Preprint typeset in JHEP style - HYPER VERSION LPT Orsay 09-07

Detours and Paths:

BRST Complexes and Worldline Formalism
arXiv:0902.0530v2 [hep-th] 1 May 2009

Fiorenzo Bastianelli ♣ , Olindo Corradini ♣,♦ and Andrew Waldron ♠

♣ Dipartimento di Fisica, Università di Bologna and INFN, Sezione di Bologna
via Irnerio 46, I-40126 Bologna, Italy
♦ Laboratoire de Physique Théorique, Université de Paris-Sud XI
Bâtiment 210, F-91405 Orsay CEDEX, France
♠ Department of Mathematics, University of California
Davis CA 95616, USA

E-mail: bastianelli,[email protected], [email protected]

Abstract: We construct detour complexes from the BRST quantization of worldline

diffeomorphism invariant systems. This yields a method to efficiently extract physical
quantum field theories from particle models with first class constraint algebras. As
an example, we show how to obtain the Maxwell detour complex by gauging N = 2
supersymmetric quantum mechanics in curved space. Then we concentrate on first
class algebras belonging to a class of recently introduced orthosymplectic quantum
mechanical models and give generating functions for detour complexes describing
higher spins of arbitrary symmetry types. The first quantized approach facilitates
quantum calculations and we employ it to compute the number of physical degrees
of freedom associated to the second quantized, field theoretical actions.

Keywords: BRST quantization, Gauge symmetry, Sigma models.

Contents
1. Introduction 1

2. Detour Complexes 2

3. Path Integrals 4

4. Maxwell Detour 5
4.1 N = 2 Supersymmetric Quantum Mechanics 5
4.2 N = 2 Spinning Particle 7
4.3 BRST Quantization 9
4.4 Counting Degrees of Freedom 11

5. Mixed Higher Spin Detour 13

5.1 Symmetric Forms and sp(2) Quantum Mechanics 13
5.2 BRST Quantization and Lie Algebra Cohomology 15
5.3 Mixed Tensors and sp(2r) Quantum Mechanics 18
5.4 Mixed Symmetry Einstein Operators 21
5.5 The sp(2r) Spinning Particle 24
5.6 Counting Degrees of Freedom 26
5.7 Lightcone Degrees of Freedom 28

6. Conclusions 30

1. Introduction
Unquestionably, gauge theories are a central pillar of modern theoretical physics.
Although they are usually presented in terms of a local symmetry of a field theoretic
action principle, it is often useful to describe them in a first quantized language. The
purpose of this Article is to present new tools to analyze gauge field theories using
a first quantized picture, and to apply them to higher spin theories. These tools
employ an algebraic construction known as the detour complex. The basic elements
of the detour complex are differential operators which form the building blocks of
the gauge theory under study. These differential operators can be represented as
quantum mechanical operators. Crucially, by forming first class constraint algebras
from them, one may consider worldline gauge theories. This procedure gives rise

–1–
to a particle model which generically is diffeomorphism invariant on the worldline,
and whose physical spectrum is related to the first quantization of the gauge field
theory. Then, the BRST quantization of the particle naturally provides a cohomo-
logical complex, out of which one builds the detour complex. Equivalence of the
BRST cohomology with the detour cohomology guarantees that the correct physical
information is properly encoded in the construction. From the detour complex one
identifies an action principle for the gauge invariant field equations and may use the
associated particle model to extract information about the quantized version of the
theory.
Our first example—which provides much intuition—is the Maxwell detour. It
describes the abelian gauge theory of differential forms and is related to the quantiza-
tion of the N = 2 spinning particle. We then consider detour complexes constructed
out of the gauging of certain orthosymplectic quantum mechanical models [1], fo-
cusing mostly on symplectic subgroups. We use them to analyze the structure of
gauge theories for bosonic fields of higher spins with arbitrary symmetry type [2]
(see also [3, 4, 5, 6, 7]; we refer the reader also to the series of higher spin review
Articles [8]).
Thus, in Section 2 we present the structure and some elements of detour com-
plexes. In Section 3 we recall the use of path integrals to treat quantum field theories
in first quantization. Section 4 contains the example of the Maxwell detour dealing
with the abelian gauge symmetries of differential forms. In Section 5 we construct de-
tour complexes for higher spin fields of arbitrary symmetry type. Finally, we present
our conclusions in Section 6.

2. Detour Complexes
As already mentioned, gauge theories are usually presented in terms of a local sym-
metry of an action principle. Other key ingredients, however, are gauge parameters,
gauge fields, field equations and Bianchi identities. These can all be packaged in a
single mathematical object known as a detour complex. Schematically
  d     δ  Bianchi 
0 −→ gauge
−→ gauge → ··· ··· → field
−→ identities −→ 0
parameters fields x
equations
(2.1)

G 

where for simplicity in the last two entries we have labeled the field space with the
equations living on that space. Recall that a sequence of operators is called a complex
when consecutive products vanish, so here

Gd = 0 = δG . (2.2)

These relations subsume the usual ones of gauge theories. Namely, if A is a gauge
field, then GA = 0 is its field equation, while if α is a gauge parameter, A → A + dα

–2–
is the corresponding gauge transformation. The relation Gd = 0 ensures that the
field equations are gauge invariant.
We call G the detour or long operator since it connects dual complexes. Op-
timally, G is self-adjoint, so that one can use it to construct an action of the
R
form S = 21 (A, GA) where the inner product is the one following naturally from
the underlying first quantized quantum mechanical model. From the action principle
viewpoint, since an arbitrary variation produces the field equations GA = 0, spe-
cializing to a gauge transformation, the operator δ, dual to d, must annihilate the
equation of motion δGA = 0. This Noether-type identity generalizes the Bianchi
identity for the Einstein tensor of general relativity, and is precisely the second rela-
tion above.
The simplest example is the Maxwell detour. In that case gauge fields are one-
forms, or sections of Λ1 M, while the gauge parameters are zero-forms Λ0 M. The
field equations and Bianchi identities are also one and zero-forms, respectively. The
differentials are the exterior derivative and codifferential while the detour operator
is simply their product. In diagrammatic notation

d δ
0 −→ ∧0 M −→ ∧1 M → · · · · · · → ∧1x
M −→ ∧0 M −→ 0
 (2.3)
δd 

Maxwell’s equations are at once recognized as δdA = 0, while gauge invariance

and the Bianchi identity follow immediately from nilpotency of d and δ. The long
operator δd connects the de Rham complex and its dual. Notice we could also not
detour, and continue the de Rham complex, the next entry being two-forms, or in
physics language, curvatures.
There are many other examples of detour complexes, for example: In four dimen-
sions the long operator δd is conformally invariant. In six dimensions, there exists
a higher order, conformally invariant detour operator δ∆d + · · · [9, 10]. Another
interesting variant is to twist the Maxwell complex by coupling to the Yang–Mills
connection of a vector bundle over M. In this case, one obtains a complex exactly
when the connection obeys the Yang–Mills equations [11, 12]. In this Article we will
concentrate on de Rham detours and their generalization to “symmetric forms”. The
key idea is to use the relation between first quantized spinning particles and geom-
etry. We obtain detour complexes by gauging these models and employing BRST
quantization. In particular, the long operator connecting a complex with its dual,
corresponds to a shift in the worldline diffeomorphism ghost number. This represen-
tation of the BRST cohomology in terms of field equations for gauge potentials is
achieved by a careful choice of ghost polarizations. After reviewing the use of path
integrals in first quantized approaches, we begin by studying generalized Maxwell
complexes.

–3–
3. Path Integrals
Just as for strings, also in particle theory a first quantized approach is often use-
ful. While the field theory language is usually appropriate, the worldline approach
can often be applied to more efficiently calculate quantum corrections, see [13] for a
review. The simplest way to introduce the worldline formalism is perhaps to recall
the example of a scalar field, whose free propagator and one-loop effective action can
be represented in terms of a quantum mechanical hamiltonian supplemented by an
integration over the Fock-Schwinger proper time. Much of the heat kernel literature
can be classified under this point of view. An integration over the proper time signals
that one is dealing with the quantum theory of a reparametrization invariant particle
system; a relativistic spinless bosonic particle for the scalar field case. Similarly a
spinning particle with N = 1 supergravity on the worldline is related to the quantum
theory of a Dirac field [14]. More generally so(N) spinning particles are related to
fields of spin N/2 [15]. Once the connection between reparametrization invariant
particle models and quantum field theories is established, it is often advantageous to
quantize the mechanical model with path integrals, i.e., summing over spinning par-
ticle worldlines. For example, this worldline approach has been used for the so(N)
spinning particle in arbitrarily curved spaces with N = 0, 1, 2 to study the effective
action for scalars [16], spin 1/2 [17], and arbitrary differential forms (including vec-
tors) [18] coupled to gravity, respectively. The cases N > 2 do not admit a coupling
to a generic curved space, but in [19] the worldline path integral has been considered
in flat space (note that conformally flat backgrounds can be treated as well [20])
where the only information contained in the one-loop effective action is the number
of circulating physical excitations.
Schematically, to compute the one-loop effective action Γ, one path integrates
over closed worldlines with the topology of the circle S 1 . Gauge fixing worldline
diffeomorphisms produces an integral over the proper time β (the circumference of
the circle). In arbitrary dimensions D, the result is
Z Z Z
−Sparticle [X] 1 ∞ dβ dD x
Γ= DX e =− D γ(x, β) (3.1)
S1 2 0 β (2πβ) 2
where the density γ(x, β) can typically be computed in a small β expansion
γ(x, β) = a0 (x) + a1 (x)β + a2 (x)β 2 + · · · (3.2)
and an (x) are called heat kernel coefficients. This expansion applies to massless
particles and generically is not convergent in the upper β limit, even after renor-
1 2
malization. (Massive particles contain an extra factor e− 2 m β which improves the
infrared behavior.) For a free theory in flat space, the only nonvanishing coefficient
is the a0 , which is constant and counts the number of physical degrees of freedom
circulating in the loop. This simplest of quantum computations is the one we focus
on in this Article.

