PHY–396 T. Problem set #3. Due September 19, 2013.

1. Consider supersymmetric QCD with Nc colors and Nf < Nc flavors. In matrix notations,
the quark chiral superfields Aif (y, θ) form an Nc × Nf matrix A while the antiquark chiral
superfields Bfi (y, θ) form an Nf × Nc matrix B. Let all the flavors be exactly massless, so
the Lagrangian is

f
Z Z  
2
L = α
d θ tr(W Wα ) + H. c. + d4 θ tr A e+2V A + B e−2V B . (1)
2

(a) Show that the classical scalar potential of the theory has form

N 2 −1
g2 X h  a
i2
Vscalar = tr λ AA† − B † B (2)
8
a=1

where g 2 = 1/Ref .

(b) Show that this potential vanishes if and only if

AA† − B † B = c × 1Nc ×Nc (3)

for some real number c. Also show that for Nf < Nc this matrix relation implies
c = 0 and hence

AA† = B † B. (4)

⋆ Lemma: Any complex square matrix X can be written as a product of two unitary
matrices U1 , U2 and a real diagonal matrix D with non-negative eigenvalues.

X = U1 DU2 . (5)

Proving this lemma is an optional exercise for the students.

1
 (c) Generalize this lemma to the rectangular matrices such as A and B. Then show that
the matrices A and B which obey eq. (4) may be written as

 
DNf ×Nf  
A = UC ×   × VA , B = VB × DNf ×Nf 0NF ×(Nc −Nf ) × UC−1
 

0(Nc −Nf )×Nf

(6)
where UC is an SU(Nc ) matrix (same gauge symmetry for A and B), VA and VB
are NF × NF unitary matrices, and D is a real ≥ 0 diagonal Nf × NF matrix, same
D = diag(d1 , . . . dNf ) for both A and B.

(d) The independent holomorphic moduli of the flat directions form an Nf × Nf matrix
M = BA. (We assume Nf < Nc .) Use eqs. (5) and (6) to show that any given
complex Nf × Nf matrix M can be written as a product BA of rectangular matrices
obeying eqs. (4). Also, argue that any two such decompositions M = BA = B ′ A′ of
the same matrix M are gauge-equivalent to each other, A′ = UA, B ′ = AU −1 for the
same U ∈ SU(Nc ).

In other words, there is a one-to-one correspondence between the classical moduli

space of SQCD and the space of complex Nf × Nf matrices M.

2. Next, an exercise in superfield Feynman rules. Evaluate the 1-loop diagram

Φ2 Φ3

(7)

Φ1 Φ4

2
(a) Dress the graph with the fermionic derivatives − 14 D 2 and − 41 D for the propagators,
then eliminate some of these derivatives according to the vertices. Then show that
after integrating d4 θ for 3 out of 4 vertices of the graph, the Feynman amplitude
R

2
 takes form

d4 q4 y 2 y ∗2
Z
iM = ×
(2π)4 (q12 + i0)(q22 + i0)(q32 + i0)(q42 + i0)
1
Z
2 2
× d4 θ4 Φ1 (θ4 )Φ2 (θ4 )D 2 D Φ3 (θ4 )Φ4 (θ4 )D 2 D δ (4) (θ4 − θ1 ) .
256 θ1 =θ4
(8)
Note: The derivatives in this formula are WRT θ4 , and each derivative acts on every-
thing to its right.

(b) Work out the action of derivatives on the δ (4) (θ4 −θ1 ) factor and show that the integral
on the second line of the amplitude (8) evaluates to

(q µ σ̄µ )α̇α
 
1 2
Z
4 2
D Φ3 D Φ4 + 4 × Dα Φ3 Dα̇ Φ4 + (q42 ) × Φ3 Φ4



d θ Φ1 Φ2 × (9)
16 2

where all the superfields depend on the same superspace coordinates θα and θ̄α̇ , and
q4µ is the momentum carried by the bottom propagator in the loop.

