aste et al., Perturbative gauge invariance: electroweak theory II 395

Altogether we obtain the following matrix equation

1
ÿ c00 ˆ i‰c03 ; b3 Š ‡ ifc3 ; b03 g : …3:3†
2cos Q
We now give the complete list of all second order conditions. We specify the
corresponding external legs, then the origin of the terms is pretty clear. Every
combination of external field operators has a corresponding one with an additional
g5 . To save space we do not write down the external legs once more for the g5 -term.
u0 F0 ee : ‰c0 ; b5 Š ˆ 0 ; ‰c00 ; b5 Š ˆ 0
u0 F3 ee : ‰c3 ; b5 Š ˆ 0 ; ‰c03 ; b5 Š ˆ 0
ege : ‰b3 ; b5 Š ˆ 0 ;
u0 Z ‰b03 ; b5 Š ˆ 0
u0 F0 nn : ‰c5 ; b6 Š ˆ 0 ; ‰c05 ; b6 Š ˆ 0
u0 F3 nn : ‰c4 ; b6 Š ˆ 0 ; ‰c04 ; b6 Š ˆ 0
ngn : ‰b4 ; b6 Š ˆ 0 ;
u0 Z ‰b04 ; b6 Š ˆ 0 (3.4)
u‡ A
egn : sin Qb1 ˆ b5 b1 ÿ b1 b6 ; sin Qb01 ˆ b5 b01 ÿ b01 b6
uÿ A
nge : sin Qb2 ˆ b2 b5 ÿ b6 b2 ; sin Qb02 ˆ b02 b5 ÿ b6 b02 (3.5)
‡ 1 5 1 1 6 01 5 01 01 6
u0 F en : sin Qc ˆ b c ÿ c b ; sin Qc ˆ b c ÿ c b
ÿ 2 2 5 6 2
u0 F 
ne : sin Qc ˆ c b ÿ b c ; sin Qc02 ˆ c02 b5 ÿ b6 c02 (3.6)
u3 W ‡ egn : cos Qb1 ˆ b3 b1 ‡ b03 b01 ÿ b1 b4 ÿ b01 b04 ;
cos Qb01 ˆ b03 b1 ‡ b3 b01 ÿ b01 b4 ÿ b1 b04
u3 W ÿ nge : cos Qb2 ˆ b2 b3 ‡ b02 b03 ÿ b4 b2 ÿ b04 b02 ;
cos Qb02 ˆ b02 b3 ‡ b2 b03 ÿ b04 b2 ÿ b4 b02 (3.7)
‡ ÿ 3 1 2 01 02 5 03 01 2 1 02
u W ege : cos Qb ˆ b b ‡ b b ÿ sin Qb ; cos Qb ˆ b b ‡ b b
ÿ ‡ 4 2 1 02 01 6
u W ngn : cos Qb ˆ ÿb b ÿ b b ÿ sin Qb ; cos Qb04 ˆ ÿb02 b1 ÿ b2 b01 (3.8)
u ‡F3 en : c1 ˆ 2…b01 c04 ‡c03 b01 ‡c3 b1 ÿ b1 c4 † ; c01 ˆ 2…b01 c4 ‡ c3 b01 ‡ c03 b1 ÿ b1 c04 †
u ÿ F3 
ne : c2 ˆ 2…ÿb02 c03 ÿ c04 b02 ÿ c4 b2 ‡ b2 c3 †;
c02 ˆ 2…ÿb02 c3 ÿ c4 b02 ÿ c04 b2 ‡ b2 c03 † ; (3.9)
u‡ F0 en : d1 c1 ˆ i…c0 b1 ‡ c00 b01 ‡ b01 c05 ÿ b1 c5 † ;
d1 c01 ˆ i…c00 b1 ‡ c0 b01 ‡ b01 c5 ÿ b1 c05 † ;
u ÿ F0 
ne : d1 c2 ˆ i…c5 b2 ‡ c05 b02 ‡ b02 c00 ÿ b2 c0 †;
d1 c02 ˆ i…c05 b2 ‡ c5 b02 ‡ b02 c0 ÿ b2 c00 † ; (3.10)
0 3 3 03 03 00 3 03 03 3
u3 F3 ee : d3 c ˆ i‰b ; c Š ÿ ifb ; c g ; d3 c ˆ i‰b ; c Š ÿ ifb ; c g
3 0 3 00 03
u3 F0 ee : d3 c ˆ i‰c ; b Š ‡ ifc ; b g ; d3 c03 ˆ i‰c00 ; b3 Š ‡ ifc0 ; b03 g ; (3.11)
5 4 4 04 04 05 4 04 04 4
u3 F3 nn : d3 c ˆ i‰b ; c Š ÿ ifb ; c g ; d3 c ˆ i‰b ; c Š ÿ ifb ; c g
4 5 4 05 04
u3 F0 nn : d3 c ˆ i‰c ; b Š ‡ ifc ; b g ; d3 c04 ˆ i‰c05 ; b4 Š ‡ ifc5 ; b04 g ; (3.12)
‡ ÿ 0 1 2 01 02 3 00 1 02 01 2 03
u F ee : d1 c ˆ i…b c ÿ b c † ÿ ic =2 ; d1 c ˆ i…b c ÿ b c † ÿ ic =2
ÿ ‡ 0 1 2 01 02 3
u F ee : d1 c ˆ ÿi…c b ‡c b †‡ic =2 ; d1 c00 ˆ ÿi…c01 b2 ‡ c1 b02 † ‡ ic03 =2 ;(3.13)
396 Ann. Phys. (Leipzig) 8 (1999) 5

