Escolar Documentos
Profissional Documentos
Cultura Documentos
Contents
1 Why Renormalize? 1
4 Calculation of δ3 and δm 9
5 Calculation of δ2 and δM 10
6 Calculation of δg 12
1 Why Renormalize?
The Lagrangian of pseudoscalar Yukawa theory is
1 2 1
L= (∂µ φ0 ) − m20 φ20 + ψ 0 i∂
/ − M0 ψ0 − ig0 ψ 0 γ 5 ψ0 φ0 . (1)
2 2
The Feynman rules for this bare Lagrangian are:
i i
= ; = ;
p /p − M0 + i p p2 − m20 + i
= −ig0 γ 5 .
1
Homer Reid: 6/8/2004 QFT Lecture Notes 2
Armed with this Lagrangian, the Feynman rules, and the LSZ reduction
formula, we can compute S-matrix elements for processes like, for example,
fermion-antifermion annihilation into a pseudoscalar:
√
hp| S |k1 , k2 i = Z2 · Z3 ·
C
k1 k2
√
= Z2 · Z 3 · C(m0 , M0 , g0 ; k1 , k2 , p) (2)
where
• Z2 and Z3 are the field-strength renormalizations for the fermion and
pseudoscalar field, and
• C is a correlation function that depends on the external momenta and on
m0 , M0 , and g0 .
The problem with formula (2) is that to get an actual number out of it we
need to know Z2 , Z3 , m0 , M0 , and g0 . We don’t know any of these things. All
we know from experiment are the physical masses and coupling constants.
There are two ways to deal with this last problem:
1. Carry out explicit diagrammatic calculations to obtain expressions for Z 2 ,
Z3 , and the physical quantities m, M , and g in terms of m0 , M0 , and g0 ;
then invert these expressions to obtain expressions for m0 , M0 and g0 in
terms of m, M , and g; then plug the resulting expressions into (2).
2. Rewrite the Lagrangian (1) as a sum of terms containing only the phys-
ical fields, masses, and coupling constants, plus additional counterterms
depending on the field-strength renormalizations and the bare masses and
coupling constants. Then do normal perturbation theory with the coun-
terterms treated as additional interaction terms. This is basically a repa-
rameterization of the unknown quantities Z2 , Z3 , m0 , M0 , and g0 into a
new set of unknown quantities, whose values we determine by demanding
that certain renormalization conditions be satisfied.
Of course, in addition to the fact that we don’t a priori know the field-
strength renormalizations or bare masses and coupling constants, there is also
the little problem that diagrams containing loops come out infinite. In our
case it turns out that this second problem is automatically remedied by the
solution we take to the first problem. But, following Weinberg, we point out
here that even if all diagrams were perfectly convergent we would still have to
renormalize the theory because, in an interacting theory, the bare masses and
coupling strengths are always different from the physical masses and coupling
strengths and from experiment we know only the latter.
Homer Reid: 6/8/2004 QFT Lecture Notes 3
iZ3
D(p) = = +··· (3)
p p2 − m 2
In other words, the physical mass is the location of the pole in p2 in the exact
propagator, and Z3 is the residue at that pole. To calculate m and Z3 in
terms of m0 , M0 , and g0 we will thus need to calculate the exact pseudoscalar
propagator.