–4–
4. Maxwell Detour
In this Section we derive the Maxwell detour complex described in Section 2. Our
method relies on BRST quantization and a careful choice of ghost polarizations. We
start by reviewing the underlying supersymmetric quantum mechanical model.

4.1 N = 2 Supersymmetric Quantum Mechanics

The physical Hilbert space of N = 2 supersymmetric quantum mechanics is the
space of differential forms Γ(∧M) and, geometrically, its quantized Noether charges
are the exterior derivative d, codifferential δ, form Laplacian ∆ and the degree
operator N [21]. In this Section, we review those results in detail.
The model is described by the action
Z n o
1 µ ∇ψ µ 1
S = dt ẋ gµν ẋν + iψ̄µ + Rµνρσ ψ̄ µ ψ ν ψ̄ ρ ψ σ , (4.1)
2 dt 2

which is invariant under rigid N = 2 supersymmetry (ε, ε̄), U(1) fermion number
symmetry (α) and worldline translations (ξ)

δxµ = iε̄ψ µ + iεψ̄ µ + ξ ẋµ ,

µ µ ∇ψ µ
µ µ µ µ ∇ψ̄ µ
Dψ = −εẋ + iαψ + ξ , D ψ̄ = −ε̄ẋ − iαψ + ξ . (4.2)
dt dt
In these formulæ D is the covariant variation: DX µ ≡ δX µ + Γµ νρ X ν δxρ which
obviates varying covariantly constant quantities. Invariance under supersymmetry
follows easily upon noting the identity
h ∇i
D, X µ = δxρ ẋσ Rρσ µ ν X ν , (4.3)
dt
using which leaves only variations proportional to five fermions that vanish thanks
to the second Bianchi identity for the Riemann tensor.
To quantize the model we work in a first order formulation

ẋµ = π µ , (4.4)

which follows from the action principle

Z n o
1 1
S = dt pµ ẋµ + iψ̄m ψ̇ m − πµ g µν πν + Rmnrs ψ̄ m ψ n ψ̄ r ψ s .
(1)
(4.5)
2 2
Here we have used the vielbein eµ m to flatten the Lorentz indices on the fermions.
Also the covariant and canonical momenta πµ and pµ are related by

πµ = pµ − iωµmn ψ̄ m ψ n , (4.6)

–5–
where ωµmn is the spin connection. Since the symplectic current pµ dxµ + iψ̄m dψ m is
expressed in Darboux coordinates, we immediately read off the quantum commuta-
tion relations
[pµ , xν ] = −iδµν , {ψ̄m , ψ n } = δm
n
. (4.7)
Motivated by geometry, we represent this algebra in terms of operators
1 ∂ ∂
pµ = µ
, ψ̄µ = µ
, ψ µ = dxµ , (4.8)
i ∂x ∂(dx )
acting on wavefunctions
D
X
Ψ = Ψ(x, dx) = Fµ1 ...µk (x) dxµ1 ∧ . . . ∧ dxµk . (4.9)
k=0

The variables dxµ are Grassmann (so we will often denote their wedge products
simply by juxtaposition) which means the coefficients Fµ1 ...µk in this expansion are
totally antisymmetric, or in other words differential forms (or sections of ∧k M).
The Noether charges corresponding to supersymmetries Q, Q̄, fermion number N
and worldline translations H are now operators acting on wavefunctions. With a
suitable normalization they equal
Q = iψ µ πµ ,

Q̄ = iψ̄ µ πµ ,

N = ψ µ ψ̄µ ,
1 i 1
H = πm π m − ωm mn πn − Rµνρσ ψ̄ µ ψ ν ψ̄ ρ ψ σ , (4.10)
2 2 2
and satisfy a N = 2 superalgebra
{Q, Q̄} = −2H , [N , Q] = Q , [N , Q̄] = −Q̄ (4.11)
with all other (anti)commutators vanishing. These are quantum results, so their
orderings matter and have been carefully arranged to (i) maintain the classical sym-
metry algebra and (ii) correspond to well known geometric operations. (This ex-
plains the explicit spin connection appearing in the Hamiltonian H.) In fact, the
charges (Q, Q̄, N , −2H) correspond precisely to the exterior derivative, codifferen-
tial, form degree and form Laplacian acting on differential forms
QΨ = dΨ , Q̄Ψ = δΨ , −2HΨ = ∆Ψ . (4.12)
Indeed the above superalgebra is precisely the usual set of relations for these operators
dδ + δd = ∆ , d N = (N − 1)d , δN = (N + 1)δ
d∆ = ∆d , δ∆ = ∆δ ∆N = N ∆ . (4.13)
Our next task is to gauge this model and obtain a one-dimensional supergravity
theory whose BRST quantization can then be studied.

–6–
4.2 N = 2 Spinning Particle
In Dirac quantization, gauging the supersymmetry and worldline translation symme-
tries of N = 2 supersymmetric quantum mechanics amounts to imposing constraints
Q = Q̄ = H = 0. This is implemented by lapse (alias worldline einbein) and gravitini
Lagrange multipliers e, χ, χ̄ in the first order action
Z n o
S = dt pµ ẋµ + iψ̄m ψ̇ m − eH − χ̄Q − χQ̄ ,
(1)
(4.14)

where H = 21 πµ π µ − 12 Rµνρσ ψ̄ µ ψ ν ψ̄ ρ ψ σ , Q = iψ µ πµ , and Q̄ = iψ̄ µ πµ are now the

classical analogues of (4.10). Integrating out the canonical momentum yields
Z n o
1 ◦µ ◦ ∇ψ µ e
S = dt x gµν x ν + iψ̄µ + Rµνρσ ψ̄ µ ψ ν ψ̄ ρ ψ σ , (4.15)
2e dt 2
where the supercovariant tangent vector
◦
x µ ≡ ẋµ − iχ̄ψ µ − iχψ̄ µ . (4.16)

The theory enjoys local supersymmetries

1 ◦ 1 ◦
δxµ = iε̄ψ µ + iεψ̄ µ , Dψ µ = − x µ ε , D ψ̄ µ = − x µ ε̄ ,
e e
δe = 2iχ̄ε + 2iχε̄ , δχ = ε̇ , δ χ̄ = ε̄˙ . (4.17)
◦ µ µ ◦µ
Invariance is easily verified by noting that D x µ = iε̄ ∇ψ
dt
+ iε ∇dtψ̄ + δe2ex .
We must now BRST quantize the model. Firstly, notice that its Dirac quantiza-
tion simply imposes the conditions

dΨ = δΨ = ∆Ψ = 0 (4.18)

on differential forms Ψ. This is the solution to the de Rham cohomology in terms of

divergence-free, harmonic forms. These conditions are interpreted as the equations of
motion for the degrees of freedom propagated by the N = 2 spinning particle model
in the target spacetime. They are given in terms of the k-form field strengths Fk
for k = 0, .., D of (4.9) satisfying the Maxwell equations dFk = δFk = 0. It is
well known that the corresponding number of physical degrees of freedom is given
PD−2 D−2

by DoF = k=0 k
= 2D−2 . However it is extremely useful to have a gauge
theoretic description of these equations, for example by introducing auxiliary/gauge
fields that can allow for a corresponding action principle. For this purpose we will
employ the more powerful BRST quantization technique. As we shall see the above
equations will correspond to the BRST cohomology at a given ghost number.
We proceed in a canonical framework and introduce fermionic worldline diffeo-
morphism ghosts with algebra
{b, c} = 1 , (4.19)

–7–
along with bosonic superghosts

[p, z ∗ ] = 1 = [z, p∗ ] . (4.20)

Wavefunctions in the BRST Hilbert space now depend also on (c, z ∗ , p∗ )

(z ∗ )s (p∗ )t  
X ∞
∗ ∗
Ψ = Ψ(x, dx, c, z , p ) = ψs,t + c χs,t , (4.21)
s,t=0
s! t!

where both ψs,t and χs,t are sections of ∧M (ungraded differential forms). The choice
to represent the BRST Hilbert space in terms of a Fock space with the above polar-
ization is one of the key points of this Article. Although, one may suspect that the
choice of polarization does not influence the cohomology of the BRST charge QBRST ,
as we shall see, it has an extremely important impact on how that cohomology is
represented. This point was first realized by Siegel, see [22]. In particular, we will
find equations of motion expressed in terms of gauge potentials. These are realized
by the the so-called long, or detour operator. For this it is crucial that we express
BRST wavefunctions as an expansion in ghost number that is unbounded below and
above. Only in this way, can we form a detour operator connecting de Rham and
dual de Rham complexes.
On the BRST Hilbert space, it is easy to construct the nilpotent BRST charge,
the result is
QBRST = c∆ + z ∗ δ + zd − zz ∗ b . (4.22)

The first three terms are the ghosts times the constraints while the final term reflects
the first class constraint algebra {d, δ} = ∆. No further terms are necessary to ensure

Q2BRST = 0 , (4.23)

as this algebra is rank 1. The other operator we shall need is the ghost number

Ngh = cb + z ∗ p − p∗ z , (4.24)

which obeys
[Ngh , QBRST ] = QBRST . (4.25)

Our task now is to compute the cohomology of QBRST , namely

QBRST Ψ = 0 , Ψ ∼ Ψ + QBRST X , (4.26)

which is the topic of the next Section.