u‡ Fÿ nn : d1 c5 ˆ i…b2 c1 ÿ b02 c01 † ÿ ic4 =2 ; d1 c05 ˆ i…b2 c01 ÿ b02 c1 † ÿ ic04 =2
uÿ F‡ nn : d1 c5 ˆ ÿi…c2 b1 ‡ c02 b01 † ‡ ic4 =2 ; d1 c05 ˆ ÿi…c02 b1 ‡ c2 b01 † ‡ ic04 =2 (3.14)
u3 F‡ en : d4 c1 ˆ b3 c1 ÿ b03 c01 ÿ c1 b4 ÿ c01 b04 ; d4 c01 ˆ ÿb03 c1 ‡ b3 c01 ÿ c1 b04 ÿ c01 b4
u3 Fÿ ne : d4 c2 ˆ ÿb4 c2 ‡ b04 c02 ‡ c2 b3 ‡c02 b03 ; d4 c02 ˆ b04 c2 ÿ b4 c02 ‡ c2 b03 ‡c02 b3 ;
(3.15)
where
1 1 1
d1 ˆ ; d3 ˆ ; d4 ˆ cos Q ÿ :
2 2 cos Q 2 cos Q
The terms with these d's and with the electroweak mixing angle obviously come
from (3.2). There are further combinations of external field operators which have
not been written down, because they give no new condition.
In case of one family, assuming b6 ˆ 0 and taking pseudounitarity into account
([1] (5.21)), the corresponding system of scalar equations has a unique solution,
which agrees with the lepton coupling of the standard model. The solution of the
above matrix equations (3.4Ð15) is not so simple. We start from the equations
(3.5). If we write these equations with matrix elements, using the fact that b5 and
b6 are diagonal (2.6), we easily conclude that b5 and b6 are actually multiples of
the unit matrix
b5 ˆ a1 ; b6 ˆ …a ÿ sin Q† 1 : …3:16†

Here a is a free parameter (the electric charge of the upper quarks or leptons)
and we have assumed that the matrices b1 ; b01 ; b2 ; b02 are nontrivial and non-diag-
onal. This is not a serious limitation because we shall see that these matrices are
essentially the unitary mixing matrices. For the leptons we assume b6 ˆ 0 instead.
This first consequence of gauge invariance is the universality of the electromag-
netic coupling: the members ek …x† and nk …x† of different generations all couple in
the same way to the photon, with a constant charge difference qe ÿ qn ˆ g sin Q,
which is the electronic charge.
Next we turn to the conditions (3.11). It is convenient to introduce the diagonal
mass matrices
me ˆ diag …mej † ; mn ˆ diag …mnj † ; j ˆ 1; . . . ng ;

where ng is the number of generations. Then (2.7) can be written as follows

i i
c1 ˆ …me b1 ÿ b1 mn † ; c01 ˆ …me b01 ‡ b01 mn † ; …3:17†
mW mW

etc. Then the first two equations in (3.11) read

1
c00 ˆ …2b03 me b3 ÿ 2b3 me b03 ‡ me b03 b3 ‡ me b3 b03 ÿ b3 b03 me ÿ b03 b3 me † ;
d 3 mZ
1
c0 ˆ …2b03 me b03 ÿ 2b3 me b3 ‡ …b3 †2 me ‡ me …b3 †2 ‡ …b03 †2 me ‡ me …b03 †2 † :
d 3 mZ
(3.18)
A. Aste et al., Perturbative gauge invariance: electroweak theory II 397