k
= +
p p p p
k−p
k1 k2
p p
+
p
k1 − p k2 − p
k
+
p p
k−p p−k
+···
The first thing to notice is that every diagram after the first contains two
factors of the bare pseudoscalar propagator D(p), so we can factor those out
Homer Reid: 6/8/2004 QFT Lecture Notes 4
and write
=
p p
k
k1 k2
2
+ · +
p
p
k−p k1 − p k2 − p
k
)
+ + ···
k−p p−k
The next thing to notice is that the thing in brackets is just the sum of all
1PI diagrams, plus two sums-of-all-1PI-diagrams with a factor of D(p) inserted
between them, plus three sums-of-all-1PI-diagrams with two factors of D(p)
inserted, et cetera, so we may write
2 ( )
= + · 1PI + 1PI 1PI +···
p p
or, in symbols,
( )
h i h i2
D(p) = D0 (p) + D02 (p) 2 2
− iM (p ) + D0 (p) − iM (p ) 2 2
+···
( )
h i h i2
2 2
= D0 (p) 1 + D0 (p) − iM (p ) + D02 (p) − iM (p )2 2
+···
D0 (p)
=
1 − D0 (p) [ − iM 2 (p2 )]
1
=
D0 (p)−1 − [−iM 2 (p2 )]
i
= 2 (4)
p − m20 − M 2 (p2 )
Homer Reid: 6/8/2004 QFT Lecture Notes 5
where
= + +···
−iM22 (p) =
k−p
d4 k Tr γ 5 i /k + M0 γ 5 i / k−/
p + M0
Z
2
= (−1)(−ig0 )
(2π)4 [k 2 − M02 + i] [(k − p)2 − M02 + i)]
(The factor of -1 comes from the fermion loop.) We can get rid of the γ 5 s in the
numerator by passing the first γ 5 through the first factor in parentheses, which
Homer Reid: 6/8/2004 QFT Lecture Notes 6
Tr / k − M0 / k−/
d4 k p + M0
Z
−iM22 (p) = +g02
(2π)4 [k 2 − M02 + i] [(k − p)2 − M02 + i)]
d4 k k · (k − p) − M02
Z
= 4g02
(2π) [k − M0 + i] [(k − p)2 − M02 + i)]
4 2 2
1
d4 k k · (k − p) − M02
Z Z
= 4g02 dx
0 (2π)4 [k 2 − 2kxp + xp2 − M02 + i]2
where
∆ = x(x − 1)p2 + M02 .
To do the integrals in d dimensions, we use the formulae
i Γ 2 − d2
d4 k 1
Z
=
(2π)4 (l2 + ∆)2 (4π)2 ∆(2−d/2)
−i d2 Γ 1 − d2
d4 k l2
Z
= .
(2π)4 (l2 + ∆)2 (4π)2 ∆(1−d/2)
Plugging in to (7),
( )
1 d d i Γ 2 − d
4g02 2Γ 1−
Z h
−iM22 (p) = −i dx 2 2
− x(x − 1)p − M02 2
.
(4π)2 0 ∆(1−d/2) ∆(2−d/2)
To simplify this a little, let’s notice that the factor in square brackets inside the
curly brackets is just ∆ − 2M02 :
( )
2 Z 1 d d i Γ 2 − d
h
4g 0 Γ 1 −
−iM22 (p) = −i dx 2 (1−d/2)2 − ∆ − 2M02 2
(8)
(4π)2 0 ∆ ∆(2−d/2)
Z 1 (d d
)
Γ 2 − d2 d
g02 2Γ 1 − 2 2Γ 2 − 2
= −i 2 dx − (1−d/2) + 2M0 (2−d/2) (9)
4π 0 ∆(1−d/2) ∆ ∆
Z 1 ( )
(d − 1)Γ 1 − d2 d
g02 2Γ 2− 2
= −i 2 dx + 2M0 (2−d/2) . (10)
4π 0 ∆(1−d/2) ∆
Homer Reid: 6/8/2004 QFT Lecture Notes 7
It is interesting to note that the first term of (10) is just what P&S get for their
calculation of M22 (p) in the original (scalar, not pseudoscalar) Yukawa theory
(P&S equation 10.33), except the sign is reversed. I am wondering if this is
related to some sort of deep fact about chiral symmetry, or if I just made a
mistake somewhere. (Note that P&S page 329 contains some major typos, in
my edition anyway, which you should correct before comparing).
In any event, the point is that with expression (10) and equation (5) we
could find an expression for m in terms of m0 , M0 , and g0 . If we found two
more such equations relating M and g to the bare constants then we could
invert the system of equations to obtain m0 , M0 , and g0 in terms of the physical
parameters and then use these values to get numbers out of equations like (2).
Obviously, this would be an unbelievable hassle, particularly if we were to try
to extend the procedure to higher orders. This motivates the development of
an alternative procedure that is much simpler.