–8–
4.3 BRST Quantization
To solve the BRST cohomology,1 we begin by requiring that Ψ is BRST closed.
Computing QBRST Ψ we find the following conditions on the differential form-valued
coefficients of the BRST wavefunction (4.21)
dψ0,t+1 = 0 , t ≥ 0,

χs−1,t+1 = δψs−1,t + 1s dψs,t+1 , t ≥ 0, s ≥ 1,

(4.27)
∆ψ0,t = dχ0,t+1 , t ≥ 0,

∆ψs,t = sδχs−1,t + dχs,t+1 , t ≥ 0, s ≥ 1.

The last pair of relations are actually not independent of the first pair save for the
special case t = 0.
We may still shift Ψ by a BRST exact term QBRST X, for which we make the
ansatz
(z ∗ )s (p∗ )t  
X ∞
X= αs,t + c βs,t . (4.28)
s,t=0
s! t!
Computing QBRST X we find equivalences/gauge invariances
ψ0,t ∼ ψ0,t + dα0,t+1 , t ≥ 0,

ψs,t ∼ ψs,t + dαs,t+1 + s(δαs−1,t − βs−1,t+1 ) , t ≥ 0, s ≥ 1,

(4.29)
χ0,t ∼ χ0,t + ∆α0,t − dβ0,t+1 , t ≥ 0,

χs,t ∼ χs,t + ∆αs,t − sδβs−1,t − dβs,t+1 , t ≥ 0, s ≥ 1.

To analyze these equations it helps to invoke the grading by ghost number. At a
given ghost number Ngh = n, there are an infinity of form fields
(ψs,s−n , χs,s−n+1) , (4.30)
indexed by s. However, using the closed conditions and exactness freedom, we can
arrange for there to be only a single independent form field at each ghost number. To
see this first examine the second equivalence relation in (4.29). By choice of βs−1,t+1
we can set
ψs,t = 0 , t ≥ 0,s ≥ 1. (4.31)
Substituting this choice in the closed conditions (4.27), we learn that
χs,t = 0 , t ≥ 1, s ≥ 1,
(4.32)
χ0,t = δψ0,t−1 , t ≥ 1 ,

1
We thank Andy Neitzke and Boris Pioline for an invaluable collaboration leading to the results
of this Section.

–9–
so the remaining independent fields are ψ0,t with t ≥ 0 and χs,0 with s ≥ 0.
From (4.27) we see that they obey the closure conditions

dψ0,t = 0 , t ≥ 1,

δdψ0,0 = 0 , (4.33)
δχs,0 = 0 , s ≥ 0.

The first and last of these are the closed conditions for the de Rham complex and
its dual, while the middle relation is the detour operator.
Now we study exactness. Firstly we note that

ψ0,t ∼ ψ0,t + dα0,t+1 , t ≥ 0, (4.34)

whose interpretation in terms of de Rham complexes is clear. Then we observe that

maintaining the gauge choice (4.31) means that further transformations must obey
1
βs,t = dαs+1,t + δαs,t−1 t ≥ 1, s ≥ 0. (4.35)
s+1
Employing this relation, then from (4.29) and using ∆ = dδ + δd we have

χ00 ∼ χ00 + δdα0,0 + d(δα0,0 − β0,1 ) = χ0,0 + δdα0,0 + d(dα1,1 ) , (4.36)

so that
χ0,0 ∼ χ0,0 + δd α0,0 . (4.37)
Again this matches the detour operator. Finally a similar manipulation for the last
equivalence in (4.29) for t = 0 and s ≥ 1 yields

χs,0 ∼ χs,0 + δ(−sβs−1,0 + dαs,0 ) − d(βs,1 − δαs,0) . (4.38)

The last term vanishes using (4.35), so calling γs ≡ −sβs−1,0 + dαs,0 (s ≥ 1), we find

χs,0 ∼ χs,0 + δγs , s ≥ 1, (4.39)

which matches perfectly the dual de Rham complex. Therefore we have proven the
equivalence of the BRST cohomology and the Maxwell detour complex

d d d δ δ δ
· · · −→ ∧M −→ ∧M −→ ∧M ∧x
M −→ ∧M −→ ∧M −→ · · ·
 (4.40)
δd 

The horizontal grading is by ghost number, increasing from left to right. The detour
occurs at ghost number zero, at exactly which point the diffeomorphism ghost number

– 10 –
 t
1

2
−2
χ
3
−1

4
0
t
−3 −2 −1 0 s
d
δ δ δ δ
ψ 1
d

d δd 2

3
d
s

Figure 1: The Maxwell complex making its way through the BRST Hilbert space. Diag-
onal lines depict form fields of equal ghost number. The first graph plots fields ψs,t and
the second χs,t .

makes its jump by one unit. In Figure 1 we depict the Maxwell detour complex
snaking its way through the BRST Hilbert space.
The physical Hilbert space identified by the Dirac quantization procedure that we
summarized earlier on is embedded in the BRST cohomology at fixed ghost number
(i.e. zero ghost number for the present case). As we have seen the same cohomology
is reproduced by the detour complex. An advantage is that the detour operator
which acts at ghost number zero is formally self adjoint and can be used to construct
a field theoretical gauge invariant action for the degrees of freedom propagated by
R R
the particle, that is AδdA ∼ F 2 as expected. The counting of degrees of freedom
for this model is well known from standard work on antisymmetric tensor fields, and
we can reproduce it using the first quantized picture in a somewhat simpler way.

4.4 Counting Degrees of Freedom

To extract quantum information one can equivalently use either the first quantized
picture of the N = 2 spinning particle or the second quantized, gauge invariant, field
theory action. The first quantized approach is quite efficient, and we use it here to
compute the number of physical degrees of freedom.
We need to evaluate the partition function of the N = 2 spinning particle on
the circle to get the one-loop effective action Γ. With euclidean conventions it reads
Z
DX DG −S[X,G;gµν ]
Γ[gµν ] = e , (4.41)
S 1 Vol(Gauge)

– 11 –
where X = (xµ , ψ µ , ψ̄ µ ) and G = (e, χ, χ̄) indicate the fields that must be integrated
over. S[X, G; gµν ] is the euclidean version of the action in (4.15)
Z 1 n o
1 ◦µ ◦ ∇ψ µ e
S[X, G; gµν ] = dτ x gµν x ν + iψ̄µ − Rµνρσ ψ̄ µ ψ ν ψ̄ ρ ψ σ . (4.42)
0 2e dt 2
The division by the volume of the gauge group implies that we need to fix the
gauge symmetries. The loop (i.e. the circle S 1 ) is described by taking the euclidean
time τ ∈ [0, 1], imposing periodic boundary conditions on the bosons (xµ , e) and an-
tiperiodic boundary conditions on the fermions (ψ µ , ψ̄ µ , χ, χ̄). The gauge symmetries
can be used to fix the supergravity multiplet to Ĝ = (β, 0, 0), where β is the leftover
modulus that must be integrated over, i.e. the proper time. As the gravitini χ and χ̄
are antiperiodic, their susy transformations (4.17) are invertible so that they can be
completely gauged away, leaving Faddeev-Popov determinants and no moduli. This
produces
Z −2 Z
1 ∞ dβ 
Γ[gµν ] = − DetA ∂τ DX e−S[X,Ĝ;gµν ] , (4.43)
2 0 β S1

where the proper time measure takes into account the effect of the symmetry gener-
ated by the Killing vector on the circle. Note that the Faddeev-Popov determinants
with antiperiodic boundary conditions (denoted by the subscript A) coming from the
local supersymmetry do not depend on the target space geometry. The overall nor-
malization (−1/2) has been inserted to match with the standard result for a single
real scalar particle.
We are interested in computing the number of physical degrees of freedom, so
we lose no generality2 by taking the flat limit gµν = δµν , and evaluate the remaining
free path integral over the coordinates xµ and their fermionic partner ψ µ , ψ̄ µ
Z D−2 Z dD x
1 ∞ dβ 
Γ[δµν ] = − DetA ∂τ D . (4.44)
2 0 β (2πβ) 2
Apart from the standard volume term and the correctly normalized proper time
factors, this result contains the degrees of freedom propagating in the loop,

DoF = (DetA ∂τ )D−2 . (4.45)

The free fermionic determinant is easily computed: the antiperiodic boundary con-
ditions produce a trace over the corresponding two-dimensional Hilbert space. Thus
DetA ∂τ = 2 and the degrees of freedom DoF = 2D−2 as expected. This first quantized
picture has been used quite extensively in [18] to describe the quantum properties of
the gauge theory of differential forms coupled to gravity.
2
For partially massless theories [23], more care is needed because there are various massless
limits, but here we are only interested in the strictly massless one.

– 12 –
5. Mixed Higher Spin Detour

In [1], supersymmetric quantum mechanical models were constructed with R-symm-

etries obeying the superalgebra osp(Q|2r). The “supercharges” of these models
transformed under the fundamental representation of osp(Q|2r), and therefore gen-
erated Q Grassmann odd supersymmetries and 2r Grassmann even symmetries. The
models were constructed in curved backgrounds. It was found that the supercharges
commuted with the Hamiltonian only if the background manifold was a locally sym-
metric space, at general values of (Q, r). For low-lying values of (Q, r) the mod-
els coincide with well known quantum mechanical theories. The osp(1, 0) models is
the N = 1 supersymmetric quantum mechanics whose single supercharge corresponds
to the Dirac operator and whose Hilbert space describes spinors. The osp(2|0) theory
reproduces the N = 2 supersymmetric quantum mechanics described in Section 4.
For both those models the locally symmetric space condition is not necessary. In this
Section we concentrate on the models with R-symmetry osp(0|2r) = sp(2r). These
models are purely bosonic. Their Hilbert spaces correspond to multi-symmetric forms
and their quantized Noether charges yield symmetrized gradient and divergence type-
operators. We will concentrate on the simplest locally symmetric space–Minkowski
space, although many of our computations should generalize easily at least to con-
stant curvature spaces. We begin with the simplest sp(2) model, which describes
totally symmetric tensors or “symmetric forms”.