Substituting this into the last two equations of (3.11), we arrive at the following
coupled matrix equations for b3 ; b03 :
d23 …me b03 ‡ b03 me † ˆ 3b03 me …b3 †2 ‡ 3…b3 †2 me b03 ‡ 3…b03 †2 me b03 ‡ 3b03 me …b03 †2
ÿ 3b3 me b03 b3 ÿ 3b3 b03 me b3 ÿ 3b03 b3 me b3 ÿ 3b3 me b3 b03
‡ me b03 …b3 †2 ‡ me b3 b03 b3 ‡ me …b03 †3 ‡ me …b3 †2 b03
‡ …b3 †2 b03 me ‡ b3 b03 b3 me ‡ b03 …b3 †2 me ‡ …b03 †3 me ; …3:19†
03 2 3 2
d23 …me b3 3 e e 3
ÿ b m † ˆ 3…b † m b ‡ 3…b † m b ‡ 3b m b b ‡ 3b m b b e 3 03 e 03 3 03 e 3 03

ÿ 3b3 me …b3 †2 ÿ 3b3 b03 me b03 ÿ 3b03 b3 me b03 ÿ 3b3 me …b03 †2

‡ me …b3 †3 ÿ …b3 †3 me ‡ me …b03 †2 b3 ÿ …b03 †2 b3 me
‡ me b3 …b03 †2 ÿ b3 …b03 †2 me ‡ me b03 b3 b03 ÿ b03 b3 b03 me : …3:20†
The coupled cubic equations (3.19Ð20) have many solutions in general. To de-
termine the solutions in the neighbourhood of the standard model, we substitute
b3 ˆ d3 …b1 ‡ x† ; b03 ˆ d3 …b0 1 ‡ y† ; b; b0 2 C ; …3:21†
and assume the matrices x; y to be small so that only terms linear in x and y must
be taken with in (3.19Ð20). Then the equations collapse to the simple form
2b0 …1 ÿ 4b02 † m ‡ …1 ÿ 12b02 † …my ‡ ym† ˆ 0 ; …3:22†
02
…1 ÿ 12b † …mx ÿ xm† ˆ 0 : …3:23†
0
Now, (3.22) yields a unique solution if b ˆ O…1† is assumed
e2
b0 ˆ ; e2 ˆ 1 ; y ˆ 0; …3:24†
2
where the last result follows by writing the vanishing anticommutator fme ; yg with
matrix elements, using mej > 0; 8j. Then (3.23) implies
…mei ÿ mek † xik ˆ 0 ; no sum over i; k …3:25†
3
that means x and b are diagonal, taking into account that the masses mej are not
degenerate. All matrices in (3.18) commute, thus

c00 ˆ 0 ; …3:26†
e
4m d3 e
c0 ˆ …b03 †2 ˆ m ; …3:27†
d3 m Z mZ
and (3.11) gives
d3 e
c3 ˆ 0 ; c03 ˆ ie2 m : …3:28†
mZ
The same reasoning can be carried through for (3.12) which leads to
d3
b04 ˆ ÿe2 ; c4 ˆ 0 ; …3:29†
2
398 Ann. Phys. (Leipzig) 8 (1999) 5

for the sign see below, and

d3 n
c04 ˆ ÿie2 m ; c05 ˆ 0 ; …3:30†
mZ
d3 n
c5 ˆ m : …3:31†
mZ
With this knowledge we turn to (3.9). Substituting c1 in the first equation by (3.17)
we arrive at
me …b1 ÿ e2 b01 † ˆ …b1 ÿ e2 b01 † mn ;
leading to
b01 ˆ e2 b1 ; …3:32†
assuming non-degenerate masses again. In the same way the second equation in
(3.9) yields
b02 ˆ e2 b2 : …3:33†
This is the chiral coupling of all fermion generations. The sign in (3.29) follows
from (3.8).
Finally, from the four conditions (3.8) it is easy to conclude
b2 ˆ 18 …b1 †ÿ1 ; …3:34†
 