This exercise shows that the momentum-integral part of a superfield Feynman graph
can have both numerator and denominator factors. The denominators come from the
propagators, while the numerators follow from the D-derivative algebra in the superfield
part of the amplitude.

3. Now, let’s count the fermionic derivatives and the powers of loop momenta in a generic
superfield Feynman graph. For simplicity, let’s focus on a Wess–Zumino-like model of one
chiral superfield Φ with a generic polynomial superpotential of degree n > 3.
Z Z Z
4 2
L = d θ ΦΦ + d θ W (Φ) + d2 θ̄ W ∗ (Φ),
(10)
m 2 y cn n
W = Φ + Φ3 + · · · + Φ .
2 6 n!

(a) Write down the superfield Feynman rules (the vertices and the propagators) for this
theory.

Now consider a generic Feynman graph with L loops, V vertices (of all kinds), P propa-
gators of the ΦΦ type, and P ′ propagators of the ΦΦ or ΦΦ types.

3
 α̇
(b) Count the fermionic derivative operators D α and D for the graph before you do the
Grassmannian integrals and use up some derivatives to close the loops via
2
D 2 D δ (4) (θ1 − θ2 )|θ1 =θ2 = 16, etc.. Show that

α̇
net #(D α ) + #(D ) = 2L + 2P − 2. (11)

(c) Now let’s do the Grassmannian integrals d4 θ2 · · · d4 θV . This process soaks up some
R R

of the fermionic derivatives, while the remaining derivatives either end up acting on
the external-leg fields in the remaining d4 θ1 integral, or else they may produce powers
R

of the loop momenta in the numerator via the anticommutators {Dα , D α̇ } = 2σαµα̇ qµ .

Calculate the maximal power of loop momenta in the numerator that may result from
this process.

(d) Calculate the superficial degree of divergence ∆ of the momentum integral of the
Feynman graph.

(e) Finally, show that

n
X
′
∆ ≤ 2 − E − P + (k − 3)Vk (12)
k=3

where E is the number of the graph’s external legs and Vk is the number of vertices
with k legs. Then use eq. (12) to argue that the WZ model with a cubic superpoten-
tial is renormalizable while theories with higher-order n ≥ 4 superpotentials are not
renormalizable.

4. The last problem is about supersymmetric QED,

Z  Z Z
α 2

4 +2eV
L = d θ Ae A + Be−2eV
B + 8 V D D Dα V + d θ mAB d2 θ̄ m∗ AB. (13)
1 2

The superfield Feynman rules for SQED will be explained in class next week. For now,
please take them for granted:

4
 • Chiral propagators:

2
i D D 2 (4)
A A = 2 × δ (θ1 − θ2 ),
p − mm∗ + i0 16
2
i mD (4)
A B = × δ (θ1 − θ2 ),
p2 − mm∗ + i0 4
(14)
i m∗ D 2 (4)
B A = 2 × δ (θ1 − θ2 ),
p − mm∗ + i0 4
2
i D 2 D (4)
B B = 2 × δ (θ1 − θ2 ),
p − mm∗ + i0 16

• Vector propagator in the Feynman gauge:

i
V V = 2
× δ (4) (θ1 − θ2 ). (15)
k + i0

• Vertices: One incoming chiral line, one outgoing chiral line of the same species, any
number n = 1, 2, 3, . . . of vector lines,

A B
V V

V V
A = i(+2e)n , B = i(−2e)n , (16)

without any superderivative factors in the numerator or denominator.

Count the superderivatives and powers of momenta in a general Feynman diagram and
show that a diagram with EC external legs of chiral superfields (A, B, A, or B), EV ex-
ternal legs of vectors, and any numbers of loops, vertices, and internal lines has superficial
degree of divergence

∆ ≤ 2 − EC . (17)

In class, I shall use this formula to prove that SQED is renormalizable.