5.1 Symmetric Forms and sp(2) Quantum Mechanics

Symmetric forms share many similarities with their totally anti-symmetric counter-
parts–differential forms. They are expressed in terms of totally symmetric tensors
and commuting coordinate differentials so that a symmetric rank s tensor ϕ(µ1 ...µs )
becomes
Φ = ϕµ1 ...µs dxµ1 · · · dxµs . (5.1)

There is an algebra of operations—gradient, divergence, metric, trace and the wave

operator first introduced by Lichnerowicz [24] and systemized in [25, 26, 1] (see [4, 27,
28, 29] for other studies) — which greatly facilitates computations when the rank s
is large or even arbitrary. In particular, it is important to note that in this algebra
(just as for differential forms) it is no longer forbidden to add tensors of different
ranks.
In flat space, there are six distinguished operators mapping symmetric tensors
to symmetric tensors:

N –Counts the number of indices

N Φ = sΦ . (5.2)

– 13 –
 tr –Traces over a pair of indices

tr Φ = s(s − 1)ϕρ ρµ3 ...µs dxµ3 · · · dxµs . (5.3)

g –Adds a pair of indices using the metric

g Φ = gµ1 µ2 ϕµ3 ...µs+2 dxµ1 · · · dxµs+2 . (5.4)

div –The symmetrized divergence

div Φ = s∇ρ ϕρµ2 ...µs dxµ2 · · · dxµs . (5.5)

grad –The symmetrized gradient

grad Φ = ∇µ1 ϕµ2 ...µs+1 dxµ1 · · · dxµs+1 . (5.6)

∆ –The Bochner Laplacian

∆ = ∇µ ∇ µ . (5.7)

The calculational advantage of these operators is the algebra they obey

[N, tr] = −2tr , [N, div] = −div , [N, grad] = grad , [N, g] = 2g ,

[tr, grad] = 2div , [tr, g] = 4N + 2D , [div, g] = 2grad ,

[div, grad] = ∆ . (5.8)

All other commutators vanish.

Symmetric forms may be interpreted as the Hilbert space of a quantum mechan-
ical model whose quantum Noether charges are given by the operators above [1]. For
flat backgrounds this theory is described by the simple action
Z  
1 µ ∗ µ
S = dt ẋ ẋµ + izµ ż . (5.9)
2

Here the complex variables (z µ , zµ∗ ) are viewed as oscillator degrees of freedom de-
scribing the index-structure of wave functions so that upon quantization
∂
z ∗µ 7→ dxµ , zµ 7→ . (5.10)
∂(dxµ )
The model’s symmetries, Noether charges and their relation to the geometric opera-
tors given above is described in detail in [1]. In particular, the sp(2) “R-symmetry”

– 14 –
is generated by the triple (tr, N+ D2 , g). The pair of operators {div, grad} transform
as a doublet under this sp(2) and their commutator produces the Hamiltonian which
corresponds to the Laplacian ∆. In this sense, {div, grad} could be viewed as pair
of “bosonic supercharges”.
Totally symmetric tensor higher spin theories can be formulated in terms of
the algebra (5.8), so this relationship between those operators and the quantum
mechanical model (5.9) yields a first-quantized worldline approach to these models.
From that viewpoint we need to construct a spinning particle model by gauging an
appropriate set of symmetries. The BRST cohomology of that one dimensional gauge
theory then yields the physical spectrum of a higher spin field theory. We construct
this spinning particle model in Section 5.5. When the symmetries that we choose
to gauge form a Lie algebra, the BRST problem becomes equivalent to one in Lie
algebra cohomology. This is the topic of our next Section.

5.2 BRST Quantization and Lie Algebra Cohomology

In this Section we formulate the theory of massless, totally symmetric higher spins [33,
34] as the Lie algebra cohomology of a very simple algebra g:
n o
gA = tr , div , grad , ∆ , (5.11)

acting on the vector space V of symmetric forms. The BRST quantization for this
algebra was first studied in [30] (see also [31, 32]).
Let us very briefly review the relationship between BRST quantization and Lie
algebra cohomology. In BRST quantization the BRST Hilbert space is expressed as
wavefunctions expanded in anticommuting ghosts
dimg
X
ΨBRST = cA1 · · · cAk ΨA1 ...Ak , (5.12)
k=0

while in Lie algebra cohomology the V -valued wavefunctions ΨA1 ...Ak are viewed as
multilinear maps g∧k → V and form the cochains of a complex. The cochain degree k
is BRST ghost number. The Chevalley–Eilenberg differential δ [35]

δ k B
ΨA1 ...Ak → g[A1 ΨA2 ...Ak+1 ] − f[A Ψ|B|A3 ...Ak+1] , (5.13)
2 1 A2
can be compactly expressed in terms of the BRST charge

1 C A B ∂
QBRST = cA gA − fAB c c , (5.14)
2 ∂cC
acting at ghost number k. The cohomology of QBRST at ghost number k equals the
Lie algebra cohomology H k (g, V ).

– 15 –
 Returning to our specific Lie Algebra (5.11), we now relate its Lie algebra coho-
mology at degree one to the massless higher spin theory. At degree one, our problem
is a very simple one: We first introduce a wavefunction for every generator
n o
ΨA = Ψtr , Ψdiv , Ψgrad, Ψ∆ . (5.15)


dimg
The closure condition ΨA ∈ kerδ yields a set of 2
differential equations following
directly from the commutation relations (5.8)

tr Ψdiv − div Ψtr = 0 ,

tr Ψgrad − grad Ψtr = 2 Ψdiv ,

tr Ψ∆ − ∆ Ψtr = 0 ,

div Ψgrad − grad Ψdiv = Ψ∆ ,

div Ψ∆ − ∆ Ψdiv = 0 ,

grad Ψ∆ − ∆ Ψgrad = 0 . (5.16)

Exactness, ΨA ∼ ΨA + XA with XA ∈ imδ, yields the gauge invariances of this set

of equations

δΨtr = tr ξ ,

δΨdiv = div ξ ,

δΨgrad = grad ξ ,

δΨ∆ = ∆ ξ . (5.17)

The fields (Ψ∆ , Ψgrad, Ψdiv ) correspond to the BRST triplet structure discussed
in [31] (denoted (C, ϕ, D)) while Ψtr is the compensator field introduced there.
Eliminating the fields Ψdiv and Ψ∆ using the second and fourth equations in (5.16),
we find that only the first and sixth of these equations are independent and give a
description of all massless, totally symmetric higher spins in terms of a pair of un-
constrained fields (Ψgrad, Ψtr ) with a single unconstrained gauge parameter ξ
1
G Ψgrad = grad3 Ψtr ,
2
 1 
tr2 Ψgrad = 4 div + grad tr Ψtr ,
4
δΨgrad = grad ξ ,

δΨtr = tr ξ . (5.18)

– 16 –
Here the operator G is given by
1
G = ∆ − grad div + grad2 tr . (5.19)
2
Although, for some contexts, a formulation of higher spin dynamics in terms of
unconstrained fields can be useful, the above system has the disadvantage that it has
terms cubic in derivatives. This problem can be removed by using some of the gauge
freedom to set the field Ψtr = 0. This yields the standard description of massless,
totally symmetric higher spins in terms of a doubly trace-free symmetric tensor and
a trace-free gauge parameter

G Ψgrad = 0 = tr2 Ψgrad ,

δΨgrad = grad ξ , tr ξ = 0 . (5.20)

It is important to note that since we kept the rank s, or in other words the eigenvalue
of the index operator N, arbitrary these relations generate the gauge invariant equa-
tions of motion for fields of any spin. We also remark that the operators {G, tr2 }
themselves generate a first class algebra.
Finally, we can formulate this system in terms of a detour complex as follows.
The field equation G Ψgrad = 0 is equivalent to the equation
 1 
G Ψgrad ≡ 1 − g tr G Ψgrad = 0 , (5.21)
4
where
1  2 2
 1  
G = ∆ − grad div + grad tr + g div − g 2∆ + grad div tr . (5.22)
2 4
We will call this operator the higher spin Einstein operator, since if Ψgrad = hµν dxµ dxν
it then produces the linearized Einstein tensor

GΨgrad = (∆hµν − 2∇µ ∇ρ hρν + ∇µ ∇ν hρρ + ηµν ∇ρ ∇σ hρσ − ηµν ∆hρ ρ )dxµ dxν . (5.23)

The higher spin Einstein operator obeys identities

div G = 0 modlef t g , (5.24)

G grad = 0 modright tr , (5.25)

where equality holds up to terms proportional to the operators g and tr acting from
the left and right, respectively. This allows us to form a detour complex
grad div
0 −→ ⊙T M/tr −→ ⊙T M/tr2 → · · · · · · → ⊙T M/
x tr2 −→ ⊙T M/tr2 −→ 0

G 
(5.26)

– 17 –
where ⊙T M/• denotes symmetric tensors modulo the relation •. The operator G
is formally self-adjoint so the equation of motion GΨgrad = 0 comes from an action
R
principle S = 12 (Φ, GΦ) where the inner product (·, ·) is the one inherited from
the underlying quantum mechanical model. The relations (5.24) and (5.25) express
the Bianchi identity and gauge invariance of this field equation. Our next task is to
generalize this construction to tensors of arbitrary symmetry types, and in particular
find compact expressions for the Einstein operators for these theories.