3 1 1
b ˆ ÿ a sin Q ; …3:35†
cos Q 4
1  1
b4 ˆ ÿ ÿ a sin Q ‡ sin2 Q† : …3:36†
cos Q 4
This means that x in (3.21) is actually zero, so that there is no other solution in the
neighbourhood of the standard model. But solutions ªfar awayº are not excluded.
All values of the b's and c's agree with the standard model [12] for an arbitrary
number ng of generations. It is not hard to check that with the results so deter-
mined all other second order conditions of gauge invariance are satisfied. To finish
this discussion we notice that pseudo-unitarity implies
b1‡ ˆ b2 ˆ 18 …b1 †ÿ1 ;
hence
1 1
b1 ˆ p V ; b2 ˆ p V ‡ ; …3:37†
2 2 2 2
where V is an arbitrary unitary matrix. This is the CKM mixing matrix [12, 13] for the
quark coupling. A similar mixing is possible for the charged leptonic currents. The re-
cently observed signals of neutrino oscillations show that this mixing probably occurs.

4 Gauge invariance at third order: axial anomalies

Adler, Bell and Jackiw [14] discovered that there exists a possibility to violate
gauge invariance at third order in the triangular graphs. This holds also true in
A. Aste et al., Perturbative gauge invariance: electroweak theory II 399

the causal approach to gauge theory ([15] sect. 5.3). The anomalous graphs
contain one axial-vector and two vector couplings (VVA) or three axial-vector
couplings (AAA) of the fermions and three external gauge fields. To have a
more compact notation, we collect all fermionic matter fields into a big vector
w ˆ …e; m; t; ne ; nm ; nt † or ˆ …d; s; b; u; c; t†; respectively. The gauge fields A;
W ‡ ; W ÿ ; Z are denoted by Ama with a ˆ 0; ‡; ÿ; 3. Then the coupling between
fermions and gauge fields in (2.1) can be written as

T1FA ˆ ig wg  m Ma wAma ‡ wg
 m g5 Ma0 wAma ; …4:1†

where Ma stands for the following matrices of matrices:

   
0 b1 0 0
M‡ ˆ ; Mÿ ˆ ; …4:2†
0 0 b2 0
 3   
b 0 b5 0
M3 ˆ ; M0 ˆ ; …4:3†
0 b4 0 b6
and similarly for the axial-vector couplings, denoted by a prime.
Each triangular graph gives rise to two diagrams which differ by a permutation
of two vertices. Therefore, we have to compute the traces
tr …Ma Mb Mc † ‡ tr …Ma Mc Mb † ˆ tr …Ma fMb ; Mc g† ;
where one or three M's must be axial-vector couplings with a prime. It is well-
known that the cancellation of the axial anomalies relies on the compensation of
these traces in the sum of the leptonic and hadronic contributions. In this way
third order gauge invariance gives a further restriction of the quark coupling. To
work this out in detail, we consider the following cases.
A. Case …a ˆ 0; b ˆ ‡; c ˆ ÿ†VVAË:
0 0
tr …M0 fM‡ ; Mÿ g† ‡ tr …M0 fM‡ ; Mÿ g†
ˆ tr ‰b5 …b1 b02 ‡ b01 b2 † ‡ b6 …b2 b01 ‡ b02 b1 †Š :

Using the results of the last section, this is equal to

e2
ˆ ng …2a ÿ sin Q† ; …4:4†
4
where ng is the number of generations, i.e. the dimension of the matrices bk, and a
is the charge of the fermions in (3.16). For leptons we have b6 ˆ 0, because the
neutrini have no electric charge, so that
aL ˆ sin Q : …4:5†

Consequently, to compensate the triangular anomaly proportional to (4.4), one

needs the compensation between leptons and quarks. We assume equal number of
families in the lepton and quark sectors. Then for three colors of quarks one must
have
2aL ÿ sin Q ‡ 3…2aQ ÿ sin Q† ˆ 0 ; …4:6†
400 Ann. Phys. (Leipzig) 8 (1999) 5

which implies
aQ ˆ 13 sin Q …4:7†

by (4.5).
In most textbooks the electric charge aQ (4.7) of the d-, s-, b-quarks is put in
and then, by requiring cancellation of the anomalies, one concludes that the num-
ber of families in the lepton and quark sectors must be equal. We have simply
reversed the argument.
B. Case …0; 3; 3†VVA :
In this case the trace is simply given by