L = L 1 + L2
where
something that looks exactly like L but with the physical fields,
L1 =
masses, and coupling constant in place of the bare quantities;
Homer Reid: 6/8/2004 QFT Lecture Notes 8
and
To be explicit,
1 2 1
L1 = (∂µ φ) − m2 φ2 + ψ i∂
/ − M ψ − igψγ 5 ψφ
2 2
and
1 2 1
L2 = δ3 ∂µ φ − δm φ2 + ψ iδ2 ∂
/ − δM ψ − iδg ψγ 5 ψφ
2 2
where
δ3 = Z 3 − 1 (13)
δm = Z3 m20 −m 2
(14)
δ2 = Z 2 − 1 (15)
δM = Z 2 M0 − M (16)
and
1/2
δg = Z2 Z3 g0 − g. (17)
i i
= =
p /p − M + i p p2 − m2 + i
× = i(/
pδ2 − δM ) × = −iδg γ 5
The first three rules are the same as the rules we wrote down above for
bare perturbation theory, but with the physical masses and coupling constant
in place of the bare masses. The remaining rules are the counterterms.
We emphasize here that all we have done is to reparameterize the unknown
quantities Z2 , Z3 , m0 , M0 and g0 into the unknown quantities δ2 , δ3 , δm , δM , and
Homer Reid: 6/8/2004 QFT Lecture Notes 9
δg . We don’t know the values of the δ quantities any better than we knew the
bare constants in the Lagrangian. To determine their values, we have to specify
renormalization conditions and adjust the values of the δ constants until these
conditions are met. In the following sections we will demonstrate this explicitly.
4 Calculation of δ3 and δm
The renormalization condition we will use to determine δ3 and δm is that the
exact pseudoscalar propagator, calculated with the renormalized Feynman rules,
have a pole at the physical mass m with residue 1. In symbols, the condition is
i
= +···
p p2 − m 2
Referring back to equations (5) and (6), we can write this condition in terms
of the pseudoscalar self-energy:
M 2 (p2 = m2 ) = 0 (18)
and
dM 2
= 0. (19)
dp2 p2 =m2
= + ×
k−p
Z 1 ( d
d
)
g2 (d − 1)Γ 1 − 2 Γ 2−
= −i 2 dx 2
+ 2M 2
+ i(p2 δ3 − δm )
4π 0 [x(x − 1)p2 + M 2 ](1−d/2) [x(x − 1)p2 + M 2 ](2−d/2)
where for the fermion loop we just plugged in the result (10) we obtained before
in bare perturbation theory, substituting M0 → M and g0 → g. Putting p2 =
m2 , condition (18) requires
Homer Reid: 6/8/2004 QFT Lecture Notes 10
( )
1 d d
g2 (d − 1)Γ 1 − Γ 2−
Z
2 2 2 2
m δ3 − δ m = 2 dx + 2M .
4π 0 [x(x − 1)m2 + M ] 2 (1−d/2) [x(x − 1)m2 + M 2 ](2−d/2)
( )
1 d d
g2 −(d − 1)x(x − 1)Γ 2 − −x(x − 1)Γ 3 −
Z
δ3 = 2
+ 2M 2 2
.
4π 2 0 [x(x − 1)m2 + M 2]
2−d/2
[x(x − 1)m2 + M 2]
3−d/2
5 Calculation of δ2 and δM
The renormalization condition fixing the values of δ2 and δM is that the exact
fermion propagator D(p) have a pole at the physical mass p = M with residue
1, i.e.
i
D(p) = = +···
p /p − M
Exactly as for the pseudoscalar propagator we can write a Dyson’s equation for
the fermion propagator:
i
D(p) =
/p − M − Σ(/p)
where
−iΣ /
p = 1PI
p−k
= + +···
k
is the sum of all irreducible insertions into the fermion propagator. Then the
renormalization conditions are
Σ / p=M =0 (20)
dΣ
= 0. (21)
d/
p
/p=M
Homer Reid: 6/8/2004 QFT Lecture Notes 11
1
/k − M
d4 k
Z Z
2
= −g dx
0 (2π)4 [k 2 − 2kxp + xp2 − (1 − x)M 2 − xm2 ]2
1
d4 k /l + x/
p−M
Z Z
= −g 2 dx
0 (2π)4 [l2 − ∆]2
d 1
g2Γ 2 −
Z
2 x/
p+M
= −i dx .