5.3 Mixed Tensors and sp(2r) Quantum Mechanics

Tensors transforming under arbitrary representations of gl(D) can be expressed either
in terms of:

1. Tensors labeled by groups of antisymmetric indices ω[µ11 ...µ1k ]···[µs1 ...µsk ] —“multi-
1 s
forms”— or schematically

⊗ ⊗···⊗ (5.27)

where s labels the number of antisymmetric columns, while the ki label the
number of boxes in each column.

or

2. Tensors labeled by groups of symmetric indices ϕ(µ11 ...µ1s )···(µr1 ...µrsr ) —“multi-
1
symmetric forms”— or schematically

⊗ (5.28)
..
.
⊗

where r labels the number of symmetric rows, while the si label the number of
boxes in each row.

In each case irreducible gl(D) representations are obtained by placing algebraic con-
straints on tensors of these types akin to the Bianchi identity of the first kind obeyed
by the Riemann tensor. Supersymmetric quantum mechanical models whose Hilbert
spaces are populated by the tensors described above were developed in [1]. The R

– 18 –
symmetry groups are O(2s) and Sp(2r) for the two respective cases. Models for
tensors with both symmetric and antisymmetric groups of indices also exist and
have an osp(2s|2r) R-symmetry. The N = 2 and sp(2) (super)symmetric quantum
mechanical models described above are the lowest lying examples of these.
Although, all the computations in this work, should in principle carry over
to both osp(Q|2r) models for arbitrary integers3 (Q, r), here we concentrate on
the sp(2r) case. There are two reasons for this choice. Firstly, because the rows
in (5.28) are symmetric, this allows us to handle arbitrarily high spins without
introducing arbitrarily quantum mechanical oscillator modes. In particular, if we
take r ≥ D wavefunctions span tensors of arbitrary type. Secondly, since only
bosonic oscillators are required, the BRST ghosts will all be fermionic leading to a
BRST wavefunction with a finite expansion in ghost modes and in turn a Lie algebra,
rather than Lie superalgebra cohomology problem.
The quantum mechanical model whose wavefunctions describe multi-symmetric
forms derives from the simple action principle
Z  
1 µ ∗ iµ
S = dt ẋ ẋµ + iziµ ż . (5.29)
2
∗
Here we have introduced 2r oscillators (ziµ , z iµ ) with i = 1 . . . r. Their kinetic term
can be written in the manifestly sp(2r) symmetric way 2i zαµ ǫαβ żβµ where zα = (zi∗ , z i )
and ǫαβ is the antisymmetric invariant tensor of sp(2r). Again, this theory can be
coupled to curved backgrounds; we refer to [1, 36] for details.
Upon quantization the oscillators can be represented in terms of sets of commut-
ing differentials [6, 1]
∂
zi∗µ = di xµ , zµi = . (5.30)
∂(di xµ )
The tensor depicted in (5.28) is then denoted

µ1 1 r r
Φ = ϕ(µ11 ...µ1s )···(µk ...µr ) d1 x 1
1 sr
· · · d1 xµs1 · · · dr xµ1 · · · dr xµsr . (5.31)
1

The sp(2) generators g, N and tr of the above Sections are promoted to r × r

matrices of operators:

gij = di xµ gµν dj xν ,

∂
Nji = dxµi ,
∂(dj xµ )
∂ ∂
trij = µ
g µν . (5.32)
∂(di x ) ∂(dj xν )
3
Note that when Q is odd, the model includes spinor fields [1].

– 19 –
The operators (gij , Nji , trij ) generate sp(2r)

[Nji , gkl ] = j
2δ(k gl)i ,
(i j) i j
[trij , gkl ] = 4δ(k Nl) + 2D δ(k δl) ,
(k
[Nji , trkl ] = −2δi trl)j ,

[Nji , Nlk ] = δkj Nli − δil Nkj . (5.33)

They correspond to “R-symmetries” of the model (5.29). Geometrically, in terms of

the picture (5.28), they count the number of indices in a given row, move boxes from
one row to another, and add or remove pairs of boxes to or from (possibly distinct)
rows using the metric tensor.
The differential operators div and grad are replaced by 2r operators corre-
sponding to the divergence and gradient acting on each row in (5.28). These form
the fundamental representation of sp(2r).

[Nji , gradk ] = δkj gradi , [Nji , divk ] = −δik divj ,

(i
[trij , gradk ] = 2δk divj) , [gij , divk ] = −2δ(ik gradj) , (5.34)

and themselves obey the supersymmetry-like algebra

[divi , gradj ] = δji ∆ . (5.35)

Before studying the Lie algebra cohomology of the above algebra, and the accom-
panying spinning particle model, in the next Sections, let us briefly discuss how
irreducible tensor representations can be obtained from the reducible ones depicted
in (5.28).
There are two pertinent notions of irreducibility for tensors. The first is with
respect to gl(D) and is obtained by studying all possible permutation symmetries.
This can be achieved using the operators Nji which move a box from row j to row i
with a combinatorial factor equaling the number of boxes in row j, for example
   
   
 ⊗   ⊗ 
   
3   
N2   = 4  . (5.36)
   
 ⊗   ⊗ 
   

Irreducible gl(D) representations correspond to Young diagrams in which the number

of boxes in each row decreases weakly (read from top to bottom). This amounts to
tensors in the kernel of all operators

{Nj>i
i }, (5.37)

– 20 –
which generate the nilradical subalgebra of gl(r).
For physical applications, often the stronger requirement of so(D) irreducibility
is placed on tensors. This amounts to additionally removing all traces and therefore
tensors in the kernel of
{Nj>i ij
i , tr } . (5.38)
This set generates the nilradical of sp(2r). It plays an important rôle in the choice
of first class algebra in the BRST construction of the next Section.

5.4 Mixed Symmetry Einstein Operators

The aim of the this Section is to derive the generalization of the Einstein opera-
tor (5.22) to higher spins of arbitrary symmetry type. This result can actually be
directly obtained by covariantizing (5.22) with respect to4 sp(2r) but here we outline
its BRST derivation to connect with our path integral techniques. (For a review of
existing available higher spin BRST techniques see [37], the unsymmetrized version
of our Einstein operator was derived by BRST techniques recently in [38]. BRST
quantization of the algebra (5.39) below was also considered in [39]).
To begin with we must choose the first class algebra generalizing (5.11). The
basic philosophy is that gauging {divi , gradi , ∆} yields the correct differential re-
lations on propagating physical modes–this is perhaps seen most easily in the path
integral approach described in Section 5.6. We also expect algebraic relations that
ensure the theory describes irreducible representations of the Lorentz group. These
will follow by also gauging the nilradical sp(2r) operators in (5.38). Hence we study
the Lie algebra cohomology of
n o
g = trij , divi , Nj>i
i , gradi , ∆ (5.39)

acting on multi-symmetric forms.

The next question we must address is what degree/ghost number cohomology
corresponds to the underlying physical system. This is resolved by recalling that
the totally antisymmetric tensor theories possess gauge for gauge symmetries. For
example, for a two-form ω with gauge transformation δω = dα, exact one-form gauge
parameters α = dβ do not act on ω. In other words, two-forms appear at degree two
in de Rham cohomology. However, multi-symmetric forms with r rows as in (5.28)
can also be expanded in multi-forms of degree r or less. This implies gauge for
gauge symmetries even for multi-symmetric forms and shows that we should study
the degree r cohomology.

2


At degree r there are dimg
r
= (r+1)
r
closure relations on dimg
r−1
fields (because
2
here dimg = (r + 1) ). These read
r B
g[A1 ΨA2 ...Ar+1 ] − f[A Ψ|B|A3 ...Ar+1] = 0 . (5.40)
2 1 A2
4
We thank Stanley Deser for this observation.

– 21 –
 dimg

(r+1)2


These fields enjoy r−1
= r−1
gauge invariances following from exactness
r−1 B
δΨA1 ...Ar = g[A1 ξA2 ...Ar ] − f ξ|B|A3 ...Ar ] . (5.41)
2 [A1 A2
Solving this system at arbitrary degree r may seem daunting, but is in fact not far
more difficult than the r = 1, sp(2) computation performed above. Let us sketch the
main ideas and then give the result.
Firstly, we note that the field
1 i1 ...ir
Ψ ≡ Ψgrad1 ...gradr = ǫ Ψgradi1 ...gradir (5.42)
r!
gives the minimal covariant field content of our final detour complex. Here, and in
what follows, we use a compact notation where field labels gradi are soaked up with
the sl(r) invariant, totally antisymmetric symbol
1
Ψi•1 ...in ≡ ǫi1 ...ir Ψ• gradin+1 ...gradir . (5.43)
(r − n)!
Now we turn to the closure relations (5.40). When one of the adjoint indices
is ∆ and all others are gradi we obtain a relation analogous to the last one in (5.16)

∆ Ψ − gradi Ψ∆i = 0 . (5.44)

The field Ψ∆i is not independent. It is eliminated by a pattern of closure relations

analogous to the sp(2) ones in (5.16)
j
divi Ψ − gradj Ψdiv i
i = Ψ∆ ,

j
trij Ψ − gradr Ψtr
r
ij = Ψ
divi
+ Ψidivj . (5.45)