tr …M0 fM30 ; M3 g† ˆ tr ‰b5 …b03 b3 ‡ b3 b03 † ‡ b6 …b04 b4 ‡ b4 b04 †Š

 
d3 1 2
ˆ ng e2 ÿ sin Q …2a ÿ sin Q† : …4:8†
cos Q 4
Due to the same factor …2a ÿ sin Q† as in (4.4) the mechanism of compensation is
the same.
C. Case …3; ‡; ÿ†VVA :
Here the trace is equal to
tr …M30 fM‡ ; Mÿ g ‡ M3 fM‡
0 0
; Mÿ g ‡ M3 fM‡ ; Mÿ g
ˆ tr ‰b03 b1 b2 ‡ b04 b2 b1 ‡ b3 …b1 b02 ‡ b01 b2 † ‡ b4 …b2 b01 ‡ b02 b1 †Š
e2 sin Q
ˆ ng …sin Q ÿ 2a† ; …4:9†
4 cos Q
with the same consequences as before.
D. Case …3; ‡; ÿ†AAA :
Now we have to compute the trace

tr …M30 fM‡
0 0
; Mÿ g ˆ tr ‰b03 b01 b02 ‡ b04 b02 b01 Š ˆ 0 :

E. Case …0; 0; 3†VVA :

Here the relevant trace is equal to
d3
tr …M02 M30 † ˆ tr ‰…b5 †2 b03 ‡ …b6 †2 b04 ˆ ng e2 sin Q…2a ÿ sin Q† :
2
Due to the same factor as in case A, the compensation between leptons and
quarks is the same. This is also true in the next case:
F. Case …3; 3; 3†VVA :

tr …M32 M30 † ˆ tr ‰…b3 †2 b03 ‡ …b4 †2 b04 Š

 
d3 3 sin Q
ˆ ng e2 sin Q ÿ …2a ÿ sin Q† :
cos Q 2
A. Aste et al., Perturbative gauge invariance: electroweak theory II 401

G. Case …3; 3; 3†AAA :

This final case is trivial:

tr …M30 †3 ˆ tr ‰…b03 †3 ‡ …b04 †3 Š ˆ 0 :

Summing up the axial anomalies completely cancel in each generation, if and only
if aQ has the value (4.7).

5 Gauge invariance at third order: tree graphs

Third order tree graphs are not covered by the general inductive proof of gauge
invariance in massless Yang-Mills theories [16]. They are part of the beginning of
the induction and need an explicit verification of gauge invariance. The latter can
easily be done in the massless Yang-Mills theory, by using the fact that all cou-
plings are of Yang-Mills type, that means proportional to fabc. But this is no longer
true for the scalar couplings in massive Yang-Mills theories [1]. Therefore, it is not
surprising that gauge invariance of third order tree diagrams fixes the last free
parameters in the scalar coupling of the electroweak theory, namely the Higgs self-
coupling (Higgs potential).
Let us first discuss a simple special case of the standard model, the U…1† Higgs
model, which contains all essential features of the scalar self-coupling. We consider
one massive gauge field W m only, in interaction with one unphysical (F) and one
physical scalar field F0 . Therefore, we let W1 ˆ W; u1 ˆ u; u~1 ˆ u~; F1 ˆ F;
mW ˆ m in (3.20) of [1] and omit the other fields. From (3.20.9, 10, 12) we then
get

g
T1 ˆ i W n …F0 @n F ÿ F @n F0 † ÿ mWn W n F0
2

m2
‡ H F0 F2 ‡ m~ uuF0 ‡ 2bF30 ; …5:1†
2m
and from (4.5.10) of [1] or (3.2.9)
m g def
T1=1 jan ˆ ÿi ufF @ m F0 ÿ F0 @ m Fg ˆ D1 ‡ D2 : …5:2†
2
The calculation and compensation of the anomalies at second order can now be
directly taken over from the appendix in [1]. Only the following two sectors 1) and
7) appear:
1) Sector …F0 ; F0 ; 1; 1†
Here we have found the two normalization terms (A.1, 2)
i 2
N1 ˆ g Wn W n F20 d…x ÿ y† ; …5:3†
4
 2 
2 mH 3b
N2 ˆ ig ÿ F20 F2 d…x ÿ y† : …5:4†
4m2 2m
402 Ann. Phys. (Leipzig) 8 (1999) 5