(4π)2 0 [x(x − 1)p2 + (1 − x)M 2 + xm2 ]
2−d/2
g 2 Γ 2 − d2
Z 1
x/
p+M
−iΣ(/p) = −i dx + i(/
pδ2 −δM ).
(4π)2 0 [x(x − 1)p + (1 − x)M 2 + xm2 ]
2 (2−d/2)
d 1
M g2Γ 2 −
Z
2 x+1
M δ2 − δM = dx .
(4π)2 0 [(x − 1)2 M 2 + xm2 ]
(2−d/2)
( )
d
g2 Γ 2 − 1
2(2 − d2 )x(x2 − 1)M 2
Z
2 x
δ2 = dx −
(4π)2 0 [(x − 1)2 M 2 + xm2 ](2−d/2) [(x − 1)2 M 2 + xm2 ](3−d/2)
Homer Reid: 6/8/2004 QFT Lecture Notes 12
6 Calculation of δg
The renormalization condition fixing the value of δg is that the exact fermion-
fermion-pseudoscalar vertex be coupled with the physical coupling strength g:
= −igγ 5 . (23)
×
= + + + ···
The renormalization condition (23) is just that the LHS of this equation equal
the first diagram on the RHS, or that the second and third diagrams on the
RHS cancel each other:
×
= − (24)
.
So let’s compute the first vertex correction diagram (omitting factors from the
external fermion and pseudoscalar lines):
p1 p1 − k p2
−igδΓ(q) = k k+q
where
l = k + xq − yk
∆ = x(x + y − 1)q 2 + (y − 1)2 M 2 + ym2 .
Getting to the last line above required some algebra and the relation p1 · q =
− 21 q 2 , which follows from q = p2 − p1 and the fact that both p1 and p2 are on-
shell. Since this diagram will be sandwiched between u(p2 ) and u(p1 ), we can
neglect the / q term in the numerator since u(p2 )(p/2 − p/1 )u(p1 ) = 0. To proceed
we use the integrals
d4 k 1 −i Γ(3 − d2 )
Z
4 2 3
=
(2π) [l − ∆] 2(4π)2 ∆3−d/2
d4 k l2 i d Γ(2 − d2 )
Z
= .
(2π)4 [l2 − ∆]3 2(4π)2 2 ∆2−d/2
Then
(
1 1 d d
−ig 3 γ 5 2 Γ(2 − 2 )
Z Z
−igδΓ(q) = dx dy
(4π)2 0 0 ∆2−d/2
)
Γ(3 − d2 )
1 2 2 2 2
− x(x + y − 1) − y q + (y − 1)M .
2 ∆3−d/2
(
1 1 d d
g3 2 Γ(2 − 2 )
Z Z
δg = − dx dy
(4π)2 0 0 [(y − 1)2 M 2 + ym2 ]2−d/2
)
2 2 Γ(3 − d2 )
−(y − 1)M .
[(y − 1) M 2 + ym2 ]3−d/2
2
Homer Reid: 6/8/2004 QFT Lecture Notes 14
0 1 3 0 (vanishes by parity)
0 2 2 3
0 3 2 0 (vanishes by parity)
0 4 0 1
1 2 0 1
Note that the pseudoscalar one-point and three-point functions must vanish,
since otherwise there would be a vacuum expectation value for the quantity
ψγ 5 ψ, which changes sign under parity transformations. I think I have this
right?
Adding up the numbers in the last column, we see that we seem to have
6 divergent constants in the theory, whereas so far we have only defined 5
renormalization parameters (δ2 , δ3 , δm , δM , δg ). So this suggests we need another
parameter somewhere. We get it by adding a term λ4!0 φ4 to the Lagrangian,
which will then give us a sixth renormalization counterterm δλ = λ0 Z32 − λ.
15