These imply
1
GΨ = gradi gradj gradk Ψktrij , (5.46)
2
where
1
G = ∆ − gradi divi + gradi gradj trij . (5.47)
2
(This operator was first derived by Labastida in [4].) As in the sp(2) case, the
field Ψktrij vanishes once we use the gauge freedom implied by exactness. However,
there are still double-trace relations satisfied by the physical field Ψ. For these we
consider the further closure relation

trij Ψldivk − divk Ψltrij − gradm Ψml

divk trij
= Ψildivk divj + Ψjl
divk divi
− Ψkl
∆ trij . (5.48)

This time the second and third terms on the left hand side can be gauged away as
can the antisymmetric part of Ψidivj so, using (5.45), we must solve the equation

trij trkl Ψ = Ψildivk divj + Ψjl

divk divi
− Ψkl
∆ trij . (5.49)

– 22 –
The antisymmetric part in k and l determines Ψkl ∆ trij
in terms of Ψildivk divj which in
turn depends on double traces of the physical field Ψ. However, symmetrizing the
above equation in any three indices causes the right hand side to vanish identically.
Therefore we learn the double trace relation

tri(j trkl) Ψ = 0 . (5.50)

Finally we must remember the closure relations coming from gauging the R-symmetries
Nj>i j>i
i . In particular, using the trivial identity δi = 0 we have

Nji Ψ − gradk ΨkNj = 0 , j > i. (5.51)

i

Gauging away the second term leaves us with the algebraic constraint

Nj>i
i Ψ = 0. (5.52)

At this point there are no further constraints on the physical field Ψ and all remaining
fields are either gauged away or algebraically dependent on Ψ. Let us gather together
the equations of motion for Ψ;

G Ψ = 0 = tri(j trkl) Ψ = Nj>i

i Ψ, (5.53)

with G given in (5.47). As a consistency check, it is not difficult to verify that the
operators {G, tri(j trkl) , Nj>i
i } themselves form a first class algebra.
These equations of motion are gauge invariant under transformations following
from exactness (5.41)
1
δΨ = ǫi1 ...ir gradi1 ξi2 ...ir ≡ gradi ξ i . (5.54)
(r − 1)!

In the last term we have employed the compact notation (5.43) also for the gauge pa-
rameters. The parameters themselves are subject to constraints that can be deduced
by carefully following which gauge freedoms were employed to remove all independent
fields save for the physical one Ψ. These read

tr(ij ξ k) = tri(j trkl) ξ m = 0 = Nj>i

i ξ k + δik ξ j>i . (5.55)

The longhand notation in the second term of (5.54) makes it clear that there are
gauge for gauge symmetries

δξgradi1 ...gradir−1 = grad[i1 ξgradi2 ...gradir−2 ] , (5.56)

coming from the Lie algebra cohomology at degree r − 1. Of course, there are further
gauge for gauge symmetries of the same form corresponding to the system being
rank r when irreducible tensors are expanded in an antisymmetric basis.

– 23 –
 To end this Section, we compute the mixed symmetry analog of the Einstein
operator (5.22). Firstly the field equation G Ψ = 0 is equivalent to
 1 1 
G Ψ ≡ 1 − gij trij + gij gkl trij trkl GΨ = 0 . (5.57)
4 48
(Here, and in the formula that follows we specialize to r = 2 for simplicity). The
mixed higher spin Einstein operator G then equals

1 
G = ∆ − gradi divi + gij divi divj + gradi gradj trij
2
1   1
− gij 2∆ − gradk divk trij − gij gradk divi trjk
4 2
1  
+ gij gkl 4∆ + gradm divm trij trkl
48
1 
− gi[j gk]l divi divj trkl + gij gradk gradl tri[j trk]l . (5.58)
6
It is self-adjoint and obeys Bianchi and gauge invariance identities

divi G = 0 modlef t gij , (5.59)

ij
G gradi = 0 modright tr , (5.60)

ensuring the existence of a detour complex analogous to (5.26). Once again, an action
R
principle S = 21 (Φ, GΦ) with inner product (·, ·) inherited from the underlying
quantum mechanics also follows immediately. Our next task is to quantize this
system. We adopt a first quantized approach which we now explain.

5.5 The sp(2r) Spinning Particle

The particle action introduced in the previous Section in equation (5.29) is invariant
under global extended supersymmetry, with Sp(2r) R-symmetry group. This action
can be used to construct locally supersymmetric particle actions that give a path
integral implementation of the quantum algebras studied previously.
Let us start from the phase space symplectic integral
Z

S = dt pµ ẋµ + iziµ ∗ iµ
ż , (5.61)

which is invariant under the action of the global transformations with symmetry
generator

1 1
G = ξH + σ̄i S i + σ i S̄i + β ij K̄ij + β̄ij K ij + αji Jij , (5.62)
2 2

– 24 –
where the classical susy generators and classical sp(2r) generators are, respectively,
given by
1 2
S̄i = p · zi∗ , S i = p · zi , H= p , (5.63)
2
K̄ij = zi∗ · zj∗ , Jij = zi∗ · z j , K ij = z i · z j . (5.64)
Here the u(r) subalgebra generated by Jij is made manifest. Note that the canonical
quantization discussed earlier simply amounts to the replacement
iS̄i → gradi , iS i → divi , −2H → ∆

K̄ij → gij , Jij → Nji , K ij → trij . (5.65)

The transformation rules for the dynamical fields can be read off from δq = {q, G}
δxµ = ξ pµ + σ̄i z iµ + σ i zi∗µ ,
δpµ = 0 ,
δz iµ = −iσ i pµ − iαji z jµ − iβ ij zj∗µ ,
∗
δziµ = iσ̄i pµ + iαij zjµ
∗
+ iβ̄ij zµj . (5.66)
Gauged actions are thus obtained by adding gauge fields coupled to the above con-
served charges. In particular, gauging all the global symmetries yields the action
Z  
µ ∗ iµ i i 1 ij 1 ij i j
S = dt pµ ẋ + iziµ ż − eH − s̄i S − s S̄i − b̄ij K − b K̄ij − aj Ji ,
2 2
(5.67)
and the transformations for the gauge fields are obtained by requiring S to be invari-
ant under local transformations Ξ(t) = (ξ(t), σ̄i (t), σ i (t), β̄ij (t), β ij (t), αji (t)), namely
δe = ξ˙ + i2σ i s̄i − i2σ̄i si ,
δsi = σ̇ i + iaij σ j − iαji sj + iβ ij s̄j − ibij σ̄j ,
δs̄i = σ̄˙i − iaji σ̄j + iαij s̄j − iβ̄ij sj + ib̄ij σ j ,



δbij = β̇ ij + i aik β jk + ajk β ik − i αki bjk + αkj bik ,



δ b̄ij = β̄˙ij − i aki β̄jk + akj β̄ik + i αik b̄jk + αjk b̄ik ,



δaij = α̇ji − i αki akj − aik αjk + i β ik b̄jk − bik β̄jk . (5.68)
However, it will be most interesting to consider partial gaugings that only involve
subalgebras of the above algebras. In fact, in many of these cases it is possible to
leave (part of) the abelian subgroup U(1)r ⊂ U(r) invariant, which would allow a
gauge-invariant Chern-Simons action
Z X
SCS = dt qi aii . (5.69)
i

– 25 –
This can be used to fix the number of indices in a particular row of a Young diagram.
Let us single out a few interesting cases

1. Gauge H, S̄i , S i , K ij , Jji (or ∆, grad, div, tr and all N’s). This amounts
to setting bij = β ij = 0 in the previous transformations rules and leaves in
particular


δaij = α̇ji − i αki akj − aik αjk (5.70)
Pr
from which it is obvious that a = aii is invariant and the unique Chern–
R i
Simons term is SCS = dt q a.

2. Gauge H, S̄i , S i , K ij , Jij≥i (or ∆, grad, div, tr and nilradical N’s). This
amounts to setting bij = β ij = 0 and aij = αji = 0 when i > j, from which

δaii = 0 , (no sum implied) (5.71)

R P
and allowed Chern–Simons are given by SCS = dt i qi aii .

We are now ready to quantize these gauged models.

5.6 Counting Degrees of Freedom

In the present Section we use the particle actions described above to compute the
number of degrees of freedom for mixed higher spin tensor multiplets. We study the
partition function
Z
DXDE
Z∼ eiS[X,E]+iSCS [E] , (5.72)
S 1 Vol(Gauge)

where X collectively denotes all dynamical fields, whereas E denotes gauge fields.
In the latter we need to carefully gauge fix all the gauge symmetries present in the
spinning particle action. We now Wick rotate to Euclidean time (periodic boundary
conditions) and use the Faddeev-Popov trick to extract the volume of the gauge
group to set a gauge choice that completely fixes all the supergravity fields up to
some constant moduli fields

E = (e, s̄i , si , b̄ij , bij , ai j ) = (β, 0, 0, 0, 0, θiδji ) , (5.73)

where âi j = θi δji is the most generic constant element of the Cartan subalgebra
of sp(2r), with θi being angles taking values in a fundamental domain. We thus have
Z ∞ Z
1 dβ dD x
Z=− DoF(q, r) (5.74)
2 0 β (2πβ)D/2