7) Sector …1; 1; 1; 1†
i 2
N10 ˆ g Wn W n F2 d…x ÿ y† ; …5:5†
4
m2
N11 ˆ ÿig2 H 2 F4 d…x ÿ y† : …5:6†
16m
These results (5.3Ð6) are valid for the U…1† Higgs model, too, because all addi-
tional couplings in the standard model do not contribute to these two sectors.
Furthermore, the final normalization term (A.20)
N20 ˆ ig2 l0 F40 d…x ÿ y† …5:7†
0
becomes now important. Until now the coupling parameters b (5.1) and l (5.7)
are arbitrary.
Gauge invariance (1.0) at third order can only be violated by local terms
 Dd…x1 ÿ x3 ; ; x2 ÿ x3 † (where D denotes a differential operator). This can easily
be seen by inserting the causal factorization of the time-ordered products and
using gauge invariance at lower orders (cf. (4.3) in [1]). We only consider the tree
diagrams
P
dQ T3 jtree ˆ i @nk T3=k
n
jtree …5:8†
k ˆ 1; 2; 3

and adopt the notations and terminology from the corresponding calculation at
second order (sect. 4 of [1]).4 There are no local terms in T3 jtree and T3=k jtree ,
because a term  d…x1 ÿ x3 ; x2 ÿ x3 † : B1 B2 B3 B4 B5 : would violate renormalizabil-
ity by power counting (terms with derivatives on d…x1 ÿ x3 ; x2 ÿ x3 † are even
worse); especially there is no freedom of normalization (cf. (4.2) in [1]). Hence
the only local terms P in (5.8) are the anomalies which are the terms
 d…x1 ÿ x3 ; x2 ÿ x3 † in @nk T3=k
n
jtree …x1 ; x2 ; x3 †. (Derivatives of the d-distribu-
kˆ1; 2; 3
tion do not appear as can be seen by power counting.) Up to permutation of the
vertices there is only one possibility to generate an anomaly: due to causal factor-
ization, e.g. T3=1 m m
…x1 ; x2 ; x3 † ˆ T1=1 …x1 † T2 …x2 ; x3 † for x1 62 …fx2 ; x3 g ‡ Vÿ †, the nor-
malization term N…2† …x2 ; x3 †  d…x2 ÿ x3 † of T2 jtree …x2 ; x3 † is contained in
m
T3=1 …x1 ; x2 ; x3 †. Hence, if @ m F0 in D1 , or @ m F in D2 (5.2), is contracted with a field
operator in N…2† …x2 ; x3 † we obtain a term

T3=1 jmtree …x1 ; x2 ; x3 †

ˆ  @ m DF …x1 ÿ x2 † d…x2 ÿ x3 † : B1 …x1 † B2 …x1 † B3 …x2 † B4 …x3 † B5 …x3 † : ‡::: :

Taking now the divergence with respect to the Q-vertex x1 , an anomaly appears
due to @m @ m DF …x1 ÿ x2 † ˆ ÿm2 DF …x1 ÿ x2 † ‡ d…x1 ÿ x2 †. Considering the external
field operators uFF30 there come anomalies from the combinations of D1 with N20
and of D2 with N2 . Gauge invariance requires the cancellation of these anomalies

…D1 ; N20 †loc ‡ …D2 ; N2 †loc ˆ 0 ;

4
Note that e.g. a term  d…x1 ÿ x2 † @m @n DF …x2 ÿ x3 † (m and n not contracted) is non-local.
A. Aste et al., Perturbative gauge invariance: electroweak theory II 403

which gives the condition

 2 
g 2 0 g 2 mH 3b
ÿi 4ig l ‡ i 2ig ÿ ˆ 0: …5:9†
2 2 4m2 2m
The factors 4 and 2 are due to the number of possible contractions between the
two members D and N. Because N20 and N2 depend on the unknown parameters
l0 and b, this is a first equation to determine these parameters.
The anomalies with field operators uFF0 Wn W n come from the terms
…D1 ; N1 †loc ‡ …D2 ; N10 †loc ˆ 0 :
This gives the condition
g g2 g i 2
ÿi 2i ‡ i 2 g ˆ 0;
2 4 2 4
which is automatically satisfied. Finally, in the sector uF0 F3 gauge invariance re-
quires
…D1 ; N2 †loc ‡ …D2 ; N11 †loc ˆ 0
and this leads to the constraint
 2 
g 2 mH 3b g 2 m2H
ÿi 2ig ÿ ‡ i 4…ÿig † ˆ 0: …5:10†
2 4m2 2m 2 16m2
From (5.9) and (5.10) we find the values of the two coupling parameters
m2H m2H
bˆ ; l0 ˆ ÿ : …5:11†
4m 16m2
The same argument applies to the whole electroweak theory: In the sector
u1 F1 F30 we get the equation (5.9) (with m ˆ mW ), and in the sector u1 F0 F31 we
find the condition (5.10) again. It is easy to see that there are no additional terms
contributing in the bigger theory. The resulting values
m2H m2H
bˆ ; l0 ˆ ÿ …5:12†
4mW 16m2W
agree with those obtained from the Higgs potential in the standard theory as
shown below. We have verified that, with these parameters, all other anomalies
from third order tree graphs cancel out, without giving further information.
For comparison with the standard model we collect all self-couplings of scalar
fields. The first order terms are contained in (3.20.10Ð12) of [1]:

m2H
V1 …F† ˆ ig F0 …F20 ‡ F21 ‡ F22 ‡ F23 † : …5:13†
4mW
The second order terms are given by the normalization terms in the appendix of [1]
m2H
V2 …F† ˆ ig2 ‰ÿ…F20 ‡ F21 ‡ F22 ‡ F23 †2 Š : …5:14†
16m2W
404 Ann. Phys. (Leipzig) 8 (1999) 5

Remembering that the second order must be multiplied by 1/2, we obtain the fol-
lowing total scalar potential
1 m2H
V…F† ˆV1 ‡ V2 ˆ ÿig2
2 32m2W
 
2 2 2 2 2 mW 2 2 2 2
 …F0 ‡ F1 ‡ F2 ‡ F3 † ÿ 8 F0 …F0 ‡ F1 ‡ F2 ‡ F3 † : …5:15†
g
To compare this with the Goldstone-Higgs potential of the standard theory we set
F1 ˆ F2 ˆ F3 ˆ 0 and add a mass term 12 m2H F20 . Omitting the factor ÿi, the po-
tential is then equal to
V…F0 † ˆ l…F40 ÿ 4aF30 ‡ 4a2 F20 † ; …5:16†
where
 2
1 gmH 2
lˆ ; aˆ mW : …5:17†
2 4mW g
This is the shifted double-well potential discussed in the introduction (1.1). Our
point is that this structure is not obtained from a clever choice of a Lagrangean
and subsequent symmetry breaking, but comes out as the necessary consequence
of gauge invariance in the massive situation.

References
[1] M. DuÈtsch, G. Scharf, Ann. Phys. (Leipzig) 8 (1999) 359
[2] C. Becchi, A. Rouet, R. Stora, Comm. Math. Phys. 42 (1975) 127, Ann. Phys. (N.Y.) 98 (1976)
287
[3] T. Kugo, I. Ojima, Progr. Theoret. Phys. Suppl. 66 (1979) 1
[4] M. DuÈtsch, T. Hurth, F. Krahe, G. Scharf, Nuov. Cim. A 106 (1993) 1029
[5] A. Connes, J. Lott, Nucl. Phys. Proc. Suppl. B 18 (1990) 29
[6] M. Dubois-Violette, R. Kerner, J. Madore, J. Math. Phys. 31 (1990) 316
[7] R. Coquereaux, G. Esposito-Farese, G. Vaillant, Nucl. Phys. B 353 (1991) 689
[8] B. Balakrishna, F. GuÈrsey, K. C. Wali, Phys. Lett B 254 (1991) 430
[9] A. H. Chamseddine, J. FroÈhlich, Phys. Rev. D 50 (1994) 2893
[10] F. Scheck, The Standard Model within Non-commutative Geometry: A Comparison of Models,
hep-th/97010073, and references therein
[11] G. Roepstorff, Superconnections and the Higgs Field, hep-th/9801040
[12] T. P. Cheng, L. F. Li, Gauge theory of elementary particle physics, Oxford University Press
1984
O. Nachtmann, Elementary particle physics: concepts and phenomena, Springer-Verlag 1990
[13] N. Cabibbo, Phys. Rev. Lett. 10 (1963) 531
M. Kobayashi, K. Maskawa, Prog. Theor. Phys. 49 (1973) 652
[14] S. Adler, Phys. Rev. 177 (1969) 2426, J. S. Bell, R. Jackiw, Nuov. Cim. A 60 (1969) 47
[15] G. Scharf, Finite quantum electrodynamics: the causal approach, second edition, Springer 1995
[16] M. DuÈtsch, Nuov. Cim. A 109 (1996) 1145