– 26 –
with
r Z
Y  r   Y r  
2π
dθi iqi θi Y
DoF(q, r) ≡ Kr e Det(∂τ − iθi ) Det(∂τ + iθi )
0 2π
|i=1 {z i=1
} | {z i=1
} | {z }
Cartan moduli for Sp(2r) + CS S̄i Si
r 
Y  Y r  r 
 Y 
× Det(∂τ + iθi )]−D Det(∂τ − 2iθi ) Det(∂τ + 2iθi )
|i=1 {z } |i=1 {z } |i=1 {z }
z,z ∗ K̄ii no sum K ii no sum
Y   Y   
× Det ∂τ − i(θi − θj ) Det ∂τ − i(θi + θj )
i6=j i<j
| {z }| {z }
U (r) step operator K̄ij , i6=j
Y  
× Det ∂τ + i(θi + θj ) (5.75)
i<j
| {z }
K ij , i6=j

being the number of degrees of freedom, and Kr−1 the number of fundamental domains
included in the integration domain. All the determinants are evaluated with periodic
boundary conditions because the fields traced over are bosonic.
The latter expression should really be understood as a generating function, in-
cluding all ingredients for all possible sp(2r) particle actions. Clearly, for each specific
gauged action, one has to pick out only those determinants that are involved in its
gauge fixing. Let us consider the example 2 of Section 5.5, which we claim corre-
sponds to the BRST cohomology of the algebra (5.39) computed in Section 5.4. For
that theory we clearly have

r Z
Y  r  Yr  
dθi iqi θi Y
2π
2−D
DoF(q, r) = Kr e Det(∂τ + iθi )] Det(∂τ − 2iθi )
i=1 0 2π i=1 i=1
Y   Y   
× Det ∂τ − i(θj − θi ) Det ∂τ + i(θi + θj ) . (5.76)
i<j i<j

Using that

Det(∂τ + iθ) = 2i sin(θ/2) (5.77)

and making the change of variables wi = e−iθi we obtain

r  I
Y 
dwi (1 − wi )2−D (wi2 − 1) Y
DoF(q, r) = Kr − q +r+1−D/2
(wj −wi )(1−wj wi ) (5.78)
i=1
2πiwi wi i i<j

– 27 –
that gives
r 
Y 
1 d ni
DoF(q, r) = Kr −
i=1
ni ! dwini
r
Y Y
× (1 − wi )2−D (wi2 − 1) (wj − wi )(1 − wj wi ) (5.79)
wi =0
i=1 i<j

with ni = qi + r + 1 − D/2 .
Let us consider a few specific examples. The simplest one is clearly r = 1 for
which we set s ≡ q + 2 − D/2 = n and, using that K1 = 1, we obtain
 
D + 2s − 4 D + s − 5
DoF(s, 1) = (5.80)
s s−1

which is precisely the dimension of a Young tableau of so(D − 2) with 1 row and s
columns.
For r = 2, we identify n2 ≡ s2 + 1 and n1 ≡ s1 , then K2 = 1 and using (5.79) we
obtain
(D + s1 − 7)!(D + s2 − 6)!(s2 − s1 + 1)!
DoF(s2 , s1 , 2) =
(D − 6)!(D − 4)!(s2 + 1)! s1! (s2 − s1 )!
× (D + s2 + s1 − 5)(D + 2s2 − 4)(D + 2s1 − 6) . (5.81)

This is the dimension of a Young tableau with s2 boxes in the first row and s1 ≤ s2
boxes in the second row.
For arbitrary r we identify nk = sk + k − 1 so that (5.79) should yield the
dimension of a generic Young tableau with sk boxes in the k−th row and s1 ≤
s2 ≤ · · · ≤ sr . In all these cases, the physical degrees of freedom correspond then
to irreducible so(D − 2) representations. In fact, for arbitrary r, we expect that
equation (5.79) is the generating function for the dimensions of irreducible so(D −2)
representations. In the next Section we obtain the same result from the second
quantized equations of motion that follow from the BRST cohomology computation
of Section 5.4.

5.7 Lightcone Degrees of Freedom

Our final computation is to verify that the path integral degree of freedom counts
match those obtained by a direct analysis of the second quantized field equations
(5.53) which we reproduce here for convenience
 1 
∆ − gradi divi + gradi gradj trij Ψ = 0 , (5.82)
2
tri(j trkl) Ψ = 0 , (5.83)

Nj>i
i Ψ = 0. (5.84)

– 28 –
These equations enjoy the gauge invariances

δΨ = gradi ξ i , where tr(ij ξ k) = tri(j trkl) ξ m = 0 . (5.85)

A detailed proof of these dimension counts, which are essentially just a higher
dimensional analog of Wigner’s original computation of unitary representations of the
Poincaré group [40], were only given rather recently in [41]. A BRST lightcone version
of this computation was given in [38]. By far the the speediest method to perform this
computation, however, is to employ lightcone gauge directly to the second quantized
Lagrangian. For completeness we present that result here. Expressing the metric as

ds2 = 2dx+ dx− + d~x 2 , (5.86)

we assume that ∂/∂x− is invertible and set it equal to unity in what follows.
Our main philosophy is to expand fields and field equations in powers of differ-
entials di x− in the x− direction. All the operators (∆, gradi , divi , gij , Nji , trij ) then
have (D − 2)-dimensional analogs operating in the ~x directions. We denote these by
hats so that
∂ b,
∆ =2 + ∆
∂x+
∂ [i,
gradi = di x− + di x+ + grad
∂x+
∂ ∂ ∂ di ,
divi = − +
+ + div
∂(di x ) ∂x ∂(di x+ )
gij = 2 d(i x− dj) x+ + g
bij ,
∂ ∂ bj ,
Nji = di x− −
+ di x+ +N i
∂(dj x ) ∂(dj x+ )
∂ ∂ b ij .
trij = 2 + tr (5.87)
∂(d(i x− ) ∂(dj) x+ )

Now we decompose the field Ψ as

Ψ(di x− ) = ψ + di x− χi , (5.88)

where ψ is independent of di x− . On the other hand the fields χi are di x− dependent,

and we focus on the term with the highest power of di x− . Examining the terms
of highest order in the field equation (5.82) coming from the grad2 tr-term, we see
that the highest order term in χi is lightcone symmetric-trace-free (i.e., annihilated
b (ij ). However, from the lightcone decomposition of gradi in (5.87), the gauge
by tr
invariance (5.85) becomes  
δΨ = di x + · · · ξ i .
−
(5.89)

– 29 –
The trace condition on the gauge parameter ξ i exactly ensures that its highest di x−
term is also lightcone symmetric-trace-free. Hence we may algebraically gauge away
the highest order term in χi . Iterating the above argument allows us to gauge away
all of χi so that
Ψ = ψ(di~x, di x+ ) . (5.90)
Our computation is completed by solving (5.82) for fields of the above form. Again
we work order by order in di x− . At highest order we learn

b ij ψ = 0 .
tr (5.91)

To study lower order terms we split

b i~x) + di x+ ψ i ,
ψ(di~x, di x+ ) = ψ(d (5.92)

where the field ψb only has (D−2)-dimensional indices. The fields ψ i are all dependent
because the next to leading order terms in (5.82) imply
∂ψ diψ = 0 .
+ div (5.93)
∂(di x+ )
This condition can always be solved in terms of the ψi in (5.92) so it remains to
gather the remaining lowest order terms in (5.82) which read
n ∂ o
2 + +∆ b ψb = 0 . (5.94)
∂x
This is simply the D-dimensional Klein–Gordon equation. Finally we still need to
impose the symmetry condition. It is not hard to see that it implies
b j>i ψb = 0 .
N (5.95)
i

Hence the independent light cone degree of freedom are described by a totally sym-
metric (D − 2)-dimensional tensors, which solve the D-dimensional wave equation
and are both (D − 2)-dimensional trace-free and irreducible. Or in other words,
the degree of freedom count is given by dimensions of so(D − 2) irreducible repre-
sentations. This shows the claimed equivalence between BRST and path integral
quantizations.

6. Conclusions
In this Article we have tackled the problem of constructing and quantizing quan-
tum field theories for tensor fields with general symmetry types using a worldline
approach. As depicted below, our starting point was a quantum mechanical (su-
per)symmetric model whose wave functions are the type of tensor fields appearing
in the desired second quantized model.

– 30 –
 (Super)symmetric
Quantum Mechanics

ւ ց
Spinning Particle BRST Detour
Path Integral ←−−→ Quantization

Thereafter we identified first class constraint operator algebras acting on the quantum
mechanical Hilbert space. From these algebras one can build a first quantized gauge
theory in two ways, either path integral methods, or BRST quantization. The former
led to a path integral representation of a spinning particle model while the latter,
using the detour complex idea, yielded the classical equations of motion of a second
quantized gauge field theory. The path integral approach gave a worldline method
for computing quantum quantities in the second quantized field theory, the simplest
of which was a count of the physical degrees of freedom. These can of course also
be computed by studying the dimensionality of the Cauchy data of the classical field
equations. Indeed we found that these two methods gave identical answers for a large
class of higher spin theories.
The quantum mechanical models we used as a starting point fall into a very
broad class of models labeled by their R-symmetry groups which are given by general
orthosymplectic supergroups. In that language our Article focused on the osp(2|0)
and osp(0|2r) models. However, it is clear that our methods can be generalized to
any of the osp(Q|2r) models. When Q is odd, these models describe spinor-tensor
fields in second quantization. The case osp(2|1) has been studied in [26] to describe
spinor valued totally symmetric tensor theories, but clearly a complete description
of fermionic second quantized models would be desirable.
There are many other directions our results lead to. The most interesting of
course, would be to shed light on self-interactions of higher spin fields. By now a
large literature exists on this subject, a consistent theme being that interactions for
higher spins requires towers of infinitely many second quantized fields (see [8] for an
extensive review of these developments). In simplest terms this points at a difficulty
gauging the number operator(s) N of our quantum mechanical models. From the
path integral viewpoint, this difficulty can be seen through the sparsity of consistent
world-line Chern–Simons terms that can be added to the worldline action.
There are two other most interesting, and in fact related, applications of our
results. These are computations of higher (second quantized) quantum amplitudes
and interactions with backgrounds. Higher amplitudes are encoded, for example, by
studying the dependence of the worldline effective action on arbitrary background
fields. It is not difficult to couple our underlying quantum mechanical models to
either background Yang–Mills or gravitational fields by twisting the connection ap-

– 31 –
pearing in the covariant canonical momentum operator. However in general this can
produce obstructions to our constraint algebras being first class. These obstructions
have been studied and explicated in [36]. The phenomenon of higher spin fields suffer-
ing inconsistencies in backgrounds is one that has been known for a long time (dating
back to work on coupling massive spin 3/2 fields, see for a thorough account [42]).
It is possible to view these obstructions to first class algebras in general spaces in
a more positive light. Namely, these algebras can be used to develop powerful invari-
ants for determining the underlying geometry of the background manifold. In turn,
when the background manifold belongs to a special class of geometries, consistency
and even enhanced symmetries and constraint algebras can result. The special rôle
played by certain geometries in string theory is an example of this phenomenon.
Another example are the Kähler higher spin models constructed in [43, 44, 45]
and (p, q)-form Kähler electromagnetism [46]. The latter of these theories follows
from the detour construction [47]. It would be most interesting to compute its path
integral quantization.

Acknowledgments

A.W. thanks INFN and the University of Bologna for the warmest hospitality during
a visit in which the bulk of this work was completed. We thank Stanley Deser,
Emanuele Latini, Andy Neitzke and Boris Pioline for collaborations during early
stages of this work. The work of F.B. and O.C. was partly supported by the Italian
MIUR-PRIN contract 20075ATT78.

References
[1] K. Hallowell and A. Waldron, SIGMA 3, 089 (2007), arXiv:0707.3164 [math.DG].
Commun. Math. Phys. 278, 775 (2008) [arXiv:hep-th/0702033]; Phys. Rev. D 75,
024032 (2007) [arXiv:hep-th/0606160];

[2] J. M. F. Labastida and T. R. Morris, Phys. Lett. B 180, 101 (1986).

[3] J. M. F. Labastida, Phys. Rev. Lett. 58, 531 (1987).

[4] J. M. F. Labastida, Nucl. Phys. B 322, 185 (1989).

[5] C. S. Aulakh, I. G. Koh and S. Ouvry, Phys. Lett. B 173, 284 (1986);
R. R. Metsaev, Phys. Lett. B 309, 39 (1993), Phys. Lett. B 354, 78 (1995), Class.
Quant. Grav. 22, 2777 (2005) [arXiv:hep-th/0412311].

[6] M. Dubois-Violette and M. Henneaux, Lett. Math. Phys. 49, 245 (1999)
[arXiv:math/9907135].

– 32 –
 [7] X. Bekaert and N. Boulanger, Commun. Math. Phys. 245, 27 (2004)
[arXiv:hep-th/0208058]; P. de Medeiros and C. Hull, Commun. Math. Phys. 235,
255 (2003) [arXiv:hep-th/0208155]; JHEP 0305, 019 (2003) [arXiv:hep-th/0303036];
Yu. M. Zinoviev, arXiv:hep-th/0211233, arXiv:hep-th/0304067; A. Campoleoni,
D. Francia, J. Mourad and A. Sagnotti, arXiv:0810.4350 [hep-th].

[8] M. A. Vasiliev, Fortsch. Phys. 52, 702 (2004) [arXiv:hep-th/0401177]; D. Sorokin,

AIP Conf. Proc. 767, 172 (2005) [arXiv:hep-th/0405069]; N. Bouatta, G. Compere
and A. Sagnotti, arXiv:hep-th/0409068; X. Bekaert, S. Cnockaert, C. Iazeolla and
M. A. Vasiliev, arXiv:hep-th/0503128.

[9] T. Branson and A. R. Gover, arXiv:math/0309085.

[10] A.R. Gover and J. Šilhan, Comm. Math. Phys. 284, 291 (2008) [arXiv:0708.3854
[math.DG]].

[11] A. R. Gover, K. Hallowell and A. Waldron, Phys. Rev. D 75, 024032 (2007)
[arXiv:hep-th/0606160].

[12] A. R. Gover, P. Somberg and V. Souček, Commun. Math. Phys. 278, 307 (2008)
[arXiv:math.DG/0606401].

[13] C. Schubert, Phys. Rept. 355, 73 (2001) [arXiv:hep-th/0101036].

[14] A. Barducci, R. Casalbuoni and L. Lusanna, Nuovo Cim. A 35, 377 (1976); L. Brink,
S. Deser, B. Zumino, P. Di Vecchia and P. S. Howe, Phys. Lett. B 64, 435 (1976);
F. A. Berezin and M. S. Marinov, Annals Phys. 104, 336 (1977).

[15] V. D. Gershun and V. I. Tkach, JETP Lett. 29, 288 (1979) [Pisma Zh. Eksp. Teor.
Fiz. 29, 320 (1979)]; P. S. Howe, S. Penati, M. Pernici and P. K. Townsend, Phys.
Lett. B 215, 555 (1988); Class. Quant. Grav. 6, 1125 (1989).

[16] F. Bastianelli and A. Zirotti, Nucl. Phys. B 642, 372 (2002) [arXiv:hep-th/0205182].

[17] F. Bastianelli, O. Corradini and A. Zirotti, Phys. Rev. D 67, 104009 (2003)
[arXiv:hep-th/0211134]; JHEP 0401, 023 (2004) [arXiv:hep-th/0312064].

[18] F. Bastianelli, P. Benincasa and S. Giombi, JHEP 0504, 010 (2005)

[arXiv:hep-th/0503155]; JHEP 0510, 114 (2005) [arXiv:hep-th/0510010].

[19] F. Bastianelli, O. Corradini and E. Latini, JHEP 0702, 072 (2007)

[arXiv:hep-th/0701055].

[20] F. Bastianelli, O. Corradini and E. Latini, JHEP 0811, 054 (2008) [arXiv:0810.0188
[hep-th]].

[21] E. Witten, J. Diff. Geom. 17, 661 (1982).

[22] W. Siegel, Int. J. Mod. Phys. A 6, 3997 (1991).

– 33 –
[23] S. Deser and A. Waldron, Phys. Rev. Lett. 87, 031601 (2001)
[arXiv:hep-th/0102166].

[24] A. Lichnerowicz, Inst. Hautes Etudes Sci. Publ. Math. 10 (1961); Bull. Soc. Math.
France 92, 11 (1964).

[25] T. Damour and S. Deser, Annales Poincare Phys. Theor. 47, 277 (1987).

[26] K. Hallowell and A. Waldron, Nucl. Phys. B 724 (2005) 453 [arXiv:hep-th/0505255].

[27] M. A. Vasiliev, Phys. Lett. B 209, 491 (1988).

[28] C. Duval, P. Lecomte, V. Ovsienko, Ann. Inst. Fourier 49, 1999 (1999),
[arXiv:math.DG/9902032].

[29] C. Duval, V. Ovsienko, Selecta Math. (N.S.) 7, 291 (2001).

[30] A. K. H. Bengtsson, Phys. Lett. B 182, 321 (1986); S. Ouvry and J. Stern, Phys.
Lett. B 177, 335 (1986); M. Henneaux and C. Teitelboim, In *Santiago 1987,
Proceedings, Quantum mechanics of fundamental systems 2* 113-152. (see
Conference Index).

[31] A. Sagnotti and M. Tsulaia, Nucl. Phys. B 682, 83 (2004) [arXiv:hep-th/0311257].

[32] G. Barnich, M. Grigoriev, A. Semikhatov and I. Tipunin, Commun. Math. Phys.

260, 147 (2005) [arXiv:hep-th/0406192].

[33] C. Fronsdal, Phys. Rev. D 18, 3624 (1978).

[34] T. Curtright, Phys. Lett. B 85, 219 (1979).

[35] C. Chevalley and S. Eilenberg, Trans. Amer. Math. Soc. 63, 85 (1948).

[36] J. Burkart and A. Waldron, arXiv:0812.3932 [hep-th].

[37] W. Siegel, “Fields,” arXiv:hep-th/9912205; A. Fotopoulos and M. Tsulaia,

arXiv:0805.1346 [hep-th].

[38] K. B. Alkalaev, M. Grigoriev and I. Y. Tipunin, arXiv:0811.3999 [hep-th].

[39] C. Burdik, A. Pashnev and M. Tsulaia, Nucl. Phys. Proc. Suppl. 102, 285 (2001)
[arXiv:hep-th/0103143].

[40] E. P. Wigner, Ann. Math. 40, 149 (1939).

[41] X. Bekaert and N. Boulanger, Commun. Math. Phys. 271, 723 (2007)
[arXiv:hep-th/0606198]; arXiv:hep-th/0611263.

[42] S. Deser and A. Waldron, Nucl. Phys. B 631, 369 (2002) [arXiv:hep-th/0112182].

[43] N. Marcus and S. Yankielowicz, Nucl. Phys. B 432, 225 (1994)

[arXiv:hep-th/9408116].

– 34 –
[44] N. Marcus, Nucl. Phys. B 439, 583 (1995) [arXiv:hep-th/9409175].

[45] F. Bastianelli and R. Bonezzi, arXiv:0901.2311 [hep-th].

[46] D. Cherney, E. Latini and A. Waldron, arXiv:0901.3788 [hep-th].

[47] D. Cherney, E. Latini and A. Waldron, in preparation.

– 35 –